Research On Parking Space Visual Recognition Technology For Automatic Parking

The advancement and development of automotive intelligent driving technology has promoted the transformation and upgrading of the automotive industry.As an important part of intelligent driving technology for cars,automatic parking technology promotes the solution to the problem of “difficult parking” in urban transportation,reduces local traffic congestion,saves parking time,and improves urban living efficiency,and promotes the development of the city to a certain extent.Traditional automatic parking technology uses ultrasonic radar to detect and identify parking spaces.It relies heavily on parked vehicles around the target parking space.When there are no parked vehicles around the target parking space,it cannot detect and identify the parking spaces.Computer vision-based automatic parking technology overcomes the shortcomings of traditional automatic parking technology and broadens the application scenarios of automatic parking technology.This paper researches and summarizes the parking space visual recognition technology for automatic parking.In order to realize the accurate identification and precise positioning of parking spaces,a parking space recognition algorithm and ranging positioning algorithm based on monocular vision are proposed,and the actual vehicle test verification is performed.The research contents of this paper mainly include the following aspects:1.The camera’s imaging model and projection principle are studied,and the camera calibration method is analyzed.The camera is calibrated by Zhang Zhengyou’s calibration method,the camera parameters are obtained,and the error analysis is performed.2.Research on the parking space recognition algorithm based on monocular vision.Aiming at the influence of poor lighting conditions and environmental disturbances on parking space recognition,a parking space visual recognition algorithm using multiple image features and integrating different image segmentation methods was proposed.The chroma feature map and brightness feature map of the parking space image are extracted through color space conversion,and they are enhanced respectively to reduce the influence of lighting conditions.The image segmentation of the feature map is performed by fusing the fuzzy C-means clustering algorithm and the region growth algorithm to extract the parking space area to reduce the impact of environmental disturbances.Finally,the parking space skeleton is extracted,the parking space line is detected,the corners of the parking space are obtained to identify the parking space.3.The distance measurement and positioning algorithm of parking spaces is studied.Based on the identification of parking spaces,the needs for ranging and positioning of parking spaces are analyzed,and the monocular ranging and positioning algorithm is used for ranging and positioning of parking spaces.By comparing and analyzing different types of monocular ranging and positioning algorithms,a ranging and positioning model suitable for parking spaces is established and experimentally verified.4.Real vehicle test and result analysis.By setting up software and hardware experimental platforms to identify and locate parking spaces in different scenarios,the effectiveness and feasibility of the parking space recognition algorithm and ranging and location algorithm proposed in this paper are verified.