Research On Real-time Parking Space And Parking Space Line Detection Algorithm Based On Vehicle-mounted Camera

In recent years,the continuous integration of automobile industry and information technology makes the unmanned driving technology develop rapidly in the industry,prompting more and more people to study it.With the ceaseless growth of domestic economy,the number of domestic cars is increasing every day.Traffic accidents happen every day,and the traffic rate of urban roads is low,which seriously affects people’s lives,and automatic driving has become an important way to solve these problems.In the process of automatic driving,automatic parking is also the key research direction.At present,the parking system based on ultrasonic and infrared radar is widely used in automatic parking system.However,when there is no object around the target parking space as a reference,the system cannot find the target empty parking space,so people begin to solve the problem from the visual point of view.In this paper,a real-time parking space and parking space line detection algorithm based on on-board camera is proposed.This method can accurately and efficiently detect idle parking spaces in different situations,and provide information for automatic parking control system,so that cars can park in parking spaces accurately.This paper mainly completes the research and implementation of the following aspects:Considering the safety of the car parking process,the visual field should be as wide as possible,In the process of camera selection,the number of cameras selected and the installation method of cameras are studied.On the basis of summarizing the previous parking space detection methods,this paper proposes an improved parking space detection algorithm based on feature combination.Firstly,different parking space scenes are classified,and different detection algorithms and judgment methods according to features are designed to determine the location of the parking space.In the process of car parking,the left,right and rear parking spaces are tested to improve the success rate of car parking.And a new coordinate system is established by taking the automobile as the center,the results detected by the first two aspects are found to correspond with the newly established coordinate system,and the corresponding distance and angle information are output,thus providing effective information for the automatic parking control system.Aiming at the research of real-time performance,starting with hardware selection and algorithm,Open MP multi-threading method is adopted to achieve real-time detection effect.The detection algorithm is transplanted to the TX2 embedded platform,and the experiment proves that the detection algorithm has good effect on the detection of parking spaces in different scenes and can eliminate the problems of shadow,obstacle occlusion and the defect of parking spaces to a certain extent,which has good reference value for further application.