Research on concrete crack recognition method based on improved YOLOv8

Timely detection of structural cracks is crucial for the safety maintenance of infrastructure. However, traditional methods have problems such as low detection accuracy, high rate of missed detection of small targets, and poor adaptability to complex lighting environments. This paper proposes an improved YOLOv8 crack detection algorithm, which enhances the local feature extraction capability by introducing a multi-scale context enhancement module (MSCAM), integrates adaptive spatial fusion (AFE) to optimize multi-scale object detection, and combines the WIoU loss function to improve environmental robustness. Experiments show that the accuracy of the improved model is 6.8% higher than that of the original model, the recall rate is 7.2% higher, and the mAP is 5.3% higher. It significantly enhances the detection ability and environmental adaptability, providing efficient and reliable technical support for structural health monitoring.