| 基于PCA-BP神经网络算法桃树叶片SPAD值高光谱估算 |
| |
| 引用本文： | 吴文强,常庆瑞,陈涛,余蛟洋.基于PCA-BP神经网络算法桃树叶片SPAD值高光谱估算[J].西北林学院学报,2019(5):134-140,224. |
| |
| 作者姓名： | 吴文强 常庆瑞 陈涛 余蛟洋 |
| |
| 作者单位： | 西北农林科技大学资源环境学院 |
| |
| 基金项目： | 国家高技术研究发展计划(863计划)项目(2013AA102401);中央高校基本科研业务项目(2452017108) |
| |
| 摘 要： | 光谱技术实现了桃树叶片SPAD(soil and plant analyzer development)值的监测,使用基于主成分分析(principal component analysis,PCA)的BP神经网络算法建立桃树叶片SPAD值光谱估算模型。分析各生育期桃树叶片SPAD值的变化及其与叶片光谱的相关关系,分析5种红边参数与SPAD值的相关性,选取相关性较高的3种红边参数,分别与SPAD值进行单因素建模;然后把红边参数和SPAD值用主成分分析、基于PCA-BP神经网络算法建模,并对估算模型进行验证,结果表明:1)5-9月,桃树叶片SPAD值呈先上升后下降的变化特征,8月达到最大;2)4个生育期所建立的3种模型均通过0.01显著性检验,其中6月估算SPAD值的模型,建模精度和验证精度均最高, R 2≥0.814;3)各生育期桃树叶片SPAD值在单因素模型中,以红边位置建立的模型估算和预测精度最高;4)各个生育期中,基于PCA-BP神经网络模型的估算效果最好,建模精度和预测精度最高, R 2最高分别为0.938和0.974;主成分分析模型次之,单因素模型最低。因此,基于红边参数建立的PCA-BP神经网络反演模型能进行桃树叶片SPAD值的准确估算,为桃树叶片叶绿素含量监测提供理论依据。
|
| 关 键 词： | 高光谱 SPAD值 红边参数 主成分分析 BP神经网络 |
Hyperspectral Estimation of Peach Leaf SPAD Value Based on PCA-BP Neural Network Algorithm |
| |
| WU Wen-qiang,CHANG Qing-rui,CHEN Tao,YU Jiao-yang.Hyperspectral Estimation of Peach Leaf SPAD Value Based on PCA-BP Neural Network Algorithm[J].Journal of Northwest Forestry University,2019(5):134-140,224. |
| |
| Authors: | WU Wen-qiang CHANG Qing-rui CHEN Tao YU Jiao-yang |
| |
| Institution: | (College of Resources and Environment,Northwest A&F University,Yangling 712100,Shaanxi,China) |
| |
| Abstract: | The monitoring of the vaue of soil and plant analyzer development (SPAD) of peach leaf were realized by spectral technology.By using the BP neural network algorithm that was based on the principal component analysis (PCA),the SPAD evaluation model for peach tree blade was established.Firstly,the changes of SPAD values of peach tree blade and its correlation with blade spectrum at each growth period were analyzed,and the correlation between five red edge parameters and SPAD value was analyzed.Finally,three red edge parameters with high correlation were selected to conduct single-factor modeling with SPAD value.Then,the red edge parameters and SPAD values were modeled using principal component analysis and PCA-BP neural network algorithm,and the estimation model was verified.1) From May to September,SPAD value of peach tree blade showed the changing characteristics of first rising and then falling,and reached the maximum value in August.2) The three models established during the four growth periods all passed the significance test at the level of 0.01,among which the models that estimated SPAD value in June had the highest accuracy in modeling and verification,and R 2 was above 0.814.3) In the single-factor model,the SPAD value of peach tree blade at each growth period had the highest estimation and prediction accuracy by using the position of red edge.4) The PCA-BP neural network model had the best estimation effect in each growth period,with the highest modeling accuracy and prediction accuracy,and the highest R 2 was 0.938 and 0.974,respectively.The principal component analysis model was in second,the single-factor model was the lowest.Therefore,the PCA-BP neural network inversion model based on red edge parameters could accurately estimate the SPAD value of peach tree blade.The results of this study would provide theoretical basis for the monitoring of chlorophyll content in peach tree blade. |
| |
| Keywords: | hyperspectrum SPAD value red edge parameters principal component analysis BP artificial neural network |
| 本文献已被 CNKI 维普 等数据库收录！ |
|
