Parking Slots Intelligent Detection And Parking Path Tracking Based On Around View System

With the transformation and upgrading of automobile industry,vehicle intelligence is becoming development trend.As one of the core technologies of driverless,automatic parking attracts the attention of major automobile manufacturers and research universities.Most of the automatic parking system products on the market with ultrasonic sensors to obtain parking slot information,This scheme highly depends on the existence and posture of adjacent vehicles and there are fewer parking scenes that can be detected.While the vehicle backup camera has a small visual field,which can’t meet the technical requirements of parking slot detection.Around view system is widely used in parking assist system due to its wide field of view and low cost.This paper focuses on the technology of parking slot detection and parking path planning and tracking method based on the around view system,The main research contents of the paper are as follows:(1)Aiming at the problem of reduced parking slot detection accuracy and stability caused by illumination changes,a parking slot detection algorithm based on deep learning was proposed.Through the calibration experiment of the fisheye camera and the vehicle and the design of a two-dimensional mapping table,around view system was built to provide original training samples.Using deep learning model to detect marking-point,and combine with Line Segment Detector to obtain complete parking slot.The test results show that the proposed method not only increases the detection types of parking slots,but also improves the detection rate and stability,the accuracy and recall reached 93.9% and 92.5% respectively.(2)Aiming at the problem that the accuracy of judging the parking slot status base on single feature is low,a judgment method combining road features and vehicle features was proposed.The image features of the vacant and occupied parking slot were analyzed,The region growing algorithm and Canny edge detector were used to extract the region growth rate and edge number of the parking slot respectively,base on the above feature information,the Naive Bayes Classifiers was used to judge the status of parking slot.By comparing and analyzing the results of parking slot state detection in different scenarios,the effectiveness of the designed algorithm was verified.(3)By analyzing the model of vehicle kinematics and parking slot requirements,A smooth and continuous three-stage parking path was designed and simulated.Aiming at the problem that the real-time control accuracy of parking path tracking is low due to external unknown disturbances,the vehicle tracking error model was built and an backstepping adaptive path tracking control algorithm was designed,Carsim-Simlink simulation were performed.The simulation results show that the controller has good tracking control performance,which improves the tracking accuracy and stability of parking.Finally,a vision sensor-based automatic parking system framework was designed,and real vehicle test were carried out for typical parallel parking scenarios,the experimental results verify the effectiveness of the designed parking path tracking control algorithm.