Image detection method for power equipment defects based on detection reference point offset analysis

Abstract

The invention relates to an image detection method for electric equipment defects based on detection reference point offset analysis. Compared with the prior art, the defects of few electric equipment defect samples and low recognition rate are solved. The invention comprises the following steps: acquisition of training image set; construction of defect image detection network; training of defect image detection network; acquisition of defect images of power equipment to be detected; detection of defect images of power equipment. Through the design of the center point prediction network, the target width and height prediction network and the center point offset network, the present invention can enlarge the detection range of electric equipment defects, not limited to the center point area of ​​the image; at the same time, it is automatically determined according to the weight value of the surrounding detection area The proportion of all loss values ​​​​in the training and learning of the surrounding detection area; it also further solves the problem of few training samples of power design defect images.

Description

technical field [0001] The invention relates to the technical field of electrical equipment, in particular to an electrical equipment defect image detection method based on detection reference point offset analysis. Background technique [0002] Current power equipment defect detection methods mainly include traditional methods and deep learning methods, and deep learning methods are basically divided into two-stage detection methods and single-stage methods according to the detection network architecture. The two-stage methods are mainly composed of convolutional neural networks and region proposal networks. , it is generally necessary to set a calibration frame with a certain area and aspect ratio in advance, and return a large number of possible area frames according to these calibration frames, and then further judge which type of these large number of possible area frames are, and the single-stage detection method is in the extraction volume After the features are integ...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to solve the defects of few defect samples and low recognition rate of electric equipment in the prior art, and provide a method for detecting electric equipment defect images based on detection reference point offset analysis to solve the above problems

Images