METHOD FOR CATADIOPTRIC VISION BASED LOCALIZATION AND MAPPING IN A PARTICLE FILTER FRAMEWORK

摘要

A method for catadioptric vision based localization and mapping in particle filter framework is provided to reduce the amount of calculation necessary for a SLAM(Simultaneous Localization And Mapping) process by alternately updating the position states of a moving object and a feature point. An omnidirectional panorama image for exploring environment is obtained at the present position of a moving object, and each position of feature points in the exploring environment is detected as an azimuth from the moving object by using the omnidirectional panorama image(400). The posterior probability distribution of the present moving object position is updated by using the detected azimuth and the prior updated probability distribution of the feature points(410). The posterior probability distribution of the feature points is updated by using the detected azimuth and the posterior probability distribution of the updated moving object position(420).