Real time lane departure warning system based on principal component analysis of grayscale distribution and risk evaluation model

Size: px
Start display at page:

Download "Real time lane departure warning system based on principal component analysis of grayscale distribution and risk evaluation model"

Transcription

1 DOI: /s Real time lae departure warig system based o pricipal compoet aalysis of grayscale distributio ad risk evaluatio model ZHANG Wei wei( 张伟伟 ), SONG Xiao li( 宋晓琳 ), ZHANG Gui xiag( 张桂香 ) State Key Laboratory of Advaced Desig ad Maufacturig for Vehicle Body (Hua Uiversity), Chagsha , Chia Cetral South Uiversity Press ad Spriger Verlag Berli Heidelberg 2014 Abstract: A techology for uiteded lae departure warig was proposed. As crucial iformatio, lae boudaries were detected based o pricipal compoet aalysis of grayscale distributio i search bars of give umber ad the each search bar was tracked usig Kalma filter betwee frames. The lae detectio performace was evaluated ad demostrated i ways of receiver operatig characteristic, dice similarity coefficiet ad real time performace. For lae departure detectio, a lae departure risk evaluatio model based o lastig time ad frequecy was effectively executed o the ARM based platform. Experimetal results idicate that the algorithm geerates satisfactory lae detectio results uder differet traffic ad lightig coditios, ad the proposed warig mechaism seds effective warig sigals, avoidig most false warig. Key words: lae departure warig system; lae detectio; lae trackig; pricipal compoet aalysis; risk evaluatio model; ARM based real time system 1 Itroductio Traffic safety o highways has bee a area of iterest for may years, ad most occurreces of vehicle accidets result from egligece ad drowsiess of the driver. Cosiderig advisory ad warig systems are more easily implemeted tha fully autoomous cotrol, ad machie visio algorithms wi proved popular i target detectio ad trackig [1], it is expected that visio based drivig assistace system ca be used to effectively decrease the umber of accidets. Particularly, a visio based lae departure warig system (LDWS) ca alert drivers to the potetial dager of departure. Curretly, may works have bee doe to develop visio based system for lae detectio ad lae departure warig. However, most of them were implemeted o persoal computer based platforms other tha embedded platforms, ad preseted limitatios i situatios like shadows, varyig illumiatio coditios, curved road model ad algorithm implemetatio complexity. Before extractig lae feature poits, iverse perspective mappig (IPM) [2 4] were usually used to compesate the distortio by perspective, but required olie computatio to perform coordiate trasformatio (ad camera calibratio). Therefore, it is reasoable to cocetrate o developig lae detectio algorithm without applyig IPM i this research, sice computig resource is rare o embedded platforms. Two mai classes of lae feature poit detectio methods have bee proposed. Quite a few of methods commoly used i LDWS [5 6] were feature drive, which was based o the edge fidig. Such methods relied o the itesity discotiuities ad thus suffered from oise effects ad irrelevat feature structure. Ofte i practice, the strogest edges like shadows were ot the lae edges, so that the detected edges do ot ecessarily fit a straight lie or a smoothly varyig model. Hece, some excellet edge operators like Sobel operator [7] ad steerable filter [8 9] were proposed, ad WANG et al [10] used Cay feature operator after applyig the self clusterig algorithm, fuzzy C mea ad fuzzy rules ito automatic brightess compesatio. Alteratively, lae regio aalysis, which ca be modeled as a classificatio problem based o color space ad texture, was usually applied ito lae detectio. I case of color, RGB ad eve CIE space [11 12] were used for classificatio by the mea ad variace of road ad o road classes. However, the apparet color is ot cosistet due to several factors such as illumiatio ad viewig geometry, thus the color statistics of the road ad o road models eed be updated to adapt the chagig coditios. Besides color space, the texture based o the amplitude of gradiet operator at some Foudatio item: Project( ) supported by the Natioal Natural Sciece Foudatio of Chia; Project(2013WK3024) supported by the Sciece ad Techology Plaig Program of Hua Provice, Chia; Project(CX2013B146) supported by the Hua Provicial Iovatio Foudatio for Postgraduate, Chia Received date: ; Accepted date: Correspodig author: SONG Xiao li, Professor, PhD; Tel: ; E mail: jqysxl@hu.edu.c

2 1634 image area has bee used for regio separatio ad used i umarked road i most cases [13 14]. Whe it comes to lae detectio i marked road, the texture aalysis based o Gabor filter was combied with color to improve feature drive lae detectio [15]. However, this approach was of high computatio complexity ad the process speed was 3 frame/s due to the high dimesio of feature vectors. So, it is ecessary to use pricipal compoet aalysis (PCA) to reduce the dimesio ad the proportio of redudacies [16]. Accordigly, to detect the lae robustly, the advatages of feature drive ad lae regio aalysis were both cosidered. After iitial detectio with Hough trasform i the Gauss pyramid of low resolutio image, i some fixed rows lay the search bars, rather tha eedig statistical iformatio of the whole regio of iterest (ROI) for EDF as Ref. [17]. I each search bar, the lae feature poit was located based o PCA of grayscale distributio ad the fitted as parabola or straight lae model. Meawhile, the ceter positio of each search bar was tracked usig Kalma filter to reduce the sesitivity to oise poit whe fittig parabola ad straight lie. I the part of lae departure warig, much literature utilized the agle based lae departure warig algorithm. LEE [17] proposed a lae departure detectio method by symmetry axis ad local maxima of EDF. JUNG ad KELBER [6] ad WANG et al [10] used the same method based o agle relatios of the boudaries. The spatial warig mechaism i Ref. [5] set warig sigals if either d M or d S was larger tha (1/4)I w, which essetially was based o agle threshold of arcta (I w /(2I h )). However, the agle based warig mechaism would cause too much false egatives due to the obvious agle variatio i most situatios. As a result, JUNG ad KELBER [18] started to focus o the lateral offset feature i lae departure. Most LDWS set warig sigals whe vehicle states reach to the predetermied agle ad offset threshold, ad these methods were likely to lead to some false positives ad egatives ad brought some disturbaces ad crucial dagers. I fact, due to the exteral eviromet ad some subjective factors of drivers, vehicle may ievitably depart the curret lae i a short time, which is acceptable i ormal drivig. Therefore, cosiderig the driver s acceptability, road regio was divided ito four zoes with differet dager degrees i this research, ad a lae departure risk evaluatio model based o the lastig time i each zoe was proposed. However, whe it comes to some drowsy but coscious driver, vehicle teds to travel i a serpetie trajectory ad drive ito some dager zoe irregularly, i which situatio the model based o lastig time may fail accordig to the departure degree. So, the model based o frequecy of travellig ito each dager zoe was put forward as a complemet to form a dual warig mechaism. 2 System framework As a typical producer/cosumer model icludig image acquisitio ad processig, the system framework is exhibited i the way of state machie flow i Fig. 1, ad the lae departure moitor ad warig mechaism, actually, should be embedded i each sub state of image processig. The trasitio coditios of each sub state i image processig represet the process results i the previous frames. CAMIF represets the dedicated camera iterface i the embedded hardware platform. I the part of image acquisitio, peripheral camera was set to output VGA format image, ad the CAMIF was cofigured to execute the widow cut fuctio to get the ROI image directly. All the cofiguratios were fiished i the iitial phase ad show i shallow history pseudo state. So did the lae iitial detectio ad locatio for both boudaries i image processig. Fig. 1 Framework of LDWS

3 Lae detectio based o PCA of grayscale distributio 3.1 Pricipal compoet aalysis of matrix Accordig to sigular value decompositio rule, ay real valued M N matrix A ca be writte as T m m p p p p A = U V = T σ 0 v 0 u L u O L (1) 0 p 1 σ 1 T p v p 1 where p=mi(m, ). The matrices U ad V are orthoormal, i.e., U T U=I ad V T V=I, ad so are their colum vectors. ui u j = vi v j = δ ij (2) The sigular values are all o egative ad ca be ordered i descedig order σ 0 σ1 L σ p 1 0 (3) Especially whe C is positive semi defiite, all eigevalues are o egative, ad C ca be costructed as the sum of a umber of outer products, i T T T T T i i ( ) C = a a = AA = U V V U = U U (4) So, Eq. (4) states λi = σ i where λ i is the eigevalues of C. From Eq. (1), aother variat ca be defied as AV = U Av = σ u (5) or j j j The equatio states that the matrix ca take ay basis vector v j ad map it i to a directio of u j with legth σ j, as show i Fig. 2. Fig. 2 Visualizatio of actio of matrix A Whe A is take as a symmetric ad positive semi defiite covariace matrix, after performig the eigevalue decompositio, which is kow as pricipal compoet aalysis, the pricipal orietatio ad magitudes of variatio of the data distributio aroud their mea ca be modeled whe Fig. 2 idicates how the pricipal compoet of the covariace matrices show the pricipal axis u j of the ucertaity ellipse correspodig to the data distributio ad how σ i = λ i show the stadard deviatios alog each axis PCA of grayscale distributio Edges usually exist at boudaries betwee regios with differet itesities, textures, or colors. The slope ad orietatio of grayscale distributio surface, which are described as gray value gradiet i mathematics, are useful for edge detectio. However, takig image derivatives highlights the high frequecies ad hece the oises are amplified, sice sigal to oise ratio is relatively lower i high frequecy sectio. So, a low pass filter was selected to smooth the ROI of image. Typically, the circularly symmetric Gauss filter is a proper oe which is idepedet of orietatio ad separable i recursive computatio. Let g X ad g Y be the first derivatives i x ad y, respectively, the the local mea value ad the covariace for x or y are X X (6) i= j= µ = w( i, j) g ( x + i, y + j ) ad σ = w( i, j)( g ( x + i, y + j ) µ ) XY X X i= j= ( g ( x + i, y + j ) µ ) (7) Y Y where w(i, j) are ormal Gaussia weights, ormalized so that w( i, j ) = 1. The covariace matrix of i= j= grayscale value gradiet is defied as σ xx σ xy C = (8) σ yx σ yy which is symmetric ad positive semi defiite. Takig PCA for matrix C, its eigevalues λ 1 ad λ 2 ca be obtaied, where λ 1 is ot smaller tha λ 2. Meawhile, it is here appropriate to take the orthoormal matrix U of C as vector [cosφ siφ] where φ resets the directio agel relative to x axis of the first pricipal compoet ad its value is 1 2 σ xy arcta, for σ xx σ yy 0 2 σ xx σ yy φ = π, for σ xx σ yy = 0 4 (9) At the areas icludig lae edge, the eigevalue alog the first pricipal axis is greater tha that alog the secod pricipal axis ad its directio agle is perpedicular to that of edge. O the cotrary, i a homogeeous regio, both eigevalues are very small. The experimetal results show that PCA of grayscale pixels reflects the gray value chages, ad this poit is explaied well i Fig. 3(a) as ellipses. I Fig. 3(a), a sub regio of a road image is show i pixels, with the widow size for PCA beig

