Research On Vehicle Detection Algorithm Based On Visual Attention Mechanism

With the increase in the number of cars as well as the limited carrying capacity of theroads, the growing phenomenon of traffic congestion and traffic accidents increased year byyear, traffic safety, environmental pollution and other serious social problems are moreserious. The traffic problem has become an important factor of obstacles and constraints to thedevelopment of the city’s economic construction. Intelligent transportation system can carryout real-time detection on road traffic information and traffic information of the currentpavement. According to the traffic flow dynamic change information, the relevantdepartments can effectively control and guide of road traffic, alleviate traffic jam, reduce theoccurrence of traffic accidents, accomplish intelligent traffic management to improve thetraffic safety and mobility.The detection and recognition of the movement of the vehicle is the most important partof intelligent transportation systems in the most basic. This article will be on the basis of ananalysis of the existing sports vehicle detection algorithm added visual attention mechanismwhich is effective to improve the shortcomings existing in the original target detectionefficiency is low, a large amount of calculation. This article will present a moving vehicledetection and recognition of vehicle detection algorithm based on visual attention mechanisms.This paper is as follows:Firstly, the article puts forward the research of visual attention mechanism based onvisual attention mechanism based learning. Extract color, brightness, direction of the threeprimary visual features on the input image through the use of Itti visual attention model.Establish the visual attention saliency maps based on these primary visual features Thesevisual attention saliency maps analysis and fusion can obtain a final area of interest view andextracts a region of interest.Secondly, this paper proposes a vehicle detection algorithm based on visual attentionmechanisms and uses the classic model of visual attention mechanisms-Itti visual attentionmodel to extract the color, brightness, direction three static initial characterizations to formthe static feature saliency map. Fuse the static feature saliency map and the motion featuremap through the background difference method to extract vehicle target. The experimentsshow that the method can be effectively extracted interest of the target object and theexperimental results are better than ordinary vehicle detection methods. Extract the contour ofthe vehicles from the interesting images obtained by the contour extraction method and then calculate the external matrixes of the vehicles and accomplish the detection and recognition ofthe moving vehicles in the video images.Thirdly, count statistics of the vehicle of the video image with the method of setting avirtual detection line on the road by analyzing video traffic flow information. Detect theinstantaneous speed of the vehicle of the video image with the method of setting virtual coilson the road. Analysis and study traffic illegal violation of phenomenon and do technicaltesting and analysis of illegal parking, illegal lane change and illegal speeding.