Lightweight road foreign object detection algorithm based on improved YOLOv8

In recent years, traffic accidents caused by foreign objects on expressways have been on the rise year by year, which have brought serious damage to vehicles or cargo. In view of the unsatisfactory detection accuracy of current road object detection algorithms, and the difficulty of deploying detection models in practical application scenarios with limited computing resources due to the large number of model parameters and high computational complexity, this paper proposes a lightweight road foreign object detection algorithm based on improved YOLOv8n. First, to enhance the detection model's accuracy for small road debris, a generalized building module CONTAINER integrated with multiple contexts was introduced to enhance the detection model's ability to extract local features of small-scale road foreign objects and accelerate convergence. Secondly, the C2f-Faster module is integrated into the detection model's backbone and neck to enhance accuracy with fewer parameters and lower complexity. Finally, in view of the limitations of immutability of border scale and weak generalization ability of CloU, Inner-IoU is used to improve the detection layer, and the scale factor ratio is added to boost boundary box regression accuracy and accelerate convergence. Through the experimental verification of ablation experiment and comparison experiment, the results indicate the proposed algorithm's average accuracy surpasses the traditional method, the enhanced algorithm model boasts an average accuracy improvement of 3%, rising from 96.0% to 99.0%, the calculation amount of the model is reduced by 1.8 from 8.2 to 6.4, and the parameter count has been decreased by 0.7, from 3.0 to 2.3. The algorithm enhances road debris detection accuracy and streamlines the calculation model, which can provide certain reference value for highway inspection and maintenance management.