Research Of Multi-Target Tracking Algorithm Based On Multi-Object Feature Obtain

Multi-target detection and tracking technology as one of the core technology of digital images, computer vision, has gained more and more attention from the scholars home and abroad, and has made some achievements in scientific research, It’s achievements have been widely used in civil and military, In civilian is mainly used in areas such as intelligent video surveillance, medical diagnosis, In military is mainly used in missile precision, airborne early warning, etc. At the same time, the stand or fall of target detection and tracking results more also directly affects the subsequent scene analysis, behavior understanding basic operations, etc.At present, The main problems of multiple target tracking is the complexity of the background, the similarity of the target color and the background color, and other aspects of mutual occlusion between multiple objectives. Considering the problem of multiple target tracking, Firstly, introduce the classical algorithms in this paper, then analysis the advantages and disadvantages of those algorithms, meanwhile for the external interference factors which influence the stability of the algorithm is analyzed. On this basis, launches research of multi-target detection and tracking algorithm which this paper put forward. Main works of this paper as follows: 1. Put forward the moving target detection algorithm based on texture feature analysis. in this paper, considering the traditional target detection algorithm in the separation of moving targets vulnerable to outside factors interfere、Self-interference and other issues. Put forward the symmetric frame differential method and background subtraction method to extract the motion area, After that by extracting and matching the texture of the motion region and background it can distinguish false targets. Through the contrast test results, this method has a certain stability and anti-jamming.2. Put forward the multi-target tracking algorithm based on multi-object feature obtain. On the basis of target detection Through comprehensive similarity matching color and geometric characteristics of the moving object to achieve data association, Then create and update the characteristic parameter information table, And through to the target state estimates, and combining the comprehensive similarity of target parameters in the target sheltered good tracking technology, Accurate tracking during mutual occlusion between the target。Finally, through the experimental comparison and analysis, the algorithm of this paper has demonstrated stable tracking capability.