图像识别的机器手抓握滑移检测系统开发

Development on robot hand grasping and slipping detecting system based on image recognition

为实现机器手抓握物体时不发生脱落,首先应检测其与被抓握物体接触面上的滑移信号。提出一种基于图像识别的机器手抓握滑移检测方法,采用中心区域匹配思想的归一化互相关算法(NCC)匹配由视觉传感器实时采集到的被抓握物体表面图像,得到被抓握物体在采集图像期间的滑移情况。实验结果表明:此系统可以准确检测被抓握物体是否发生滑移及滑移的方向和大小,具有高准确度、高灵敏度等优点。

To achieve the object to be grasped by robot without shedding,the first thing to do is detecting the slipping signal between the object and the robot. A method for slipping signal detection based on image recognition is presented,use normalized cross-correlation algorithm（ NCC） to match surface image of object real-time collected by the visual sensor,slipping case during collecting images of grasped object robot can be obtained by comparing these two images. Experimental results indicates that this method can accurately detect the slip signal of the object which grasping by robot and the direction and size of slipping,the method possesses advantages of high accuracy,and high sensitivity etc.