| 无人机飞行高度对植被覆盖度和植被指数估算结果的影响 |
| |
| 引用本文： | 何勇,杜晓月,郑力源,朱姜蓬,岑海燕,许丽佳.无人机飞行高度对植被覆盖度和植被指数估算结果的影响[J].农业工程学报,2022,38(24):63-72. |
| |
| 作者姓名： | 何勇 杜晓月 郑力源 朱姜蓬 岑海燕 许丽佳 |
| |
| 作者单位： | 1. 浙江大学生物系统工程与食品科学学院，杭州 310058；2. 农业农村部光谱检测重点实验室，杭州 310058；;3. 四川农业大学机电学院，四川 625014 |
| |
| 基金项目： | 浙江省重点研发计划（021C02023） |
| |
| 摘 要： | 将无人机与多种成像传感设备相结合可实现田间作物表型信息的全面获取。针对田间复杂环境下无人机搭载多种成像传感设备在不同飞行高度处提取的作物信息具有差异性的问题，本研究着重探究了无人机搭载两种成像传感设备获取图像时，不同飞行高度对估算植被覆盖度以及植被指数结果的影响。首先为防止外界环境变化对获取图像质量造成干扰，通过最近邻插值算法将无人机飞行高度为25 m处获取的两个多光谱和可见光图像数据集分别退化为十个不同地面分辨率的模糊图像数据集，以模拟无人机在不同飞行高度中获取的作物图像。然后获取50m高度处的无人机图像数据集通过皮尔逊相关性分析验证模拟数据集的有效性。最后采用随机森林模型估算不同数据集中的植被覆盖度，分类精度大于91%。结果发现，当植被覆盖度小于二分之一时，随着地面分辨率的降低该指标不断被低估，反之则被高估。飞行高度50 m的真实图像与模拟图像估算植被覆盖度结果的相关系数r为0.992 8，两者具有强相关性，模拟图像估算得到的植被覆盖度变化具备参考意义。植被指数估算结果中，首先对无人机图像数据集进行辐射校正、阈值分割等图像预处理，然后根据公式计算得到植被指数，最后通过假设性检验对十个图像数据集计算得出的植被指数进行分析。结果发现，可见光植被指数在飞行高度61 m时具备显著性差异，多光谱植被指数在十个高度下均没有显著性差异，因此为保证无人机获取数据的准确性与完整性，建议当无人机搭载本文的两种相机获取作物信息时建议飞行高度不高于61 m。本研究为研究者利用无人机搭载多传感设备获取作物信息设定合适的飞行高度、减小作业成本提供参考。
|
| 关 键 词： | 无人机 植被覆盖度 植被指数 飞行策略 |
| 收稿时间： | 2022/9/5 0:00:00 |
| 修稿时间： | 2022/11/11 0:00:00 |
Effects of UAV flight height on estimated fractional vegetation cover and vegetation index |
| |
| He Yong,Du Xiaoyue,Zheng Liyuan,Zhu Jiangpeng,Cen Haiyan,Xu Lijia.Effects of UAV flight height on estimated fractional vegetation cover and vegetation index[J].Transactions of the Chinese Society of Agricultural Engineering,2022,38(24):63-72. |
| |
| Authors: | He Yong Du Xiaoyue Zheng Liyuan Zhu Jiangpeng Cen Haiyan Xu Lijia |
| |
| Institution: | 1. College of Biosystems Engineering and Food Science, Zhejiang University, Hangzhou 310058, China; 2. Key Laboratory of Spectroscopy Sensing, Ministry of Agriculture and Rural Affairs, Hangzhou 310058, China;; 3. College of Mechanical and Electrical Engineering, Sichuan Agricultural University, Sichuan 625014, China |
| |
| Abstract: | Abstract: Accurate crop information acquisition is very important for real-time monitoring of crop growth status, and is also the critical for crop breeding and agricultural precision management. With the development of remote sensing (RS) technology and Unmanned Aerial Vehicle (UAV) flight technology, big phenotypic data of various plants especially image data can be collected in a large area. Different sensing devices are often used for the purpose of obtaining accurate and comprehensive information on crop growth. The diversity of sensing devices brings with it the need to establish multiple parameters adapted to the environment and the cooperation between them. There is an urgent need to develop effective approaches to setting parameters to explore UAV flight height that can ultimately be used for multi-sensor combinations in this process. Multi-sensor based on the UAV can obtain phenotypic data within the effective range under this parameter. The multispectral (MS) camera and high-resolution RGB camera are simultaneously mounted on the UAV to acquire images at a speed of 2.5m/s. To guarantee the possibility of successful image stitching, the heading overlap and the side overlap are set to 75% and 60%, respectively. The initial flight heights are set to 25m and 50m in order to exclude the wind field generated by the high-speed rotation of the UAV''s paddles from disturbing the crop. UAV at low flight height can obtain images with high ground resolution (GR). These images are degraded by image processing algorithms into a series of images with different GR, which are used to simulate the crop images obtained by the UAV at different flight heights. This is important to reduce the impact of environmental changes such as light intensity on image quality. The random forest (RF) algorithm has the feature of not requiring a large number of training samples, so this method is used to calculate the fractional vegetation cover (FVC) of every sample plot in this paper. The results show that the classification accuracy is greater than 91%. Pearson correlation analysis is also performed using real and simulated images at a flight height of 50m to verify the feasibility of simulated images for FVC estimation. The vegetation indices (VIs) are the combination of spectral reflectance, which can be used to enhance vegetation information and weaken non-vegetation information to a certain extent. we calculated five VIs for validating our study according to the formula. The results demonstrate that FVC will show regular changes according to the GR because of the image blending phenomenon. When the FVC is less than half, the indicator is constantly underestimated with the decrease of the GR, otherwise, it is overestimated. The Pearson correlation coefficient between the real and simulated images at a flight height of 50m is 0.9928, which has a high correlation. And the FVC estimated using the real image also has the same regular variation as above. Significant differences in RGB VIs occurred at a ground resolution of 15 mm/pixel, i.e., flight height of 61 m, and no differences at 12 mm/pixel, i.e., flight height of 50 m. However, the MS VIs were not significantly different at both ground resolutions. The spatial resolution of the multispectral sensor used in this paper is lower than that of the RGB camera, and the calculated minimum ground resolution is 43 mm/pixel, which is not as fine as that of the RGB image. Therefore, some vegetation information is confused in the multispectral images under the flight height gradient of this study. The image blending effect of the acquired RGB images in a certain flight height range within 61m has less impact on the acquisition of spatial information. When we use these two devices to collect crop information at the same time, it is possible to set the flight height no higher than 61 meters. It provides a reference for researchers to obtain crop information to set a suitable flying height and reduce operating costs by using multi-sensor equipment carried by UAVs. The changes of spectral and spatial information in higher flight heights can be investigated when there is an experimental demand according to the method of this paper. |
| |
| Keywords: | UAV Fractional Vegetation Cover Vegetation Index Flight strategy |
|
| 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 |
|
