Research On Container Positioning And Box Number Identification Technology Based On Image Processing

With the development of computer technology and the increasing demand for automated production,image processing is more popular than traditional methods in solving practical engineering problems.This paper studies the container positioning and box number identification technology based on image processing.The main research contents include the use of binocular stereo vision technology to obtain the three-dimensional coordinates of the container position,and the automatic identification of the container number by the artificial neural network system.Through the research of the subject,it is expected to provide technical support for the automation of container loading and unloading operations.For the camera calibration work,this paper builds a test bench consisting of a binocular camera,a calibration board and a fixed bracket.The binocular camera is calibrated using two different specifications of calibration boards.By comparing the average error values of corner detection feedback from Matlab calibration toolbox,a more suitable chessboard specification for camera calibration is selected.And finally the internal and external parameters of the binocular camera are obtained.In order to realize the positioning function of container,this paper firstly performs image preprocessing,distortion correction and stereo correction on the image pairs acquired by the binocular camera so that the pair of images to be matched satisfies the limit constraint.Then,the region-based local matching algorithm is used to perform stereo matching on the image pairs to obtain the parallax of the disparity map and the corresponding feature points,and the depth information of the feature points is calculated by using the triangulation method.In order to ensure the accuracy of depth information extracted,it is necessary to obtain a disparity map with good effects.However,in the binocular stereo vision experiment,the disparity map acquired by the same stereo matching algorithm often fails to achieve the effect of stereo matching on the international standard map.Aiming at this problem,this paper analyzes some factors affecting the stereo matching effect,and studies the influence of illumination variation and camera height feature on the stereo matching effect through contrast experiments,and finds out the lighting conditions and the height of the camera for the experiment of depth information extraction.Analysis of the disparity map obtained by stereo matching shows that the disparity map obtained by the illumination condition of sunny natural light is the best;when the camera placement height is 2~3 times of the baseline distance,the complete target object image pair can be obtained,and the disparity map works best.For the box number recognition function,a container box character segmentation algorithm suitable for different arrangement methods is proposed.The algorithm can automatically distinguish the box number information of horizontal and vertical alignment,and divide and store the characters.At the same time,the algorithm can handle the case of character sticking and erase the box of the last character.Finally,this paper establishes a training set of numbers and uppercase English letters,constructs andtrains BP neural network in Matlab environment,and recognizes the box number characters in parallel.After testing the box number images of different arrangement modes,the result shows that this neural network can accurately identify the box number information.In this paper,an experimental test system is established.The correctness of the method is verified by a numerical example test.The research work has certain reference significance for the application of image processing technology.