【正文】 c monitoring ang enforcement application[J].Image and vision Computing，2004：485501 [16] Zhu Z,Xu G,Yang B,Shi D,:Realtime image tracking for automatic traffic monitoring[J].Image and vision Computing，2000：781794[17] Haag of plex driving maneuvers in traffic. imageseguences[J] Image and vision Computing，1998：517527[18] Pai C,Tyan H,Liang Y,Liao H detectiob and tracking at crossroads[J]. pattern Recognition，2004：10251034[19] 羅軍輝，馮平．．機械工業(yè)出版社，2005：1257[20] 王耀南，李樹濤，毛建旭．計算機圖像處理與識別技術(shù)．高等教育出版社，2001：734[21] 張錚，王艷平，薛桂香．數(shù)字圖像處理與機器視覺．人們郵電出版社，2010：1127[22] 章毓晉．圖像分割[M]．科學出版社，2001：2344[23] 馮俊萍，趙轉(zhuǎn)萍，徐濤．基于數(shù)學形態(tài)學的圖像邊緣檢測技術(shù)[J] ．航空計算技術(shù)，2004：78134[24] 鄭江濱．視頻監(jiān)視方法研究博士論文．西北工業(yè)大學，2002：4—13[25] 王賓，潘建壽等．基于Matrox卡的視頻圖像運動檢測[J]．西北大學學報（自然科學網(wǎng)絡版），2004：72103[26] 涂虬．智能視覺監(jiān)視中目標檢測與跟蹤算法研究[D]．華中科技大學，2010：125[27] 李江春．運動目標檢測與跟蹤算法的研究與實現(xiàn)[D]．吉林大學，2009：1322[28] Haritaoglu,L．S．Davis．D．Harwood．w4：who?when?where?what?a real time system for detecting and tracking people[C]．In Proceedings of the European Conference on Computer Vision，1998，877892 [29] N．Diehl．Objectoriented motion estimation and segmentation in image sequence[J].Signal Processing：Image Communication，1991：2356[30] 徐晶，方明，楊華民．計算機視覺中的運動檢測與跟蹤．國防工業(yè)出版，2012：2546[31] Richard Szeliski ．計算機視覺——算法與應用．艾海軍，興軍亮．清華大學出版社，2012：1134[32] 張國云，郭龍源，胡文靜．計算機視覺與圖像識別．科學出版社，2012：1932[33] 楊威，張?zhí)镂模畯碗s景物環(huán)境下運動目標檢測的新方法[J]．計算機研究與發(fā)展，1998：724728[34] Fukunage L．D．Hostetler．The estemation of the gradient of a density funtion with application in pattern recognition[J]．IEEE Transaction on Information Theory，1975：3240[35] 余靜，游志勝．自動目標識別與跟蹤技術(shù)研究綜述[J] ．計算機應用研究，2005：1215[36] charles k. Chui，guanrong chen．卡爾曼濾波及其實時應用．戴洪德，周紹磊戴邵武譯．第四版．科學出版社，2013：1336[37] 安建虎，劉健，肖陽輝．基于自適應參考模板的相關跟蹤算法[J]．計算機工程與應用，2003：103144[38] D．A．Montera,S．k Rogers，D．W．Ruck,w．Dennis，M．E．Oxley． Object tracking through adaptive correlation[J]．Optical Engineering，1994，3：294302．[39] 王茜蓓，彭中，劉莉．一種基于自適應閾值的圖像分割算法[J]．北京理工大學學報，2003，23(4)：521524[40] Threshold Selection Method from GrayLevel IEEETransactions on SystemsMan and Cybemetics，1979：6266[41] 韓恩奇，王蕾． 圖像分割的閾值法綜述[J]．系統(tǒng)工程與電子技術(shù)，2002：91102[42] 龔聲蓉，劉純平，王強．數(shù)字圖像處理與分析[M]．清華大學出版社，2006：1443[43] 王賓．視頻序列中運動目標檢測與跟蹤有關問題的研究．碩士論文．西北大學電子系，2004：1233[44] 章毓晉．圖像工程（下冊）——圖像理解與計算機視覺[M] ．清華大學出版杜，2000：819[45] 蘇金鵬，阮沈勇．MATLAB實用教程電子．工業(yè)出版社，2005：1139附錄A 英文原文Abstract.This paper describes a vision system that recognizes moving targets such as vehicles and pedestrians on public streets. This system can: (1)classify targets {vehicle, pedestrian,others}and,for “vehicles,” discriminate vehicle typesand (2) estimate the main colors of targets. According to theinput images to the system,the categories of targets are set as mule (golf cart for workers),sedan,van,truck,pedestrian (single or plural),and other (such as noise). Their colors are set as six color groups{red,orange, yellow。 green。 blue,light blue。 white,silver,gray。 dark blue,dark gray,black。dark red,dark orang}. In this experiment, we collected images of targets from 9:00 . to 5:00 . on sunny and cloudy days as system training samples. The recognition ratio was %under the condition that both the recognition results of typeand color agreed with the operator’s judgment. In addition, the system can detect predefined specific targets such as delivery vans, post office vans, and police cars by bining recognition results for type and color. The recognition ratio for specific targets was %. For the classification and estimation of targets,we employed a statistical linear discrimination method(linear discriminant analysis, LDA) and a nonlinear decision rule (weighted Knearest neighbor rule, KNN). IntroductionRecently, along with increasing demand for security, a growing number of monitoring cameras has been appearing on publicstreets. This paper presents a vision system that can recognize moving targets such as vehicles and pedestrians on publicstreets. The targets of this system are moving outdoors. For this reason, the information gained from the targets is influenced by several factors:Shadows cast by other objects Direction of sunlightReflection of other objects on the targets Obscuring of targets by other objectsTherefore, it is not always possible to obtain valid information simultaneously for both category classification and target color estimation. For example, the shapes of targets could be broken or the color of targets could be affected by strong sunlightreflection or shadows of other objects. To resolve these problems, we present a scheme that can determine the tracks of targets and at the same time independently classify the categories of targets as well as estimate their colors. Moreover,by paring each result throughout the tracking sequence,the scheme can automatically choose discriminant results with the largest matching probability as the final decision. Figure 1shows a sample input image used in this system.Among previous research efforts, many reports[1,2,3,4] have concentrated on the search for recognition systems that are suitable for the lanes of a highway to check moving vehicles running in one direction and, for toll stations, to check stopped vehicles. However, no studies have sought to classify and categorize moving targets and estimate their colors and consequently identify specific vehicles on general thoroughfares.FigA1 Input image to the system and its segmentation Trucks Vans SedansFigA2 Sample images of targets extracted from area ALimin et al[5] expressed the 3D contour of an automobile using 29 structure parameters to estimate a target’s 3D orientation and to distinguish the type of automobile from its 2D image. They reported their primary results. Wu et al[6]also used 29 structural parameters to express the 3D geometry of an automobile, train a neural network, and dinstinguish six types of automobiles from their 2D images with 91% pro