Research On Convenient Vision Recognition And Measurement System Of Sheet Metal Parts Based On Inverse Perspective Transformation

In the context of intelligent manufacturing,the research group proposes the cloud platform project of intelligent industrial supply chain in order to balance supply and demand,improve the utilization rate of production factors,and promote the transformation of intelligent manufacturing.The platform aims at the reuse of surplus materials in the medium plate processing industry.It integrates the information collection,tracking and intelligent typesetting of surplus materials in the platform.In order to collect and input the contour and size area information of surplus material,the visual measurement technology is introduced as the efficient support.In this paper,the technology is studied in three fields: image segmentation and recognition,inverse perspective transformation and image contour extraction.Finally,a convenient identification and measurement system of sheet metal and parts based on reverse perspective is developed.With the help of label and inverse perspective transformation technology,it can realize the contour extraction,the calculation of area and other size information of the image taken at any angle.And it is verified by examples that the effect can meet the actual engineering requirements of the project.The specific research work of this paper is as follows:(1)An image segmentation algorithm based on optimized Canny operator is proposed.(2)The paper compares PCA algorithm with PCA-SIFT algorithm and choose the best as the label recognition algorithm of this system.Meanwhile the matching results are optimized by RANSAC algorithm.It provides a good initial condition for the study of inverse perspective transformation.(3)A sub-pixel corner detection algorithm for parallelogram-like labels is proposed.Based on it,the four-point inverse perspective transformation algorithm is implemented.The experimental results show that the measurement error of the image transformed by this algorithm is less than 4%,which meets the engineering requirements.It provides convenient image processing conditions for subsequent contour extraction and area calculation.(4)An optimized snake GVF algorithm based on improved initial contour is proposed to automatically extract the contour of the object in the image.As for the image with complex background,an image segmentation algorithm based on Grabcut is provided to separate foreground and background,which is convenient for snake GVF active contour algorithm to extract contour.Meanwhile an algorithm of automatic corner pick-up is developed in this paper,which can automatically find the best corner in a certain area of mouse click,to meet the needs of human-computer interaction.(5)On the basis of the above research contents and conclusions,this paper designs the framework of the convenient visual recognition and measurement system of sheet metal and parts based on reverse perspective,puts forward the overall flow chart of the system,and develops the system on the platform of MATLAB 2016 a.By cases testing,the results show that the contour extraction is effective and the overall measurement error is less than 5%,which meets the engineering requirements.So it can provide a solid data base for subsequent processes in the cloud platform project.