Research And Simulation Based On Feature Matching Target Tracking

Target tracking algorithm and the related technology is an important part of the field of computer vision tracking. The importance and necessity of the research are becoming the focus of the research. With the large-scale development of civilian industry and military industry, especially in the city safety monitoring, intelligent traffic surveillance, medical imaging, weapons guidance navigation, combat command system, high resolution digital video image sequence of target tracking technology is already growth explosive. In the practical application scenarios due to ground maneuvering target in complex background environment(such as buildings, plants occlusion), as well as the target itself turning, rotation and target scale. How to meet the above requirements, while ensuring the tracking stabilization, is the focus and difficulty of the current research.Tracking algorithm for maneuvering targets on the ground, many scholars paid great effort, also receiving great achievements, but according to the specific object and its motion tracking algorithm research, due to the scene and target different certain differences.This paper set about from the feature matching method and combining with the feature point selection method, to study and research of related algorithm, combined with the actual problem, for the ground maneuver target in the condition of random speed, through changing the light area, partial shade area and target their turn or rotate and target image acquisition equipment scale, etc., the local optimization and improvement.The main work of this paper has the following five aspects:(1) According to the practical problem, a kind of algorithm which can track the target of the mark is able to track the target in the complex background. Maneuvering targets on the ground in the condition of random speed, through the light change area, partial cover area and the target and the target image acquisition equipment turn or rotate its to zoom in, four cases, able to track to mark the algorithm of target.According to the practical problem, a kind of algorithm which can track the target of the mark is able to track the target in the complex background, such as the speed change, the angle change, part of occlusion and the scale of the four kinds of the moving target.(2) In-depth study the feature points matching algorithm, combined with feature point matching, FAST matching algorithm, BRISK feature matching algorithm, solve the correlation algorithm in for tracking maneuvering targets on the ground, the single frame time cost. Based on this problem, we proposes a feature cluster tracking algorithm and a consensus feature matching algorithm.(3) Surround on the long term tracking of the target of model-free model. In real life and work, ground target shape, posture, etc., there is a big difference, by manually labeled goal although it can save large amount of prior knowledge learning, but for the mastery of the target profile and shape decreases. To describe the model-free feature point, through the screening of effective feature points, to achieve the long-term tracking target.(4) In view of the high resolution image sequence, the single frame image resolution is 1600?1200 pixels, 4:3 aspect ratio, on the selected both have stable target tracking,.(5) CCS simulation. Rewrite the program code for the C language, and in the CCS hardware simulation platform for simulation.