Research On 3D Overrun Detection Technology Of Three-dimensional Garages Based On Binocular Vision

With the rapid development of economy in China,the number of private cars has increased exponentially.For the sake of the solution of the parking problem of numerous vehicles,the stereo garage emerges as the times require.It is necessary to carry out overloading detecting and three-dimensional profile detection of the whole vehicle in advance to ensure the security of the stereo garage and the vehicles in storage.Besides,the pressure sensor can ideally aim at overweight detecting.Infrared detection is still widely applied to the overloading detecting for the three-dimensional size of the vehicle.However,the detection results have low accuracy and are susceptible to light interference.The emergence of binocular vision technology brings a new solution to this problem.In order to satisfy the needs of the overloading detecting of the stereo garage,this thesis proposes an overweight judgement method by using binocular vision technology to detect the size of the whole vehicles in storage to ensure the security of the vehicles and the stereo garage.This thesis aims at the specific scene of the stereo garage,combined with binocular vision technology and image processing technology to pay attention to the arithmetic of the overweight detecting system.Firstly,this thesis completes the binocular camera’s calibration to obtain the internal and external parameters and distortion coefficient.Since the Harris corner detection in the calibration algorithm is more likely to occur misdetection points and be time-consuming during the midsection procedure,researches have been carried out to improve the defects.The results indicate that the improved Harris corner detection algorithm has excellent detection performance and can effectively inhibit the misdetection points and shorten the time of corner detection.Then attention is paid to solve the possible noise problem of the stereo garage by merging and improving the edge detection algorithm of Canny arithmetic operators and modified morphological transform operator.Experimental results show that this algorithm has good anti-noise ability.Besides,it achieves the effective suppression of noise and obtains a better quality edge contours.Then attention is paid to solve the possible noise problem of the stereo garage by presenting the edge detection algorithm which merges and improves Canny arithmetic operators and modified morphological transform operator.This algorithm achieves the effective suppression of noise and obtains a better quality edge contours.Besides,this algorithm has been proved to have good anti-noise ability through experiments.Thirdly,this thesis studies the most critical stereo matching technology for binocular ranging and proposes an improved Non-Local Cost Aggregation(NLCA)algorithm.The improved algorithm of this thesis overcomes the high mismatch rate of the single algorithm by combining Census transform and edge detection technology and remain the high robustness of the Census transform.While the algorithm has been proved to obtain a more accurate disparity map through multiple experiments,it has higher matching performance and lower mismatch rate compared with other algorithms.Finally,by inversely calculating the world coordinates and solving noncoplanar line,the three-dimensional size of the vehicle to be inspected is obtained,and the overloading determination has been carried out.