基于全方位视觉的移动机器人动态目标识别和定位研究-机械工程专业论文.docx

???????????????????????お ???????????????????????お DYNAMIC TARGET RECOGNITION AND POSITIONING BASED ON THE COMPREHENSIVE VISION OF MOBILE ROBOT ABSTRACT The lenses panorama of the three-dimensional visual image system which based on the fish-eye-lens is widely used for visual navigation of micro robot and close big view objects recognition and positioning. Fisheye lenses, which the angle of view closes 180 ° , can achieve close or super close big field three-dimensional visual perception. This visual perception techniques is widely concerned and hot researched in recent years. This paper studies the optical flow method identify maneuvering target which can detect moving object in the scene which not know any of the information. The maneuvering target tracking algorithm suitable for any available state space model of nonlinear non-guassian system ,which based on the classical theory of particle filter ,and its precision is very close optimal estimation. With the support of the 863 National Program of China—Based on the omni-directional mobile robot dynamic visual target detection and navigation?NO. 2007AA04Z229??the paper has conducted a deep research in the dynamic target recognition and positioning based on the comprehensive vision of mobile robot. There are the main achievements and the conclusions obtained as follows: Firstly, established the all-around vision image which based on the fisheye lenses aberrations correction system. Based on the cube projection model to rectify the fisheye images, and use the fisheye lenses aberrations correction system which based on the cube projection model for fisheye images on the calibration. Secondly, the improvement of optical flow method can search the characteristic points which matched in the image, detect the moving target of image sequences. Using the characteristic of target, through the particle filter forecast in subsequent frames of the motion parameters , analyzing the matching matrices between neighboring frame