Browsing by Subject "UAV"

Now showing 1 - 3 of 3

Effects of different ground segmentation methods on the accuracy of UAV-based canopy volume measurements
(2024) Han, Leng; Wang, Zhichong; He, Miao; He, Xiongkui; Han, Leng; College of Science, China Agricultural University, Beijing, China; Wang, Zhichong; Tropics and Subtropics Group, Institute of Agricultural Engineering, University of Hohenheim, Stuttgart, Germany; He, Miao; College of Science, China Agricultural University, Beijing, China; He, Xiongkui; College of Science, China Agricultural University, Beijing, China
The nonuniform distribution of fruit tree canopies in space poses a challenge for precision management. In recent years, with the development of Structure from Motion (SFM) technology, unmanned aerial vehicle (UAV) remote sensing has been widely used to measure canopy features in orchards to balance efficiency and accuracy. A pipeline of canopy volume measurement based on UAV remote sensing was developed, in which RGB and digital surface model (DSM) orthophotos were constructed from captured RGB images, and then the canopy was segmented using U-Net, OTSU, and RANSAC methods, and the volume was calculated. The accuracy of the segmentation and the canopy volume measurement were compared. The results show that the U-Net trained with RGB and DSM achieves the best accuracy in the segmentation task, with mean intersection of concatenation (MIoU) of 84.75% and mean pixel accuracy (MPA) of 92.58%. However, in the canopy volume estimation task, the U-Net trained with DSM only achieved the best accuracy with Root mean square error (RMSE) of 0.410 m 3 , relative root mean square error (rRMSE) of 6.40%, and mean absolute percentage error (MAPE) of 4.74%. The deep learning-based segmentation method achieved higher accuracy in both the segmentation task and the canopy volume measurement task. For canopy volumes up to 7.50 m 3 , OTSU and RANSAC achieve an RMSE of 0.521 m 3 and 0.580 m 3 , respectively. Therefore, in the case of manually labeled datasets, the use of U-Net to segment the canopy region can achieve higher accuracy of canopy volume measurement. If it is difficult to cover the cost of data labeling, ground segmentation using partitioned OTSU can yield more accurate canopy volumes than RANSAC.
Implementation and improvement of an unmanned aircraft system for precision farming purposes
(2016) Geipel, Jakob; Claupein, Wilhelm
Precision farming (PF) is an agricultural concept that accounts for within-field variability by gathering spatial and temporal information with modern sensing technology and performs variable and targeted treatments on a smaller scale than field scale. PF research quickly recognized the possible benefits unmanned aerial vehicles (UAVs) can add to the site-specific management of farms. As UAVs are flexible carrier platforms, they can be equipped with a range of different sensing devices and used in a variety of close-range remote sensing scenarios. Most frequently, UAVs are utilized to gather actual in-season canopy information with imaging sensors that are sensitive to reflected electro-magnetic radiation in the visual (VIS) and near-infrared (NIR) spectrum. They are generally used to infer the crops’ biophysical and biochemical parameters to support farm management decisions. A current disadvantage of UAVs is that they are not designed to interact with their attached sensor payload. This leads to the need of intensive data post-processing and prohibits the possibility of real-time scenarios, in which UAVs can directly transfer information to field machinery or robots. In consequence, this thesis focused on the development of a smart unmanned aircraft system (UAS), which in the thesis’ context was regarded as a combination of a UAV carrier platform, an on-board central processing unit for sensor control and data processing, and a remotely connected ground control station. The ground control station was supposed to feature the possibility of flight mission control and the standardized distribution of sensor data with a sensor data infrastructure, serving as a data basis for a farm management information system (FMIS). The UAS was intended to be operated as a flexible monitoring tool for in-season above-ground biomass and nitrogen content estimation as well as crop yield prediction. Therefore, the selection, development, and validation of appropriate imaging sensors and processing routines were key parts to prove the UAS’ usability in PF scenarios. The individual objectives were (i) to implement an advanced UAV for PF research, providing the possibilities of remotely-controlled and automatic flight mission execution, (ii) to improve the developed UAV to a UAS by implementing sensor control, data processing and communication functionalities, (iii) to select and develop appropriate sensor systems for yield prediction and nitrogen fertilization strategies, (iv) to integrate the sensor systems into the UAS and to test the performance in example use cases, and (v) to embed the UAS into a standardized sensor data infrastructure for data storage and usage in PF applications. This work demonstrated the successful development of a custom rotary-wing UAV carrier platform with an embedded central processing unit. A modular software framework was developed with the ability to control any kind of sensor payload in