4 1636 eccetricity of a ellipse, c is the gray ormalizatio factor, p(i, j) presets the gray value of pixel. By itegratig trace(σ) ad gray value ito the defiitio i Eq. (10), it becomes a powerful meas to distiguish the regios with lae edge from the homogeeous oes. As for the gray ormalizatio factor c, T is a real value betwee 0 ad 255. The larger the T is, the more likely the pixel poit is to be a lae feature poit. So, the lae feature poit extractio algorithm (PCA EA) based o PCA of gray distributio is available with some threshold TG through OTSU [19]. 3.3 Lae trackig usig Kalma filter I order to extract lae edge iformatio efficietly, thirtee search bars, whose width is about 2 times of that of lae, were set i several fixed rows for both sides of curret laes, as show i Fig. 4. Due to camera perspective, the bar should be shorter at the top of ROI (far field), ad loger at the bottom(ear field), as the red bar shows. I iitial state, a Hough trasform was Fig. 3 Display of PCA of gray value chages as ellipses (a) ad trace value (b) 5 5 pixels. Complemetally, the trace of auto correlatio matrix for each pixel is described i Fig. 3(b), i which trace values reach their peaks at the lae edges, with the origi of coordiates locatig i the top left poit of Fig. 3(a). I reality, λ2 is merely small i compariso to λ1 due to oise, makig the ellipses alog lae edge be evolved ito some lies as show i blue rectagle i Fig. 3(a), while the others i the homogeeous regios arrow dow to some poits i orage rectagle because of λ1 ad λ2 beig small. More tha that, the gray average value o lae edges is much greater tha that of other pixels i a search bar, which ca ehace the detectio robustess. With the help of the elliptical visualizatio of covariace matrices, the behavior of grayscale distributio has bee aalyzed. The agle f oly poits out the directio of gray value chages, without idicatig the extet of gray value chages. I order to locate the lae feature poit exactly, the edge stregth of each pixel is defied as l l 1 T = (1-2 )trace( å ) p (i, j ) = 1 e 2 p (i, j ) c l1 c With e = l12 - l22 l1 ad it presets (10) the Fig. 4 Detectio result based o PCA of grayscale distributio ad Kalma trackig i day time (a), raiy day (b) ad ight time (c)

5 1637 implemeted i the Gauss pyramid of low resolutio image to reduce computig complexity, ad the the itersectio poit of the fixed row ad Hough lie was take as the system state vector for trackig. Suppose that the positio of lae feature poit ca be obtaied every costat time T (30 frame/s) ad laes move i lie with costat speed, so the liear discrete model of each lae feature poit ca be defied as X = A X + w Zk = H k X k + v k k + 1 k + 1, k k k where X = [ x x & ] k k k (11) ad it represets the positio ad velocity of lae feature poit alog X axis at time k; 1 T A k+1,k is the trasitio matrix ad A k + 1, k = 0 1 i the track algorithm; Z k is the measure output; H k is the observatio model that maps the true space to the observed space ad H k =[1 0]; w k ad v k are kow as process oise ad measuremet oise, all of which coform to Gauss distributio. I Fig. 4, the ceter of each detectio bar was tracked usig Kalma filter. The blue pixels are cadidate lae feature poits determied by PCA of grayscale distributio ad edge stregth. Cosiderig the gray average value of blue pixels was higher tha that of red pixels of about 17%, the ivalid cadidate feature poit ca be easily igored, thus leavig the valid feature poit of max value T aroud by red circles. The feature poits of each lae were subsequetly fitted as dashed/ solid parabola or straight lie accordig to the trackig result. As demostrated i Fig. 4, the whole detectio algorithm ca get satisfactory lae boudary uder various illumiatio coditios. Followig Ref. [2], the groud truth pixels about lae markigs ca be geerated efficietly through TS images. As the umber of pixels correspodig to lae markigs is small (about 1% 5%), so oly the left part of ROC makes sese to lae detectio algorithm, which is the reaso DSC (C DS ) is used for complemetary aalysis. DSC is defied as the harmoic mea of precisio rate (PR) ad recall rate (also amed TPR) at differet threshold T G N = + = C R R N N TP with R P DS P TP TP + FP C DS 2 N TP = 2 N + N + N TP FP FN (13) DSC curves imply the overlap betwee the groud truth pixels ad the extractio feature poits provided by differet algorithms, typically takig the 43rd percetile local threshold (PLT), media symmetrical local threshold (MLT SLT)[20] as compariso object for the PCA EA. The valuatio is performed alog a referece database [21] icludig su reflectio, o uiform shadows ad the result is show i Fig. 5. I Fig. 5, it ca be cocluded that the performace of MLT SLT is close to that of PCA EA from ROC 3.4 Lae detectio performace evaluatio The essece of lae detectio is classifyig the image pixels ito two classes: lae feature poits ad o lae markig poits, which make it be a classifier evaluatio problem. Classificatio accuracy is the usual idex to evaluate the classifier performace ad whe it comes to the biary classificatio problem, two classical evaluatio tools are used: receiver operatig characteristic (ROC) ad dice similarity coefficiet (DSC). ROC curves are obtaied by plottig TP rate (TPR) versus FP rate (FPR) at differet local gray or gradiet threshold T G, which are defied as R TP N N = = N N N N TP FP, R FP TP + FN FP + TN (12) where N TP correspods to the umber of true positive, N TN to true egative, N FP to false positive, FN to false egative. So, the larger the area uder the ROC curve, the better the detectio ad classificatio algorithm. Fig. 5 ROC (a) ad DSC (b) curves for lae feature poit extractio algorithm

6 1638 curves, ad the latter whose true positive rate reaches to 0.95 at most time is better tha the other two. Meawhile, the DSC curves show that PCA EA has the biggest peaks tha MLT SLT ad PLT at optimal threshold 45 aroud, ad the width of peak o PCA EA is larger tha MLT SLT, which meas the PCA EA is less sesitive o the selectio of threshold value ad has strog robustess. Figure 6 shows the digital waveform display of image acquisitio ad trasfer to processor, startig at blue colum bar 1 ad edig at blue colum 2. The digital sigals are show i white waveform. Table 1 describes the real time performace of the mai compoets of algorithm, idicatig the satisfactory requiremets of 30 frame/s. Fig. 7 Partitio of road regio without itetioal directio sigal, the biggest dager occurs. Some warig sigals should be set accordig to the dager degree. Fig. 6 Digital waveform of image acquisitio through CAMIF Table 1 Real time performace Algorithm Image acquisitio Lae feature poits locatio usig PCA Bar trackig usig Kalma filter UI update Total 4 Lae departure detectio algorithm Time/s I geeral, the dagerous lae departure situatios result from the vehicle gettig close to or eve crossig the lae boudaries, but they ca be classified ito two differet forms, i differet lastig time ad i differet frequecies. Therefore, road was divided ito some regios of differet security levels firstly ad the the risk evaluatio models based o the lastig time ad frequecy were proposed to detect these two departure situatios respectively sice the relative lateral offset l 0 /W ca be got usig the liear part of the lae boudary model [18]. Suppose the width of sigle lae is b, the zoes Z 1, Z 2, Z 3, Z 4 were respectively, defied as safety zoe, trasitio zoe, alert zoe, dager zoe as show i Fig. 7. Safety zoe is of the highest security, whose width is b 2 b 1. Trasitio zoe with the width of b 1 is ext to the ier side of the lae lies ad has lower security, so vehicle ruig i this zoe with log lastig time ad ruig ito this zoe frequetly meas some potetial safety hazard. Vehicle beig i alter zoe of width b 2 meas pressig lie ad has bigger potetial safety hazard tha that of trasitio zoe. Oce the vehicle rus over the alter zoe ad ito dager zoe 4.1 Defiitio of risk evaluatio model The four zoes have their correspodig security level: Z 1 >Z 2 >Z 3 >Z 4, ad the risk evaluatio fuctio based o lastig time is expressed as: R t =f(z i, t) where i=1, 2, 3, 4, R t is the risk coefficiet, Z i is the divided zoe ad t is the correspodig lastig time i zoe Z i. The risk evaluatio model based o lastig time is defied i Fig. 8(a). Similarly, the risk evaluatio coefficiet based o frequecy R N ca be modeled as show i Fig. 8(b), ad N i (i=1, 2) is trasitio threshold for N. Fig. 8 Risk evaluatio model based o lastig time (a) ad frequecy (b) 4.2 Risk comprehesive evaluatio algorithm At a give piece of time, the timestamp ad frame stamp of vehicle drivig ito or out of zoe Z i (i=

