Research On Pedestrian Detection Technology Based On Video Surveillance

Recent years witnessed the popularity of the application of video surveillance equipment, which make researchers’ enthusiasm in the field of machine vision based on various kinds of video monitoring technology research. Among which, pedestrian detection technology has a broad application prospect. At present, it has emerged a group of classic algorithms. However, the attitude, shield, environmental factors will influence detection accuracy, speed, etc, so there is still large room to improve.This essay proposes a pedestrian detection algorithm based on multi-channel feature by analyzing the predecessors’ research results, aiming at various problems existing in the classical algorithm. We build a multi-channel feature based on the improved histogram of oriented gradient, the PCA-SIFT and the feature of the color space. Multi-channel feature has the capacity of fast speed extraction, strong characterization and a good recognition effect. Experimental simulation, and comparation with algorithm based on the histogram of oriented gradient of the experimental results show that the proposed algorithm in detection rate and detection accuracy has a better detection performance, and is superior to histogram of oriented gradient algorithm.Furthermore, this essay builds a pedestrian detection system based on visual monitoring system using multi-channel feature algorithm. Visual Studio 2010 and OpenCV have provided a platform for the development. The system is mainly composed of environment initialization, read the video, extract the region of interest, pedestrian detection and so on several parts. The main innovation point is designed a pedestrian detection algorithm to improve the detection rate of the region of interest. In the experimental link, we compared the system designed in this paper with OpenCV computer vision library integration in the direction of the histogram of oriented gradient algorithm. The experimental results show that the designed system in such aspects as speed, accuracy, miss rate is better than that of the histogram of oriented gradient algorithm. In the practical application, it shows good performance.