Research And Implementation Of Pedestrain And Bicycle Detection Based On Vision

With the continuous development of the socio-economic and urbanization, the number of car, the basic transport vehicle, is increasing, that leads to terrible traffic safety problems. Pedestrians and vehicles are the main participants in the traffic behavior.Detection and tracking algorithm based on monocular vision of pedestrian and vehicle is researched and implemented in this paper. In real life, there are many kinds of vehicle.This paper just does one step depth-study of bicycling pedestrian on the basis of pedestrian detection. Characteristics of bicycling pedestrian contain not only the characteristics of pedestrian but also more complicated, and the speed of Bicycling pedestrian is faster than the pedestrian. Therefore, the study for bicycle pedestrian has important implications.This article mainly finished the research and implementation:First, the article used image preprocessing, and edge detection technique to extract effective edge contour, and utilized pedestrians and cycling pedestrian contour feature to segmentation the region of interest. Second, the article completed the cascade classifier design which was based on Adaboost algorithm by learning and training the collected positive and negative pedestrian and cycling pedestrian, and the cascade classifier can effectively detect and identify the pedestrian and bicycle pedestrian goals in the scope of the candidate region. Third, this thesis implemented the tracking of pedestrian objects based on CamShift algorithm, under the HSV color environment, according to extraction of the color column diagram of pedestrians detected. In order to reduce the influence of factors just as object shelter and noise, on the basis of CamShift tracking, this paper uses the moving target prediction algorithm to improve the tracking accuracy of the target position. Therefore, this article used the CamShift and Kalman filter combination algorithm for object tracking prediction.The experimental results show that:the segmentation algorithm based on the edge feature can be good at extracting the interesting region of pedestrian or bicycle pedestrian. Recognition algorithm has high detection rate. Object tracking that combined with prediction algorithm has high accuracy and fast speed.