| 基于无人机平台的柑橘树冠信息提取 |
| |
| 引用本文： | 束美艳,李世林,魏家玺,车荧璞,李保国,马韫韬.基于无人机平台的柑橘树冠信息提取[J].农业工程学报,2021,37(1):68-76. |
| |
| 作者姓名： | 束美艳 李世林 魏家玺 车荧璞 李保国 马韫韬 |
| |
| 作者单位： | 中国农业大学土地科学与技术学院，北京 100193 |
| |
| 基金项目： | 内蒙古科技重大专项（2019ZD08） |
| |
| 摘 要： | 为了快速获取柑橘树冠信息,提升柑橘园精准管理,该研究基于无人机平台获取了柑橘数码和多光谱影像,分析了无人机影像反演柑橘树冠信息的效果。首先利用无人机数码影像及分水岭算法进行柑橘单木分割,然后构建柑橘树冠层高度模型,提取柑橘株数、株高、冠幅投影面积等结构参数信息,进而利用无人机多光谱影像获取柑橘的8种常用植被指数,采用全子集分析法筛选柑橘冠层氮素含量的敏感植被指数,构建基于多元线性回归的冠层氮素遥感反演模型,进行以冠幅为基本单元的柑橘树冠层氮素含量遥感制图。研究结果表明:柑橘的单木识别准确率在93%以上,召回率在95%以上,平均F值为96.52%;柑橘树的反演株高与实测株高具有较强的相关性,决定系数R2为0.87,均方根误差为31.9cm;单株冠幅投影面积与人工绘制的冠幅面积的决定系数,除果园A在12月的结果较低(R2为0.78)外,其余均在0.94及以上;采用全子集分析法筛选的柑橘冠层氮素敏感植被指数为归一化植被指数(NDVI)、绿色归一化植被指数和冠层结构不敏感指数,所建立的多元回归模型的决定系数R2达0.82,均方根误差为0.22%,相对误差为6.59%。综上,无人机影像在柑橘树冠参数信息提取方面具有较好的应用效果,能够快速有效地提取柑橘树冠参数信息。该研究可为使用无人机平台进行果园精准管理提供技术支撑。
|
| 关 键 词： | 无人机 多光谱 柑橘 株高 冠幅投影面积 冠层氮素含量 |
| 收稿时间： | 2020/11/19 0:00:00 |
| 修稿时间： | 2021/12/15 0:00:00 |
Extraction of citrus crown parameters using UAV platform |
| |
| Shu Meiyan,Li Shilin,Wei Jiaxi,Che Yingpu,Li Baoguo,Ma Yuntao.Extraction of citrus crown parameters using UAV platform[J].Transactions of the Chinese Society of Agricultural Engineering,2021,37(1):68-76. |
| |
| Authors: | Shu Meiyan Li Shilin Wei Jiaxi Che Yingpu Li Baoguo Ma Yuntao |
| |
| Institution: | College of Land Science and Technology, China Agricultural University, Beijing 100193, China |
| |
| Abstract: | Citrus fruit, one of the most important economic crops, is playing an important role in the industrial development of modern agriculture in rural China. However, the management mode of most orchards in China is currently undeveloped and extensive, particularly with high dependence on labor force, as well as insufficient scientific and technological support. In recent years, the unmanned aerial vehicle (UAV) monitoring technology has become a significant way to quickly extract the structural parameters in the growth of field crops at the park scale, due to its flexibility, low cost, and high resolution imaging. This study aims to construct a monitoring system for the citrus canopy structure and nutrition information using the UAV digital and multi-spectral remote sensing, to get he with the single tree segmentation. The UAV digital images and watershed algorithm were used to segment the structural dataset of citrus canopy, and then the canopy height model of citrus trees was established to extract the plant height using digital surface module. Structural parameters were also calculated, such as the number of citrus trees, and canopy projection area at the park scale. In addition, the UAV multispectral images were used to obtain eight common vegetation indexes, thereby to predict the nitrogen content of canopy in the citrus trees. The whole subset analysis was used to screen the sensitive vegetation index for the nitrogen content of canopy in the citrus trees. The inversion model of canopy nitrogen was constructed using the multiple linear regression. The remote sensing mapping was carried out to estimate the nitrogen content of citrus canopy in park scale. The results showed that: (1)Since the planting density of fruit trees was low in the experimental area, there was a certain distance between trees that can be clearly distinguished. The watershed image processing was selected to segment the single tree of height model for a citrus canopy. The overall identification accuracy, recall rate, and F value of the fruit trees were above 93 %, 95 %, and 94 %, respectively, indicating that the model was well suitable to monitor the number of fruit trees in the park. (2) The canopy structure parameters of individual fruit trees were obtained in the individual tree segmentation. There was a strong correlation between the plant height of citrus trees extracted by the canopy height model and the measured value, where the R2 = 0.87, and RMSE = 31.9cm. (3) Using the watershed segmentation, the extracted projection area of crown width per plant achieved a high correlation with the artificial sketching area. The correlation coefficient was more than 0.93 in most cases, except that of orchard A lower than 0.78 in December. Meanwhile, the extraction accuracy of the model depended greatly on the single tree segmentation. (4) In full subset analysis, the sensitive vegetation indexes were selected to determine the nitrogen content of citrus canopy, including the normalized difference vegetation index (NDVI), green normalized difference vegetation index (GNDVI), and simple ratio pigment index(SIPI), where the R2 and RMSE of the model were 0.82 and 0.22%, respectively. The data demonstrated that the nitrogen content of most fruit trees in orchard B was in the suitable range, while there was excessive application of nitrogen fertilizer in orchard A. Therefore, the UAV technology can greatly contribute to extract the physical and chemical parameters of citrus canopy, further to improve the level of accurate management of citrus on the large-scale orchard. |
| |
| Keywords: | UAV multi-spectral citrus plant height crown projection area canopy nitrogen content |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《农业工程学报》浏览原始摘要信息 |
| 点击此处可从《农业工程学报》下载免费的PDF全文 |
|
