[目的]叶绿素含量可以用来评价棉花的长势情况，快速、准确和大面积监测棉花叶绿素含量，有助于实现精准农业。[方法]分别用0～2阶(步长为0。2)的分数阶微分处理和1～10尺度下的小波变换对田间测定的陆地棉和海岛棉等2种棉花的高光谱反射率进行处理，提高棉花叶绿素含量反演精度。通过分析不同处理方式的光谱与叶绿素含量之间的相关性，筛选得出敏感波段;并运用支持向量机回归和随机森林回归模型分别构建棉花叶绿素含量高光谱估算模型。[结果](1)在全波段范围内，2种棉花325～1 075 nm光谱反射率曲线整体变化趋势基本相同，其反射率均随着叶绿素含量的增加而增大。(2)经连续小波变换和分数阶微分变换后，2种棉花高光谱数据和叶绿素含量的相关性有所增强。使用随机森林回归和小波能量系数7对陆地棉叶绿素含量的反演效果最好，建模集决定系数(coefficient of determination，R2)为0。931，均方根误差(root mean square error，RMSE)为 0。782，剩余预测偏差(residual prediction deviation，RPD)为 2。162;使用随机森林回归和小波能量系数6对海岛棉叶绿素含量的反演效果最佳，建模集R2为0。932，RMSE为1。198，RPD为2。687。[结论]本研究可为棉花叶绿素含量遥感估算提供技术参考。
Hyperspectral inversion of chlorophyll content in cotton based on spectral processing method
[Objective]Assessing cotton growth status through chlorophyll content offers a swift,accurate,and extensive monitoring of cotton development,which aids in precision farming.[Methods]To enhance the accuracy of chlorophyll content evaluation in cotton,fractional-order differentiation ranging from 0 to 2(with a step size of 0.2)and wavelet transform within scales from 1 to 10 to process the hyperspectral reflectance data collected from both upland cotton and sea island cotton fields were employed.By analyzing the correlation between different spectral processing techniques and chlorophyll content,sensitive spectral bands were identified.Subsequently,support vector machine regression(SVR)and random forest regression(RFR)models were employed to construct hyperspectral estimation models for cotton chlorophyll content.[Results](1)In the wavelength range from 325 to 1 075 nm,the spectral reflectance curves of the two cotton species show similar overall trends,with reflectance increasing with the increase in chlorophyll content.(2)Following continuous wavelet transform and fractional-order differentiation,the correlationship between hyperspectral data and chlorophyll content improved for both cotton species.Inversion models revealed that using RFR and wavelet energy coefficient 7 had the best results for upland cotton chlorophyll content estimation,with a coefficient of determination(R2)of 0.931,root mean square error(RMSE)of 0.782,and residual prediction deviation(RPD)of 2.162.Similarly,for sea island cotton,employing RFR and wavelet energy coefficient 6 resulted in the most effective chlorophyll content estimation,with the R2 of 0.932,RMSE of 1.198,and RPD of 2.687.[Conclusion]This study provides technical insights for remotely estimating chlorophyll content in cotton plants.
upland cottonsea island cottonchlorophyll contentcrown height spectrumcontinuous wavelet transformfractional order differential
万方数据
维普
