An Image Based Detection of Pedestrian Crossing

Safe traveling of the blind and vision-disabled people is a trouble in their daily lives. Pedestrian crossing area is an important traffic sign which should be recognized in the image-based blind aid devices. This paper proposes a method for extracting pedestrian crossing based on image processing, which contains bipolarity testing, morphological operations, edge detection and radon transform techniques. By introducing a parameter "bipolarity" which represents the gray level contrast in an image, areas with strong contrast were selected. Morphology processing approaches were used to analysis and process noises in bipolarity image. According to the corresponding relationships between an image and its radon transform result, pedestrian crossing features, such as number and edge of pedestrian crossing stripes were extracted in transform domain. This algorithm was proved to be effective with 96.2% accuracy under the test of 54 real crossing images.