Infrared image power equipment segmentation method based on generative confrontation network

Abstract

Segmentation method of infrared image power equipment based on generative confrontation network, including the following steps: S1: Prepare network input data; S2: Build generative confrontation model: Cycle-Gan; S3: Design thermal imaging image semantic segmentation model; S4: Design segmentation network loss function. In the present invention, how to eliminate the difference between different modal data for thermal imaging images is fully utilized by using existing pre-trained segmentation models, and using generative confrontation networks to establish connections between different modal data, so that thermal imaging images can be fully utilized Based on the advantages of the existing natural image semantic segmentation model, a thermal imaging image semantic segmentation model with better performance is realized.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a method for segmenting infrared image power equipment based on a generative adversarial network. Background technique [0002] For power systems, technologies such as intelligent inspection robots and drones are gradually replacing inefficient manual inspections. In order to more comprehensively monitor the abnormality of equipment in the power system, in addition to collecting natural images under normal lighting conditions, thermal imaging images collected by infrared detection and diagnosis technology under unsatisfactory lighting conditions are also used in practical applications. The collected thermal images contain information on whether the equipment is abnormal, and it is particularly important to use artificial intelligence technology to realize automatic abnormal detection of thermal images. To achieve intelligent analysis of thermal imaging images, semantic ...

AI Technical Summary

Problems solved by technology

[0005] However, due to the structural differences between natural color images and thermal imaging images, it is a challenging task to directly process them with widely adopted technical schemes in natural images.

The two main technical difficulties are: the distribution of thermal imaging images and natural images is very different, and there is currently no system for processing thermal imaging images; the deep learning model widely used at present requires reasonable pre-training is necessary, but there is no such parametric model for thermal imaging images

[0006] The existing thermal imaging semantic segmentation system directly builds a depth model and uses the labeled images of thermal imaging to learn it, which does not make good use of the advantages of the depth model currently researched on natural images.

Images