Research On Detection Method Of Crimping Defect Of Tension Clamp Based On Image Recognition

The tension clamp is an indispensable component for connecting insulators and wires in overhead transmission lines.The crimping defect of the tension clamp will directly affect the safety of line operation and even lead to large-scale paralysis of the power grid.Therefore,improving the detection efficiency and detection accuracy of crimping defects of tension clamps in the operation and maintenance of transmission lines is the key to preventing accidents.In this paper,the intelligent detection method of the crimping defect of the tension clamp is researched,and a detection method based on X-ray image recognition is proposed,which can complete the automatic identification of the crimp defect of the tension clamp anti-slip groove.The acquisition and marking of X-ray images for the crimping defects of the anti-skid grooves of the tension clamps are completed,and the image samples are amplified through geometric transformation and pi Xel transformation,and the images are denoised through median filtering.The image sharpening method and histogram equalization method are used to enhance the edge information and contrast of the image.The built detection model is used for verification.The results show that the image enhancement can effectively improve the detection accuracy.For the X-ray image of the tension clamp,Laplace The edge e Xtraction effect of the operator is better than that of the Sobel operator and the Prewitt operator.Based on the analysis of the current target detection algorithm,the model based on YOLOv3 is used to detect the defect of the tension clamp.Aiming at the unbalanced proportion of positive and negative samples in the training process of the YOLOv3 model,Focal Loss was used to improve the YOLOv3 loss function to reduce the influence of negative samples and indistinguishable samples on the detection results.The recognition accuracy of the test data set is 89.65%,and the detection speed is 20 images per second,which achieves good recognition accuracy and detection efficiency.In order to further improve the detection accuracy,a hierarchical detection method of automatic image cutting is proposed.The image is pre-detected and the anti-skid groove area is located by the model based on Mobile Net-YOLOv4,the automatic cutting of the anti-skid groove area is realized according to the positioning coordinates,and the model based on YOLOv3 is used to detect the defects of the cut image.The recognition accuracy is further improved with a slight decrease in detection speed.