Research And Realization Of Car Side Parking Environment Perception Based On Line Structured Light

In recent years,the advent of the Internet has provided powerful technical support for the development of the automobile industry,making the automatic parking system an important direction for intelligent vehicle research.In the process of automatic parking,the key to parking is the accurate detection of parking spaces.The rapid development of machine vision has made vision-based measurement an important aspect of 3D measurement.And the three-dimensional measurement of line structured light is widely used in industrial production due to its characteristics of non-contact,high precision and real-time.It has become a research hotspot to three-dimensional information of a cross section of the measured object is solved by using the three-dimensional visual measuring principle of the line structured light.At the same time,the development of cutting-edge technologies such as intelligent control technology and computer technology has made the technology of unmanned smart car an important direction for the development of science and technology in the future.Automatic parking technology is another key technology used in the research of unmanned smart car.Combined with the current research on the application of unmanned car,this paper studies and realizes a method of car-side-parking environment perception based on line structured light.The main contents include:(1)The overall scheme of car side parking environment-detecting system is designed and the car side parking environment-detecting system based on the line structured light visual sensor is set up.Based on the working principle of line structured light vision sensor,the structure scheme of line structured light sensor + camera is designed;According to the actual working conditions,we select the camera,line structured light projector,filters and other hardware,we also install and fix the line structured light visual sensor;researching and designing the overall detection program of car side parking environment-detecting system.(2)The triangulation principle of line structured light vision sensor is analyzed.A measuring model based on perspective projection of the car side parking environment sensing system is established.An integrated calibration method based on planar checkerboard for the car side parking environment sensing system is proposed and implemented.This method calibrates the parameters of the camera and light plane equations at one time which reduces the experimental content and improves the calibration efficiency of the system.By analyzing the accuracy of the integrated calibration of the car side envirornment-detecting system,it is verified that the system integrated calibration method has high accuracy and good practicability.The accuracy of the integration calibration of the lateral parking environment sensingsystem is analyzed,which validating the high accuracy and good practicality of the method of the system integration calibration.(3)A full-resolution sub-pixel coordinate extraction algorithm of light-stripe center based on Zhang thinning method and gray-scale centroid method is proposed and implemented.Based on the analysis of the characteristics of line structured light stripe,Zhang thinning method is used to obtain the initial value of the light-stripe center for the parking image.Drawing on the idea of gradient centroid method,the full-resolution extraction algorithm based on grayscale centroid method is adopted to realize the sub-pixel coordinate extraction of the light-stripe center.According to the experimental results,it is verified that the algorithm has the characteristics of fast speed and moderate accuracy of extraction,and meets the real-time requirements of the car side parking environment sensing system.(4)For the visualization of the detecting results,another sub-pixel coordinate extraction algorithm of Hessian matrix based on ROI and threshold method is proposed and implemented.The center-extraction algorithm of line structured light stripe using the Hessian matrix method is studied.Two problems exist in the Hessian matrix method.One is the high requirement of the image quality,the other is the inefficient computation.Thus,a fast sub-pixel center extraction method based on ROI,threshold method and Hessian matrix method,is adopted.According to the experimental results,it is verified that the algorithm has the characteristics of high extraction accuracy and good robustness,and moderate extraction speed which meets the requirements of 3D reconstruction in the environment-detecting system of car side parking.(5)The software design of the environment-detecting system of car side parking is given,and the algorithm based on MFC and OpenCV is realized.The standard box is used to verify the measurement accuracy of environmental detection,and verify the feasibility and correctness of the proposed method by the experiment of 3-D reconstruction.