Robot target recognition and tracking based on panoramic vision

Abstract

Aiming at the problem of the algorithm of target recognition based on SIFT(Scale invariant feature transform) feature matching is poor at celerity and the algorithm of target tracking based on the target's color feature is easily affected by outside. A target recognition algorithm based on SIFT which based on color guide and a target tracking algorithms based on compressed sensing theory is presented. According to the color image information such as pre-preprocessing identification, identification of the target region, Obtained according to pretarget template, after recognition of the pre-image for SIFT feature matching, can be more accurately identify the target, and determines the target pixel coordinates in the image according to the determined target pixel coordinate. Used compressed sensing target tracking algorithm, you can target the implementation of tracking. The experiments prove that these two algorithms can be applied to panoramic visual object recognition and tracking.