摘要
棉花打顶作为棉花田间管理中的关键环节,打顶作业的质量将影响棉花后续的生长,精准识别定位棉株对棉花机械化打顶具有积极的意义。简要概述机器视觉技术在国外农业领域的应用研究,总结对比国内棉花打顶领域的定位技术,并对双目立体视觉技术、BP(Back Propagation)神经网络、卷积神经网络在棉株识别定位上的研究进行介绍,讨论分析机器视觉在棉花打顶方面存在的识别速度慢、算法冗余、作业环境制约等问题,并对棉株顶叶的分类识别、打顶机的改进优化、化学打顶剂的精准喷施以及实现多功能作业等具体研究方向进行展望,为后续棉花智能化打顶研究提供参考。
Cotton topping is the key link in cotton field management.The quality of the topping operation will affect the subsequent growth of cotton.Accurately identifying and locating cotton plants has positive significance for mechanized cotton topping.This article briefly summarizes the application research of machine vision technology in the foreign agricultural field and summarizes&compares the positioning technology in the domestic cotton topping field by introducing the research of binocular stereo vision technology,BP(Back Propagation)neural network,and convolutional neural network in cotton plant identification and positioning.It also discussed and analyzed machine vision problems in the cotton topping,such as slow recognition speed,algorithm redundancy,and operating environment constraints.The classification and recognition of cotton top leaves,the improvement and optimization of cotton topping machines,the precise spraying of cotton topping agents,and the realization of multi-functional operations are expected to reference subsequent research on intelligent cotton topping.
作者
刘海涛
伊丽丽
兰玉彬
韩鑫
崔立华
Liu Haitao;Yi Lili;Lan Yubin;Han Xin;Cui Lihua(School of Agricultural Engineering and Food Science,Shandong University of Technology,Zibo,255000,China;Shandong Agricultural Aviation Intelligent Equipment Engineering Technology Research Center,Zibo,255000,China;Shandong Lüfeng Agricultural Group Co.,Ltd.,Binzhou,256600,China)
出处
《中国农机化学报》
北大核心
2021年第6期159-165,共7页
Journal of Chinese Agricultural Mechanization
基金
国家现代农业产业技术体系棉花产业体系田间管理机械化岗位专家项目(CARS—15—22)
山东省引进顶尖人才“一事一议”专项经费资助项目(鲁政办字[2018]27号)
教育部产学合作协同育人项目(教高司函〔2020〕6号)
山东理工大学校级本科教学研究与改革项目(教务函〔2020〕37号)。
关键词
棉花
机器视觉
智能打顶
双目立体视觉
神经网络
cotton
machine vision
intelligent topping
binocular stereo vision
neural network