| 本文已被：浏览 794次 下载 447次 |
码上扫一扫！ |
| 基于反射光谱全波段与双波段甘蔗叶片叶绿素预测模型比较 |
|
陈晓1, 李修华1, 王策1, 周永华1, 丁永军2, 农梦玲3, 艾矫燕1
|
|
|
| (1.广西大学 电气工程学院, 南宁 530004;2.兰州城市学院 信息工程学院, 兰州 730070;3.广西大学 农学院, 南宁 530004) |
|
| 摘要: |
| 以甘蔗品种新台糖22号（ROC22）叶片为研究对象，针对全波段和双敏感波段处的反射率分别建立甘蔗叶片叶绿素含量的预测模型，对比各模型的精度。全波段方面，以可见-近红外光谱反射率为输入量，提取出前5个主成分后，分别采用多元线性回归（MLR）与BP神经网络（BPNN）方法建立全波段模型M1与M2；敏感波段方面，选择731和785 nm这2个敏感波段及由二者计算出的植被指数为输入量，建立一元线性回归（SLR）模型M3、MLR模型M4以及BPNN模型M5。研究结果表明：M1与M2的预测值与实测值间的决定系数R2分别为0.792 4和0.892 9；M3、M4、M5的R2分别为0.821 2、0.840 1和0.848 2；BPNN模型精度高于线性回归模型；虽M5的精度稍低于M2的精度，但M5只包含2个敏感波段信息，具有更高的工程应用价值。 |
| 关键词: 甘蔗 光谱反射率 叶绿素含量 双波段 BP神经网络 |
| DOI：10.11841/j.issn.1007-4333.2018.08.13 |
| 投稿时间：2017-09-03 |
| 基金项目:国家自然科学基金项目（31401290；31360291）；广西自然科学基金项目（2015GXNSFBA139261）；广西研究生教育创新计划项目资助 |
|
| Comparison of sugarcane leaf chlorophyll prediction models based on full and dual reflected spectral bands |
|
CHEN Xiao1, LI Xiuhua1, WANG Ce1, ZHOU Yonghua1, DING Yongjun2, NONG Mengling3, AI Jiaoyan1
|
| (1.College of Electrical Engineering, Guangxi University, Nanning 530004, China;2.College of Information Engineering, Lanzhou City University, Lanzhou 730070, China;3.College of Agriculture, Guangxi University, Nanning 530004, China) |
| Abstract: |
| A sugarcane variety ROC22 was studied to build different chlorophyll prediction models based on full-band and dual-band reflectance spectra in the visible-near infrared range,and the accuracies of each model were compared and discussed.Two different ways were considered to explore accurate and applicable models:The first way was to build multiple linear regression (MLR) model (M1) and BP neural network (BPNN) model (M2) with full-band (all of the spectral data) input;For the consideration of engineering applicability,the second way was to build single linear regression (SLR) model (M3),MLR model (M4),BPNN model (M5) with dual-band.M1 and M2 had R2 between the predicted and measured values of 0.792 4 and 0.892 9,respectively.M3,M4 and M5 had R2 of 0.821 2,0.840 1 and 0.848 2,respectively.The results proved that the BPNN model had higher accuracy than the linear regression models although M2 had the highest accuracy.M5 which had the second highest R2 and just needed 2 sensitive bands information was more suitable for engineering application. |
| Key words: sugarcane spectral reflectance chlorophyll content dual-band BP neural network |
