摘要
冬枣的水分是冬枣分级的重要指标,为了提高检测效率,需要做到快速大批量检测。但是传统的高光谱反演模型研究都在暗箱中进行,并不能做到大批量和快速检测。文章对南疆十一团的冬枣所采取的光谱进行了平场校正,提取5×5感兴趣区域,使用了4种光谱预处理方法和2种波长算法使用偏最小二乘回归(PLSR)建立最优模型。模型的校正标准偏差(RMSEC)为1.0683 (g/100 g),预测标准偏差(RMSEP)1.0805 (g/100 g),校正决定系数(RC2)为0.8013,预测决定系数(Rp2)为0.8053,RPD为2.3346,相关系数(r)为0.9156。将所建立的最优模型应用到冬枣的每一个像素点,利用伪彩色编码技术进行假彩色合成。将冬枣表面的水分差异通过不同的颜色表示出来,能够有效地对南疆冬枣的水分进行快速无损检测。
The moisture content of winter jujube is an important index for the classification of winter jujube.However,the traditional hyperspectral inversion model research is carried out in a dark box,so it can’t be detected in large quantities and fast.In this paper,the spectrum of winter jujube from the southern Xinjiang was plane-corrected,5×5 regions of interest were extracted,and four spectral preprocessing methods and two wavelength algorithms were used to establish the optimal model using partial least square regression (PLSR).The calibration standard deviation (RMSEC) of the model was 1.0683 (g/100 g),the prediction standard deviation (RMSEP) 1.0805(g/100 g),the calibration determination coefficient (RC2) was 0.8013,the prediction determination coefficient (Rp2) was 0.8053,the RPD was 2.3346,and the correlation coefficient (r) was 0.9156.The optimal model is applied to every pixel of winter jujube and the false color synthesis is carried out by using the false color coding technique.The difference of water content on the surface of winter jujube can be expressed by different colors,which can effectively detect the water content of winter jujube in southern Xinjiang quickly and nondestructively.
作者
李伟
罗华平
索玉婷
王长旭
徐嘉翊
LI Wei;LUO Huaping;SUO Yuting;WANG Changxu;XU Jiayi(School of Mechanical and Electrical Engineering,Tarim University,Alar 843300,China;The Key Laboratory of Colleges&Universities under the Department of Education of Xinjiang Uygur Autonomous Region,Alar 843300,China)
出处
《食品科技》
CAS
北大核心
2020年第2期329-333,共5页
Food Science and Technology
基金
国家自然科学基金项目(11964030,11464039,11164023)。
关键词
可视化
高光谱成像
反演
visualization
hyperspectral imaging
the inversion
