Semi‐automatic parking slot marking recognition for intelligent parking assist systems

Abstract

This paper proposes a semi‐automatic parking slot marking‐based target position designation method for parking assist systems in cases where the parking slot markings are of a rectangular type, and its efficient implementation for real‐time operation. After the driver observes a rearview image captured by a rearward camera installed at the rear of the vehicle through a touchscreen‐based human machine interface, a target parking position is designated by touching the inside of a parking slot. To ensure the proposed method operates in real‐time in an embedded environment, access of the bird's‐eye view image is made efficient: image‐wise batch transformation is replaced with pixel‐wise instantaneous transformation. The proposed method showed a 95.5% recognition rate in 378 test cases with 63 test images. Additionally, experiments confirmed that the pixel‐wise instantaneous transformation reduced execution time by 92%.