弱光环境下仓库搬运机器人抓取控制方法  

Grasping Control Method of Warehouse Handling Robot in Low Light Environment

作　　者：许尧 操蓉蓉[1] 翟志敏 汪立立 XU Yao;CAO Rongrong;ZHAI Zhimin;WANG Lili(Guoneng Shenwan Anqing Power Generation Co.,Ltd.,Anqing 246000,China;School of Electronic Science&Engineering,Southeast University,Nanjing 210096,China)

机构地区：[1]国能神皖安庆发电有限责任公司,安徽安庆246000 [2]东南大学电子与工程学院,江苏南京210096

出　　处：《测控技术》2024年第6期8-13,20,共7页Measurement & Control Technology

基　　金：国家能源集团安庆电厂智能仓储管理系统的研究和应用科技创新项目(KJXM2023-003)。

摘　　要：为精准、稳定、可重复地完成抓取动作,研究了弱光环境下基于稳定轻量级网络的仓库搬运机器人抓取控制方法。首先,针对搬运环境弱光图像,基于稳定轻量级编/解码网络提取抓取区域弱光特征并进行融合处理,获得正常光抓取区域特征;其次,在深度分离融合提取层中,通过处理正常光抓取区域特征,重构深层特征,从而恢复特征提取时丢失的细节信息;再次,在网络输出层内输入重构特征,输出仓库搬运机器人手爪的抓取位姿参数;最后,通过手眼标定得到的搬运目标图像,采集相机坐标系与机器人坐标系的坐标转换关系,将抓取位姿参数转换成仓库搬运机器人抓取控制量,完成对仓库搬运机器人的抓取控制。实验证明,该方法可有效提取搬运目标抓取区域特征,并可有效预测仓库搬运机器人抓取位姿,能完成仓库搬运机器人抓取控制,且抓取精度较高。To achieve precise,stable,and repeatable grasping actions,a grasping control method for warehouse handling robot based on stable lightweight network in low light environment is studied.Firstly,for handling low light images in the environment,a stable and lightweight encoding/decoding network is used to extract low light features in the grasping area and perform fusion processing to obtain normal light grasping area features.Secondly,in the deep separation fusion extraction layer,by processing the normal light grasping area features,deep features are reconstructed to recover the lost detail information during feature extraction.Then,reconstruction features are input into the network output layer,and the grasping pose parameters of the warehouse handling robot's gripper are output.Finally,the coordinate transformation relationship between the camera coordinate system and the robot coordinate system is obtained through handling target image obtained by hand-eye calibration.The grasping pose parameters are converted into grasping control variables for the warehouse handling robot,completing the grasping control of the warehouse handling robot.The experiment proves that this method can effectively extract the features of the grasping area of the handling target,and effectively predict the grasping posture of warehouse handling robot,complete the grasping control of warehouse handling robot,and have high grasping accuracy.

关 键 词：弱光环境 稳定轻量级网络 仓库搬运机器人 抓取控制 弱光增强 手眼标定 

分 类 号：TP242.6[自动化与计算机技术—检测技术与自动化装置]