基于高光谱图像的东北稻米品种快速分类  

Rapid classification of northeast rice varieties based on hyperspectral imagery

作　　者：司刚正 岳鑫 吕众 杨珩 王盛楠 李凤娇 宋少忠 温昌礼 谭勇[1] SI Gangzheng;YUE Xin;LYU Zhong;YANG Heng;WANG Shengnan;LI Fengjiao;SONG Shaozhong;WEN Changli;TAN Yong(College of Science,Changchun University of Science and Technology,Changchun Jilin 130022,China)

机构地区：[1]长春理工大学理学院,吉林长春130022

出　　处：《太赫兹科学与电子信息学报》2020年第4期687-691,共5页Journal of Terahertz Science and Electronic Information Technology

基　　金：吉林省大学生创新训练资助项目(201910186055);吉林省科技发展计划资助项目(20190303108SF);吉林省发改委创新能力建设资助项目(2019C053)。

摘　　要：针对稻米品种无损鉴别需求,利用高光谱技术分析了3种稻米样品的光谱图像特征,实现了利用液晶可调滤波器(LCTF)光谱相机对3种稻米的探测、分类与鉴别。通过高光谱相机采集稻米样品的VIS/NIR光谱图像,运用Matlab软件及ENVI软件对高光谱图像进行处理分析,获得各样本的相对反射率曲线,结合图像阈值分割技术,得到各波段光谱图像的二值图像。结合图像及数据,分析不同品种稻米的光谱差异,发现稻米于480~550nm波段有较为明显的特征峰,品种之间光谱差异明显,且不同品种稻米的二值图像明暗占比不同,以此完成稻米品种的分类与鉴别。研究结果表明,光谱图像的相对反射率和二值图像在稻米品种快速分类与识别的应用中具有较好的应用前景。Based on the nondestructive identification requirements of rice varieties, the spectral image features of three rice samples are analyzed by hyperspectral technique, and the detection, classification and identification of three kinds of rice using Liquid Crystal Tunable Filter(LCTF) spectral camera are realized. The VIS/NIR(Visible/Near-Infrared) spectral images of rice samples are collected by hyperspectral camera, and the hyperspectral image is processed and analyzed by Matlab and ENVI software. The relative reflectance curves of each sample are obtained. By using image threshold segmentation technology, the spectral images of each band are obtained. Combining the images and data, the spectral differences of different varieties of rice are analyzed. It is found that the rice had a distinct characteristic peak in the 480-550 nm band. The spectral differences between different varieties are obvious, and the ratios of the brightness of the binary images for different varieties of rice are different as well. The results show that the relative reflectivity and binary image of spectral images have good prospects in the application of rapid classification and identification of rice varieties.

关 键 词：光谱图像法 无损检测 高光谱图像技术 阈值分割 

分 类 号：TN919[电子电信—通信与信息系统]