Detection Technology Of Metal Complex Surface Defects Based On Machine Vision

The quality of complex metal surfaces not only determines the quality evaluation of metal workpieces,but also affects the performance of end products.Therefore,the detection of complex metal surfaces has always been the focus of research in relevant departments.At present,most of the current metal complex surface detection methods are mainly manual and have great limitations.Therefore,it is necessary to seek new methods that can replace manual labor and can perform real-time detection in factories.Machine vision technology based on image processing technology is widely used in industry.A large number of applications in this provide new avenues for this.In this regard,this paper proposes a detection method for metal complex surface defects on the basis of researching machine vision technology.The main contents are as follows.1)Owing to the slow speed and low efficiency of manual identification of metal surface defects,a defect partition algorithm based on PCC-CFSFDP clustering was studied.Firstly,the background difference method and multi-constraint noise reduction model are used to locate the suspected defect targets,and then the density peak clustering algorithm and region extraction algorithm based on potential cluster centers are used to merge or separate the defects,and the images to be processed are divided into multiple small defect areas,so as to achieve the partition goal.The experimental results show that the method can realize the partitioning of the image,and improve the partitioning speed to a certain extent,which lays a foundation for the subsequent detection.2)In view of the problem of low accuracy of defect target extraction caused by background texture interference in complex metal surface defect detection,a defect extraction algorithm based on regional texture suppression and segmentation is studied.In this method,the background texture of the region is weakened by the adaptive texture suppression algorithm based on the fuzzy operator,and then the defect segmentation and labeling algorithm is used to segment and label the suppressed defect image to realize the extraction of the defect target.The experimental results indicate that the method can efficaciously cope with the problem of large error of traditional threshold segmentation on the basis of weakening the background texture,so that the defect target can be accurately extracted.3)Since the recognition accuracy of defect multi-classification algorithm is not high,this thesis proposes a defect classification method based on decision tree SVM.The algorithm uses the geometric and texture characteristics of the image to extract the defect features,and optimizes the feature vector by using the principal component analysis method.On this basis,a multiclassifier based on decision tree is constructed to achieve accurate classification of defects.The test results reveal that the method can availably divide the three typical types,and has a high accuracy rate,and can reduce false detection and missed detection.4)Experiments verify the defect detection method.Combined with the technical index requirements and detection content of defect detection,the effectiveness and universality of the detection method are proved through experiments by taking metal workpieces with different surfaces as samples.Experiments show that this method can realize automatic detection of complex metal surface defects,and its detection rate is over 99.5%,which basically achieves the characteristics of zero missed detection,low false detection rate,and fast detection speed,which fulfils the requirements of actual industrial inspection.