Design and Implementation of real-time road surface detection System Based on YOLOv8-GD Algorithm

The main content of this study is to improve the accuracy and speed of road potholes detection, and establish a system with real-time road surface detection and integrated data reporting. In response to some drawbacks of traditional methods, this paper proposes a lightweight road pothole detection model based on improved YOLOv8-GD. By replacing the neck feature fusion network in the YOLOv8 model with the aggregation and distribution mechanism (GD), the cross layer fusion ability of the model for feature information of potholes and the generalization ability of the model are improved. For the bumpy data-set collected in complex road environments, the author applied different exposure and noise reduction treatments to the images in the data-set. Based on the original data-set, different treatments were applied to each data-set. The trained model on this basis has higher recognition accuracy and makes the system's model training more robust. evaluation results demonstrate promising performance across various IoU thresholds, indicating its effectiveness in accurately localizing fruits within bounding boxes. Additionally, YOLOv8-GD exhibits substantial improvements in detection performance, achieving high precision and recall rates. The system has improved accuracy and efficiency, making it more suitable for the deployment and application of road inspection and mobile detection equipment in practical scenarios.