Automatic Parking System Research Based On Visual Environment Perception

As a convenient means of transportation,automobile has become the most common way of travel in People's Daily life.However,with the increase of the number of vehicles in cities,parking space resources in cities become increasingly strained,and parking difficulties become more and more prominent.Automatic parking system is not only an effective means to solve the problem of parking difficulty,but also an important part of intelligent cars.Most of the existing automatic parking systems are based on ultrasonic radar,millimeter wave radar and other ranging sensors for environment perception,to obtain less information from the environment and low intelligence.This paper studies the automatic parking system based on visual environment perception,designs the parking slot recognition method based on visual sensor and the vehicle pose estimation method based on visual and wheel speed sensor information fusion.The overall architecture of the parking system is designed,the parking system is simulated and tested on a real vehicle.Firstly,the theoretical basis is studied,and the pose description is analyzed.Firstly,the pose problem is decomposed into two parts: rotation and translation.Homogeneous coordinates are used to describe the pose in two-dimensional space.This paper introduces the relationship between the camera imaging model and the four coordinate systems in the imaging process,and studies the imaging principle of the fish-eye lens used in the follow-up research of this paper.Secondly,the method of parking slot recognition based on vision sensor is studied.Image distortion and inverse perspective transformation can be realized by preprocessing.The shape and size of different kinds of parking space lines were determined according to the relevant standards,and the characteristic position shape of the parking space,namely the position of the parking slot angle,was selected as the target position of image matching according to the standards.Sobel operator was used to extract the gradient between the horizontal and numerical directions of the image,and the matching template was made from the standard parking slot angle image.The bilinear interpolation method was used to make the template adaptive to rotation.Thesimilarity between the template pixels and the image pixels to be measured is expressed by cosine distance,and the matching similarity is obtained by weighting the matching values of the template pixels with gaussian convolution.The termination condition of the algorithm is designed to improve the efficiency of the algorithm.The validity and timeliness of the recognition algorithm are verified by testing the algorithm by selecting parking slot images of different scenes.Then,the problem of vehicle pose estimation in the process of automatic parking tracking control is studied.The vehicle position estimation method based on the wheel-speed sensor is studied firstly.The displacement and orientation of the vehicle are calculated by using the wheel-speed sensor information as the input of the vehicle motion model.After that,the visual information and wheel speed sensor information are fused,the observation information of the visual sensor is used as the observation equation of the system,and the Taylor expansion function is used to linearize the matrix function of the vehicle motion system model and observation model,and the sensor information is fused by the extended Kalman filtering method.Simulation results show that the proposed method is effective in vehicle pose estimation.Finally,the architecture of the automatic parking system is designed.The system is integrated and simulated based on Matlab/Simulink.The vehicle kinematics model,wheel speed sensor model,visual sensor model vehicle pose estimation model,Relative tracking error model and the parking control model based on incremental PD control are designed.Through simulation test,the vehicle pose estimation based on information fusion and the tracking of preset trajectory are realized,and the effectiveness of the system function is verified.The feasibility and practicability of the proposed method are verified by a real car test.