Research On Autonomous Parking System Research And Design Based On 360-degree Panoramic Vision

With the improvement of people’s living standards and the progress of the automobile industry,automobiles has became the most convenient means of transportation for modern people.In recent years,the number of cars in China has increased sharply,which brings convenience as well as difficulties such as traffic jams and parking difficulties.The parking problem has became the most prominent problem.The automatic parking system can effectively solve this problem,so it has became a research hotspot in recent years.This paper mainly studies the automatic parking system based on panoramic vision.It makes a systematic introduction to the background of parking assist system,the research situation of the automatic parking system and the market status,which shows the significance of this article.At the same time,this paper designs three major functional modules to realize the automatic parking function,and the automatic parking system is divided into three modules:parking space recognition、path planning and path tracking^[1].The main design of each part is as follows:The first part is the parking space recognition module.In order to achieve a good detection effect,it is necessary to obtain the information of the parking space marking lines around the vehicle.Firstly,this article uses a panoramic view system which generate a real-time panoramic bird’s eye view by using four cameras installed around the body to complete the detection and recognition of parking spaces.Secondly,This article introduces in detail the process of camera model coordinate system conversion,multi-camera joint calibration and image stitching of four camera systems.These methods solve the problems of wide-angle distortion and four camera stitching and fusion.Generating a good panoramic picture by using the real car experiment.Thirdly,This article use the panoramic bird’s-eye view images for parking space detection and recognition.This paper overviews of Hough transform,LSD（Line Segment Detection）line detection based on line detection parking recognition method,and finally uses PSDL（Learning-based Parking-slot Detection）method to detect parking spaces.Compared with the traditional method of using straight line detection to identify parking spaces,the recognition shows better experimental results of PSDL through using these three methods to detect the actual parking space.The second part is the path planning module.The paper establishs the kinematics model of the vehicle at low speed by using the Ackerman steering principle.Based on the kinematics model,the arc tangency method and the arc tangent line method are simplified and determine the relevant parameters of the minimum parking space in parallel parking conditions and the constraints on the starting point of the vehicle.In the case of vertical parking,the paper use the combination of single arc and double arc to plan parking.Then,the vehicle decides to use the single arc or double arc method by measuring the side distance from the garage.This paper uses matlab software to simulate the proposed path planning method.Simulation experiments show that:under the premise that the parking space meets the minimum parking space,the vehicle can be successfully parked in the parking space,and the generated path is simple and easy to control.Finally,in the path tracking module.The control process based on preview tracking and dead reckoning is introduced in detail.The paper uses PD control based on preview to track the vehicle’s path.This paper build a joint simulation platform,and use Simulink and Carsim joint simulation to simulate the path tracking method to verify the feasibility of the method.