real-time. The sensors can be triggered and their measurements are retrieved, fused together with the carrier’s navigation information, logged and broadcasted to a ground control station. The setup was used as basis for further research, focusing on information generation by sophisticated data processing. For a first application of predicting the grain yield of corn (Zea mays L.), a simple RGB camera was selected to acquire a set of aerial imagery of early- and mid-season corn crops. Orthoimages were processed with different ground resolutions and were computed to simple vegetation indices (VI) for a crop/non-crop classification. In addition to that, crop surface models (CSMs) were generated to estimate the crop heights. Linear regressions were performed with the corn grain yield as dependent variable and crop height and crop coverage as independent variable. The analysis showed the best prediction results of a relative root mean square error (RMSE) of 8.8 % at mid-season growth stages and ground resolutions of 4 cm px −1 . Moreover, the results indicate that with on-going canopy closure and homogeneity accounting for high ground resolutions and crop/non-crop classification becomes less and less important. For the estimation of above-ground biomass and nitrogen content in winter wheat (Triticum aestivum L.) a programmable multispectral camera was developed. It is based on an industrial multi-sensor camera, which was equipped with bandpass filters to measure four narrow wavelength bands in the so-called red-edge region. This region is the transition zone in between the VIS and NIR spectrum and known to be sensitive to leaf chlorophyll content and the structural state of the plant. It is often used to estimate biomass and nitrogen content with the help of the normalized difference vegetation index (NDVI) and the red-edge inflection point (REIP). The camera system was designed to measure ambient light conditions during the flight mission to set appropriate image acquisition times, which guarantee images with high contrast. It is fully programmable and can be further developed to a real-time image processing system. The analysis relies on semi-automatic orthoimage processing. The NDVI orthoimages were analyzed for the correlation with biomass by means of simple linear regression. These models proved to estimate biomass for all measurements with RMSEs of 12.3 % to 17.6 %. The REIP was used to infer nitrogen content and showed good results with RMSEs of 7.6 % to 11.7 %. Both NDVI and REIP were also tested for the in-season grain yield prediction ability (RMSE = 9.0–12.1 %), whereas grain protein content could be modeled with the REIP, except for low-fertilized wheat plots. The last part of the thesis comprised the development of a standardized sensor data infrastructure as a first step to a holistic farm management. The UAS was integrated into a real-time sensor data acquisition network with standardized data base storage capabilities. The infrastructure was based on open source software and the geo-data standards of the Open Geospatial Consortium (OGC). A prototype implementation was tested for four exemplary sensor systems and proved to be able to acquire, log, visualize and store the sensor data in a standardized data base via a sensor observation service on-the-fly. The setup is scalable to scenarios, where a multitude of sensors, data bases, and web services interact with each other to exchange and process data. This thesis demonstrates the successful prototype implementation of a smart UAS and a sensor data infrastructure, which offers real-time data processing functionality. The UAS is equipped with appropriate sensor systems for agricultural crop monitoring and has the potential to be used in real-world scenarios.
UAV remote sensing for high-throughput phenotyping and for yield prediction of Miscanthus by machine learning techniques
(2022) Impollonia, Giorgio; Croci, Michele; Ferrarini, Andrea; Brook, Jason; Martani, Enrico; Blandinières, Henri; Marcone, Andrea; Awty-Carroll, Danny; Ashman, Chris; Kam, Jason; Kiesel, Andreas; Trindade, Luisa M.; Boschetti, Mirco; Clifton-Brown, John; Amaducci, Stefano
Miscanthus holds a great potential in the frame of the bioeconomy, and yield prediction can help improve Miscanthus’ logistic supply chain. Breeding programs in several countries are attempting to produce high-yielding Miscanthus hybrids better adapted to different climates and end-uses. Multispectral images acquired from unmanned aerial vehicles (UAVs) in Italy and in the UK in 2021 and 2022 were used to investigate the feasibility of high-throughput phenotyping (HTP) of novel Miscanthus hybrids for yield prediction and crop traits estimation. An intercalibration procedure was performed using simulated data from the PROSAIL model to link vegetation indices (VIs) derived from two different multispectral sensors. The random forest algorithm estimated with good accuracy yield traits (light interception, plant height, green leaf biomass, and standing biomass) using 15 VIs time series, and predicted yield using peak descriptors derived from these VIs time series with root mean square error of 2.3 Mg DM ha−1. The study demonstrates the potential of UAVs’ multispectral images in HTP applications and in yield prediction, providing important information needed to increase sustainable biomass production.