The YOLOv11 Lightweight Algorithm Based on Improved Pipeline Corrosion Detection

Aiming at the problems such as low efficiency, high cost and difficulty in real-time response in the traditional corrosion detection methods of drainage pipes, this paper proposes a lightweight corrosion detection algorithm based on YOLOv11. This method is based on YOLOv11n and designs two core improvement modules: Firstly, the efficient and lightweight StarNet network is adopted to replace the original backbone structure, significantly reducing the number of parameters and computational complexity while maintaining a good feature extraction ability; Secondly, a new type of lightweight detection head, Detect_LSCD, was proposed. Combined with depth-separable convolution and scale adjustment modules, efficient and accurate corrosion area detection at multiple scales was achieved. Based on the self-built dataset of pipeline corrosion images, this paper conducts a systematic experimental evaluation of the improved model. The experimental results show that this model achieves significant parameter optimization and improvement in computational efficiency while maintaining a high detection accuracy. The model parameters were reduced by 34.6% and the computational cost was decreased by 33.3%, while achieving excellent detection performance of 95.9% for [email protected] and 74.5% for [email protected]:0.95, providing an efficient and accurate solution for pipeline corrosion detection. The improved algorithm proposed in this paper has a simple and efficient structural design, excellent performance, good deployment adaptability and real-time detection ability, and is suitable for the intelligent detection task of pipeline corrosion in resource-constrained environments. (Abstract)