Edge intelligent vision for road disease detection

Road disease detection is crucial to ensure traffic safety and extend the service life of the road. In order to realize automatic, fast, efficient and high-precision road disease detection, we propose an efficient road disease detection method with edge intelligent vision. First of all, using the advanced Yolov8s as the baseline model, we replace the neck PAFPN of the Yolov8s model with the weighted bi-directional feature pyramid network (BiFPN) and design a weight-sharing detection head for road disease characteristics such as cracks, potholes, and bumps. Then, an improved network structure Yolov8s-BE is constructed. Two public road disease datasets are used to train and verify the performance of the improved Yolov8s-BE. Finally, the trained model is deployed on the edge intelligent terminal development kit, and two RGB-D cameras are also integrated to develop a portable road disease image acquisition and real-time detection system. The experimental results verify the detection ability of the improved Yolov8s-BE for pavement disease. Compared with the baseline model Yolov8s, the presented yolov8s-BE achieved a achieves a significant improvement. Compared with the latest method Yolov11, the improvements are also slightly better. In addition, the developed edge intelligent road detection system also achieved stable, reliable and efficient practical applications.