Monocular-vision-based parking assistance for constrained environment

This paper presents a parking assistance system using visual information. The system input is an image taken by a single camera fixed to the vehicle. In the input image, we detect feature points corresponding to parking slots, among the keypoints detected by SIFT algorithm. The vehicle's location is estimated using detected feature points, and the vehicle's motion is expected using state of vehicle's wheels. Then, the system displays where the vehicle is expected to move in the parking lot based on the current location and the current wheel status. The experimental results show that the presented system can assist a driver in parking the vehicle.