改进特征波段选取和混合集成建模的东北粳稻叶绿素含量估算

Chlorophyll Content Estimation of Northeast Japonica Rice Based onImproved Feature Band Selection and Hybrid Integrated Modeling

利用光谱信息快速、无损和准确的检测水稻冠层叶片叶绿素含量,对水稻的长势评估、精准施肥、科学管理都具有非常重要的现实意义。以东北粳稻为研究对象,以小区试验为基础,获取关键生长期的水稻冠层高光谱数据。首先采用标准正态变量校正法(SNV)对光谱数据进行预处理,针对处理后光谱数据,以随机蛙跳(RF)算法为基础,结合相关系数分析法(CC)和续投影算法(SPA),提出一种融合两种初选波段的改进型随机蛙跳算法(fpb-RF)筛选叶绿素含量的特征波段,并分别与标准RF,CC和SPA方法进行对比。以提取的特征波段作为输入,结合线性模型和非线性模型各自优势,提出一种高斯过程回归(GPR)补偿偏最小二乘(PLSR)的叶绿素含量混合预测模型(GPR-P):利用PLSR法对水稻叶绿素含量初步预测,得到叶绿素含量的线性趋势,然后利用具有较好非线性逼近能力的GPR对PLSR模型偏差进行预测,两者叠加得到最终预测值。为了验证所提方法优越性,以不同方法提取的特征波段作为输入,分别建立PLSR、最小二乘支持向量机(LSSVM)、BP神经网络预测模型。结果表明:相同预测模型条件下,改进fpb-RF算法提取特征波段作为输入可较好的降低模型复杂性、提高模型预测性能,各模型测试集的决定系数(R_(P)^(2))和训练集的决定系数(R_(C)^(2))均高于0.7047。另外,在各算法提取特征波段进行建模时,GPR-P模型的R_(C)^(2)和R^(2)P均高于0.7553,其中,采用fpb-RF方法提取的特征波段作为输入建立的GPR-P模型预测精度最高,R_(C)^(2)和R_(P)^(2)分别为0.7815和0.7796,RMSEC和RMSEP分别为0.9041和0.9283 mg·L^(-1),可为东北粳稻叶绿素含量的检测与评估提供有价值的参考和借鉴作用。

Using spectral information to detect chlorophyll content in rice canopy leaves quickly,non-destructively and accurately has a great practical significance for rice growth evaluation,precise fertilization and scientific management.In this paper,japonica rice in northeast China is taken as the research object,and rice canopy hyperspectral data of key growth stages are obtained through plot experiments.Firstly,the standard normal variate(SNV)is used to preprocess the spectral data,based on the processed spectral data and the random frog(RF)algorithm,by combining a correlation coefficient analysis method(CC)and the successive projections algorithm(SPA),an improved random frog algorithm(fpb-RF)is proposed,which combines two primary bands to select the feature bands of chlorophyll content,It is compared with the standard RF,CC and SPA methods,respectively.A hybrid prediction model(GPR-P)with gaussian process regression(GPR)compensation partial least squares regression(PLSR)is proposed:PLSR method is used to preliminarily predict the chlorophyll content in rice to obtain the linear trend of chlorophyll content,and then the GPR with good nonlinear approximation ability is used to predict the deviation of PLSR model,then the final prediction value is obtained by superposition of two outputs.To verify the superiority of the proposed method,with the feature bands by different extraction methods as inputs,PLSR,Least Square Support Vector Machine(LSSVM)and BP neural network prediction models are respectively established.The results show that under the same prediction model conditions,the improved fpb-RF algorithm can better reduce the complexity and improve the model’s prediction performance by extracting feature bands as input.Both the determination coefficient(R_(P)(2))of the test set and the determination coefficient(R_(C)^(2))of each model’s training set are higher than 0.7047.In addition,the R_(C)^(2)and R_(P)^(2)of the proposed GPR-P model are both higher than 0.7553 when each algorithm extracts feature bands.Among them,the GPR-P model with the input of the feature band extracted by the fpb-RF method has the highest prediction accuracy,R_(C)^(2)and R_(P)^(2)are 0.7815 and 0.7796 respectively,RMSE-C and RMSE-P are 0.9041 and 0.9283 mg·L^(-1)respectively,which provides a valuable reference for the detection and evaluation of chlorophyll content in northeast japonica rice.