基于深度学习和图像处理的水果收获机器人抓取系统  

Fruit Harvesting Robot Grabbing System Based on Deep Learning and Image Processing

作　　者：董戈 Dong Ge(Henan Logistics Vocational College,Zhengzhou 450003,China)

机构地区：[1]河南物流职业学院,郑州450003

出　　处：《农机化研究》2021年第3期260-264,共5页Journal of Agricultural Mechanization Research

基　　金：河南省教育科学“十三五”规划2018年度一般课题(〔2018〕-JKGHYB-0577)。

摘　　要：首先,介绍了水果收获机器人抓取系统的总体架构;然后,利用深度学习对水果目标识别进行了研究,实现了一套基于卷积神经网络的目标检测算法;接着,利用图像处理技术实现了对目标物体定位的功能,可以引导水果收获机器人完成对目标水果的采摘。实验结果表明:水果收获机器人抓取系统对水果坐标的计算误差较小,且具备较强的水果识别和定位能力。It first introduces the overall architecture of the fruit harvesting robot's grabbing system,then studies the fruit target recognition by using the deep learning,realizes a set of target detection algorithm based on convolution neural network,and then realizes the target object positioning function by using the image processing technology,which can guide the fruit harvesting robot to finish the target fruit picking.In the actual experiment,the calculation error of fruit coordinate of the robot grabbing system is small,which shows that the system has strong fruit recognition and positioning ability.

关 键 词：水果收获机器人 深度学习 目标识别 图像处理 

分 类 号：S225[农业科学—农业机械化工程] TP242[农业科学—农业工程]