基于高光谱小波分维的茎瘤芥镉含量估算模型

【目的】定量揭示三峡库区特色蔬菜茎瘤芥瘤茎期食用部分瘤茎镉含量与茎瘤芥生长早期叶片光谱信息之间的相关关系，为快速、精准预测茎瘤芥食用部分镉含量提供理论指导，以实现在茎瘤芥生长早期阶段掌握瘤茎食用品质。【方法】以三峡库区重庆段土壤镉含量和国家土壤环境质量标准为背景，通过盆栽模拟试验，设置0、0.1、0.2、0.3、1 mg·kg^-1 5个不同镉添加浓度处理。于茎瘤芥4个生育时期同步测定叶片光谱反射率、不同部位镉含量等信息，基于Daubechies函数2-8小波基的第5尺度提取茎瘤芥膨大前期光谱高频信息，结合相关分析与回归分析构建茎瘤芥瘤茎期食用部分瘤茎镉含量的预测模型。【结果】茎瘤芥不同生育时期叶片和瘤茎期食用部分瘤茎对镉均有较强的富集作用，且随着土壤镉处理浓度的增大而增加；瘤茎镉含量与不同生育时期叶片镉含量存在极显著的相关性，相关系数均在0.9以上。不同生育时期的茎瘤芥叶片光谱反射率均能较好地响应土壤镉胁迫，尤其是膨大前期对镉的光谱响应最明显；膨大前期的分维系数与不同镉胁迫下叶片镉含量均达到显著相关水平，有较大的预测瘤茎镉含量的潜力。基于膨大前期“db5”尺度构建的预测模型R2达到0.929，RMSE为1.0540，是预测瘤茎镉含量的最优模型。【结论】茎瘤芥对镉有较强的富集作用，能通过食物链进一步对人体健康构成潜在威胁，需要对其食品安全问题特别关注；小波分维方法能较好地提取重金属污染下的叶片光谱弱信息；基于“db5”建立的预测茎瘤芥食用部分瘤茎镉含量的幂函数模型，即Y=-0.029+32.878X^-7.656（R²=0.929，RMSE=1.0540，P＜0.01）是最优的估算模型，在茎瘤芥生长早期能较好地实现对其成熟时期食用部分瘤茎镉含量的预测，为提早掌握其食用品质奠定了一定的理论基础。

Hyperspectral Estimation of Cadmium Content in Tumorous Stem Mustard Based on the Wavelet-Fractal Analysis

【Objective】The aim of this study was to reveal the relationship between the cadmium contents of tumorous stem at mature period and leaf spectral reflectance at the early stages of tumorous stem mustard (Brassica juncea var. tunida), a famous local vegetable in the Three Gorges Reservoir Region. The results could provide theoretical guidance for predicting cadmium contents of tumorous stem of tumorous stem mustard rapidly and accurately, which was helpful to know the edible quality in the early growth stage of vegetable. 【Method】A pot experiment was conducted to set up 0, 0.1, 0.2, 0.3 and 1 mg·kg^-1 with different concentrations of cadmium in soil (calculated according to dry soil weight) based on the present situation of soil cadmium contamination in the main planting area of tumorous stem mustard. The leaf spectral reflectance and cadmium contents were synchronously determined in the main four growth periods of vegetable. Wavelet-Fractal Analysis was used to obtain detailed information about leaf spectral reflectance based on db2-db8 wavelet bases under the fifth dimension. Using correlation and regression analysis, non-linear models were constructed for estimating cadmium contents of tumorous stem of tumorous stem mustard advanced in the early growth stage. 【Result】There was a relatively intense accumulation of cadmium in the leaves and tumorous stem of tumorous stem mustard, which increased with the increased cadmium concentrations in soil. There was a significant positive correlation between the cadmium contents of the tumorous stem and the leaves. The leaf spectral reflectance in four growth stages, especially in the early enlargement stage of tumorous stem, could well respond to the contaminated cadmium in soil. The fractal dimension coefficients of leaf spectral reflectance based on "db2-db8" were significantly negatively correlated with the leaf cadmium contents in the early enlargement stage of tumorous stem. The models were simulated based on the relationships among the cadmium concentration in the tumorous stem at mature period of tumorous stem mustard, the leaf fractal dimension coefficients in the early enlargement stage of tumorous stem and the contaminated cadmium in the soil. The model based on "db5"was the optimal simulation, which R² was 0.929 and Root Mean Square Error (RMSE) was 1.0540. 【Conclusion】The results showed that cadmium could accumulated in leaf and tumorous stem of tumorous stem mustard, which could threaten the human health and the food safety, so it should be paid more attentions. Wavelet-Fractal Analysis was suitable to assess the situation of heavy metal contamination of vegetable due to its sensitivity for subtle spectral information. The optimal model was established based on wavelet db5 of leaf spectral reflectance, that was Y=-0.029+32.878X^-7.656, with R² and RMSE was 0.929 and 1.0540, respectively, which could accurately predict the cadmium content of tumorous stem at the mature stage by using the spectral data from the early growth stage, and provide a theoretical foundation for knowing the edible quality in the early growth stage of tumorous stem mustard.