基于图像处理技术和神经网络实现机采茶分级.pdfVIP

茶叶科学 20 17，37 （2 ）：182~ 190 Journal of Tea Science 投稿平台： 基于图像处理技术和神经网络实现机采茶分级 * 吴正敏，曹成茂 ，谢承健，吴佳胜，胡汪洋，汪天宇 安徽农业大学工学院，安徽 合肥 230036 摘要：为解决名优绿茶采摘环节的瓶颈问题，提出对机采大宗绿茶进行分级的思路。现有绿茶机采设备采摘 的鲜叶一般只能制作普通的大宗绿茶，鲜叶存在混杂、破碎率高和老梗叶等问题，本文基于Labview vision 、 图像处理技术和神经网络算法分析机采绿茶成品的凸包面积、 凸包周长、长轴长度、短轴长度等形态特征并 对样本进行分类，实现从机采大宗绿茶中分选出名优绿茶。其中样本的形态特征采用工业 CCD 摄像头获取； 用户界面用 Labview 自定义开发设计，数据交互方便，开发周期短。茶叶样本试验结果表明：该方案机采绿 茶成品的分级正确率可以稳定在90% 以上。本研究为进一步研究机采茶分级设备提供了良好的理论基础。 关键词：机采绿茶；分级；神经网络；图像处理技术 中图分类号：TS272.3 ；S23 文献标识码：A 文章编号：1000-369X （2017 ）02- 182-09 Grading of Machine Picked Tea Based on Image Processing Technology and Neural Network * WU Zhengmin, CAO Chengmao , XIE Chengjian, WU Jiasheng, HU Wangyang, WANG Tianyu College of Engineering, Anhui Agricultural University, Hefei 230036, China Abstract: To solve the picking problem of famous green tea, a new technology was proposed to classify machine picked tea in this paper. Fresh tea leaves plucked by machine are often mixed, with tea stalks and have a high broken rate and thereby only suitable for making general green tea. The convex hull area, convex hull perimeter, long axial length, short axial length and other morphological features of machine pluck ed tea leaves were analyzed by Labview vision, image processing technology and neural network to screen high quality tea. Industrial CCD camera with appropriate optical system was used to collect object classification features . User interface was developed by Labview, which can realize the data interaction, convenien