In order to detect the Cd stress levels of rice growing in natural environment fast and accurately, based on wavelet transform technology in the visible light and near-infrared region (NIR), a method of identifying stress levels of rice under Cd pollution was explored. The hyperspectral data, biochemical parameters and heavy metals concentration in folium were collected for the rice growing in natural farmlands. Wavelet transform of hyperspectral reflectance (350～1 300 nm) was performed by using Daubechies 5 mother function and wavelet energy coefficients of spectral reflectance were extracted. In addition, the model between wavelet energy coefficient and Cd content was established. The result showed that the wavelet coefficients of the fifth decomposition level (d5) proved successful for detecting Cd pollution of rice; the singularity range of rice located in the region around 550～810 nm of spectral signal under Cd pollution； and the singularity amplitude was 0. 04; The centre of modulus maxima located at 700 nm. Regression model based on third level wavelet energy coefficient can estimate the Cd content of rice accurately with the coefficient of determination (R2) of 0. 958, and root mean square error (RMSE) of 0. 122. It can be concluded that the singularity analysis technology applying wavelet transform to reflectance has been shown to be very promising in detecting rice under Cd pollution effectively, and wavelet energy coefficients can estimate Cd content of rice, and provide important reference for detecting other metal-induced stress on crop.%为了快速、准确地探测自然环境下水稻镉污染胁迫状况,提出了一种基于可见光-近红外光谱小波分析技术的快速识别和估算水稻镉污染的方法.根据野外实测水稻高光谱数据、水稻叶片主要生化参数及重金属含量等数据,利用Daubechies小波系的db5小波函数对350～1 300 nm水稻高光谱反射率进行9层分解,并提取小波能量系数进行镉含量回归建模.结果显示:第5层小波分解(d5)的奇异范围为550～810nm,奇异幅度为0.04,模极大值的中心位于700 nm处,对识别水稻镉污染效果最佳;以第3层小波能量系数作为自变量的回归模型对水稻镉含量估算精度最高,其决定系数R2高达0.958,均方根误差RMSE为0.122.小波奇异性分析可以较准确的诊断水稻镉污染胁迫状况,基于小波能量系数的建模能有效估算水稻镉污染胁迫水平.
展开▼
