Bin-Picking中无纹理工件的分割  

Segmentation of Texture-Less Object for Bin-Picking

作　　者：刘瑶 赵慧[1,2] 伍世虔 陈彬 LIU Yao;ZHAO Hui;WU Shi-qian;CHEN Bin(School of Machinery and Automation,Wuhan University of Science and Technology,Hubei Wuhan 430081,China;Institute of Robotic and Intelligent System,Wuhan University of Science and Technology,Hubei Wuhan,430081,Chi-na;Shool of Information Science and Engineering,Wuhan University of Science and Technology,Hubei Wuhan 430081,China)

机构地区：[1]武汉科技大学机械自动化学院,湖北武汉430081 [2]武汉科技大学机器人与智能系统研究院,湖北武汉430081 [3]武汉科技大学信息科学与工程学院,湖北武汉430081

出　　处：《机械设计与制造》2022年第9期278-281,287,共5页Machinery Design & Manufacture

基　　金：国家自然科学基金面上项目(61775172)。

摘　　要：为了解决Bin-Picking系统中无纹理工件的分割问题,提出了一种将颜色域和深度域信息融合的分割方法。该方法首先提取颜色域和深度域的边缘特征,然后基于二者的边缘特征,分别通过形态学预处理获取标记图,利用标记图辅助分水岭分割。融合二者的分割信息获得初分割结果,最后通过区域生长法以颜色域分割结果为边界优化分割区域。实验结果表明该算法适用于Bin-Picking视觉检测环节中无纹理目标的分割,对于复杂场景可获得较好的目标潜在位置区域,具有较高的鲁棒性和实时性。In order to solve the problem of segmentation for texture-less object in Bin-picking system,a novel segmentation algo⁃rithm based on a mixture of the segmentations on the RGB image and depth map is proposed in the paper.Firstly,The algorithm extracts the edges of RGB image and depth map.Then,with the assistance of these two types of edges,the segmentations of the RGB image and the depth image are obtained respectively by the morphological reconstruction watershed approach.Further⁃more,fusing the RGB image and depth map as integrated segmentation.Finally,the integrated segmentation is refined by seed⁃ed region growth in which the RGB segmentation information is treated as growth boundary.The experimental results show that the method have robustness and the real ability for the Bin-picking visual detection,and it can get better potential location re⁃gions for the complicated situations.

关 键 词：Bin-Picking 无纹理工件 机器视觉 分水岭 目标分割 

分 类 号：TH16[机械工程—机械制造及自动化] TP391.41[自动化与计算机技术—计算机应用技术]