Progressive Adaptive Feature Network for Radar Image Denoising Method Using Multiple Attention

With the acceleration of urbanization, urban road safety has become the focus of attention, requiring timely detection and maintenance of urban roads. Complex environmental factors and interference sources produce varying degrees of image noise, affecting the image clarity of ground-penetrating radar and causing great obstacles to subsequent road detection. Denoising ground-penetrating radar images can provide powerful help for urban road safety and sustainable development. The image clarity and detail recovery after denoising are directly related to the accuracy of urban road safety detection in the later stage. This paper proposes a progressive adaptive feature network based on multiple attention, which is designed for the problem of radar image denoising in complex environments. This method combines a progressive multi-stage network module fusion strategy and makes full use of multiple attention mechanisms to successfully restore clear radar images. Experimental results show that by capturing noise characteristics at multi-scale resolutions. This approach greatly improves the model's denoising robustness and generalization capabilities, and is an effective method applied to ground-penetrating radar.