Classification and feature band extraction of diseased citrus plants based on UAV hyperspectral remote sensing
-
摘要:目的
结合传统与现代农业病虫害监测的优缺点，探索通过无人机高光谱遥感技术检测出患病的柑橘植株、通过人工田间调查方式判断其患病种类及患病程度的病虫害监测方法。
方法使用无人机获取原始高光谱图像，经过光谱预处理和特征工程后，采用连续投影算法提取对柑橘患病植株分类贡献值最大的特征波长组合，基于全波段使用BP神经网络和XgBoost算法、基于特征波段使用逻辑回归和支持向量机算法，建立分类模型。
结果基于全波段的BP神经网络和XgBoost算法的ROC曲线下面积(Area under curve，AUC)分别为0.883 0和0.912 0，分类准确率均超过95%；提取出698和762 nm的特征波长组合，基于特征波长使用逻辑回归和支持向量机算法建立的分类模型召回率分别达到了93.00%和96.00%。
结论基于特征波长建模在患病样本分类中表现出很高的准确率，证明了特征波长组合的有效性。本研究结果可为柑橘种植园的病虫害监测提供一定的数据和理论支撑。
Abstract:ObjectiveCombined with the advantages and disadvantages of traditional and modern agricultural pest monitoring, the method of monitoring pest and disease were discussed, which detected the diseased citrus plants by UAV hyperspectral remote sensing technology and judged the disease species and disease degree by artificial field investigation.
MethodThe original hyperspectral images were obtained by UAV. After spectral preprocessing and feature engineering, continuous projection algorithm was used to extract the feature wavelength combination which contributed the most to the classification of citrus diseased plants. Finally, the BP neural network and XgBoost algorithm were used based on the full band, and the logistic regression and support vector machine algorithm were used to establish the classification model based on the characteristic band.
ResultThe AUC scores of BP neural network and XgBoost were 0.8830 and 0.9120 respectively, and the accuracy rates of both methods were over 95%. The feature wavelength combination of 698 and 762 nm was extracted. Based on this characteristic band, the recall rates of logistic regression and support vector machine algorithm were 93.00% and 96.00% respectively.
ConclusionThe model based on characteristic band shows high accuracy in the classification of disease samples, which proves the effectiveness of characteristic wavelength combination. This result can provide some data and theoretical support for monitoring diseases and pests in citrus plantations.
-
-
![]()
图 1 试验区域及样本标注
粉、红、蓝、黄、白圆圈标记分别代表1、2、3、4级和患病未定级的黄龙病植株，三角形标记为缺素植株；没有标记的植株为完全健康植株
Figure 1. Experiment area and sample marking
The pink, red, blue, yellow and white circle markers represented the plants with Huanglongbing disease of grades 1, 2, 3, 4 and indefinite respectively, and the triangular markers represented the plants lacking in nutrients; The plants without markers were the complete healthy plants.
表 1 BP神经网络结构
Table 1 The structure of BP neural network
层(类型)
Layer (Type)输入节点数
Number of input node输出节点数
Number of ouput node激活函数
Activation function输入层(全连接) Input layer(Full connect) 125 32 ReLu 隐含层1(全连接) Hidden layer 1(Full connect) 32 32 ReLu 隐含层2(全连接) Hidden layer 1(Full connect) 32 16 ReLu 输出层(全连接) Output layer(Full connect) 16 2 Sigmoid 
下载: 导出CSV
表 2 BP神经网络和XgBoost测试结果(混淆矩阵)
Table 2 The test results of BP neural network and XgBoost (Confusion matrix)
模型 Model 项目 Item 预测为患病 Predicted illness 预测为健康 Predicted healthy 总计 Total BP神经网络
BP neural network真实为患病 True illness 78 22 100 真实为健康 True healthy 7 493 500 总计 Total 85 515 600 XgBoost 真实为患病 True illness 85 15 100 真实为健康 True healthy 13 487 500 总计 Total 98 502 600
