面向港口设备的目标检测与装备调度算法  

Algorithm of target recognition and equipment scheduling for port equipmen

作　　者：丁峰 徐晓强 方超凡 DING Feng;XU Xiao-qiang;FANG Chao-fan(Wuhu Port and Businss Co.,Ltd.,Wuhu 241000,Anhui Province,China)

机构地区：[1]芜湖港务有限责任公司,安徽芜湖241000

出　　处：《信息技术》2025年第3期128-132,共5页Information Technology

基　　金：2021年安徽省高等学校质量工程重点项目(2021jy-xm0105)。

摘　　要：为提升港口大型设备的工作效率,文中设计了一种面向港口设备的智能化云控制系统,并在该系统中嵌入了基于目标检测网络的设备识别算法和结合遗传算法的装备调度策略。采用YOLOv5作为目标检测网络原型,改进了网络算法结构,并且在网络训练过程中通过剪枝和量化压缩了网络体积。目标检测实验结果显示,相较于传统的YOLOv5,优化后算法的目标检测评价参数mAP提升了0.16,网络实现速度缩短约5.49ms。针对港口设备调度建立问题模型和目标函数,采用改进遗传算法求解最短时间、最短路径的最优解,装备调度实验结果表明,通过所提算法得到的结果更接近实际最优数据。In order to improve the efficiency of large-scale port equipment,an intelligent cloud control system for port equipment is introduced in this paper.The equipment identification algorithm based on target detection network and equipment scheduling strategy combined with genetic algorithm are embedded in the system.YOLOv5 is adopted as the target detection network prototype,the network algorithm structure is improved,and the network volume is reduced by pruning and quantization during network training.The target detection experiment results show that compared with traditional YOLOv5,the optimized YOLOv5 in this paper improves the target detection evaluation parameter mAP by 0.16.Network implementation speed is 5.49 ms faster.The problem model and objective function of port equipment scheduling are established,and the optimal solution of shortest time and shortest path is obtained by using genetic algorithm.The experiment results of equipment scheduling show that the optimal solution of equipment combined with genetic algorithm is closer to the actual data than the traditional algorithm.

关 键 词：港口大型设备 神经网络 目标检测 YOLO算法 调度规划 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP391.41[自动化与计算机技术—控制科学与工程]