Motion Target Detection And Tracking Based On Video Sequences

Motion target detection and tracking based on video sequences is a hot topic in the field of computer vision,also the key issues to be solved in a lot of applications which based on vision. Moving target accurate detection and tracking is the premise for valid advanced subsequent behavior recognition and decision-making, cornerstone of high-level vision. Moving target detection and tracking technology play a very important role in the applications of human-computer interaction, automatic driving, safety monitoring, missile guidance and intelligent transportation systems. But in the practical application of this technology still faces many difficulties: the moving of camera and mutation of light will increase the difficulty of target detection, and the complex background, the changing prospect will affect the performance of tracking. To solve these problems, this paper made the following works:In this paper, many algorithms for moving object detection in video sequences were analyzed, and the frame difference, background subtraction made were carried out experiments. The background subtraction method for extracting background was implemented on the basis of this, and used in the subsequent experiments.The MeanShift tracking was implemented in the process of researching tracking algorithms, the gradient for target as the distribution characteristics of the algorithm, constructed the optimal function for MeanShift algorithm, using the convergence rate for target tracing. SIFT algorithm detailed analysis subsequently, the improved SIFT algorithm is proposed based it's insufficient. The improved SIFT algorithm was combined with background frame difference method, and experimented on moving target tracking system. This method is also applied based on SIFT feature information to determine priority for retention method, has a good effect on the target blocking and deformation.