7 1639 2, 3, 4) ca be got ad so is the lastig time for each zoe. It is worth otig that: 1) The actual lastig time for trasitio zoe was a combiatio of the time drivig ito trasitio zoe, alter zoe ad dager zoe; 2) The actual lastig time for alter zoe was a combiatio of the time drivig ito alter zoe ad dager zoe; 3) The lastig time for zoe Z i (i=2, 3, 4) was reset to zero whe vehicle just ruig i zoe Z 1. I the lae departure course, the risk evaluatio coefficiet based o lastig time for every zoe was computed oce vehicle touch ay zoe boudary, ad the the average of all the coefficiet values was take as the actual risk evaluatio coefficiet. Similarly, the risk evaluatio coefficiet based o frequecy ca be computed. So, the risk comprehesive evaluatio value ca be achieved ad updated from the bigger of the two coefficiets whe vehicle touch ay zoe boudary. Furthermore, some warig sigs should be issued oce the evaluatio value exceeds threshold R th (about 0.3) which is obtaied by empirical aalysis ad slightly chages with vehicle speed, ad the the lastig time ad frequecy of each zoe was reset to zero. Figure 9 shows a example of lae departure i which vehicle travel at speed of km/h. Providig that b 1 =b/10, b 2 =b/7, the gree, blue, red lie correspodig to their couterparts i Fig. 7 are show i Fig. 9 where lae width is described as a stadard uit. The black curve alog 1800 m travelig distace represets vehicle s relative lateral offset l 0 /W, which was recorded at 30 frame/s i 92 s. It is clear that vehicle experieced several severe left ad right lae departure evets. I real world experimet parameters of lastig time ad frequecy were set as: t 0 =1 s, t 1 =10 s, t 2 =12 s, N 1 =16, N 2 =20. Table 2 shows a example of the departure lastig time for each zoe i the first 30 s time rage which is overlaid by gree trasparet layer i Fig. 9. I half a miute occur four left departure evets ad oe right departure evet. For example, i the 1st left departure evet, after ruig ito Z 4 with 1.71 s, R t reaches to 0.694, which is got from: [( t 0 )/(t 2 t 0 )+( t 0 )/ (t 1 t 0 )+1]/3. The frequecies i half a miute for vehicle ruig ito zoe Z 2, Z 3, Z 4 are 5, 4, 2 respectively, so just for the whole experimet R N reaches to 0.675, which is (5 2/N /N 2 +1)/3. Actually warig sigs issued at the 1st, 5th case. Whe it comes to lae departure mechaism based o agle symmetry axis ad local maxima [17], the detectio mechaism issued warig sigals every time whe vehicle s relative lateral offset cross the blue lae boudaries i Fig. 9, which lead to some disturbace to the driver. Complemetally, after icreasig agle threshold slightly, the detectio mechaism teds to miss departure detectio whe the relative lateral offset cross ito Z 4, brigig much potetial risk without warig. 5 Experimets Fig. 9 vehicle relative lateral offset i some severe lae departure evets A hardware platform with a embedded warig system was implemeted. Sice ARM processors are characterized by their cost effective, low power capabilities, ad high performace, the core processor S3C6410 was selected, which was ARM11 based 32 bit RISC CPU of 667 MHz. The power cosumptio is about 1.5 W esured by a high efficiecy DC/DC coverter with wide 4 V to 60 V iput voltage rage. The CMOS sesor OV9650 was cofigurable to extract the VGA images i 30 frame/s through stadard SCCB iterface. Furthermore, to optimize memory utilizatio, the search bar was represeted i ru legth ecodig; some variables i courses like Gauss low pass filterig ca recursively be accumulated ad restored i memory buffers, istead of set a block of memory to read/write Table 2 Lastig time i lae departure experimet Lastig time(s)/r t Z 2 /R t Z 3 /R t Z 4 /R t Z 3 /R t Z 2 /R t 1st (left) 0.76/0 0.80/ / / / d (left) 0.70/0 0.38/ / / / rd (left) 0.65/0 0.84/ / / / th (left) 0.12/0 0/0 0/0 0/0 0/0 5th (right) 0.21/0 0.24/0 1.92/ / /0.448

8 1640 them i every step. With the help of DMA chael ad the widow cut fuctio i CAMIF of S3C6410, the whole processig performace for ROI of a image size of is o less tha 25 frame/s. I highway experimets, three videos captured i daytime, as well as two videos captured i ight time, were radomly selected icludig some complex road types ad high dyamic rage scees, which ca be see from the sapshots i Fig. 10. The average detectio rate of lae detectio ca be more tha 99% uder differet traffic ad lightig coditios. Table 3 shows the statistics of the detectio rate over the five videos, ad the experimet results of lae detectio ad lae departure detectio are elaborated respectively, icludig the high dyamic brightess situatios i aother 16 videos. 5.1 Frame based daytime evaluatio I the first video referred i the ormal situatio, the lae markigs were well paied ad o u favorable pheomea occurred, so lae successful detectio rate ca reach up to 100%. The warig sigals were also correctly set. The warig mechaism was issued whe the risk comprehesive evaluatio value is more tha 0.31 cosiderig vehicle speed is about 80 km/h. The system issued warig sigal at the 502d frame ad stopped at the 563rd frame. Accordig to itermediate record, this warig is triggered through warig mechaism based o lastig time, because vehicle rus ito the dager zoe with 0.8 s. The secod video depicted a lae chagig situatio. Whe chagig laes, the curret lae lie to cross was vertical to image bottom ad the other two lae lies was almost symmetrical to the vertical lie, so by meas of this poit, the successful detectio ad trackig rate ca be achieved up to 99.88%. The system started warig at the 597th ad 1334th frame. The approachig situatio, which meas the host vehicle was approachig the frot vehicle ad the overtakes the frot vehicle, was demostrated i the third video. Ufortuately, lae boudaries were partly occluded by the vehicle i frot, thus resultig i usatisfactory detectio ad trackig results. This problem ca be solved by adjustig the parameters of lae chagig model. Fig. 10 Sapshots of videos i differet weather coditios: (a) Day time (b) Raiy time (c) Night time Table 3 Frame based successful detectio rate i differet situatios Time Frame Successful Failed Detectio rate/% Normal Lae chage Approachig Total No street light Well lighted Total High dyamic brightess Day Night Day & ight Video Total

9 1641 Table 4 Miss detectio rate ad false warig rate of frames i diverse situatios Groud truth Detectio result Straight lae Left curved lae Right curved lae Total No departure No departure 4503 frames (100%) 4257 frames (92.71%) 3745 frames (95.44%) frames (96.05%) Departure 0 frame 335 frames (7.29%) 179 frames (4.56%) 514 frames (3.95%) Departure No departure 0 frame 27 frames (1.26%) 165 frames (6.53%) 192 frames (3.02%) Departure 1709 frames (100%) 2102 frames (98.74%) 2359 frames (93.47) 6170 frames (96.98%) 5.2 Frame based ighttime evaluatio The 4th video shows o street light situatio i ighttime. I this case, the oly light source was from the host vehicle, ad the system was susceptible to the head light of vehicles i opposite directio. So, the successful detectio rate dropped to 98.64% sice the detectio of lae markigs was difficult i this case. However, the system ca still set warig sigals correctly at the 375th ad 962d frames. I the well lighted situatio, the system successfully detected ad tracked the lae with high efficiecy. After a chai of mild departure itesios i a short time, the system triggered the warig sigals correctly whe R N was more tha 0.33 cosiderig vehicle speed is about 70 km/h. 5.3 High dyamic brightess evaluatio I additio to evaluatig the detectio accuracy o some videos, some fragmets of high dyamic brightess situatios, like shadows, direct sulight, suddely pullig out of the tuel, were further validated i the other 16 videos. Owig to the strategy of lae trackig, the system obtaied a satisfactory detectio rate up to 98.86%. Whe it comes to direct sulight for a log time, the detectio performace degeerated more or less. This problem was ufamiliar ad ca be solved by automatic exposure cotrol, automatic gai cotrol ad icreasig the umber of fixed rows i ear field, sice image i ear field was much less likely to be affected by direct sulight. 5.4 Evet based warig evaluatio I order to validate the algorithm for detectig lae departure evets, videos i both day time ad ight time uder differet situatios were segmeted for departure evet. Furthermore, these evets were classified ito three categories: i straight laes, i left curved laes ad i right curved laes. To demostrate the performace of the proposed algorithm, the miss detectio rate ad the false warig rate were obtaied for all kid of video segmets. The detectio results are listed i Table 4. For departure evets from straight lae, the missdetectio rate ad the false warig rate are 0%. For departure evets from left curved lae, the missdetectio rate is 1.26% ad the false warig rate is 7.29%, which is higher tha that of right curved lae of 4.56%. This is due to the deviatio of the host vehicle from the ceter of lae, especially whe host vehicle with driver o the left side travel o the left curved laes it is accustomed to deviatig to the left side lae boudary. Noetheless, the whole mechaism still provide some miss detectios ad false alarms maily because of some road with sharp curved directio ad large lae fittig error o roads with severe oisy factors. However, the system possesses satisfactory performace sice the miss detectio rate ad false warig rate are o more tha 4% i etire situatios ad the whole performace ca be improved by more robust lae detectio method ad cosiderig curved model ito road regio partitio. 6 Coclusios 1) Aalysis results o auto geerated TS images idicate that the proposed lae detectio algorithm provides a accurate locatio ad fit to lae boudaries, showig better results tha other similar algorithms. 2) The dual mechaism warig algorithm based o lastig time ad frequecy is valid to detect uiteded lae departure uder various coditios, ad its missdetectio rate ad false warig rate are both acceptable. 3) The average processig speed is o less tha 25 frame/s o embedded platform with S3C6410 processor. Future work will adjust the warig trigger timig to get more comfortable drivig experiece ad some sophisticated strategies like matched filter would be itegrated ito the lae detectio approach to get more accurate lae fittig results. Refereces [1] JEFFREY D, ANTHONY G. Autoomous drivig A practical roadmap [C]// 2010 SAE Covergece, Detroit, Michiga, Uited States: SAE, 2010: [2] AMOL B, MONSON H, MARK T. A ovel lae detectio system with efficiet groud truth geeratio [J]. IEEE Trasactios o Itelliget Trasportatio Systems, 2012, 13(1): [3] STEPHAN S, SARATH K, ALEN A, GAMINI D. Robust lae detectio i urba eviromets [C]// Proceedigs of the 2007 IEEE/RSJ Iteratioal Coferece o Itelliget Robots ad Systems. Sa Diego, CA, USA: IEEE Press, 2007: [4] MOHAMED A. Real time detectio of lae markers i urba streets

