基于PointNet神经网络的蜂鸣器智能抓取实现  

Intelligent grasping of buzzer based on PointNet neural network

作　　者：曹馨予 CAO Xinyu(Xi’an Traffic Enginering Institute,Xi’an 710000,China)

机构地区：[1]西安交通工程学院,西安710000

出　　处：《自动化与仪器仪表》2024年第10期325-329,共5页Automation & Instrumentation

基　　金：西安交通工程学院2022年度校级中青年基金项目《改革开放以来钢琴作品的“中国精神”表达研究》(2022KY-66)。

摘　　要：位姿估计是实现智能抓取的关键技术,然而电子钢琴蜂鸣器总体规格偏小,定位要求高。为了提高蜂鸣器的位姿估计精度,以及智能抓取的准确率,研究基于点网络,设计了一种由特征提取模块和分类预测模块组成的蜂鸣器识别网络,由多层全连接层构成的分类预测模块对得到的点云全局特征进行处理,并将输出的位姿信息输入抓取控制系统,从而实现蜂鸣器的智能抓取。结果表明,研究所提方法对位置估计的误差均值不超过3 mm,对旋转估计的误差均值不超过2°,对蜂鸣器的分类准确率在85%左右,在x、y、z方向的位置误差也均在2 mm左右甚至更小达到了较优的位姿估计准确度。研究可以节省电子钢琴蜂鸣器生产线的人力和生产成本,对提高电子钢琴的生产效率具有重要意义。Pose estimation is the key technology to realize intelligent grasping.However,the general specification of electronic piano buzzer is small and the positioning requirement is high.In order to improve the pose estimation accuracy of the buzzer and the accuracy of intelligent grasping,a buzzer recognition network composed of feature extraction module and classification prediction module was designed based on the point network.The classification prediction module composed of multi-layer fully connected layers processed the obtained global features of the point cloud and input the output pose information into the grasping control system.Thus realize the intelligent capture of the buzzer.The results show that the mean error of the proposed method for position estimation is less than 3mm,the mean error of rotation estimation is less than 2°,the classification accuracy of the buzzer is about 85%,and the position errors in the x,y and z directions are also about 2mm or even less,achieving a better pose estimation accuracy.This research can save manpower and production cost of electronic piano buzzer production line.

关 键 词：PointNet神经网络 蜂鸣器 位姿估计 智能抓取 

分 类 号：G433[文化科学—教育学] TP29[文化科学—教育技术学]