Parking slot detection system based on structural similarity index

Abstract

A quest of vacant parking space in the public area can indirectly lead to traffic congestion which can be troublesome for drivers in terms of time efficiency. This study is expected to assist drivers to get the available parking slots information in a real-time manner and support the parking control systems by constantly updating the information of vacant parking slots positions in public areas. A vision-based parking slots recognition method is proposed to identify occupied areas by vehicle which is divided into two main parts: setup configuration and object detection. Canny Edge and Hough Line Transform are used to achieve line detection for parallel parking slot marking; contour extraction and bounding rectangular are then applied for an initial parameter to form a reference area as a region of interest (ROI). Moreover, Structural Similarity Index Measurement (SSIM) exploits the reference image and target image to identify whether the area is empty or occupied by vehicle depending on structure comparison. Experimental result shows, from 50 sample images of parking slots attained by surveillance camera, the detection accuracy of 92% and precision of 89% are obtained using selected features with tuning SSIM threshold level of 0.4.