10 1642 [C]// IEEE Itelliget Vehicles Symposium. Eidhove, The Netherlads: IEEE Press, 2008: [5] HSIAO P, YEH C, HUANG S, FU L. A portable visio based real time lae departure warig system: day ad ight [J]. IEEE Trasactios o Vehicular Techology, 2009, 58(4): [6] JUNG C, KELBER C. Lae followig ad lae departure usig a liear parabolic model [J]. Image ad Visio Computig, 2005, 23: [7] LIU Xi, XU Xi, DAI Bi. Visio based log distace lae perceptio ad frot vehicle locatio for full autoomous vehicles o high roads [J]. Joural of Cetral South Uiversity, 2012, 19: [8] JOEL C, MOHAN M. Video based lae estimatio ad trackig for driver assistace: Survey, system, ad evaluatio [J]. IEEE Trasactios o Itelliget Trasportatio Systems, 2006, 7(1): [9] WANG Y, NAIM D, ALIN A. A ovel system for robust lae detectio ad trackig [J]. Sigal Processig, 2012, 92: [10] WANG J, LIN C, CHEN S. Applyig fuzzy method to visio based lae detectio ad departure warig system [J]. Expert Systems with Applicatios, 2010, 37: [11] CHENG H, JENG B, TSENG P, FAN K. Lae detectio with movig vehicles i the traffic scees [J]. IEEE Trasactios o Itelliget Trasportatio Systems, 2006, 7(4): [12] MA Chao, XIE Mei. A method for lae detectio based o color clusterig [C]// 2010 Third Iteratioal Coferece o Kowledge Discovery ad Data Miig. Phuket, Tailad: IEEE Press, 2010: [13] BUI T, NOBUYAMA E. A local soft votig method for texture based vaishig poit detectio from ustructured road images [C]// SICE Aual Coferece. Akita, Japa: SICE, 2012: [14] RASMUSSEN C, KORAH T. O vehicle ad aerial texture aalysis for visio based desert road followig [C]// Proceedigs of IEEE Computer Society Coferece o Computer Visio ad Patter Recogitio (CVPR). Sa Diego, CA, USA: IEEE Press, 2005: [15] RODOLFO T, MIGUEL T. A compariso of gradiet versus color ad texture aalysis for lae detectio ad trackig [C]// The 6th Robotics Symposium i Lati America (LARS). Valparaiso, Chile: IEEE Press, 2006: 1 6. [16] MANISH H, LIU J, JOHN F. Image texture aalysis: Methods ad comparisos [J] Chemometrics ad Itelliget Laboratory Systems, 2004, 72: [17] LEE J. A machie visio system for lae departure detectio [J]. Computer Visio ad Image Uderstadig, 2002, 86: [18] JUNG C, KELBER C. A lae departure warig system usig lateral warig usig u calibrated camera [C]// The 8th Iteratioal IEEE Coferece o Itelliget Trasportatio Systems. Viea, Austria: IEEE Press, 2005: [19] XU X, XU S, JIN L, SONG E. Characteristic aalysis of Otsu threshold ad its applicatios [J]. Patter Recogitio Letters, 2011, 32: [20] POLLARD E, GRUYER D, TAREL J, IENG S, CORD A. Lae markig extractio with combiatio strategy ad comparative evaluatio o sythetic ad camera images [C]// The 14th Iteratioal IEEE Coferece o Itelliget Trasportatio Systems. Washigto, DC, USA: IEEE Press, 2011: [21] VEIT T J, NICDLLE P. ROMA (ROadMAkigs). [EB/OL]. de doees de referece/article/ roma road markigs (Edited by DENG Lü xiag)

SPH4U Transmission of Waves in One and Two Dimensions LoRusso

SPH4U Transmission of Waves in One and Two Dimensions LoRusso Waves travelig travellig from oe medium to aother will exhibit differet characteristics withi each medium. Rules A wave of fixed frequecy will have a shorter wavelegth whe passig from a fast medium to

More information

THE EFFECTS OF COUPLED INDUCTORS IN PARALLEL INTERLEAVED BUCK CONVERTERS

THE EFFECTS OF COUPLED INDUCTORS IN PARALLEL INTERLEAVED BUCK CONVERTERS THE EFFECTS OF COUPED INDUCTORS IN PARAE INTEREAVED BUCK CONVERTERS Paul Seria, David Fi ad Geoff Walker pseria@itee.uq.edu.au School of Iformatio Techology ad Electrical Egieerig Uiversity of Queeslad

More information

THE LATENT DEMAND METHOD

THE LATENT DEMAND METHOD THE LATENT DEMAND METHOD Bruce W. Ladis, Russell M. Otteberg, Vekat R. Vattikuti SCI, Ic., 18115 U.S. Highway 41North, Suite 600, Lutz, FL 33549, USA Email: bladis@sciworld.et Travel patters i a metropolita

More information

number in a data set adds (or subtracts) that value to measures of center but does not affect measures of spread.

number in a data set adds (or subtracts) that value to measures of center but does not affect measures of spread. Lesso 3-3 Lesso 3-3 Traslatios of Data Vocabulary ivariat BIG IDEA Addig (or subtractig) the same value to every umber i a data set adds (or subtracts) that value to measures of ceter but does ot affect

More information

University of California, Los Angeles Department of Statistics. Measures of central tendency and variation Data display

University of California, Los Angeles Department of Statistics. Measures of central tendency and variation Data display Uiversity of Califoria, Los Ageles Departmet of Statistics Statistics 13 Istructor: Nicolas Christou Measures of cetral tedecy Measures of cetral tedecy ad variatio Data display 1. Sample mea: Let x 1,

More information

Load Calculation and Design of Roller Crowning of Truck Hub Bearing

Load Calculation and Design of Roller Crowning of Truck Hub Bearing Sed Orders for Reprits to reprits@bethamsciece.ae 106 The Ope Mechaical Egieerig Joural, 2015, 9, 106-110 Ope Access Load Calculatio ad Desig of Roller Crowig of Truck Hub Bearig Xitao Xia,1, Shujig Dog

More information

The structure of the Fibonacci numbers in the modular ring Z 5

The structure of the Fibonacci numbers in the modular ring Z 5 Notes o Number Theory ad Discrete Mathematics Vol. 19, 013, No. 1, 66 7 The structure of the iboacci umbers i the modular rig Z J. V. Leyedekkers 1 ad A. G. Shao 1 aculty of Sciece, The Uiversity of Sydey

More information

ASSESSMENT SCORING SYSTEM OF ROAD SAFETY INFRASTRUCTURE

ASSESSMENT SCORING SYSTEM OF ROAD SAFETY INFRASTRUCTURE ASSESSMENT SCORING SYSTEM OF ROAD SAFETY INFRASTRUCTURE Case Study of Mihag District i Shaghai, Chia Yue Wag, Daocheg Su 2 School of Traffic ad Trasportatio Egieerig, Togi Uiversity, Shaghai, Chia ABSTRACT

More information

Human-Robot Interaction: Group Behavior Level

Human-Robot Interaction: Group Behavior Level SPECOM'006, St. Petersburg, 5-9 Jue 006 Huma-Robot Iteractio: Group Behavior Level Lev Stakevich, Deis Trotsky Sait Petersburg State Techical Uiversity 9, Politechicheskaya, St. Petersburg, 955, Russia

More information

Research Article. Relative analysis of Taekwondo back kick skills biomechanics based on 3D photograph parsing. Mingming Guo

Research Article. Relative analysis of Taekwondo back kick skills biomechanics based on 3D photograph parsing. Mingming Guo Available olie www.jocpr.com Joural of Chemical ad Pharmaceutical Research, 203, 5(2):64-69 Research Article ISSN : 0975-734 CODEN(USA) : JCPRC5 Relative aalysis of Taekwodo back kick skills biomechaics

More information

Footwork is the foundation for a skilled basketball player, involving moves

Footwork is the foundation for a skilled basketball player, involving moves The Complete Book of Offesive Basketball Drills Basic Footwork ad Cuts Drills 2 Equipmet 16 chairs (or less) Persoel The etire team Basic Footwork ad Cuts Drills Two coaches How to Ru the Drill Five feet

More information

A SECOND SOLUTION FOR THE RHIND PAPYRUS UNIT FRACTION DECOMPOSITIONS

A SECOND SOLUTION FOR THE RHIND PAPYRUS UNIT FRACTION DECOMPOSITIONS Fudametal Joural of Mathematics ad Mathematical Scieces Vol., Issue, 0, Pages -55 This paper is available olie at http://www.frdit.com/ Published olie November 9, 0 A SECOND SOLUTION FOR THE RHIND PAPYRUS

