This study presents an improved YOLOv8n-based detection method to address the issue of false detections of dam cracks that is caused by low-quality surveillance images

limited effective samples

and interference from complex backgrounds. This model is trained using a dataset comprising 193 real-world crack images featuring complex engineering backgrounds

and enhanced by modifying the mosaic data augmentation mechanism and incorporating negative sample training targeted at the objects that were often falsely detected. Numerical experiments demonstrate that under small-sample training conditions

the YOLOv8n model achieves a mean Average Precision (mAP) of 89.2%

meeting the requirements of general engineering applications. After negative sample training

the mAP increases to 92.5%

and the false detection rate is reduced by 10.1%

providing an effective solution to the false detection problem in complex background scenarios. Our findings indicate that the YOLOv8n model is well-suited for dam surveillance images of suboptimal quality

and that the negative sample training strategy significantly improves detection accuracy. This approach offers a novel solution to crack identification in hydraulic projects

practically significant for engineering applications.