Power equipment infrared image segmentation method based on deep learning and image gradient

Abstract

The invention discloses a power equipment infrared image segmentation method based on deep learning and image gradient, and the method comprises the steps: carrying out the feature mining of an infrared image through deep learning, completing the first segmentation, combining the temperature features of the infrared image, completing the second segmentation through the gradient of the image, and effectively improving the boundary segmentation precision. Areas of key parts of the electrical equipment can be separated, original infrared image information of the electrical equipment is reserved,and direct and accurate data support is provided for diagnosis and analysis of infrared images; and the diagnosis efficiency and the intelligent level of the infrared image can be improved, the laborcost is reduced, and the efficiency and the accuracy of power equipment fault analysis work are improved.

Description

technical field [0001] The invention relates to a method for segmenting infrared images of power equipment based on deep learning and image gradients. Background technique [0002] Power equipment is a basic part of the power system, and its safe and stable operation is the prerequisite for ensuring the safe and stable operation of the power grid. Due to the long-term operation of power equipment, various failures often occur due to the influence of design, process and operating environment. According to statistics, in the early stage of power equipment failure, the temperature will change abnormally, usually manifested as abnormal heating, so The operating status of the equipment can be judged by detecting and analyzing the temperature of the electrical equipment. Infrared detection technology has the advantages of non-contact detection, no electromagnetic interference, fast long-distance detection, intuitive results, etc., and can quickly and easily judge the health statu...

AI Technical Summary

Problems solved by technology

Chinese patent CN108898077A discloses a recognition method of infrared heat map of electric power equipment and an infrared recognition system of electric power equipment, wherein the recognition method of infrared heat map of electric power equipment disclosed uses deep learning to recognize texture and heat distribution, and has relatively low noise Strong anti-interference ability, but this method can only get the probability of power equipment in the infrared heat map, and the accuracy of segmentation is low by identifying power equipment through the probability threshold; Chinese patent CN109784348A discloses an infrared power equipment identification and online diagnosis method And its system, the power equipment identification method involved in it, generates training data by establishing a virtual three-dimensional model of power equipment with infrared characteristics, and constructs a deep neural network model for power equipment identification. This method needs to establish a virtual three-dimensional model, which is difficult to operate And the workload is heavy; the calculation of image recognition is large, which reduces the efficiency of diagnosis; the accuracy of image recognition depends on the quality of the 3D model, and there is uncertainty; 》[Zou Hui, Huang Fuzhen, etc. Multi-target positioning of power equipment infrared images based on improved FAsT-Match algorithm, Chinese Journal of Electrical Engineering, 2017, 37(2): 591-598] Using the approximate simulation between infrared images and visible light images This method can better solve the problem of over-segmentation, but it needs to analyze the power equipment template, visible light image and infrared image at the same time, and needs to calculate the affine transformation parameters in advance, which leads to this problem. The method takes a long time; Chinese patent CN108564565A discloses a deep learning-based multi-target positioning method for power equipment infrared images and the document "Research on Infrared Image Segmentation Technology for Power Equipment Based on Mask R-CNN" [Wu Kehe, Wang Minjian, etc. Based on Mask Research on Infrared Image Segmentation Technology of Power Equipment Based on R-CNN, Computer and Digital Engineering, 2020, 2(48): 417-422], using the Faster R-CNN and Mask R-CNN network architectures to construct deep learning models respectively, and realizing the feature Extraction and area classification reduce the amount of calculation and improve recognition efficiency, but there are problems with poor segmentation of small targets and the accuracy of feature extraction depends on the quality of infrared images

Images