Terahertz + infrared small target recognition technology for ablation defects in high-voltage cable buffer layer

To address the issue of high concealment of erosion defects in high-voltage cable buffer layers, making it difficult to determine the severity of defects and locate them, a terahertz+infrared small target recognition technology for erosion defects in high-voltage cable buffer layers is proposed. Using the terahertz imaging system and the infrared imaging system, the terahertz images of the buffer layer inside the cable and the temperature distribution images of the cable surface are acquired. The wavelet threshold denoising method is used to denoise the collected terahertz and infrared images to improve the image quality. The image transformation parameters were used to register the terahertz images and infrared images. The Faster RCNN network is selected to extract the ablation defect features of the registered terahertz images and infrared images using convolutional neural network, and the lightweight feature fusion network is used to fuse the extracted feature maps. Determine candidate features of ablation defects through the region of interest pooling layer and transmit them to the fully connected layer. The fully connected layer uses the Softmax function to output the ablation defect recognition results. The experimental results show that this technology can effectively identify and locate ablation defects, utilizing the advantages of complementary multi-source information to improve the reliability of defect identification.