A Color-Based Target Localization Method and Its Implementation on Embedded Auto-Mobile Robot System

In this paper, a target localization method based on color recognition and connected component analysis is presented. The raw image is converted to HSI color space through a lookup table, followed by a line-by-line scan to find all the connected domains. By checking the size of each domain, most pseudo targets can be omitted and, meanwhile, the target position would be calculated. Owing to the absence of extra filtering process, the real-time performance of the whole system is greatly improved. Experiments on our ARM system demonstrate that the robot could trace multiple targets under a mean frame rate of 15FPS, including decoding process, which satisfies the requirement of real-time video processing on auto-mobile robot systems.