高级检索
    朱怡, 吴永波, 周子尧, 张智雯, 安玉亭. 基于高光谱数据的互花米草营养成分反演[J]. 北京林业大学学报, 2020, 42(9): 92-99. DOI: 10.12171/j.1000-1522.20200008
    引用本文: 朱怡, 吴永波, 周子尧, 张智雯, 安玉亭. 基于高光谱数据的互花米草营养成分反演[J]. 北京林业大学学报, 2020, 42(9): 92-99. DOI: 10.12171/j.1000-1522.20200008
    Zhu Yi, Wu Yongbo, Zhou Ziyao, Zhang Zhiwen, An Yuting. Inversion of nutrient components of Spartina alterniflora based on hyperspectral data[J]. Journal of Beijing Forestry University, 2020, 42(9): 92-99. DOI: 10.12171/j.1000-1522.20200008
    Citation: Zhu Yi, Wu Yongbo, Zhou Ziyao, Zhang Zhiwen, An Yuting. Inversion of nutrient components of Spartina alterniflora based on hyperspectral data[J]. Journal of Beijing Forestry University, 2020, 42(9): 92-99. DOI: 10.12171/j.1000-1522.20200008

    基于高光谱数据的互花米草营养成分反演

    Inversion of nutrient components of Spartina alterniflora based on hyperspectral data

    • 摘要:
        目的  粗蛋白、粗纤维和粗脂肪含量是决定草本质量的指标之一,研究以江苏大丰麋鹿国家级自然保护区(核心三区)内的互花米草为对象,结合高光谱数据和实测数据,找到光谱数据与互花米草营养成分的关系,建立相应的营养成分估测模型。
        方法  在2018年3、5、7和10月,利用ASD Field Spec Pro FR2500高光谱仪采集互花米草叶片的光谱数据,同时采集互花米草叶片,测定其粗蛋白、粗纤维和粗脂肪含量,对原始反射率及其一阶导数和互花米草粗蛋白、粗纤维和粗脂肪含量实测值之间进行相关性分析,在350 ~ 1 364 nm和1 411 ~ 1 799 nm的范围内选取与互花米草各营养成分含量的相关系数绝对值大于0.7的波段,同时构建选定的8个植被指数,进行建模分析,筛选出最优模型。
        结果  粗蛋白的估测模型以673 nm和1 740 nm处的光谱一阶反射率的多元线性估测最优(R2 = 0.917,RMSEP = 1.344);粗纤维的估测模型以1 738 nm处的光谱一阶反射率的线性估测最优(R2为0.741,RMSEP为1.708);粗脂肪含量的估测模型以1 734 nm和880 nm处的光谱一阶反射率的多元线性估测最优(R2 = 0.737,RMSEP = 0.343)。
        结论  在构建的估测模型中,粗蛋白和粗纤维的估测模型对互花米草粗蛋白和粗纤维含量的估测结果极好,粗脂肪含量的估测模型可以对互花米草粗脂肪含量做出很好的预测。研究结果为评估互花米草营养质量提供参考和技术依据。

       

      Abstract:
        Objective  Contents of crude protein, crude fiber and crude fat are major indexes to determine the food quality of herb. The spectral reflectances and nutritional values of Spartina alterniflora were collected in Dafeng Pere David Deer Reserve in Jiangsu Province, China (core area III), and the relationship between the spectral data and the nutrient components of the plant was studied to obtain nutritional component estimation models.
        Method  In March, May, July and October 2018, the hyperspectral reflectances of S. alterniflora were collected by an ASD Field Spec Pro FR2500 high spectrometer. Meanwhile, the crude protein, crude fiber and crude fat contents of S. alterniflora leaves were determined by field collection and lab work. The correlation analysis was obtained from comparison of the hyperspectral original reflectances, its first derivative reflectances and the measured values of crude protein, crude fiber and crude fat contents of S. alterniflora. In the range of 350−1 364 nm and 1 411−1 799 nm, the bands with the absolute correlation coefficient of each nutrient component contents of S. alterniflora higher than 0.7 were selected, and 8 vegetation indexes were constructed for modeling analysis, and the optimal model was selected.
        Result  Multiple linear estimation with spectral first-order reflectances at 673 nm and 1 740 nm was the best for estimating crude protein content (R2 = 0.917, RMSEP =1.344), while linear estimation model with spectral first-order reflectances centered at 1 738 nm obtained optimal estimation result on crude fiber content (R2 = 0.741, RMSEP = 1.708). The crude fat content was best estimated at 1 734 and 880 nm by multiple linear estimations with first-order reflectances (R2 = 0.737, RMSEP = 0.343).
        Conclusion  Among the established estimation models, models of crude protein and crude fiber have excellent estimation results for the crude protein content and crude fiber content of S. alterniflora, the estimation model of crude fat can better predict the crude fat content of S. alterniflora. The study provides a reference and technical basis for evaluating the nutritional quality of the S. alterniflora.

       

    /

    返回文章
    返回