基于TOI-Net的高精度货车超载智能判别方法  

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

作　　者：梁健 康杰虎 赵宗扬 吴斌[1] 王雪森 Liang Jian;Kang Jiehu;Zhao Zongyang;Wu Bin;Wang Xuesen(State Key Laboratory of Precision Measurement Technology and Instruments,Tianjin University,Tianjin 300072,China;Tianjin Municipal Engineering Design&Research Institute,Tianjin 300051,China)

机构地区：[1]天津大学精密测量技术与仪器国家重点实验室,天津300072 [2]天津市政工程设计研究总院有限公司,上海300051

出　　处：《仪器仪表学报》2024年第5期319-328,共10页Chinese Journal of Scientific Instrument

基　　金：天津市交通运输科技发展计划项目(2022-09);国家自然科学基金(52375545,62371339)项目资助。

摘　　要：针对货车超载运输为道路安全带来巨大威胁,而目前主流的货车超载判别方法存在判别效率低、监管范围小、检测成本高的问题,提出了一种基于TOI-Net的高精度货车超载智能判别方法。首先,设计了针对于超载判别任务的货车行驶轨迹图像生成方法,可将多维度货车行驶轨迹时空数据转化为货车行驶轨迹图像,在降低数据复杂性的同时实现了特征的聚合;然后,设计了一个高精度货车超载智能判别模型TOI-Net,其由RepVGG模块和位置注意力模块组成,能够充分挖掘货车行驶轨迹数据中的超载信息特征,高效完成超载判别任务。在货车超载数据集上的实验结果表明,所提方法的超载判别准确率为96.1%,且性能指标均高于主流识别网络,实现了对于货车超载行为的精确、快速和全面的判别。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.

关 键 词：智能交通 人工智能 货车超载判别 卷积神经网络 注意力机制 

分 类 号：U491.2[交通运输工程—交通运输规划与管理] TH39[交通运输工程—道路与铁道工程]