Efficient Pavement Damage Detection Algorithm Based on Lightweight Feature Enhancement Network

Efficient detection of pavement damage in complex natural environments is crucial for intelligent maintenance of transportation infrastructure. To address challenges related to structural complexity, resource utilization, and detection accuracy, this paper presents a novel pavement damage detection algorithm based on a lightweight feature enhancement network. The model integrates the ghost module into YOLOv5s to improve feature extraction efficiency and creates a more lightweight backbone network. SimAM is then employed to enhance the ability to localize and perceive the damage target. Additionally, the position loss function is replaced with Wise-IoUv3, optimizing gradient distribution to further improve detection accuracy. Experimental results demonstrate that the proposed model achieves a size of only 8.03 M, average detection frames per second of 144, and a mean average precision (mAP) of 84.1%, outperforming YOLOv5s by 4.4%. It also reduces parameters and GFLOPs by 41.4% and 44.3%, respectively, while increasing detection speed by 37.9%. Compared to YOLOv7-tiny, the model improves accuracy by 6.7%, mAP by 3.4%, recall by 0.2%, and F1 score by 0.5%. Additionally, the average precision for four damage types—longitudinal cracks, transverse cracks, alligator cracks, and potholes—improves by 3.2%, 2.7%, 6.5%, and 1.0%, respectively. These findings confirm that the proposed algorithm not only meets real-time detection requirements but also provides lightweight and high-precision pavement damage detection, enabling broader applications in intelligent transportation monitoring systems.