A Novel Lightweight Disease Detection Algorithm Used in Ground-Penetrating-Radar

To solve the problems of excessive parameters in deep learning models of Ground Penetrating Radar (GPR) disease detection, this paper proposes a lightweight disease detection algorithm based on YOLOv8. A lightweight model is obtained by replacing its backbone network with the GhostHGNetV2 network. The SimAM attention mechanism module is used to improve the ability to perceive different levels of feature representation. Additionally, the WIOU is used as the loss function to improve the convergence and accuracy of the algorithm detection performance. Experimental results demonstrate that the proposed algorithm reduces the parameters of original networks by 25.3%, model size by 28.8%, and computational complexity (GFLOPs) by 20.3%. The effectiveness of the proposed lightweight disease detection algorithm is verified.