基于机器视觉的车辆检测与跟踪方法研究-检测技术与自动化装置专业论文.docxVIP

万方数据 万方数据 Abstract With China’s urbanization and high speed economic development, the number of vehicles in our country is rising significantly year after year. These tools give us much convenience,but at the same time it also brings a series of problems such as environmental pollution, traffic congestion and traffic accidents. In this context, intelligent traffic control system based on machine vision came into being. In this system, the vehicle target detection, tracking and behavior analysis are the most basic parts. From the most commonly used detection and tracking algorithms to start,this article mainly makes the following aspects of work: The traditional Gaussian Mixture Model algorithm only sets a fixed background learning rate, when the environment is mutated, the traditional algorithm is difficult to return to normal in a short time. Aiming at this problem, this article first designs a environmental change judgment factor, its value is the number of foreground pixels divided by the total number of pixels, then according to the factor value to adjust the background learning rate. Experimental results show that this method can help the detection system quickly overcome the effects caused by the environment mutation. The CamShift algorithm is easy to generate the problem of tracking failure when the target motion state changed, aiming at this problem,this article introduces the Kalman Filter based on CamShift algorithm. This article first uses Kalman Filter to estimate the motion states and then obtains the target’s pre-estimated location in the current frame, then runs CamShift algorithm at the estimated position to arrive the true position of the target. Experimental results show that this method can effectively help CamShift algorithm eliminate the effects brought by the changes of target motion state. The traditional CamShift algorithm requires people use hands to set which target to be tracked when tracking targets,and it can not realize multi target tra