Research And Implementation Of Garage Parking Position Detection Based On Binocular Vision

The automatic parking technology plays an important role in the field of driving assistance because of the advantages of time saving,good safety and high success rate.The existing automatic parking system mainly uses ultrasonic or laser radar to detect the obstacle in the rear of the vehicle,which relies on object reflection to obtain the information of the target location.Therefore,it lacks the detection ability of the plane targets such as parking position line and so on,Monocular vision can detect planar targets,but it is difficult to get the distance between target and vehicle directly,then there are obvious security risks.In order to realize the vehicle's identification and measurement of the target parking position line,improve the success rate of the automatic parking,the subject research used binocular vision technology,researched and designed the method of identification and measurement of the parking position line,and the rationality and validity of the research design are verified in the underground parking lot.The subject completed the following research design work:(1)Based on binocular vision measurement model,the factors that improve the detection precision were analyzed,The key parameters of the binocular camera are designed,the platform which is used to collect the real vehicle parking image data was builted,and multi-group parking images were collected under the underground parking environment.(2)In order to extract the image features of the parking space,the best ROI region extraction method was studied,the source of noise interference was analysised for the target image,the quality of the image was improved by denoising pretreatment method,and the edge image of the target parking space is obtained.(3)In order to make the imaging plane of both sides of the camera completely coplanar,the image distortion of wide-angle camera was studied,the single camera calibration was used to dislodge the image distortion for both sides of the camera,and the stereo correction was used for the right and left camera.(4)The extraction and matching algorithm of the image feature was determined,the interference characteristics were screened out by designing feature screening methods,the discrimination strategy of the parking space was developed.The matching,the parallax calculation and the ranging of the binocular interest points in the target parking spaces were come true.In the end,many times image acquisition experiments were carried out on underground parking lot,the experimental results verified the reasonable,feasible and effective of parking space detection method that was proposed in this article.