報告人 : 林福城 指導老師 : 陳定宏 1 From Res. Center of Intell. Transp. Syst., Beijing Univ. of Technol., Beijing, China By Zhe Liu ; Yangzhou Chen ; Zhenlong Li Appears in: Computer Science and Information Engineering, 2009 WRI World Congress on
Outline 1.Introduction 2.Moving object detection 2-1.Conscutive image difference 2-2.Backgrout difference 3.Moving object tracking 3-1.Review of Tracking Algorithm 3-2.Camshift Multiple Vehicle Tracking 4.Traffic Parameters Estimation 5.Experimental Results 6.Conclusions 2
1.Introduction Traffic management and information systems: 1.Inductive loop detectors 2.Visual surveillance systems Our approach specifies three sub process: 1. Vehicle Extraction : Consecutive image difference Background difference 2. Vehicle Tracking 3. Traffic Parameter Estimation 3
2.1 Consecutive Image Difference D(x,y) is the difference image. Mask(x,y) is the image after binarization. 4
2.2 Background Difference(1) It assume a moving object would not stay at the same position for more than half of n frames. 5
2.2 Background Difference(2) 6
2.Moving Object Detection Conclusion 1.Consecutive2.Background Easily realizedGood The change of scene luminance Good Extract preciseGood Process timeGood After morphological processEqual 7
3-1.Review of Tracking Algorithm 1.Tracking based on a moving object region: Size, Color, Shape, Velocity, Centroid 2.Tracking based on an active contour of a moving object 3.Tracking based on a moving object model 4.Tracking based on selected features of moving objects : Corner 8
3-2.CamShift Multiple Vehicle Tracking 9
Mean-Shift Object Tracking General Framework: Target Localization Search in the model’s neighborhood in next frame Start from the position of the model in the current frame Find best candidate by maximizing a similarity func. Repeat the same process in the next pair of frames Current frame …… ModelCandidate 10
Mean-Shift Object Tracking Target Representation Choose a reference target model Quantized Color Space Choose a feature space Represent the model by its PDF in the feature space Kernel Based Object Tracking, by Comaniniu, Ramesh, Meer 11
Mean-Shift Object Tracking PDF Representation Similarity Function: Target Model (centered at 0) Target Candidate (centered at y) 12
4.Traffic Parameters Estimation 1.Vehicle count 2.Vehicle average speed 3.Vehicle size 13
Experimental Results 14
Conclusions In this paper, we have presented methods for detecting and tracking multiple vehicles in an outdoor environment. Each detected vehicle is assigned a camshift tracker which can effectively track object with different size and shape under different illumination conditions. The method fails to handle long slow moving vehicle queue. 15