Improved YOLOv5 model based asphalt pavement disease detection algorithm

ObjectiveThe traditional manual inspection and early image detection methods have some problems in the health and safety assessment on asphalt pavement, e.g., low efficiency, high false detection rate for small targets and complex cracks. In order to meet this challenge, this paper proposes an improved YOLOv5 model based asphalt pavement disease detection algorithm, aiming to meet the needs of rapid and accurate assessment on large-scale road network.MethodFirst, context augmentation module was selected to replace SPPF module to adapt to the irregular pavement disease target shape on the basis of YOLOv5 model. Simultaneously, the model size was reduced. Second, the coordinate attention mechanism was embedded in the backbone network; and the efficient multi-scale attention mechanism was introduced into the feature fusion network to improve the feature extraction ability of pavement disease targets. Finally, the coordinate convolution was introduced in front of the detector to better perceive the change of spatial feature information of disease target.ResultThe improved asphalt pavement disease detection algorithm can effectively improve the average precision value and FPS value of network, reaching 89.28% and 53.36 respectively.Compared with the original model, the improved parameters are almost the same, and the comprehensive performance is superior to other models, which significantly reduces the false detection and missed detection in complex scenes. It shows stronger generalization ability and robustness for diseases with variable scales and irregular shapes.ConclusionThis study proves the effectiveness of the improved strategy, improving the detection accuracy and environmental adaptability while maintaining real-time. It provides a reliable detection scheme for road automatic inspection and intelligent maintenance, and has practical application value for road health assessment.