Improvement of YOLOv8 in the Application of Concrete Defect Detection

混凝土缺陷检测对于保障建筑结构的安全性和耐久性至关重要。传统方法如超声波检测和射线扫描存在效率低、成本高或对小尺度缺陷敏感度不足等问题。近年来，基于深度学习的目标检测算法(如YOLO系列)在该领域展现出巨大潜力。文章针对YOLOv8在混凝土缺陷检测中的应用进行了改进，提出了SPPE_DCNv4主干网络、BiFormer注意力机制以及FASFF_Head检测头的协同优化方案。实验结果表明，改进后的模型在精确率、召回率及mAP等关键指标上显著优于原始YOLOv8及其他版本，特别是在处理微小不规则缺陷时表现出色。该研究为混凝土缺陷的智能化检测提供了高效解决方案，并为未来在更复杂场景中的应用奠定了基础。Concrete defect detection is crucial for ensuring the safety and durability of building structures. Traditional methods, such as ultrasonic testing and radiographic scanning, suffer from low efficiency, high costs, or insufficient sensitivity to small-scale defects. In recent years, deep learning-based object detection algorithms (e.g., YOLO series) have shown great potential in this field. This paper presents an improved YOLOv8 model for concrete defect detection, featuring a SPPE_DCNv4 backbone network, BiFormer attention mechanism, and FASFF_Head detection head. Experimental results demonstrate that the improved model significantly outperforms the original YOLOv8 and other versions in key metrics such as precision, recall, and mAP, particularly in detecting small and irregular defects. This study provides an efficient solution for intelligent concrete defect detection and lays the foundation for future applications in more complex scenarios.