Figures from this paperAbstractFigures from this paperAbstractFigures from this paperAbstractFigures from this paper

Abstract

The automatic detection of pavement cracks plays a crucial role in road maintenance. However, existing detection methods often suffer from large model sizes and slow detection speeds, limiting their application in pavement crack detection systems. To address this, this paper proposes a pavement crack detection method based on an improved lightweight YOLOv8 model. First, the ShuffleNetV2 lightweight network is used to replace the C2f module in YOLOv8, improving the model's inference speed and thus meeting the requirements for real-time detection. Moreover, the Convolutional Block Attention Module (CBAM) is integrated to improve the model's ability to detect small cracks by enhancing focus on both channel and spatial features. Finally, the head structure of YOLOv8 is optimized to better handle challenges posed by complex pavement textures and variations in lighting conditions. The experimental results demonstrate that the size of the improved model is 3.8MB, representing a reduction of 74%, 39%, and 34% compared to YOLOv5, YOLOv8, and YOLOv10, respectively. Additionally, the number of parameters in the model is reduced by 43% compared to the original YOLOv8 model, and this smaller parameter size significantly enhances the detection speed of the model. Overall, the model demonstrates significant advantages in crack detection speed and model size, highlighting its potential for practical applications.