Research On Key Technology Of Weld Defect Detection Based On Machine Learning

Digital image processing has been studied for many years,many related theories have been mature.The quality of the weld directly affects the safety of the product.To ensure the safety of the product,it is necessary to detect the defect of the weld.X-ray imaging is always the first choice because of its low cost and fast imaging speed.However,due to the influence of imaging technology,the X-ray weld image has many disadvantages,such as noise,low contrast,subjectivity and low recognition efficiency,which make the current detection method low accuracy and recognition speed.Based on this,this paper focuses on how to improve the automatic level of X-ray weld image defect detection and the accuracy of defect recognition,focusing on the problem of defect recognition accuracy based on the framework of intelligent learning algorithm.First of all,through the analysis of the image preprocessing process,it is found that it has a significant impact on the later defect recognition accuracy.Aiming at the problem of multi noise and low scale of weld image,image enhancement technology and mathematical morphology processing method are used to effectively enhance the image and reduce the image noise without changing the basic features of the image.The experimental results show that the ideal results are obtained.Based on the characteristics of multiple defects in X-ray images,a continuous edge based weld defect detection algorithm is proposed to detect the defects and cut the areas.By accurately locating the defect position to mark the weld area,the subsequent defect detection is limited to a small area,which can greatly reduce the time required for identification;in view of the characteristics of the lack of rich colors and the large number of defects in the detected image,the edge detection method is used to initially detect the image to realize the segmentation of the background of the continuous edge area;for the defect-free area,there are also continuous edge areas.Feature data is extracted from the picture,and the classifier is used to judge whether the data extracted from the picture is a defective picture,which further improves the efficiency of defect recognition.Through the analysis of several typical defect detection algorithms,the results show that the algorithm can effectively extract the defect area in the X-ray weld image.Defect detection and identification is an important basis for weld quality assessment.For the part of defect identification,the convolution neural network algorithm based on seamcarving transformation is proposed to identify and classify the defects.The appropriate network structure is selected to train the classification model and test the defects.Based on the selected optimal model,the determined defect detection and classification steps are used to train and test the test set.Aiming at the problem that the accuracy of the test is affected by the lack of training samples,the method of neuron random deactivation is adopted to suppress the phenomenon of over fitting.Based on the above research,this paper implements the defect detection and recognition algorithm,and verifies the accuracy of the algorithm on the test set.The simulation results show that the accuracy of the proposed algorithm for weld defect detection and recognition is 95% and above,and the running time is controlled within 1 second.