More information

SYMMETRY AND VARIABILITY OF VERTICAL GROUND REACTION FORCE AND CENTER OF PRESSURE IN ABLE-BODIED GAIT

SYMMETRY AND VARIABILITY OF VERTICAL GROUND REACTION FORCE AND CENTER OF PRESSURE IN ABLE-BODIED GAIT SYMMETRY AND VARIABILITY OF VERTICAL GROUND REACTION FORCE AND CENTER OF PRESSURE IN ABLE-BODIED GAIT Yu Wag ad Kazuhiko Wataabe Graduate School of Educatio, Hiroshima Uiversity, Hiroshima, Japa, wagyu@hiroshima-u.ac.jp

More information

DFC NIST DIGITAL MASS FLOW CONTROLLERS. DFC with optional LCD readout PROG RS485. Programmable Mass Flow Controller with Digital Signal Processing

DFC NIST DIGITAL MASS FLOW CONTROLLERS. DFC with optional LCD readout PROG RS485. Programmable Mass Flow Controller with Digital Signal Processing Programmable Mass Flow Cotroller with Digital Sigal Processig Microprocessor drive digital flow cotrollers allow oe to program, record, ad aalyze flow rates of various gases with a computer via a RS-485

More information

ICC WORLD TWENTY ( WORLD CUP-2014 )- A CASE STUDY

ICC WORLD TWENTY ( WORLD CUP-2014 )- A CASE STUDY INTERNATIONAL JOURNAL OF MATHEMATICAL SCIENCES AND APPLICATIONS Volume 5 Number 1 Jauary-Jue 215 ICC WORLD TWENTY2-214 ( WORLD CUP-214 )- A CASE STUDY Bhavi Patel 1 ad Pravi Bhathawala 2 1 Assistat Professor,

More information

Catenary Analysis and Calculation Method of Track Rope of Cargo Cableway with Multiple Loads

Catenary Analysis and Calculation Method of Track Rope of Cargo Cableway with Multiple Loads MATEC Web of Cofereces 8, 01008 (016) DOI: 10.1051/ mateccof/016801008 DME 016 Cateary Aalysis ad Calculatio Method of Track Rope of Cargo Cableway with Multiple Loads Jia Qi a, Ju Che, Liag Qiao, Jiacheg

More information

This report presents an assessment of existing and future parking & traffic requirements, for the site based on the current development proposal.

This report presents an assessment of existing and future parking & traffic requirements, for the site based on the current development proposal. CR166916b02 20 Jue 2017 Melida Dodso Melida Dodso Architects PO Box 5635 Hughes ACT 2605 Email: mdodso@melidadodsoarchitects.com.au Dear Melida, Project: Dickso Apartmets Lowrie St Re: Statemet o Parkig

More information

SPEED OF SOUND MEASUREMENTS IN GAS-MIXTURES AT VARYING COMPOSITION USING AN ULTRASONIC GAS FLOW METER WITH SILICON BASED TRANSDUCERS

SPEED OF SOUND MEASUREMENTS IN GAS-MIXTURES AT VARYING COMPOSITION USING AN ULTRASONIC GAS FLOW METER WITH SILICON BASED TRANSDUCERS SPEED OF SOUND MEASUREMENTS IN GAS-MIXTURES AT VARYING COMPOSITION USING AN ULTRASONIC GAS FLOW METER WITH SILICON BASED TRANSDUCERS Torbjör Löfqvist, Kęstutis Sokas, Jerker Delsig EISLAB, Departmet of

More information

A Comparison of MOEA/D, NSGA II and SPEA2 Algorithms

A Comparison of MOEA/D, NSGA II and SPEA2 Algorithms Iteratioal Joural of Egieerig Treds ad Applicatios (IJETA) Volume 4 Issue 6, Nov-Dec 7 RESEARCH ARTICLE A Comparo of MOEA/D, ad Algorithms Rajai [1], Dr. Aad Prakash [] Asstat Professor of Computer Sciece

More information

Range St. Dev. n Mean. Total Mean % Competency. Range St. Dev. n Mean. Total Mean % Competency

Range St. Dev. n Mean. Total Mean % Competency. Range St. Dev. n Mean. Total Mean % Competency CSWE Core Competecies - MSW 2013 MSW Program Dual Degree MSW/JD Program* *Due to the curret limited umber of Dual Degree studets, the MSW outcomes data represets outcomes for both programs. 1 : Idetify

More information

Policy sensitivity analysis of Karachi commuters

Policy sensitivity analysis of Karachi commuters Policy sesitivity aalysis of Karachi commuters A. Q. Memo 1 & K. Sao 2 1 Departmet of Civil Egieerig, NED Uiversity of Egieerig ad Techology, Karachi, Pakista 2 Departmet of Civil ad Evirometal Egieerig,

More information

Electrooculogram Signals Analysis for Process Control Operator Based on Fuzzy c-means

Electrooculogram Signals Analysis for Process Control Operator Based on Fuzzy c-means (IJACSA) Iteratioal Joural of Advaced Computer Sciece ad Applicatios, Vol. 6, No. 9, 5 Electrooculogram Sigals Aalysis for Process Cotrol Operator Based o Fuzzy c-meas Jiagwe Sog Raofe Wag* Guaghua Zhag

More information

DEPARTMENT OF ELECTRICAL AND COMPUTER ENGINEERING, THE UNIVERSITY OF NEW MEXICO ECE-238L: Computer Logic Design Fall Notes - Chapter 6.

DEPARTMENT OF ELECTRICAL AND COMPUTER ENGINEERING, THE UNIVERSITY OF NEW MEXICO ECE-238L: Computer Logic Design Fall Notes - Chapter 6. PARTMNT OF LCTRICAL AN COMPUTR NGINRING, TH UNIVRSITY OF NW MXICO C-238L: Computer Logic esig Fall 23 RGISTRS: Notes - Chapter 6 -bit Register: This is a collectio of '' -type flip flops, where each flip

More information

Obstacle Avoidance for Visually Impaired Using Auto-adaptive Thresholding on Kinect s Depth Image

Obstacle Avoidance for Visually Impaired Using Auto-adaptive Thresholding on Kinect s Depth Image Obstacle Avoidace for Visually Impaired Usig Auto-adaptive Thresholdig o Kiect s Depth Image Muhamad Risqi Utama Saputra, Widyawa, Paulus Isap Satosa Departmet of Electrical Egieerig ad Iformatio Techology

More information

Daytime Road Marker Recognition Using Grayscale Histogram and Pixel Values

Daytime Road Marker Recognition Using Grayscale Histogram and Pixel Values Vol.8/No.1 (2016) INTERNETWORKING INDONESIA JOURNAL 11 Daytime Road Marker Recogitio Usig Grayscale Histogram ad Pixel Values Zamai Md Sai, Hadhrami Abd Ghai, Rosli Besar, W.S. Loi, Member IEEE Abstract

More information

Patrick Boston (Leeds University) and Mark Chapman (Edinburgh University)

Patrick Boston (Leeds University) and Mark Chapman (Edinburgh University) Iversio o requecy-depedet AVO data or luid properties Patric Bosto (Leeds Uiversity) ad Mar Chapma (Ediburgh Uiversity) Summary A umber o methods to estimate the requecy-depedece o relectivity have bee

More information

WIND TUNNEL EXPERIMENT ON THE EFFECT OF WIND ON SMOKE EXHAUST SYSTEMS FOR A HIGH RISE BUILDING

WIND TUNNEL EXPERIMENT ON THE EFFECT OF WIND ON SMOKE EXHAUST SYSTEMS FOR A HIGH RISE BUILDING WIND TUNNEL EXPERIMENT ON THE EFFECT OF WIND ON SMOKE EXHAUST SYSTEMS FOR A HIGH RISE BUILDING K.C. Chug Mechaical Egieerig, Natioal Yuli Uiversity of Sciece ad Techology Touliu City, Yuli 4, Taiwa ABSTRACT

More information

Available online at ScienceDirect. Procedia Engineering 113 (2015 )

Available online at  ScienceDirect. Procedia Engineering 113 (2015 ) Available olie at www.sciecedirect.com ScieceDirect rocedia Egieerig 3 (205 ) 30 305 Iteratioal Coferece o Oil ad Gas Egieerig, OGE-205 The twi spool efficiecy cotrol Michuri A.I. a *, Avtoomova I.V. a

More information

P h o t o g r a p h i c L i g h t i n g ( 1 1 B )

P h o t o g r a p h i c L i g h t i n g ( 1 1 B ) 9 1 5 9 P h o t o g r a p h i c L i g h t i g ( 1 1 B ) 30S/30E/30M A Photography Course 9 1 5 9 : P h o t o g r a p h i c L i g h t i g ( 1 1 B ) 3 0 S / 3 0 E / 3 0 M Course Descriptio This course focuses

More information

The Analysis of Bullwhip Effect in Supply Chain Based on Strategic Alliance

The Analysis of Bullwhip Effect in Supply Chain Based on Strategic Alliance The Aalysis of Bullwhip Effect i Supply Chai Based o Strategic Alliace Rui Xu, Xiaoli Li, Xiaomi Sog, Gag Liu School of Busiess, Hubei Uiversy, Wuha, 430062, P. R. Chia, wdxj826@126.com Abstract. Bullwhip

More information

Modelling Lane Changing Behaviour of Heavy Commercial Vehicles

Modelling Lane Changing Behaviour of Heavy Commercial Vehicles Modellig Lae Chagig Behaviour of Heavy Commercial Vehicles Modellig Lae Chagig Behaviour of Heavy Commercial Vehicles Sara Moridpour 1, Majid Sarvi 2, Geoff Rose 3 1 Post Graduate Studet, Moash Uiversity,

