| With the transportation industry continuous developing and car owners increasing, traffic accidents occur frequently. Traffic accidents have a lot of harmfulness. The most dangerous situation is that a pedestrian crosses the road. Therefore a crossing pedestrian detection is of great significance to driving safely.This essay summarizes a variety of pedestrian detections and identification algorithm as well as designs a new crossing pedestrian detection algorithm on the basis of the existing algorithms. Taking the mutual blocked when a large number of pedestrians crossing the road into account, I improve the crowd detection algorithm. The system can alert drivers whether it is a single crossing pedestrian or crowd.Work done in this article:Making use of template matching reaffirmed the region of interest to solve inaccurate positioning of machine learning not only improves the detection rate but also plays a role in accurate position. In the template matching process, we extract the legs feature image with less noise, and layer template with the actual tilt angles of the legs.Innovative features of this system:The system has higher recognition rate as well as the accuracy location of pedestrian positioning and can recognize a person’s specific attitude, real-time.Innovation of this paper:1. Connected biological body structural characteristics and human leg periodic movement, the leg is split. The system extracts the image of the characteristics of a directional and gray difference combination of pedestrian leg.2. According to the human leg swing angle, we layer template samples. This method is effective to solve the template matching real-time problem.3. A new line detection method is presented. We vote a straight-line by its actual angle. Taking advantage of it, we improve crowd detection and we get a better detection result.The system has been installed in the test car and detected the actual road. It gets better detection results. The system will launch the product. |
PDF Full Text Request