In order to improve adaptability for environment and reduce influence of illumination and stains on road image

an algorithm for lane detection based on visual sensor is proposed. Firstly

effect of illumination and stains is analyzed

and improve region growing method based on priori knowledge

including projection principle

then segment image and divide boundary region of road

then points set of lane feature is accurately obtained using fused edge detecting data

Hough transform is used to extract straight line to match road straight line model in lane detection. Experimental results indicate this algorithm effectively decrease interference by illumination and stains and improves robustness

accuracy rate is 97 %.