摘要
为了实现机器人臂手系统的目标抓取,采用Kinect对目标信息进行实时检测.首先,采用张正友棋盘标定法完成对Kinect内外参数的标定.其次,利用深度信息进行深度分割,滤除大部分干扰背景,再通过颜色与形状特征实现目标的识别与定位.将识别对象的3维坐标通过以太网发送至机械臂控制台,随后机械臂移动至目标位置.最后采用变积分PID算法控制灵巧手接触力,保证响应的快速性及精密性,实现灵巧手的精细抓取.通过设计一套完整的实验系统验证了该方法的有效性.
To realize automatic object grasping by a robot arm hand system, Kinect is used for real-time detection of the object. The Zhang Zhenyou chessboard method is applied to calibrate the intrinsic and external parameters of the Kinect. Depth segmentation is done to filter out most of the background interference, and identification and location of object are achieved based on the color and shape features. The object＇s 3D coordinates is sent to the manipulator console to locate the target position through TCP/IP communication. A changing integration PID algorithm is applied to achieve fast and accurate grasp by controlling pressure on the dexterous hand. An experiment system is developed to verify effectiveness of the proposed methods.
出处
《上海大学学报（自然科学版）》
CAS
CSCD
北大核心
2016年第4期421-431,共11页
Journal of Shanghai University:Natural Science Edition
基金
上海市产学研合作项目(CXY-2013-28)
关键词
Kinect标定
深度分割
变积分PID
灵巧手
Kinect calibration
depth segmentation
changing integration PID
dexterous hand
