Methodological Study on the Influence of Truck Driving State on the Accuracy of Weigh-in-Motion System

The weigh-in-motion (WIM) system weighs the entire vehicle by identifying dynamic forces of each axle on road. load is very important to detect total weight vehicle. Different drivers have different driving behaviors, and when large trucks pass through weighing detection area, state may affect accuracy system. This paper proposes YOLOv3 network model as basis for this algorithm, which uses feature pyramid (FPN) idea achieve multi-scale prediction deep residual (ResNet) extract image features, so a balance between speed accuracy. In paper, spatial pooling (SPP) cross stage partial (CSP) are added original improve learning ability convolutional neural make more lightweight. Then detection-based target tracking method with Kalman filtering + RTS (rauch–tung–striebel) smoothing used truck status information (vehicle trajectory speed). Finally, effective size in states statistically analyzed. experimental results show that has high real-time performance extraction, can be analyze influence accuracy, provides theoretical support personalized correction WIM At same time, it beneficial assist existing traffic accurately provide highway health management decision making providing reliable monitoring data.