改进YOLOV5模型的智能锯树机器人系统设计
2025-08-06殷洪海何律汪大海
摘 要:针对在锯树工程测量中遇到的环境复杂,耗时费力且效率不高等测量问题,设计了一种改进的深度学习算法(You Only Look Once, YOLO)模型的智能锯树机器人系统。通过无线传感网络,构建了无线传感器锯树工程测量拓扑结构,实现锯树工程测量供配电系统全方位数据监测;建立窄带物联网(Narrow Band Internet Of Things,NB IOT)分布式结构,提出采用改进YOLOV5算法对障碍物识别,使用超声波传感器和实验室虚拟仪器集成环境(Laboratory Virtual Instrument Engineering Workbench,LabVIEW)以及摄像机完成机器人的自主越障仿真操作,大大提高智能锯树机器人通信能力和计算能力。实验结果表明,改进后的YOLOV5算法模型在经过50000次迭代后,其准确率较改进前增加了9个百分点,召回率较改进前增加了8个百分点,在近距离障碍物识别上,准确率可达97.15%,且距离越近,其障碍物识别效果和自主越障效果越好。
关键词:智能锯树机器人;YOLOV5;超声波传感器;NB IOT;CSMA;障碍物
中图分类号:TP39文献标识码:A
Improved Design of Intelligent Sawing Robot System
Based on YOLOV5 Model
YIN Honghai,HE Lv,WANG Dahai
(Changzhou Jinling Electric Power IndustryCo., Ltd., Changzhou ,Jiangsu 213000, China)
Abstract:In view of the complex environment, time consuming and inefficient measurement problems encountered in sawing engineering measurement, this paper designs an intelligent sawing robot system based on an improved deep learning algorithm (You Only Look Once, YOLO) model. Through wireless sensor network, the wireless sensor sawing engineering measurement topology is constructed to realize the comprehensive data monitoring of power supply and distribution system of sawing engineering measurement. The narrow band internet of things (NB IOT) distributed structure is established, and the improved YOLOV5 algorithm is proposed to identify obstacles. Ultrasonic sensors, laboratory virtual instrument engineering workbench (LabVIEW) and cameras are used to complete the autonomous obstacle crossing simulation operation of the robot. Greatly improve the communication ability and computing power of intelligent sawing robot. The experimental results show that after 50,000 iterations of the improved YOLOV5 algorithm model, its accuracy rate has increased by 9 percentage points and the recall rate has increased by 8 percentage points compared with that before the improvement. In the short distance obstacle recognition, the accuracy rate can reach 97.15%, and the closer the distance, the better the obstacle recognition effect and autonomous obstacle clearing effect.
Key words: intelligent tree sawing robot; YOLOV5; ultrasonic sensor; NB IOT; CSMA; obstacles
在农业和林业领域,由于环境复杂性和任务的特殊性,传统的机器人技术往往无法满足需求[1-2]。特别是在树木锯除领域,由于树木的不规则形状和复杂的生长环境,传统的机器人很难进行精确的操作和移动。
文献[3]提出使用YOLOV3算法识别障碍物,通过对目标物和障碍物进行分类、存放和编号,可有效检测出障碍物和目标物的类别及实际位置,且识别准确率高。但未涉及对障碍物的跨越,且所需数据量和算力较大。文献[4]提出改进人工势场法实现自动避障,通过构建被控对象模型,分析避障运动学,获取环境参数,实现对环境的建模,从而实现对自动避障轨迹的规划,但该方法不适用于环境复杂的避障。……
