通过研究铅污染胁迫下水稻的光谱高频组份的分维数来诊断水稻铅污染胁迫水平。根据实验区水稻冠层实测ASD高光谱数据和同步获取的农田土壤重金属含量数据，利用Daubechies小波系中的“Db5”母小波对水稻的350—1300 nm波段范围进行小波分解得到第5层高频组份(d5)，并采用盒维法计算d5的分维数，最后采用模糊数学建立d5的分维数与污染胁迫水平的数学模型。结果表明：d5能有效地探测到铅污染胁迫的光谱弱信息，并实现不同污染水平水稻高光谱信号的分离；d5分维数在同一污染水平的年际相对变化率小于4%，不同污染水平的区分度大于75%，即高、中和低污染水平水稻高光谱d5分维数的86.7%、75%和91.7% 分别集中在1.160—1.200，1.220—1.275和1.280—1.320三个区间；采用升(降)半梯形分布的隶属度函数建立水稻高光谱d5分维数与其污染胁迫水平的数学模型，并进行了模型精度检验，其判别精度大于90%。小波变换、分形分析和模糊数学三者相结合有效地实现了光谱弱信息提取、度量及建模，达到水稻重金属污染胁迫状况监测的目的，也为作物其他环境胁迫弱信息的动态识别与精确度量提供借鉴意义。

How to extract and calculate subtle spectral feature information of crop under various environmental-induced stresses from hyperspectral remote sensing is crucial for the application of remote sensing in monitoring agricultural pollution. The objective of this paper is to monitor the stress levels of rice under the Pb pollution. Hyperspectral data and heavy metal content were collected in the fi eld experiment. The fi fth level high—frequency component (d5) was obtained by performing wavelet transform to hyperspectral refl ectance (350-1300 nm) and the fractal dimension of d5 was also calculated. Then the relationship between fractal dimension of d5 and different stress levels of rice was established by the fuzzy logic model. The results showed that: (1) d5 can effectively distinguish the stress levels of rice Pb pollutions; (2) the annual relative variation ratio for fractal dimension of d5 was below 4%, and the classifi cation accuracy of fractal dimension of d5 was above 75%. Namely, fractal dimension values of d5 for rice under high, medium and low pollutions have percentage of 86.7% between 1.160 and 1.200, 75% between 1.220 and 1.275, 91.7% between 1.280 and 1.320, respectively. (3) the high (low) semi-trapezoidal functions were carried out to construct a model to detect stress levels of rice.