Research On Pedestrian Detection Technology Based On Multi-channel Features

Pedestrian detection in machine vision, image detection and pattern recognition, etc,is used. It is based on image processing, optical electronics, computer and testing, and other modern technology together, constitutes an integrated system. The pedestrian detection technology has been widely applied in many fields, such as motor vehicle intelligent auxiliary system, intelligent monitoring system, advanced man-machine interactive system, etc. Pedestrians as a test subject, under the specific environment and the prerequisites, to analyze and understand the pedestrian has the vital significance, it is also the first premise of accurate and efficient pedestrian detection. Because pedestrians with nonrigid properties, however, in the actual detection at the same time,will be affected by the complicated background, light and shade, recognition and extraction of pedestrians increasing the difficulty, the pedestrian detection is a challenging research topic.First, this paper introduces the pedestrian detection research status at home and abroad, analyzes its research value and urgency. According to the existing pedestrian detection system, summarize the important and difficult pedestrians Detection technology exists, and according to the detected needs, research and selection based on a combination of performance characteristics of the detection system.Secondly, to extract the characteristic information of pedestrian accurate and clear, in the image preprocessing, weighted average method is used to analyse the image gray level stretch to enhance image contrast, median filter for removing noises, image histogram equalization enhancement, between 2 d the most kinds of threshold segmentation, fast accurate extraction under test target, so that it is separated from background region.Finally, according to the test results and the experimental analysis of the existing features, this paper proposes a pedestrian detection algorithm based on multi channel characteristics, to overcome the past in dealing with pedestrian detection feature small color image information and can not replace the shortcomings of weak features. This method compared with other single feature pedestrian detection method has good skillsand robustness, and can be achieved in a variety of environments better detection results.