基于激光雷达识别卡车位姿的方法研究  

Research on Truck Pose Recognition based on Laser Radars

作　　者：冯仁宇 陈明 张文婷 田华亭 FENG Renyu;CHEN Ming;ZHANG Wenting;TIAN Huating(Shanghai Tobacco Group Co.,Ltd.,Shanghai 200082,China;Yunnan KSEC Intelligent Equipment Co.,Ltd.,Kunming 650500,China)

机构地区：[1]上海烟草集团有限责任公司,上海200082 [2]云南昆船智能装备有限公司,云南昆明650500

出　　处：《新技术新工艺》2024年第12期26-30,共5页New Technology & New Process

基　　金：云南省重点研发计划基金项目(202202AC080001)。

摘　　要：在成品托盘发运的无人化操作中,为实现物流链路的全自动化与智能化,自动导引车(AGV)必须克服卡车货物位置不确定、搬运路径不固定的挑战。卡车姿态识别是实现货物精准定位与自动装卸的关键环节。传统的3D对象检测方法虽然能够在单一目标和固定场景下提供较为有效的解决方案,但在应对多车型卡车下的复杂识别需求时存在局限性。为此,提出了一种基于点云特征识别的卡车位姿计算方法。通过多线激光雷达获取车厢外部的三维点云数据,结合RANSAC算法拟合卡车模型,并提取平面轮廓特征,实现了AGV对卡车实时姿态的精确计算。该方法显著提升了对多类卡车姿态的识别精度,解决了物流行业中多样化卡车姿态定位的技术难题,具备应对复杂场景下自动装卸任务的能力。In the automated operation of finished pallet delivery,in order to achieve full automation and intelligence in the logistics chain,it required automatic guided vehicles(AGV)to address challenges related to uncertain truck cargo positions and variable transportation routes.Truck pose recognition was a critical factor for precise cargo positioning and automatic loading and unloading.While traditional 3D object detection methods offer relatively effective solutions in single-target and fixed-scene scenarios,they were limited in addressing the complex recognition requirements posed by multiple truck models.To overcome these limitations,it was proposed that a truck pose calculation method based on point cloud feature recognition.By utilizing multi-line laser radars to acquire three-dimensional point cloud data of the truck’s exterior,the method employed the RANSAC algorithm for truck model fitting and extracted planar contour features to enable precise realtime pose calculation.This approach significantly enhanced the accuracy of truck pose recognition across various truck types,effectively solved the challenges of diversified truck pose positioning in the logistics sector,and had the ability to handle automatic loading and unloading tasks in the complex scene.

关 键 词：自动导引车 激光雷达 物流自动化 三维点云 位姿计算 装车技术 RANSAC算法 

分 类 号：TP242.2[自动化与计算机技术—检测技术与自动化装置] TP751[自动化与计算机技术—控制科学与工程] TP391.9