Autonomous mobile robot motion control in non-structured environments based on real-time video processing

This paper describes a method for mobile robot position estimation based on feature tracking through the sequence of images provided by a conventional video camera. The block-based correspondence algorithms selects candidates through grey level matching and local coherence analysis; a DSP-based multiprocessing image processing system is used to achieve real time performance. Point-to-point correspondences are used to estimate the motion parameters of a mobile robot. Motion-based procedures to eliminate false matching have been implemented. The method has been applied to a mobile robot in non-structured and outdoor environments.