A pole-mounted device metal point defect detection method and device based on image recognition

Abstract

The application discloses a kind of based on image recognition's column device metal point defect detection method and device, method includes: the sample image of column device metal point is collected to form training set;Improved YOLOv8 neural network model and convolutional neural network model are established and are trained respectively;The video data of column device to be detected is obtained, video data is carried out lens segmentation, key frame extraction and superposition sliding segmentation, and obtains segmentation feature map;The segmentation feature map is input to YOLOv8 neural network model and carries out column device identification;Column device feature map is respectively input to YOLOv8 neural network model and convolutional neural network model, and respectively obtain the first probability that metal point has defect and the second probability that metal point does not produce defect;According to the first probability and the second probability, whether metal point has defect is determined;The method can improve the accuracy of metal point defect identification.

Images