Visual Position And Tracking Of Dynamic Object Based On Computer Vision

Target visual location and tracking is a technology which combines target detecting, extracting,identifying, locating and tracking in the continuous dynamic image sequence, through the technologywe can obtain the target information such as location, size, speed, acceleration and trajectoryreal-timely, and achieve motion control, medical diagnostics, industrial inspection, three-dimensionalreconstruction and navigation of high-level tasks. Considering its great application value, more andmore people devote themselves to the domain. But the complexity of the practical applicationenvironments such as similar color interference, illumination changes and occlusion increase thedifficulty of development.This paper has a systematic study of the conventional image processing, target detection, targettracking, Kalman filtering and Particle filtering algorithms, and provide some improved algorithmsbased on the analysis and comparison of the characteristics of the traditional. The main ideas of theimproved algorithms are as follows: Firstly, we introduces the Kalman filter or Particle filter into thetraditional Camshift algorithm to enhance the prediction performance of the algorithm, and theimproved algorithm can solve the tacking failure problem of the traditional Camshift when the targetis seriously occluded; Secondly, in order to cure drawbacks of traditional Camshift algorithm whenthe target is interfered by similar color, the improved algorithm integrated particle filter, the targetcolor and contour information into the Camshift algorithm to increase the anti-jamming capability,and can solve the similar color interference effectively. Meanwhile, the Camshift algorithm can’tachieve the re-entry tracking after the target missing, in order to resolve this problem, the improvedalgorithm provide a global searching method and can track the re-entry target.Finally, the proposed improved Camshift algorithm based on particle filter and multiple featuresis carried on the computer for simulation and the improved Camshift algorithm based on Kalman filteris transplanted into the robot experimental platform, experimental results show that, the improvedalgorithms are simple and effective, have good tracking results.