Design And Implementation Of Real-time Video Target Tracking System

Video target tracking is one of the popular research directions of computer vision,and is widely used in transportation,medicine,military and other fields.Currently,target tracking is developing rapidly,but in the scenarios of fast moving,occluding or losing targets,the target tracking effect needs to be improved.Therefore,how to achieve effective real-time target tracking has become an urgent engineering problem.This paper designs and implements a real-time video target tracking system,which has a good performance in real-time and long-term target tracking,and can complete the engineering requirements for acquiring video information and tracking specific targets.The main work of the thesis is as follows:(1)Design the overall scheme of a video target tracking system.First analyze the functional requirements and performance requirements of the target tracking system to clarify the engineering requirements.Then design the system framework to ensure that the work logic is reasonable.Also select the appropriate development platform and programming language.Next,design the board control function to complete the control of the working state of the video capture card.Finally,the specific functions of each module of the system are designed in detail.(2)Improved traditional ASMS algorithm for target tracking.Aiming at the real-time target tracking implementation requirements of the system,this paper chooses the less complex ASMS algorithm as the main research object,and proposes an improved scheme for the shortcomings of the algorithm.In the actual scenes such as fast moving and target occlusion,the ASMS algorithm has poor tracking accuracy and is easy to cause target tracking deviation.Moreover,the algorithm has no re-detection mechanism,and the target loss will cause tracking failure and cannot achieve long-term target tracking.Aiming at the above problems,this paper proposes the decision strategy of overlapping rate of adjacent frame target regions.When the target moves fast,the algorithm increases the search range according to the moving direction to ensure that the search box covers the target.At the same time,by using the movement information of the target in the video sequence,the initial value of the current frame target position estimation is reset,so that the setting of the value in the algorithm is more reasonable.Add the re-detection mechanism and adopt the backward position consistency check method.If the forward and backward position estimates are inconsistent,it indicates that the target tracking is biased,and the search window can be adjusted.If the target is lost,the re-detection mechanism is enabled,and the SURF feature matching strategy is used to perform target detection near the target lost position.When the target is re-detected,the target position and size information is updated,and the target tracking can be continued,which improves the robustness of the target tracking algorithm.Finally,the improved target tracking algorithm performs algorithm performance evaluation on the international standard dataset(OTB100 and VOT2016)test platforms.The results show that compared with the traditional ASMS algorithm,the improved algorithm has better tracking accuracy and robustness,especially in the scenes of target occlusion,fast moving,target loss.And the algorithm frame rate is up to 89 fps,which meets the requirements of real-time video target tracking.(3)Complete real-time video target tracking system.In this paper,the improved ASMS algorithm is used as the target tracking algorithm of the system image processing module and the target detection is introduced.So the tracker is initialized by using the detected target position information to realize the single target tracking function of the collected video.The system provides a human-computer interaction interface,and the operator can control the entire workflow through system software.Finally,the experimental results show that the video target tracking system designed and implemented in this paper can complete video acquisition and target tracking.Under the condition of video resolution of 768?576,the system running frame rate reaches 26 fps,which meets real-time performance and has certain engineering applications.