Mobile tracking robot will play an important role in our future life. Target recognition, distance estimation and motion estimation are three primary functions of this kind of robot, respectively. Consequently, the irrelevant three technologies, pattern recognition, binocular vision and fast block-matching algorithm, make up of a novel approach of computer vision navigation for mobile tracking robot. This novel approach is analyzed in this paper with emphasis on distance estimation and motion estimation. Extended discussion of several interesting observations are also presented. Experimental results show the validity of the approach. Integrating the key components together, a simplified verification frame based on LEGO Mindstorms NXT is designed to check the three technologies combination.