More information

Characteristics of CNG Bubbles in Diesel Flow under the Influence of the Magnetic Field

Characteristics of CNG Bubbles in Diesel Flow under the Influence of the Magnetic Field Joural of Applied Fluid Mechaics, Vol. 10, No., pp. 149-155, 2017. Available olie at www.jafmolie.et, ISSN 1735-3572, EISSN 1735-345. DOI: 10.1889/acadpub.jafm.73.243.27148 Characteristics of CNG Bubbles

More information

8.5. Solving Equations II. Goal Solve equations by balancing.

8.5. Solving Equations II. Goal Solve equations by balancing. 8.5 Solvig Equatios II Goal Solve equatios by balacig. STUDENT BOOK PAGES 268 271 Direct Istructio Prerequisite Skills/Cocepts Solve a equatio by ispectio or systematic trial. Perform operatios usig itegers,

More information

Characterization of Refrigeration System Compressor Performance

Characterization of Refrigeration System Compressor Performance Purdue Uiversity Purdue e-pubs Iteratioal Refrigeratio ad Air Coditioig Coferece School of Mechaical Egieerig 2002 Characterizatio of Refrigeratio System Compressor Performace A. Mackese Uiversity of Wiscosi-Madiso

More information

Basic Gas Spring Theory

Basic Gas Spring Theory Calculatig the iitial force The iitial force of the gas sprig ca be calculated as the sealed area of the pisto rod or the pisto (depedig o desig) multiplied by the pressure iside the gas sprig. The larger

More information

EMSBS/EMST. Drill For Machining Ultra-Deep Minute Holes FEATURES. For ultra-deep drilling of miniature holes. New chip stopper controls chip flow.

EMSBS/EMST. Drill For Machining Ultra-Deep Minute Holes FEATURES. For ultra-deep drilling of miniature holes. New chip stopper controls chip flow. EMSBS/EMST Drill For Machiig Ultra-Deep Miute Holes FEATURES For ultra-deep drillig of miiature holes. New chip stopper cotrols chip flow. EMSBS: Deep hole mirco-drills with TH Coatig. EMST: Micro- starter

More information

Adaptive Neuro-Fuzzy control of an unmanned bicycle

Adaptive Neuro-Fuzzy control of an unmanned bicycle Adaptive euro-fuzzy cotrol of a umaed bicycle A. Shafiekhai, M. J. Mahoob, ad M. Roozegar 3 Ceter for Mechatroics ad Automatio, School of Mechaical Egieerig College of egieerig, Uiversity of Tehra Tehra,

More information

Extensible Detection and Indexing of Highlight Events in Broadcasted Sports Video

Extensible Detection and Indexing of Highlight Events in Broadcasted Sports Video Extesible Detectio ad Idexig of Highlight Evets i Broadcasted Sports Video Dia W. Tjodroegoro 1, Yi-Pig Phoebe Che 2, Bih Pham 3 1 School of Iformatio Systems, Queeslad Uiversity of Techology, Brisbae,

More information

Hazard Identificaiton of Railway Signaling System Using PHA and HAZOP Methods

Hazard Identificaiton of Railway Signaling System Using PHA and HAZOP Methods www.ijape.org Iteratioal Joural of Automatio ad Power Egieerig (IJAPE) Volume 2 Issue 2, February 2013 Hazard Idetificaito of Railway Sigalig System Usig PHA ad HAZOP Methods Jog Gyu Hwag *1, Hyu Jeog

More information

ELIGIBILITY / LEVELS / VENUES

ELIGIBILITY / LEVELS / VENUES ELIGIBILITY / LEVELS / VENUES 10U - SQUIRT MINOR '09 & MAJOR '08 Eligibility: Top six teams i the league at each level will qualify based o regular seaso league play. Format: Divisioal crossover with semi-fial

More information

The Prediction of Dynamic Strain in Leaf-Type Compressor Valves With Variable Mass and Stiffness

The Prediction of Dynamic Strain in Leaf-Type Compressor Valves With Variable Mass and Stiffness Purdue Uiversity Purdue e-pubs Iteratioal Compressor Egieerig Coferece School of Mechaical Egieerig 1972 The Predictio of Dyamic Strai i Leaf-Type Compressor Valves With Variable Mass ad Stiffess M. Moavei

More information

ELIGIBILITY / LEVELS / VENUES

ELIGIBILITY / LEVELS / VENUES ELIGIBILITY / LEVELS / VENUES 10U - SQUIRT MINOR '09 & MAJOR '08 Eligibility: Top six teams i the league at each level will qualify based o regular seaso league play. Format: Divisioal crossover with semi-fial

More information

Computation of the inviscid drift force caused by nonlinear waves on a submerged circular cylinder

Computation of the inviscid drift force caused by nonlinear waves on a submerged circular cylinder csnak, 2011 Iter J Nav Archit Oc Egg (2011) 3:201~207 http://dx.doi.org/10.3744/jnaoe.2011.3.3.201 Computatio of the iviscid drift force caused by oliear waves o a submerged circular cylider Hyeok-Ju Koh

More information

ELIGIBILITY / LEVELS / VENUES

ELIGIBILITY / LEVELS / VENUES ELIGIBILITY / LEVELS / VENUES 10U - SQUIRT MINOR '08 & MAJOR '07 Eligibility: Top six teams i the league at each level will qualify based o regular seaso league play. Format: Divisioal crossover with semi-fial

More information

Series 600 Accessories

Series 600 Accessories Series 600 Geeral These accessories are a rage of devices for completig a peumatic circuit. These valves, with their special fuctios, are iserted betwee two valves, betwee a valve ad a cylider, or followig

More information

XFM DIGITAL MASS FLOW METER. XFM with Profibus Interface. XFM without. Readout. XFM with. Readout. Option

XFM DIGITAL MASS FLOW METER. XFM with Profibus Interface. XFM without. Readout. XFM with. Readout. Option BULLETIN EM20180402 Desig Features Supports up to 23 Egieerig Uits (icludig User Defied). Stores calibratio data for up to 10 gases. Programmable Totalizer idicates total gas quatity. High ad low gas flow

More information

ELIGIBILITY / LEVELS / VENUES

ELIGIBILITY / LEVELS / VENUES ELIGIBILITY / LEVELS / VENUES 10U - SQUIRT MINOR '08 & MAJOR '07 Eligibility: Top six teams i the league at each level will qualify based o regular seaso league play. Format: Divisioal crossover with semi-fial

More information

draft final report NGSIM Arterial-Lane Selection Mode Federal Highway Administration Cambridge Systematics, Inc.

draft final report NGSIM Arterial-Lane Selection Mode Federal Highway Administration Cambridge Systematics, Inc. NGSIM Arterial-Lae Selectio Mode draft fial report prepared for Federal Highway Admiistratio prepared by Cambridge Systematics, Ic. with Itelliget Trasportatio Systems Laboratory Massachusetts Istitute

More information

ANALYSIS AND MODELING TIME HEADWAY DISTRIBUTIONS UNDER HEAVY TRAFFIC FLOW CONDITIONS IN THE URBAN HIGHWAYS: CASE OF ISFAHAN

ANALYSIS AND MODELING TIME HEADWAY DISTRIBUTIONS UNDER HEAVY TRAFFIC FLOW CONDITIONS IN THE URBAN HIGHWAYS: CASE OF ISFAHAN TRANSPORT ISSN 648-442 prit / ISSN 648-348 olie 2 Volume 26(4): 375 382 doi:.3846/648442.2.635694 ANALYSIS AND MODELING TIME HEADWAY DISTRIBUTIONS UNDER HEAVY TRAFFIC FLOW CONDITIONS IN THE URBAN HIGHWAYS:

More information

Traffic conflicts at roundabouts: risk analysis under car-following conditions

Traffic conflicts at roundabouts: risk analysis under car-following conditions Itersectios Cotrol ad Safety 143 Traffic coflicts at roudabouts: risk aalysis uder car-followig coditios G. Guido, A. Vitale & V. Gallelli Departmet of Tow ad Coutry Plaig, Uiversity of Calabria, Italy

More information

GFC NIST MASS FLOW CONTROLLERS. Typical Stainless Steel GFC Mass Flow Controller. Design Features. General Description. Principles of Operation

GFC NIST MASS FLOW CONTROLLERS. Typical Stainless Steel GFC Mass Flow Controller. Design Features. General Description. Principles of Operation BULLETIN EM20180402 Model thermal Mass Flow Cotrollers are desiged to idicate ad cotrol set flow rates of gases. The combies the characteristics ad accuracy of covetioal mass flow devices ito a uique compact

More information

T I reduction of traffic fatalities and injuries in 2010 [l].

