Enhanced YOLOv8-based detection of surface damages on conveyor belts with improved accuracy and efficiency

Damage to the surface of composite conveyor belts is a prevalent issue that can lead to functional failure. In this study, the YOLOv8 algorithm is enhanced. The convolutional block attention module attention mechanisms and deformable convolutions are introduced, and a small-object detection layer is added. The Focaler-CIoU loss function is also employed. These enhancements are designed to enhance the recognition of damage. In the context of complex operational conditions, the findings indicate a mean recognition accuracy of 93.8% for conveyor belt surface damage, thereby substantiating the efficacy of the enhanced algorithm.