Pedestrian Detection And Tracking

Pedestrian detection and tracking based on the video is one of the key technologies in computer vision and image processing. It has important application value and broad application prospect in intelligent video surveillance, smart cars, images and video retrieval and multimedia analysis. So, pedestrian detection and tracking becomes a popular research issue.The main works of the thesis are summarized as follows:First, pedestrian detection:analyzing Histograms of Oriented Gradients (HOG) feature, Support Vector Machine (SVM) classifier, and Intersection Kernel Support Vector Machines (IKSVM) classifier; Then constructing two kinds of framework—HOG combined with SVM classifier and HOG combined IKSVM. The experiments in INRIA show the good performance of the algorithms.Second, pedestrian tracking:detailed introducing Meanshift algorithm and Kalman filter and completing the tracking method based on Meanshift and Kalman filtering. In the end of this chapter, we show the test results of a standard video library.In addition, human counting based on head feature extraction: extracting the heads’contour feature in vertical view image and introducing the preprocessing, head detection, head tracking and human counting. We test the performance of algorithm in some video captured in natural scenes.Finally, the thesis also completed a project about human detection and segmentation of image. Using certain algorithms in the field of image processing, this project has good performance in practice.In conclusion, we construct pedestrian detection and tracking, human counting based on head feature extraction, human detection and segmentation in image, three systems effectively and efficiently.