T I reduction of traffic fatalities and injuries in 2010 [l]. Visiosese; A Advaced Lateral Collisio Warig System Tim va Dijck ad Geert A.J. va der Heijde t.vadiick~studct.utwete.l e.a.i.vafiderheijde~s~det.utwete.l Researchcetre ADA, Cetre for Trasport Studies, Faculty

More information

Capacity of Shared-Short Lanes at Unsignalised Intersections

Capacity of Shared-Short Lanes at Unsignalised Intersections Nig Wu itle ad Abstract Author's Address apacity of Shared-Short aes at Usigalised Itersectios Nig Wu ABSA he calculatio procedures i recet highway capacity mauals do ot eactly treat ared/ort laes at usigalized

More information

HEEL RETENTION SYSTEMS

HEEL RETENTION SYSTEMS TECHNOLOGY OVERVIEW the whole may be greater tha the sum of its parts. But the parts are pretty cool too. Our costat pursuit of excellece is deeply embedded i our desig, techology ad materials. Every attribute

More information

DAMAGE ASSESSMENT OF FIBRE ROPES FOR OFFSHORE MOORING

DAMAGE ASSESSMENT OF FIBRE ROPES FOR OFFSHORE MOORING RECOMMENDED PRACTICE DNV-RP-E304 DAMAGE ASSESSMENT OF FIBRE ROPES FOR OFFSHORE MOORING APRIL 2005 FOREWORD (DNV) is a autoomous ad idepedet foudatio with the objectives of safeguardig life, property ad

More information

n UL Listed and FM Approved for n Solenoid control n Quick pressure relief valve 73Q n Pressure sustaining & reducing valve 723

n UL Listed and FM Approved for n Solenoid control n Quick pressure relief valve 73Q n Pressure sustaining & reducing valve 723 Pressure Relief/Sustaiig Valve Prioritizig pressure zoes Esurig cotrolled pipelie fill-up Prevetig pipelie emptyig Pump overload & cavitatio protectio Safeguardig pump miimum flow Excessive lie pressure

More information

PERFORMANCE TEAM EVALUATION IN 2008 BEIJING OLYMPIC GAMES

PERFORMANCE TEAM EVALUATION IN 2008 BEIJING OLYMPIC GAMES XV INTERNATIONAL CONFERENCE ON INDUSTRIAL ENGINEERING AND OPERATIONS MANAGEMENT The Idustrial Egieerig ad the Sustaiable Developmet: Itegratig Techology ad Maagemet. Salvador, BA, Brazil, 06 to 09 October

More information

Chilled Mirror Dew Point Instrument

Chilled Mirror Dew Point Instrument 102 The Relatioship Betee Gas Ad Pressure (SF6) Whe the gas is i the temperature belo the codesig poit, de poit measuremet is ot available This gas codeses o the mirror, the figure belo idicates the relatioship

More information

Analytical and numerical prediction of water sorptivity in cement-based materials

Analytical and numerical prediction of water sorptivity in cement-based materials Life Cycle Predictio ad Maagemet of Cocrete tructures, apporo, March Aalytical ad umerical predictio of water sorptivity i cemet-based materials Licheg WANG Dr. Dalia Uiversity of Techology, Dalia Chia

More information

securing your safety

securing your safety securig your safety LIFE POINT PRO DEFIBRILLATOR AN AUTOMATED EXTERNAL DEFIBRILLATOR (AED) CAN BE THE DIFFERENCE BETWEEN LIFE AND DEATH. Statistics show that early 200,000 people die each year as a result

More information

GFC NIST MASS FLOW CONTROLLERS. Typical Stainless Steel GFC Mass Flow Controller. Design Features. General Description. Principles of Operation

GFC NIST MASS FLOW CONTROLLERS. Typical Stainless Steel GFC Mass Flow Controller. Design Features. General Description. Principles of Operation Model thermal Mass Flow Cotrollers are desiged to idicate ad cotrol set flow rates of gases. The combies the characteristics ad accuracy of covetioal mass flow devices ito a uique compact desig at low

More information

Introduction to Algorithms 6.046J/18.401J/SMA5503

Introduction to Algorithms 6.046J/18.401J/SMA5503 Itroductio to Algorithms 6.046J/8.40J/SMA5503 Lecture 4 Prof. Charles E. Leiserso Quicsort Proposed by C.A.R. Hoare i 962. Divide-ad-coquer algorithm. Sorts i place lie isertio sort, but ot lie merge sort.

More information

Outline. Changing needs in Urban Traffic. Introduction The starting point Basic principles Analysis Facts Changes Context Solutions Key messages

Outline. Changing needs in Urban Traffic. Introduction The starting point Basic principles Analysis Facts Changes Context Solutions Key messages Chagig eeds i Urba Traffic Rob Methorst Chair COST 358 Pedestrias Quality Needs Cetre for Trasport ad Navigatio 27th of October 2011 Outlie Itroductio The startig poit Basic priciples Aalysis Facts Chages

More information

Gait-Event-Based Synchronization Method for Gait Rehabilitation Robots via a Bio-inspired Adaptive Oscillator

Gait-Event-Based Synchronization Method for Gait Rehabilitation Robots via a Bio-inspired Adaptive Oscillator TBME-0485-205 Gait-Evet-Based Sychroizatio Method for Gait Rehabilitatio Robots via a Bio-ispired Adaptive Oscillator Gog Che, Peg Qi, Member, IEEE, Zhao Guo ad Haoyog Yu, Member, IEEE Abstract I the field

More information

AN ESTIMATION OF INTER-ZONAL BUS DEMAND ON COLOMBO - KALUTARA TRANSPORT CORRIDOR

AN ESTIMATION OF INTER-ZONAL BUS DEMAND ON COLOMBO - KALUTARA TRANSPORT CORRIDOR AN ESIMAION OF INER-ZONAL BUS DEMAND ON COLOMBO - KALUARA RANSPOR CORRIR Sama J. Widaapathiraage ad Amal S. Kumarage Departmet of Civil Egieerig, Uiversity of Moratuwa, Sri Laka. ABSRAC- I the past, it

More information

Simulation Study of a Bus Signal Priority Strategy Based on GPS/AVL and Wireless Communications

Simulation Study of a Bus Signal Priority Strategy Based on GPS/AVL and Wireless Communications Simulatio Study of a Bus Sigal Priority Strategy Based o GPS/AVL ad Wireless Commuicatios Che-Fu Liao* Ceter for rasportatio Studies ad the Itelliget rasportatio Systems Istitute Uiversity of Miesota 200

More information

Coal Pulveriser. Global Solutions

Coal Pulveriser. Global Solutions Coal Pulveriser. Global Solutios > Worm Gears > Gear Uits > Coupligs & Freewheels Global Coal Pulveriser Solutios Reold is a global egieerig group with over 130 years of experiece i the desig ad maufacture

More information

Headfirst Entry - Diving and Sliding

Headfirst Entry - Diving and Sliding 5/31/2011 Safe divig board use, safe slide use, usi Headfirst Etry - Divig ad Slidig POOLALARMS.COM HOME POOL SAFETY PRODUCTS POOL SAFETY ARTICLES & REPORTS POOL SAFETY LINKS & RESOURCES POOLALARMS.COM

More information

ANALYSIS AND ASSESSMENT OF LITHUANIAN ROAD ACCIDENTS BY AHP METHOD

ANALYSIS AND ASSESSMENT OF LITHUANIAN ROAD ACCIDENTS BY AHP METHOD ISSN 1822-427X/eISSN 1822-4288 2018 Volume 13 Issue 3: 238 260 https://doi.org/10.7250/bjrbe.2018-13.414 ANALYSIS AND ASSESSMENT OF LITHUANIAN ROAD ACCIDENTS BY AHP METHOD MARIUS JAKIMAVIČIUS * Dept of

More information

Controlling noise at work

Controlling noise at work Health ad Safety Cotrollig oise at work The Cotrol of Noise at Work Regulatios 2005 Guidace o Regulatios This is a free-to-dowload, web-friedly versio of L108, (Secod editio, published 2005). This versio

More information

Maximizing. Potential in the 4x100 Relay. 4x100 - Entire Race. A Review of Relay Basics. NFHS Relay Zone Measurements

Maximizing. Potential in the 4x100 Relay. 4x100 - Entire Race. A Review of Relay Basics. NFHS Relay Zone Measurements Potetial i the 4x Relay Preseted by Bria FitzGerald Rio Mesa High School 19 FACA Track & Field Cliic Jauary -12 Daytoa Beach Maximizig A Review of Relay Basics Basic Premise: Get the bato safely aroud

More information

Polynomial functions have graphs that are smooth and continuous. c) Use your knowledge of quadratic functions to sketch the graph.

Polynomial functions have graphs that are smooth and continuous. c) Use your knowledge of quadratic functions to sketch the graph. Math 165 4.1 Polyomial fuctios I class work Polyomial fuctios have graphs that are smooth ad cotiuous 1) For the fuctio: y 2 = x 9. a) Write the fuctio i factored form. b) Fid the x-itercepts or real zeros

More information

Version IV: April a publication from

Version IV: April a publication from Versio IV: April 2013 a publicatio from AWRF Associated Wire Rope Fabricators Recommeded Practice ad Guidelie Disclaimer for AWRF Recommeded Guidelie Swager Safety Guide Associated Wire Rope Fabricators

More information

Impacts of Typhoon on Wave Height at Bangkhuntien Shoreline

Impacts of Typhoon on Wave Height at Bangkhuntien Shoreline America Joural of Evirometal Scieces 6 (4): 308-315, 2010 ISSN 1553-345X 2010 Sciece Publicatios Impacts of Typhoo o Wave Height at Bagkhutie Shorelie 1 Prasertsak Ekphisutsutor, 1 Prugcha Wogwises, 2

More information

Transit Development Plan December 2011

Transit Development Plan December 2011 Trasit Developmet Pla December 2011 Table of Cotets Chapter Oe - Ivetory Itroductio... 1-1 Data Sources... 1-1 Settig the Stage... 1-2 Service Area... 1-3 Fixed Route Service Routes... 1-4 Overall Ridership

More information

GENETICS 101 GLOSSARY

GENETICS 101 GLOSSARY GENETICS 101 This documet is iteded to provide a basic uderstadig of caie geetics to the America Eskimo Dog (AED) ower ad breeder. It is simplified from scietific facts ad uderstadig. It is ot iteded to

More information

A Comparative Investigation of Reheat In Gas Turbine Cycles

A Comparative Investigation of Reheat In Gas Turbine Cycles Proceedigs of the Iteratioal Gas Turbie ogress 3 Tokyo November -7, 3 IGT3Tokyo TS-86 A omparative Ivestigatio of Reheat I Gas Turbie ycles K. Sarabchi Departmet of Mechaical egieerig Uiversity of Tabriz

More information

The Application of Simulating WAves Nearshore Model for Wave Height Simulation at Bangkhuntien Shoreline

The Application of Simulating WAves Nearshore Model for Wave Height Simulation at Bangkhuntien Shoreline America Joural of Evirometal Scieces 6 (3): 299-307, 2010 ISSN 1553-345X 2010 Sciece Publicatios The Applicatio of Simulatig WAves Nearshore Model for Wave Height Simulatio at Bagkhutie Shorelie 1 Prasertsak

More information

ELIGIBILITY / LEVELS / VENUES

ELIGIBILITY / LEVELS / VENUES ELIGIBILITY / LEVELS / VENUES 10U - SQUIRT MINOR '08 & MAJOR '07 Eligibility: Top six teams i the league at each level will qualify based o regular seaso league play. Format: Divisioal crossover with semi-fial

More information

JOBST Elvarex Plus Feel the difference in seamless 3D!

