摘要
针对现阶段国内外小麦不完善粒研究仍存在检测识别率不高、检测条件苛刻、检测时间过长等问题,开发了一套基于图像识别的小麦不完善粒快速检测技术,对图像采集、关键硬件、机器视觉和深度学习等做了一系列研究.研究结果表明,开发的小麦不完善粒图像检测设备对一个标准小麦样品检测时间能够控制在10 min以内,对不完善粒的平均识别率达到80%以上.相较于传统的人工检测,检测周期大大缩短,同时避免了不同质检员之间的感官误差,对于小麦快速收购、储存管理和品级鉴定具有重要意义,同时对于农业自动化检测技术的发展也具有积极意义.
In identification of the unsound kernels of wheat at present,there are still some problems such as low detection recognition rate,harsh detection condition and long detection time.A set of rapid detection technology based on image recognition has been developed,and a series of experiments about Image acquisition,key hardware,machine vision and Deep learning were made.The results show that the detection time of one standard wheat sample is less than ten minutes through the testing equipment,the average recognition rate of the unsound kertnels of wheat is over 80%.The detection cycle is greatly shortened compared to the traditional manual detection.Meanwhile,the sensory error between different quality inspectors is avoided.It is of great significance for wheat rapid acquisition,storage management and grade identification.and it has positive significance for the development of agricultural automation detection technology.
作者
李晓亮
石恒
董德良
司建中
陈领
曹婷翠
Li Xiaoliang;Shi Heng;Dong Deliang;Si Jianzhong;Chen Ling;Cao Tingcui(Sinograin Chengdu Storage Research Institute Co.Ltd,610091;Sinograin Luoyang Depot Co.Ltd;College of Electronics and Information,Engineering,610065)
出处
《粮食储藏》
2019年第5期46-51,共6页
Grain Storage
基金
国家重点研发计划(2017YFD0401404)
中国储备粮管理集团有限公司2016年科技项目(小麦不完善粒自动检测技术及仪器的研发)
关键词
小麦不完善粒
机器视觉
深度学习
人工检测
the unsound kernels of wheat
machine vision
deep learning
traditional manual detection
作者简介
李晓亮,通讯地址:成都市青羊区广富路239号32栋。
