论文部分内容阅读
作物叶绿素含量的估测可以为精准农业提供技术支持。该文利用PROSAIL模型模拟了不同叶绿素水平下的大豆冠层光谱反射率,而后针对多期实测高光谱及相应的叶绿素数据,在对响应波段进行小波能量系数提取的基础上,分别采用多元线性回归、BP神经网络和RBF神经网络、以及偏最小二乘法进行估算,并进行了比较分析。研究结果表明,基于小波分析的三种回归模型都取得了较好的估算效果,验证模型的R2分别为0.634,0.715,0.873和0.776,其中RBF神经网络方法和基于高斯核函数的PLS模型精度最好,能够全面稳定地估算叶绿素含量。
Crop chlorophyll estimates provide technical support for precision agriculture. In this paper, we used the PROSAIL model to simulate the spectral reflectance of soybean canopy at different chlorophyll levels. Based on the hyperspectral and corresponding chlorophyll data measured in different periods, we extracted the energy coefficients of the wavelet using the multiple linear regression , BP neural network and RBF neural network, as well as partial least squares method to estimate, and carried out a comparative analysis. The results show that all the three regression models based on wavelet analysis have achieved good estimation results. The R2 of validation model is 0.634, 0.715, 0.873 and 0.776, respectively. The accuracy of RBF neural network and PLS model based on Gaussian kernel function is the best Good, can be a comprehensive and stable estimate of chlorophyll content.