Research On Object Recognition And Localization Algorithm Of Industrial Robot Grinding Based On Vision

With the vigorous development of the third industrial revolution,human gradually from mechanization,electrification era to a more advanced automation,intelligent era,and machine vision due to its reliability,efficiency,continuity and non-contact and other advantages,is one of the main ways to intelligent industrial robots.In order to improve the working efficiency of industrial production line,it is necessary to automatically identify and locate a large number of workpieces.However,there are many workpieces with similar shapes on the production line,and the features extracted from similar workpieces will also be similar,which leads to the misidentification and inaccurate positioning of the workpieces.To solve the above problems,this thesis proposes a fast template matching algorithm based on the inner contour pixel projection,and studies the recognition and location algorithm of similar workpiece.This topic has mainly completed the following work:1.According to the research of the existing workpiece recognition and localization algorithm and the characteristics of the target workpiece in this paper,the overall scheme of the machine vision system is designed,and the hardware required by the system is selected;Based on Zhang Zhengyou's calibration method,a simple and fast calibration method for monocular camera is proposed.2.Considering the existing edge detection technology is difficult to maintain the continuity of the edge while removing the influence of noise in the image and scratches on the surface of the workpiece on edge detection,an improved edge detection technology based on second-order differential operator and adaptive morphology is proposed.Firstly,an adaptive weighted fusion algorithm of Gaussian filter and bilateral filter is optimized,which can enhance the image denoising effect while maintaining the continuity of image edges.Secondly,an adaptive morphology method of multi-structural elements was designed to process the image.On the basis of keeping the geometric shape of the workpiece unchanged,the influence of scratches on the subsequent edge detection was eliminated.Finally,the second order differential operator is used to detect the edge of the preprocessed image.The experimental results show that the improved edge detection algorithm is effective in removing the scratches on the workpiece surface.Compared with the traditional differential operator,the edge clarity and peak signal-to-noise ratio(PSNR)are greatly improved,which lays a good foundation for the subsequent workpiece recognition and location.3.In view of the traditional template matching based recognition and location algorithm is easy to cause misidentification of similar workpiece,and the matching time is serious,a template matching method of similar workpiece recognition and location is proposed.First,the image is transformed by perspective,and its perspective transformation matrix is obtained.Then the outer contour of the workpiece is extracted and identified.Finally,a histogram statistical method based on the pixel projection of the inner contour is proposed to segment the inner contour of the workpiece,extract the feature point vectors of the inner contour and the template respectively,judge the similarity between the feature vectors with Euclide distance,and identify the type of the workpiece.Based on the perspective transformation matrix and the experimental results of camera calibration,the pixel coordinates of the inner contour of the workpiece to be polished are converted into world coordinates,so as to guide the robot to carry out grinding operation.4.According to the experimental requirements,the experiment platform of the workpiece identification and positioning system is reasonably built.In the environment of Visual Studio 2015,the human-machine interface software of the workpiece identification and positioning is developed by using the Qt framework,which facilitates the experimental operation and results display.The accuracy and target positioning accuracy of the proposed algorithm are verified by experiments.The experimental results show that the improved algorithm meets the requirements of real-time performance and positioning accuracy for guiding the robot to carry out grinding in industrial environment.