JOBST Elvarex Plus Feel the difference in seamless 3D! JOBST Elvarex Plus Feel the differece i seamless 3D! Custom-made flat-kit compressio garmets for the maagemet of lymphatic diseases of the hads ad feet 3D Comfort, Compressio ad Techology JOBST Elvarex

More information

Active Travel The Role of Self-Selection in Explaining the Effect of Built Environment on Active Travel

Active Travel The Role of Self-Selection in Explaining the Effect of Built Environment on Active Travel Active Livig Research Buildig Evidece to Prevet Childhood Obesity ad Support Active Commuities RESE ARCH brief Fall 2009 activelivigresearch.org Active Travel The Role of Self-Selectio i Explaiig the Effect

More information

Operating Instructions SURGICAL POWER & ACCESSORIES

Operating Instructions SURGICAL POWER & ACCESSORIES SURGICAL POWER & ACCESSORIES Table of Cotets PAGE 2 PAGE 3-5 PAGE 6 Itroductio Warig Traiig Setup Illustratio Drill Motor Setup Quick Coectors Motor / Foot Cotrol Coectios Attachmet Istallatio Straight

More information

Influences of Teaching Styles and Motor Educability on Learning Outcomes of Volleyball

Influences of Teaching Styles and Motor Educability on Learning Outcomes of Volleyball Asia Social Sciece; Vol., No. 5; 14 ISSN 1911-17 E-ISSN 1911-5 Published by Caadia Ceter of Sciece ad Educatio Iflueces of Teachig Styles ad Motor Educability o Learig Outcomes of Volleyball Syamsuddi

More information

Acta Geophysica vol. 57, no. 4, pp DOI: /s

Acta Geophysica vol. 57, no. 4, pp DOI: /s Acta Geophysica vol. 57, o. 4, pp. 98- DOI:.478/s6-9-7-4 Structure of Numerically Simulated Katabatic ad Aabatic Flows alog Steep Slopes Evgei FEDOROVICH ad Ala SHAPIRO School of Meteorology, Uiversity

More information

WATER VALVES REPLACEMENT PROJECT

WATER VALVES REPLACEMENT PROJECT WATER VALVES REPLACEMET PROJECT BROWSVILLE, TEXAS BID o B026-17 PUBLIC B R O W S V I L L E UTILITIES B OARD Brownsville - Board of Directors RAFAEL VELA...CHAIR RAFAEL S. CHACO...VICE-CHAIR MARTI C. ARAMBULA...SECRETARY/TREASURER

More information

7.0 Nonmotorized Facilities

7.0 Nonmotorized Facilities 7.0 Nomotorized Facilities 7.1 Sectio Overview This sectio describes the existig coditios ad ay idetified future impacts with the project o omotorized facilities withi the study area. Data were collected

More information

» WYOMING s RIDE 2013

» WYOMING s RIDE 2013 » WYOMING s RIDE 2013 WYOMING S RIDE 2013 preseted by First Iterstate Bak of Wyomig bikemswyomig.org 1 We are people who wat to do somethig about ow.» Our Wyomig Bike MS Ride Two-Day Wyomig Ride: To view

More information

Climbing/Rappelling NATIONAL STANDARDS BOY SCOUTS OF AMERICA

Climbing/Rappelling NATIONAL STANDARDS BOY SCOUTS OF AMERICA 2007 Climbig/Rappellig NATIONAL STANDARDS BOY SCOUTS OF AMERICA These stadards apply to coucil ad district climbig/rappellig activities. Climb O Safely, No. 20-099B, applies to uit activities. Toppig Out,

More information

Travel Demand Management Implementation in Bandar Lampung

Travel Demand Management Implementation in Bandar Lampung Travel Demad Maagemet Implemetatio i Badar Lampug Rahayu Sulistyorii 1, a, 1 Lampug Uiversity Istitute Techology of Sumatera a Keywords: Traffic

More information

"The twisting movement of any hoof should, for physiological reasons, not be hindered by Shoeing." (Lungwitz 1884)

The twisting movement of any hoof should, for physiological reasons, not be hindered by Shoeing. (Lungwitz 1884) Volume 15: Issue 2 Shoeig for Rotatioal Deviatio of the Equie Limb by Michael J. Wildestei CJF, FWCF (Hos) Examiatio of a horse prelimiary to shoeig should be made while the aimal is at rest ad afterwards

More information

A Different View on TBM Face Equilibrium in Permeable Ground

A Different View on TBM Face Equilibrium in Permeable Ground A Differet View o TBM Face Equilibrium i Permeable Groud Tiago Gerheim Souza Dias Ghet Uiversity, Ghet, Belgium Adam Bezuije Ghet Uiversity, Ghet, Belgium ad Deltares, Delft, the Netherlads ABSTACT The

More information

Seated valves (PN 16) VF 2-2-way valve, flange VF 3-3-way valve, flange

Seated valves (PN 16) VF 2-2-way valve, flange VF 3-3-way valve, flange Data sheet Seated valves (PN 16) VF 2-2-way valve, flage VF 3-3-way valve, flage Descriptio Features: Bubble tight desig 15-80, 200-300 Sap mechaical coectio together with AMV(E) 335, AMV(E) 435 Dedicated

More information

A Data Envelopment Analysis Evaluation and Financial Resources Reallocation for Brazilian Olympic Sports

A Data Envelopment Analysis Evaluation and Financial Resources Reallocation for Brazilian Olympic Sports WSEAS TRANSACTIONS o SYSTEMS Reato Pescarii Valério, Lidia Agulo-Meza A Data Evelopmet Aalysis Evaluatio ad Fiacial Resources Reallocatio for Brazilia Olympic Sports RENATO PESCARINI VALÉRIO LIDIA ANGULO-MEZA

More information

The University of New South Wales. (Image Image 4) - (Image Image 2)

The University of New South Wales. (Image Image 4) - (Image Image 2) DISAR (Differetial ISAR) DISAR is a ISAR data processig i which topographic patter is carefully removed. ISAR topography vs DISAR chage i topography DISAR Applicatios Subsidece (udergroud miig, oil/water

More information

TIME DOMAIN SIMULATIONS OF A COUPLED PARAMETRICALLY EXCITED ROLL RESPONSE IN REGULAR AND IRREGULAR HEAD SEAS

TIME DOMAIN SIMULATIONS OF A COUPLED PARAMETRICALLY EXCITED ROLL RESPONSE IN REGULAR AND IRREGULAR HEAD SEAS 8 th Iteratioal Coferece o 349 TIME DOMAIN SIMULATIONS OF A COUPLED PARAMETRICALLY EXCITED ROLL RESPONSE IN REGULAR AND IRREGULAR HEAD SEAS S. Ribeiro e Silva, T. Satos ad C. Guedes Soares Uit of Marie

More information

Application of a Statistical Method to investigate patterns of beach evolution in the vicinity of a seawall

Application of a Statistical Method to investigate patterns of beach evolution in the vicinity of a seawall Joural of Coastal Research SI 95-99 ICS (Proceedigs) Polad ISSN 79-8 Applicatio of a Statistical Method to ivestigate patters of beach evolutio i the viciity of a seawall J. M. Horrillo-Caraballo ad D.E.

More information

Overview. Creating Walkable, Bikeable Communities Developing Effective Active Transportation Projects and Programs. Add photo or photos

Overview. Creating Walkable, Bikeable Communities Developing Effective Active Transportation Projects and Programs. Add photo or photos Creatig Walkable, Bikeable Commuities Developig Effective Active Trasportatio Projects ad Programs Add photo or photos Sposored by Caltras ad [local sposors] Overview Effective itervetios that ca improve

More information

Introductory Rules of Officiating Small Sided Games Under 6 &Under 8 HANDBOOK

Introductory Rules of Officiating Small Sided Games Under 6 &Under 8 HANDBOOK Itroductory Rules of Officiatig Small Sided Games Uder 6 &Uder 8 O F F I C I A L S HANDBOOK 1 T his hadbook is iteded to be used i cojuctio with the US Youth Soccer Small-sided Game Official s classroom

More information

-H- Note. Flow control valve. Key features

-H- Note. Flow control valve. Key features Flow cotrol valves Flow cotrol valves Key features Fuctio Flow cotrol or oe-way flow cotrol valves regulate the pisto speed of peumatic drives durig advace ad retur strokes. This is doe through suitable

More information

Cooling of a Reciprocating Compressor through Oil Atomization in the Cylinder

Cooling of a Reciprocating Compressor through Oil Atomization in the Cylinder Purdue Uiversity Purdue e-pubs Iteratioal Compressor Egieerig Coferece School of Mechaical Egieerig 2010 Coolig of a Reciprocatig Compressor through Oil Atomizatio i the Cylider Rodrigo Kremer Embraco

More information