Design And Implementation Of Vehicle Detection And Tracking System Based On Video

With the rapid development of science and technology,automobile has become an indispensable means of transportation in our life,but it brings convenience to people’s life at the same time,it also brings a series of problems.With the increase of the number of automobiles,traffic congestion and traffic accidents occur one after another.In order to solve these problems,it is necessary to deal with the traffic environment intelligently,so as to achieve intelligent transportation.The basis of intelligent processing is to detect and track the moving vehicles in the surveillance video,so as to realize the practical functions of classifying the vehicles,judging whether the vehicles are running on the prescribed lanes and tracing the vehicles causing accidents.Based on the above background and actual project requirements,this paper proposes and implements a vehicle detection and tracking system based on video image.The specific contents of this paper are as follows:Firstly,the main scheme of vehicle detection and tracking system is designed.The system requirements are divided into two aspects: performance and function.Then,through the research and discussion of the target tracking field,the system functions are set.The long and short video processing is used as the basis of classification,and the tracking method is planned.Secondly,the main tools of system development are introduced.Finally,the specific system development plan is formulated.Secondly,according to the design scheme,the short-time tracking system is designed.The key idea of video tracking method design is to obtain the optimal value by iteration and random sampling and resampling of particles.The Kalman and particle filter principles are systematically studied and analyzed.On this basis,the design method of the system is proposed,and the design ideas of the main parameters are explained.At last,the main design block diagrams of two filtering and tracking methods are given.Thirdly,According to the requirement of long-time tracking,a method of combining tracking and detection is proposed.On this basis,SIFT is taken as the starting point to improve the tracking accuracy.Long-time video tracking will lead to poor tracking effect due to adverse environmental factors such as occlusion.For this reason,TLD(Tracking-Learning-Detection)framework tracking idea is introduced.The idea is applied to the system to enhance the stability of vehicle tracking system to deal with adverse environment and achieve the goal of long-term tracking.The improvement idea is put forward to make the robustness of the tracking system to deal with the adverse conditions such as strong change of illumination and camera rotation to a higher level.Finally,the system is programmed and tested.The subordinate relation of each module of the system to the program class is taken as the starting point,and the system programming idea is improved,and the core code is explained.Then,a set of detailed system testing scheme is worked out to test the system function in detail,and the test results are analyzed systematically in order to evaluate the system performance.The system test shows that the vehicle detection and tracking system based on video image can reach the standard of intellectualization,high efficiency and representativeness.The system runs stably and has high accuracy and practical application value.For the field of intelligent transportation,it provides a good solution and operation platform for vehicle and road information processing.