基于傅里叶描述子的平面物体表征与抓取规划  

Planar Object Representation and Grasp Planning Based on Fourier Descriptor

作　　者：许少秋 陈根良[1] 唐楚禹 XU Shaoqiu;CHEN Genliang;TANG Chuyu(School of Mechanical Engineering,Shanghai Jiao Tong University,Shanghai 200240,China)

机构地区：[1]上海交通大学机械与动力工程学院,上海200240

出　　处：《机械与电子》2022年第7期61-68,共8页Machinery & Electronics

基　　金：国家科技重大专项(2019YFA0709001)。

摘　　要：针对传统抓取规划方法难以计算复杂轮廓物体、计算量大耗时长的弊端,以及基于深度学习的抓取规划方法物理模型简单、训练集制作耗时长的缺陷,提出了一种基于傅里叶描述子的平面物体表征及抓取规划方法。对物体进行图像采集和轮廓提取,将以散点形式表示的轮廓转化为傅里叶描述子,提取高阶项从而还原物体轮廓,构造了轮廓的参数方程。根据轮廓参数方程和接触点位置,使用不同接触模型构造抓取映射矩阵,采用凸包推进算法计算不同抓取位姿在力力矩空间上的抓取性能指标。基于抓取性能指标,采用一种改进的粒子群方法进行抓取点规划,在二指夹持器和四指夹持器模型上进行了仿真验证。Aiming at the disadvantages of traditional grasp planning methods,such as difficulty in calculating objects with complex contours,large consumption of computation time,as well as the disadvantages of grasp planning methods based on deep learning,such as simple physical model and long time consuming in making training set,a method of planar object representation and grasp planning based on Fourier descriptor is proposed.The image of the object is collected and the contour is extracted.The contour represented by scatter points is transformed into Fourier descriptor,and the highorder terms are extracted to restore the object contour,so the parameter equation of the contour is constructed.According to the contour parameter equation and the positions of contact points,the grasping mapping matrix is constructed using different contact models,and the grasping quality measures of different grasp poses in forcemoment space are calculated by convex hull propulsion algorithm.Based on the grasping quality measure,an improved particle swarm optimization method is used to plan the grasping points,and the method is verified by simulation on both twofinger gripper and fourfinger gripper models.

关 键 词：轮廓提取 傅里叶描述子 凸包推进 抓取规划 

分 类 号：TP241.3[自动化与计算机技术—检测技术与自动化装置] TP242.6[自动化与计算机技术—控制科学与工程]