Vision Based Monte Carlo Localization For Mobile Robot

Mobile robot automatic localization is the premise and basis for real intelligent mobile robot. Robot self-localization is closely related with the type of sensors, because environmental information obtained from the sensors carried by robot is used to achieve the mobile robot position. Vision-based mobile robot localization is paid more and more attention because of the unique advantages of vision sensor. But the visual localization also exists many suspects and difficulties which is caused by complexity and instability of visual information processing to be overcome. In order to achieve accurately the robot localization, using effectively visual information, monocular vision-based mobile robot probabilistic localization is studied in the unknown environment. The major research contents are as follows:Firstly, this paper summarizes the background and significance of the research, and analyzes the status and existing problems of the vision-based mobile robot localization algorithms in detail, and establishes a variety of mobile robot system models, these systems model are the basis of a platform for the vision-based mobile robot localization problem.Secondly, introducing SIFT algorithm which is the most popular local feature extraction algorithm. SIFT algorithm is improved to reduce the number of redundant feature points. The performance of the improved SIFT algorithm is verified by the simulation experiments.Thirdly, the hierarchical image matching algorithm for the real-time and accuracy requirement to process image is proposed in the field of image appearance-based mobile robot localization, then combining with Monte Carlo algorithm to achieve the mobile robot global localization accurately. The process of mobile robot localization is accelerated by the hierarchical matching, which combined the global feature with the local feature. The performance of the hierarchical image matching algorithm is verified by the simulation experiments, and experimental results that mobile robot localization method combined with Monte Carlo algorithm are effective.Finally, a square-root cubature particle filter based Monte Carlo Localization algorithm is proposed to alleviate the sample degeneracy and impoverishment of particle filters in robot Monte Carlo Localization. The importance sampling and resampling are improved in the proposed algorithm to alleviate the sample degeneracy and impoverishment of particle filters in mobile robot MCL. The performance of the improved MCL algorithm is verified by the simulation experiments.