铜离子胁迫下玉米叶片污染信息的SM-DMFD探测模型

SM-DMFD model for detecting pollution information of corn leaves stressed by copper ions

通过设置不同浓度铜离子(Cu2+)胁迫梯度的玉米盆栽实验,在采集了不同类型玉米叶片的光谱数据以及测定其Cu2+含量的基础上,通过建立叶片光谱变异的信息甄别模型(SM),并采用光谱谐波分析(HA)预处理、经验模态分解(EMD)、离散小波多层分解、量规法分形维数(DMFD)计算等方法,构建了玉米叶片光谱污染信息的SM-DMFD探测模型。同时,利用常规的绿峰高度(GH)、红边位置(REP)、红边最大值(MR)、红边一阶微分包围面积(FAR)和盒维数法(BDM)等植被重金属污染信息监测方法进行应用结果比较与分析,实验结果表明:SM-DMFD模型应用所得DMFD值与所测玉米叶片中Cu2+含量的相关系数R和拟合判定系数R2分别达到0.986 0和0.972 3,说明该模型能够有效甄别光谱间差异信息并能判别叶片的污染程度,从而验证了SM-DMFD模型在玉米的Cu2+污染信息监测方面具有更好的有效性和优势。而且,通过对玉米污染信息不同探测方法的监测结果进行可视化,也验证了SM-DMFD模型监测效果最为理想。

Potted corn experiments were set up with different Cu^2＋ stress gradients in this study, based on the collected spectra of different types of corn leaves and the measured Cu^2＋contents in the leaves. An SM-DMFD model for detecting the pollution information of corn leaf spectra was constructed by building a screening model（SM） of leaf spectral variant information and using some spectral analysis methods such as the harmonic analysis（HA）preprocessing, empirical mode decomposition（EMD）, discrete wavelet multilayer decomposition,and divider method fractal dimension（DMFD）methods. The application result of the SM-DMFD model was analyzed and compared with the results obtained by some conventional methods such as the green-peak height（GH）, red edge position（REP）, maximum value of red edge（MR）, first derivative area of red edge（FAR）, and box dimension method（BDM）for monitoring plant heavy metal pollution information. The analyzed and compared results show that the correlation coefficient R of the DMFD values obtained using the SM-DMFD model and the Cu^2＋ contents in corn leaves and the fitting determination coefficient R2 reached 0.986 0 and 0.972 3, respectively. These coefficients indicated that the SM-DMFD model can effectively discriminate the variant information between the different spectra and can judge the pollution degree of corn leaves, thus verifying that the SM-DMFD model has better effectiveness and superiority in monitoring Cu^2＋pollution information of the corn. Meanwhile, the model was also verified as having the ideal monitoring effect through visualizing the monitoring results of corn pollution information detected via the different methods.