The Improved YOLO-V5 Based Automatic Non-parking Overloading Detection Method

Overloading of vehicles would cause serious damage to national public property, and further, it would also pose a serious threat to the personal safety of citizens. How to detect the vehicle overloading is particularly important, which has attracted a great deal of attention. However, it exhibits two major drawbacks, that are low detection efficiency and the poor detection effect, in the existing overloading detection methods, and thus the losses caused by the vehicle overloading cannot be effectively avoided. To improve both the detection efficiency and effect, in this paper we propose an automatic overloading detection method without parking, in which the computer vision technology and improved YOLO-V5 real-time detection algorithm have been utilized. To the authors best knowledge, there have not been reported works that explicitly address non-parking overloading detection. Three major steps are employed, leading to a high accuracy detection method. Firstly, the images collected in the actual road conditions are selected as the source of the data set. Secondly, in terms of the evaluation criteria, the model is trained for many times and the parameters are adjusted by the authors. Finally, the optimal parameters are obtained, which greatly improves the efficiency and effect of the detection. The test results in actual road conditions show that the method proposed in this paper has a high recognition rate and a fast recognition speed, which can greatly meet the actual application requirements.