Technology And System For Detecting Surface Defects Of Complex Curved Workpieces Based On Machine Vision

At this stage,the detection of surface defects on large and complex curved work pieces,especially image acquisition,still uses manual acquisition.Many times in order to achieve accurate image acquisition,workers need to adjust their postures.In the long term,workers’ physical health is greatly damaged.Time-consuming and labor-intensive,and inefficient,far from meeting the requirements of modern industrial automation In addition,although manual inspection can detect defects in the image,the location of the defect cannot be accurately marked in the work piece,so the worker cannot quickly locate and repair it,causing repetition of work and waste of time In view of the above problems,this paper studies the surface defect detection system for automatic detection.In order to realize the automatic detection of surface defects of large and complex curved work pieces,the main work completed in this paper is as follows:First,in order to realize the automatic collection of large-scale curved surface work piece images,an image acquisition method based on depth of field and field of view is proposed.Through the analysis and calculation of the three parameters of the depth of field,field of view,and resolution of the industrial camera,the relationship between the depth of field and the field of view under the premise of meeting the resolution requirements is obtained.Based on this,the 3D point cloud of the work piece is meshed to obtain The three-dimensional coordinates of the work piece of the camera shooting point are set to transform the coordinate system,adjust the position and attitude of the camera at the end of the robotic arm,and realize the path planning of image acquisition Experiments show that this method can accurately capture the image of the work piece along the contour of the surface of the collection part while ensuring that the camera and the surface normal coincide exactly.Secondly,for the problem of defect recognition and extraction,image noise is used to eliminate noise interference of the image,and image enhancement and image sharpening algorithms are used to separate features from the background.Support Support Vector Machine(SVM),Multilayer Perceptron(MLP),Gaussian Mixed Model(GMM).and K-NN.Study the learning process of these four classification models A large number of defectsamples are selected for experiments.All of them have obtained good classification results.After analyzing and comparing time and detection accuracy,this paper Gaussian mixture model is mainly used for surface defect recognition and extraction.Finally,for the defect location problem,the feature coordinates are used to extract the position coordinates of the target features,and then the defect information is reflected in the three-dimensional coordinate system of the work piece through coordinate transformation.Through experimental verification,this method can realize the positioning from image defects to the work piece surface.