| 基于多时相和多角度的烤烟烟碱密度遥感监测 |
| |
| 引用本文： | 杨艳东,贾方方,刘国顺,彭桂新,于建春.基于多时相和多角度的烤烟烟碱密度遥感监测[J].中国烟草学报,2019,25(2):40-47. |
| |
| 作者姓名： | 杨艳东 贾方方 刘国顺 彭桂新 于建春 |
| |
| 作者单位： | 1.河南农业大学/烟草行业烟草栽培重点实验室/河南省生物炭工程技术研究中心, 河南省郑州市农业路63号 450002 |
| |
| 基金项目： | 河南中烟工业有限责任公司项目ZYKJ201416 |
| |
| 摘 要： | 目的 快速、准确获取烟株烟碱含量，实时把控烟叶品质和精准施肥。 方法 以不同区域、品种、氮肥处理连续两年的田间试验为基础，采集了烤烟整个生育期的多角度光谱反射率，选用观测天顶角和方位角组合形成的光谱数据构建烟碱密度的植被指数，建立最佳观测角度下的烟碱密度监测模型。 结果 （1）30°天顶角同210°方位角的角度组合为最佳观测角度；（2）RVI（1630，1740）和NDVI（1630，1740）为监测烟碱密度的最佳敏感光谱参数；（3）构建的两种预测模型BP神经网络和支持向量机，其R2分别为0.944和0.996，模型验证的均方根误差P-RMSE分别为0.858和0.011，两者均具有较高的精准度和普适性。 结论 表明多角度、多时相遥感在大田烟碱密度的监测中具有良好的应用前景。
|
| 关 键 词： | 多角度 高光谱 烟碱密度 BP神经网络 支持向量机 |
| 收稿时间： | 2018-09-14 |
Monitoring of nicotine density in flue-cured tobacco based on multitemporal and multiangular remote sensing |
| |
| Affiliation: | 1.Henan Agricultural University, Key Laboratory for Tobacco Cultivation/Henan Engineering Research Center for Biochar, Zhengzhou 450002, China2.Shangqiu Normal University, Shangqiu 476000, China3.Technology Center, China Tobacco Henan Industrial Co., Ltd, Zhengzhou 450016, China |
| |
| Abstract: | Rapid and accurate measuring of nicotine content in tobacco plants is of great significance for the control of tobacco quality and precise fertilization. As nicotine is in a non-uniform distribution in tobacco plants, current single-angle and single-period remote sensing observation methods may neglect nicotine information of middle and lower leaves or only obtain single period information of vegetation, which is difficult to meet the needs of quantitative remote sensing. In contrast, multi-angle, multi-temporal remote sensing can provide directional information and time information of vegetation, which helps to understand the characteristics of two-way reflection and improve the ability to quantitatively invert vegetation structure and physiological parameters. This research was carried out in different regions, varieties and nitrogen fertilizers in Xuchang Tobacco Science and Education Park of Henan Agricultural University, Songxian County of Luoyang City and Lushi County of Sanmenxia City for two consecutive years from 2016 to 2017. Data were collected from root-grown period of flue-cured tobacco (30d after transplanting), prolonged period (55d after transplanting) and mature period (75d after transplanting); for each treatment, 3 strains with uniform growth and typical representativeness were selected. Spectral reflectance of 73 angles per tobacco was collected. The total number of samples was 175, 200, and the total number of samples per single angle was 2, 400. Half of the random selection was made for modeling and verification. Nicotine density vegetation index was constructed using spectral data formed by different combinations of zenith angle and azimuth angle. Quantitative relationship between nicotine density normalized vegetation index (NDVI) and ratio vegetation index (RVI) was systematically analyzed by decremental fine sampling method. Results showed that the combination angle of 30° zenith angle and 210° azimuth was the best observation angle for nicotine canopy density; band combinations of R2 > 0.8 of RVI were (1, 596~1, 610, 1, 757~1, 761), (1, 609~11, 672, 1, 733~1, 760) and (1, 682~1, 691, 1, 716~1, 722); band combinations of R2 > 0.8 of NDVI were (1, 598~1, 605, 1, 758~1, 761), (1, 610~1, 668, 1, 734~1, 755) and (1, 683~1, 691, 1, 717~1, 722); and the best sensitive spectral parameters for monitoring nicotine density were RVI (1, 630, 1, 740) and NDVI (1, 630, 1, 740). A random sample of 120 samples from a total sample of single angle were used to build a prediction model, while the remaining 120 samples were used to validate the model. BP neural network and SVM (support vector machine), as nonlinear nicotine density prediction models, were constructed by selecting the top 20 RVI values and the first 20 NDVI with the best angle of 30°-210° values as independent variables, with the R2 being 0.944 and 0.996, respectively, and the RMSE of the model verification being 0.858 and 0.011, respectively. Both models had high precision and universality. This study proves that multi-angle and multi-temporal remote sensing has a good application prospect in the monitoring of nicotine density of tobacco. |
| |
| Keywords: | |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《中国烟草学报》浏览原始摘要信息 |
|
点击此处可从《中国烟草学报》下载全文 |
|
