Development of a Vehicle-View Autonomous Parking Space Recognition Model Using Mask R-CNN and YOLO Models

A deep-learning model that allows autonomous cars to recognize available parking spaces when attempting to search for spaces to park in parking lots is proposed herein. The proposed deep-learning model is designed to perform spatial recognition, i.e., partitioning the driving space and available parking space into segments from the perspective of the vehicle during autonomous driving. We propose the utilization of the Mask R-CNN model represented by a two-stage detection model in a computer vision field and the YOLO model represented by a one-stage detection model for the recognition of available parking spaces. The effectiveness of the proposed model is confirmed via a parking-space-recognition experiment in large indoor parking lots using the output results of the abovementioned two models.