High-precision intelligent identification method of truck overload based on TOI-Net

Truck overload transportation is an enormous threat to road safety. Currently, the main identification method for truck overload has low identification efficiency and a small scope of supervision. To address these problems, this article proposes a truck overload identification method based on deep learning. Firstly, a method is designed for generating truck trajectory images specifically for the overload determination task, which can transform multidimensional spatiotemporal truck trajectory data into truck trajectory images, reducing data complexity while aggregating features. Then, we design a high-accuracy truck overload intelligent identification model TOI-Net, which consists of RepVGG modules and location attention modules. It can fully extract overload information features from truck trajectory data and efficiently complete the overload checkpoints task. Experiments are implemented on the truck overload dataset. The results show that the accuracy of the proposed method for overload identification is 96. 1%, with performance metrics higher than mainstream recognition networks, achieving precise, rapid, and comprehensive identification of overload behavior. © 2024 Science Press. All rights reserved.