Sensor Fusion-Based Parking Assist System

This paper proposes a parking assist system that fuses around view monitor (AVM) image, ultrasonic sensor, and in-vehicle motion sensor. The proposed system recognizes various types of parking slot markings using AVM image sequences and classifies occupancies of the detected parking slots using ultrasonic sensors. Once a desirable parking slot is selected by a driver, its position is continuously tracked by fusing AVM images and motion sensor-based odometry. Experimental results show that the proposed system can reliably detect and track various types of parking slot markings.