A detection and classification method for power grid inspection insulators

Abstract

The invention discloses a method for detecting and classifying insulators in power grid inspections. Positive samples and negative samples are determined through a target prior model, and parameters of the target prior model flow between multiple levels through feature sharing, which greatly improves the performance of insulators. The detection accuracy is shortened while the search process is shortened; the target prior mechanism is used to mine and match irrelevant negative samples, thereby guiding the process of detecting insulators to be better identified in the source data, improving the stability of generating positive samples, and achieving the use of prior knowledge. Effective identification of insulators.

Description

technical field [0001] The invention relates to the field of image recognition, in particular to a detection and classification method for power grid inspection insulators. Background technique [0002] Nowadays, people's demands for production and life are constantly increasing, and the scale of power system construction is also expanding. In terms of network monitoring of power systems, aerial photography can monitor most power lines and equipment, and has become a very important monitoring method, among which the inspection of transmission lines is an important part. As an important part of transmission lines, insulators play an important role in ensuring the safe operation of power systems. [0003] In the power inspection task, it is necessary to detect and classify defects for different types of insulators. With the continuous expansion of the power system, a large amount of aerial survey image data comes one after another, and manual measurement methods cannot be use...

AI Technical Summary

Problems solved by technology

[0007] Aiming at the above-mentioned deficiencies in the prior art, a method for detecting and classifying power grid inspection insulators provided by the present invention solves the problem that the existing convolutional neural network-based method for detecting and classifying power grid inspection insulators is not sufficient enough to extract information from multi-scale The semantic information of the convolutional feature map is efficiently used in target detection, and the detection and recognition ability of the target is not efficient enough

Images