Learning-Based Parking-Slot Detection

The vision-based parking-slot detection has caught much attention in the academic research and actual applications for its promising performance. In this paper, we propose a vision-based parking-slot detection algorithm, which can detect parallel parking-slots and vertical parking-slots simultaneously by detecting their T-shaped markings. Further more, we can also detect some standard 45-degree and 60-degree inclined parking-slots. Our method is based on the adaboost algorithm in machine learning, and achieve more than 95% accuracy on our test set. The proposed method has a certain tolerance for lighting changes and poor clarity of some parking-slots. At the same time, we propose a secondary detection method based on deep learning for the case where the two internal corners of the parallel parking- slot don’t use standard marking lines, and determine if the parking-slot is occupied.