Abstract: A solar tracking approach based on fish-eye images is proposed.At first,by processing the(H,I,S) components,the high light areas are enhanced.Then by rectifying the fish-eye images into the perspective ones,the sun areas are recognized according to the contour,offset and magnitude of transformed areas;At last,the azimuth angle and the elevation angle are derived from the sun area barycentre.Experimental results show that our proposed approach can not only locate the solar orientation accurately in ordinary environment,but it is also workable under occlusion and image distortion.