下载: 导出CSV
表 3 特征波长下LR和SVM分类结果(混淆矩阵)
Table 3 The results of LR and SVM test bases of feature bands(Confusion matrix)
模型 Model 项目 Item 预测为患病 Predicted illness 预测为健康 Predicted healthy 总计 Total LR 真实为患病 True illness 93 7 100 真实为健康 True healthy 44 456 500 总计 Total 137 463 600 SVM 真实为患病 True illness 96 4 100 真实为健康 True healthy 61 439 500 总计 Total 157 443 600
下载: 导出CSV
表 4 4种模型评估得分
Table 4 The evaluation scores of four models
评价指标
Evaluating
indicator全波段 Full-wave band 特征波段 Feature band BP Net XgBoost LR SVM ACC 0.951 7 0.953 3 0.915 0 0.891 6 Recall 0.780 0 0.850 0 0.930 0 0.960 0 Precision 0.917 6 0.867 3 0.678 8 0.611 5 F1-score 0.843 8 0.858 5 0.784 9 0.747 1 AUC 0.883 0 0.912 0 0.921 0 0.919 0
下载: 导出CSV
-
[1] DENG X, LAN Y, HONG T, et al. Citrus greening detection using visible spectrum imaging and C-SVC[J]. Comput Electron Agr, 2016, 130: 177-183. doi: 10.1016/j.compag.2016.09.005
[2] 陈波, 姚林建. 光谱检测技术在柑橘黄龙病诊断中的研究进展[J]. 赣南师范大学学报, 2018, 39(6): 69-72. [3] 兰玉彬, 邓小玲, 曾国亮. 无人机农业遥感在农作物病虫草害诊断应用研究进展[J]. 智慧农业, 2019, 1(2): 1-19. [4] 纪景纯, 赵原, 邹晓娟, 等. 无人机遥感在农田信息监测中的应用进展[J]. 土壤学报, 2019, 56(4): 1-13. [5] 黄文江, 张竞成, 师越, 等. 作物病虫害遥感监测与预测研究进展[J]. 南京信息工程大学学报(自然科学版), 2018, 10(1): 30-43. [6] LI X, LEE W S, LI M, et al. Spectral difference analysis and airborne imaging classification for citrus greening infected trees[J]. Comput Electron Agr, 2012, 83: 32-46. doi: 10.1016/j.compag.2012.01.010
[7] KUMAR A, LEE W S, EHSANI R J, et al. Citrus greening disease detection using aerial hyperspectral and multispectral imaging techniques[J]. J Appl Remote Sens, 2012, 6(1): 63542. doi: 10.1117/1.JRS.6.063542
[8] 兰玉彬, 朱梓豪, 邓小玲, 等. 基于无人机高光谱遥感的柑橘黄龙病植株的监测与分类[J]. 农业工程学报, 2019, 35(3): 92-100. doi: 10.11975/j.issn.1002-6819.2019.03.012 [9] LAN Y, HUANG Z, DENG X. Comparison of machine learning methods for citrus greening detection on UAV multispectral images[J]. Comput Electron Agr, 2020, 171: 10524. doi: 10.1016/j.compag.2020.105234
[10] 李修华, 李民赞, LEE W S, et al. 柑桔黄龙病的可见−近红外光谱特征[J]. 光谱学与光谱分析, 2014, 34(6): 1553-1559. doi: 10.3964/j.issn.1000-0593(2014)06-1553-07 [11] 尚方信, 郭浩, 李钢, 等. 基于One-class SVM的噪声图像分割方法[J]. 计算机应用, 2019, 39(3): 874-881. doi: 10.11772/j.issn.1001-9081.2018071494 [12] TAX D M J, DUIN R P W. Support vector data description[J]. Mach Learn, 2004(54): 45-66.
[13] CHAWLA N V, BOWYER K W, HALL L O. SMOTE: Synthetic minority over-sampling technique[J]. J Artif Intell Res, 2002(16): 321-357.
[14] 成忠, 张立庆, 刘赫扬, 等. 连续投影算法及其在小麦近红外光谱波长选择中的应用[J]. 光谱学与光谱分析, 2010, 30(4): 949-952. doi: 10.3964/j.issn.1000-0593(2010)04-0949-04 [15] 高洪智, 卢启鹏, 丁海泉, 等. 基于连续投影算法的土壤总氮近红外特征波长的选取[J]. 光谱学与光谱分析, 2009, 29(11): 2951-2954. doi: 10.3964/j.issn.1000-0593(2009)11-2951-04 [16] 吴迪, 金春华, 何勇. 基于连续投影算法的光谱主成分组合优化方法研究[J]. 光谱学与光谱分析, 2009, 29(10): 2734-2737. doi: 10.3964/j.issn.1000-0593(2009)10-2734-04 [17] CHEN T, GUESTRIN C. XgBoost: A scalable tree boosting system[C]//The 22nd ACM SIGKDD International Conference. ACM, 2016: 785-794.
[18] 王术波, 韩宇, 陈建, 等. 基于深度学习的无人机遥感生态灌区杂草分类[J]. 排灌机械工程学报, 2018, 36(11): 1137-1141. [19] 高林, 杨贵军, 于海洋, 等. 基于无人机高光谱遥感的冬小麦叶面积指数反演[J]. 农业工程学报, 2016, 32(22): 113-120. doi: 10.11975/j.issn.1002-6819.2016.22.016
