Multilayer convolutional neural network-based power transmission line equipment image defect detection method and system

Abstract

The invention discloses a multilayer convolutional neural network-based power transmission line equipment image defect detection method and system. The method comprises the steps of performing modularpreprocessing on an original training set image, and inputting the image subjected to the modular preprocessing into a multilayer convolutional neural network model for performing training; for the image subjected to the modular preprocessing, selecting different training set sizes and training parameters, repeating the step 2 for performing multi-time experiments, performing comparative analysison classification accuracy and efficiency, selecting out optimal training parameters and performing storage; and performing filtration through a judger formed based on an environment structure and priori knowledge, and correcting false detection and missing detection information to obtain a final image defect detection result. The method and the system have the beneficial effects that a machine learning model with multiple hidden layers is built, and valuable expressive features are learnt from a large amount of data, so that the classification or prediction accuracy is improved.

Description

technical field [0001] The invention belongs to the technical field of pattern recognition, and in particular relates to a method and system for detecting image defects of overhead transmission line equipment based on a multi-layer convolutional neural network. Background technique [0002] High-voltage transmission lines are the main way of power transmission. Regular inspection of transmission lines, timely detection and elimination of defects and hidden dangers, and prevention of major accidents are of great significance to the power system. Transmission line inspection robots can integrate the latest mechatronics, data visualization, and intelligent identification technologies, and use autonomous or remote control methods to partially replace people approaching overhead transmission line equipment for visible light or infrared detection, and compare and trend inspection data Analysis and timely detection of potential accidents and precursors of failures in power grid ope...

AI Technical Summary

Problems solved by technology

[0010] In online operations, robots often face direct sunlight or encounter complex background interference such as the field and urban areas, resulting in problems such as motion blur, low contrast, and various defect shapes in the collected images, resulting in the accuracy of defect extraction cannot be guaranteed; In addition, the effective description of artificial features and the accurate selection of artificial features are often very difficult, requiring heuristic methods and very professional knowledge, largely relying on personal experience, and requiring features to be invariant to rotation, scaling, and translation. Therefore, the above-mentioned defect detection method based on artificial features is less effective for online inspection operations.

Images