Development of a loaded vehicle enforcement system using learning-based image processing analysis

Accidents caused by falling loads of poorly loaded vehicles are gradually increasing. The accidents caused by the fall of a load are one of the major road accidents that can damage roads, vehicles, and humans. In this study, we intend to develop a loaded vehicle enforcement system using deep learning techniques. In this study, using Yolo V3, Mask RCNN, and poor loading labeling data, a deep learning method was used to develop a system that quantitatively scored and regulated the degree of loading defects in poorly loaded vehicles. In particular, in the case of poorly loaded labeling data, the system was developed in a way that the accuracy was gradually improved by using the transfer learning method to control the system. The results of this study are expected to reduce the manpower consumed in the existing crackdown on unloading faults by reporting the faulty viscosity of the unloading vehicle by scoring. This study is also expected to be able to reduce the cost.