关键词: 斑马线, 双极系数, 形态学, Radon变换

Abstract: A new method to effectively detect pedestrian crossing in the image-based blind aid devices is proposed.Candidate regions of a crossing are selected by careful evaluation of the “bipolarity” which represent the gray level contrast in an image，and the noise of bipolarity image is eliminated by using morphology processing approaches.According to the corresponding relationships between an image and its Radon transform result，pedestrian crossing features，such as number and length of pedestrian crossing stripes，angles in vision field，are extracted in transform domain.This algorithm is proved to be effective with 96.2% accuracy under the test of 54 real crossing images.

Key words: pedestrian crossing, bipolarity, morphology, Radon transform