摘要
目标的快速精准定位是实现基于视觉的机械手在复杂工况下对工件进行自动抓取的关键技术,基于单目视觉与激光传感技术,提出一种采用AdaBoost算法配合多激光传感技术实现对堆叠工件快速精确定位的方法。该方法采用基于局部二值模式特征的AdaBoost算法结合多尺度策略对视觉获取的RGB图片进行检测,获取工件水平位置信息;结合主动跟踪激光传感系统,获取堆叠工件表面的法向量,确定工件的空间位姿。搭建了硬件实验平台,在此实验平台上开发了一套视觉与激光结合的堆叠工件定位抓取系统,并以非标零件为实验对象,在堆叠情况下进行多组实验,在模拟车间自然光照环境下,工件的识别率为96.4%,平均耗时为18 ms,工件定位的平均坐标偏差为1.17 mm,法向量平均偏差为1.39°,机器人抓取率为98.2%。实验结果表明:该方法定位准确、速度快,在自动化生产线上具有可行性。
Fast and precise positioning of target is the key technology to realize automatic grasping of workpiece under complex working conditions by vision-based manipulator.Based on monocular vision and laser sensing technology,a fast and accurate positioning method for stacked workpieces using AdaBoost algorithm and multi-laser sensing was proposed.This method used AdaBoost algorithm based on Local Binary Pattern(LBP)features and multi-scale strategy to detect the RGB images acquired by vision and obtain the horizontal position information of the workpiece.Combining with active tracking laser sensor system,the normal vector of workpiece surface was obtained,and the workpiece’s spatial position and posture were determined.A hardware experiment platform was built,and a workpiece positioning and grabbing system based on vision and laser fusion was developed on this platform.The non-standard parts were used as experimental objects to carry out multi-group experiments under stacking conditions.In simulated workshop natural lighting environment,the recognition rate of workpiece is 96.4%,the average detection time is 18 ms,the average coordinate deviation of workpiece recognition is 1.17 mm,the average deviation of normal vector is 1.39°,and the grasping rate of robot is 98.2%.The experimental results show that the method is accurate,fast and feasible in the automatic production line.
作者
周伟亮
王红军
刘磊
董力中
邹湘军
Zhou Weiliang;Wang Hongjun;Liu Lei;Dong Lizhong;Zou Xiangjun(College of Engineering,South China Agricultural University,Guangzhou 510642,China)
出处
《现代制造工程》
CSCD
北大核心
2019年第12期135-139,98,共6页
Modern Manufacturing Engineering
基金
国家重点科技计划项目(2017YFD0700100)
广东省科技计划项目(2016B090912005)
广东省公益与能力建设项目(2016A010102013)
关键词
堆叠工件定位
ADABOOST算法
局部二值模式特征
多尺度策略
主动跟踪
多激光传感
positioning of stacked workpieces
AdaBoost algorithm
Local Binary Pattern(LBP)features
multi-scale strategy
active tracking
multi-laser sensing