Method and device for detecting foreign matters on telegraph pole based on deep learning

Abstract

The invention discloses a method and device for detecting foreign matters on a telegraph pole based on deep learning, and the method comprises the steps: selecting a preset number of pictures containing the telegraph pole and the foreign matters, carrying out the expansion processing of the pictures, obtaining a preset multiple of picture libraries, and marking the position information of the telegraph pole and the foreign matters in each picture as a sample library; taking the feature map extracted by the sixth convolutional layer of the SSD object detection algorithm as a seventh type of grid, combining the grid with the original six types of grids, taking the combined grid as the input of a final loss function, and training to obtain a detection model of the telegraph pole and the foreign matters; finally, detecting the telegraph pole and the foreign matter in the input picture at the same time through the detection model obtained through training; calculating the first proportion of the overlapping area of the telegraph pole and the foreign matter to the foreign matter area, removing the foreign matter with the first proportion lower than a preset threshold value, so that the detection fault-tolerant rate is increased. According to the method, the process is simplified, and the precision of the model on small-object foreign matters is improved.

Description

technical field [0001] The invention belongs to the field of machine vision, and in particular relates to a method and device for detecting foreign objects on utility poles based on deep learning. Background technique [0002] Usually, foreign objects on utility poles (such as bird's nests, balloons, kites, etc.) may have a great impact on power transmission lines, such as causing short circuits in wires and causing damage to power supply in large areas. Therefore, it is very necessary to detect and remove the foreign matter on the utility pole. Traditional detection methods require staff to work on the spot, which consumes a lot of manpower and material resources, and the scope of monitoring is also very limited. [0003] Using deep learning to detect the position of the target object in the image is a hot technical means at present. Through the image data collected outdoors, combined with the ability of machine deep learning and recognition, automatic recognition and det...

AI Technical Summary

Problems solved by technology

[0004] Although the existing technology has proposed some technical solutions based on deep learning to detect orbital foreign objects, the existing deep learning has relatively low detection accuracy for smaller-scale objects, and the accuracy of foreign object detection is not high.

Images