基于高阶谱法作物重金属污染元素判别与污染程度诊断

Discrimination and Diagnosis of Copper and Lead Heavy Metal Pollution Elements and Their Pollution Degrees Based on High-order Spectral Method

基于不同铜离子(Cu^(2+))和铅离子(Pb^(2+))胁迫梯度下玉米叶片光谱微分数据,结合高阶谱估计与灰度-梯度共生矩阵(Gray gradient co-occurrence matrix,GGCM)的特征提取方法,提出了Cu^(2+)和Pb^(2+)污染定性分析、污染元素种类识别和污染程度诊断的方法。首先,测量了不同胁迫梯度下玉米叶片光谱数据以及叶片中富集的Cu^(2+)、Pb^(2+)含量;然后,利用高阶谱估计的ARMA模型参数法对各类玉米叶片微分光谱数据序列进行双谱估计,得到bisp_rts和bisp_qs矩阵及其相应的双谱三维图,从而可以直观可视地定性分析玉米是否已受Cu^(2+)和Pb^(2+)污染,辨别出Cu^(2+)或Pb^(2+)污染的元素类别;最后,构造bisp_rts和bisp_qs矩阵相应的GGCM,通过提取各GGCM的纹理参量特征值,诊断玉米叶片受Cu^(2+)和Pb^(2+)的污染程度。实验结果表明:高阶谱估计可以定性分析玉米老叶(O)、中叶(M)、新叶(N)是否已受Cu^(2+)和Pb^(2+)污染,也可辨别出O、M叶片所受Cu^(2+)或Pb^(2+)污染的元素类别;bisp_rts矩阵的灰度分布不均匀性(T1)、能量(T2)特征值均能反映O、M叶片中Pb^(2+)含量的变化,能较好地诊断O、M叶片中Pb^(2+)的污染程度,而bisp_qs矩阵的小梯度优势(T3)特征值能反映O、M叶片中Cu^(2+)含量的变化,能较好地诊断O、M叶片中Cu^(2+)的污染程度。

It has always been a hot topic on using hyperspectral data to analyze in-depth crop heavy metal pollution. Some methods were put forward for qualitatively analyzing copper ion（ Cu^（2＋）） and lead ion（ Pb^（2＋）） pollution,discriminating the kinds of pollution elements and diagnosing their pollution degrees combined with the feature extraction methods of the higher-order spectral estimation and the gray gradient co-occurrence matrix（ GGCM） based on derivative spectral data of the corn leaves stressed by different Cu^（2＋）and Pb^（2＋）concentrations. Firstly,the spectral data of the corn leaves were collected and the Cu^（2＋）,Pb^（2＋）contents in the leaves were measured,which the potted corns were cultivated and stressed by different Cu^（2＋）or Pb^（2＋）concentrations. Then,the bisp_rts and bisp_qs matrixes and their bi-spectral 3 D graphs were obtained by the bi-spectral estimation（ BSE） of differential spectral data sequences of various corn leaves that the BSE was carried out by using the ARMA model parameter method of higher order spectral estimation,so that a corn leaf was analyzed visually and qualitatively to have been polluted or not by Cu^（2＋）and Pb^（2＋）,and the kind of the pollution element could be discriminated to be Cu^（2＋）or Pb^（2＋）. Finally,the GGCMs were constructed which were corresponded to the bisp_ rts or bisp_ qs matrixes,the Cu^（2＋）and Pb^（2＋）pollution degrees of corn leaves could be diagnosed by extracting the texture parameter eigenvalues of each GGCM. The experimental results showed that it can not only qualitatively analyze whether the old（ O）,middle（ M） and new（ N） leaves of corn were polluted by Cu^（2＋）and Pb^（2＋）,but also correctly discriminate the O and M leaves were polluted by which one of the tow element based on the higher-order spectral estimation; the un-uniformities of gray distribution（ T1） and energy（ T2） eigenvalues of the bisp_rts matrix could reflect the changes of Pb^（2＋）content in O and M leaves,so the T1 and T2 might well diagnose the pollution degree of Pb^（2＋）in O and M leaves,and the small gradient advantage（ T3） eigenvalue of the bisp_qs matrix could reflect the changes of Cu^（2＋）content in O and M leaves,so the T3 might well diagnose the pollution degree of Cu^（2＋）in O and M leaves.