Abstract：According to the requirement of automatic parking system for identifying the parking slot by visual method, a parking slot recognition algorithm was proposed based on image gradient and Gaussian weighted cosine distance. The acquired parking space image was dedistorted, and the inverseperspective transformation was performed to generate a top view. The gray processing and the image filtering processes were performed. The gradient of the preprocessed image was extracted by the Sobel operator, and the parking space position of parking slot was selected. An image gradient template was created, and a rotation template was created by bilinear interpolation to detect parking spaces at different angles. The Gaussian weighted cosine distance was used as the similarity measure to detect the position of parking slot in the image. The termination condition of the algorithm was designed to improve the speed of the algorithm. The experimental results show that the multiple parking spaces can be recognized under different working conditions by the proposed algorithm with fast recognition speed and not easy affecting by cracks on the road surface.
