Enhanced Infrared Defect Detection for UAVs Using Wavelet-Based Image Processing and Channel Attention-Integrated SSD Model

In this paper, we develop a defect target detection algorithm based on image processing and feature matching to address background noise in the detection of defects in infrared images of Unmanned Aerial Vehicle (UAVs), as well as to improve real-time monitoring capabilities. We begin by constructing an infrared defect image processing model using wavelet multilayer decomposition, which effectively suppresses texture information within the complex background by reconstructing low-frequency and high-frequency coefficient images. To further enhance defect feature extraction, we segment and filter the reconstructed infrared image to obtain a clearer representation of infrared defect features. We also design a feature point matching algorithm that integrates SURF (Speeded-Up Robust Features) for feature extraction and target image matching. Finally, to further improve the real-time performance and accuracy of defect detection, we enhance the SSD (Single Shot MultiBox Detector) target detection algorithm by introducing the Channel Attention Mechanism (CAM) to create the SSD-CA model. Experimental results demonstrate that the proposed defect detection model achieves a defect detection accuracy of 99.4% on both the IT-UAV and MVTecAD datasets, with a recall rate of 97.68% after 3,000 iterations. Moreover, the results indicate that the network incorporating the channel attention mechanism significantly improves detection performance compared to the network without it, with the defect detection index improving by 1.11%, and the false detection and leakage rates reduced by 0.69% and 0.44%, respectively. These findings suggest that the SSD-CA model proposed in this paper not only effectively extracts and enhances defect features but also intelligently suppresses non-essentia feature information in the background, thereby enabling real-time automatic detection of infrared image defects. This provides robust technical support for UAV inspection and maintenance in complex environments.