Research On Object-tracking Method Of Portable Mobile Robot

Real-time tracking dynamic object in moving background is a study field of Automatic Target Recognition(ATR)which include segment target, pick-up character and recognize target. Although some achievements have been made in the studying of moving object detecting and tracing, how to detect and trace moving object in complex background and meet demand of real-time processing is still a hotspot of research in this domain because moving object oftenstands in the diverse surroundings.First, in this paper, the object-tracking methods existing classified, and some representative methods are analyzed, from following major issues: arithmetic content, hypothesis qualification, transcendental knowledge, calculating quantity, realizing difficulties and improved idea, from which proposed the selecting rule of object-tracking methods in different scene.Second, According to the moving background of robot object-tracking, we combined the accelerating algorithm with template-updating algorithm in correlation-matching, then, proposed a template-updating tracking algorithm with character restrict. It's basic idear is based on coarse correlation-matching and applied figure distinguishing arithmetic to rectify tracking point, which is the rule of template-updating. It can cut down calculation on correlation-matching and enhance veracity of object-recognition. The experiment result show that the algorithm is simple and steady ,it can be real-time realized by soft ware, At the same time, it can over come partial disturb of barrier and has a good value in practice.At last, in order to acquire 3D spatial and motion parameters of moving target, the limitation of second focus method to get the depth of target is analyzed by experimentation according to perspective projection model of target feature point. then, according to the land object, we build a robot perspective projection model based on the principle of object 3D information restoration, and proposed a arithmetic of 3D coordinate restorarion of object feature points. At the same time, we completed object movement parameters estimate and object tracking and prediction, and presented experimental result.