Biography

I am a researcher and expert for utilizing unmanned aerial vehicles (UAV) and unmanned ground vehicles (UGV) for sensor measurements in agriculture. I focus amongst others on hyperspectral remote sensing, photogrammetry, image processing, geo-information, programming, prototyping, and multivariate statistics. My field of research comprises both grain and forage production.

CV

Competences: 
Multi- and hyperspectral remote sensing in agriculture, UAV, UGV, GNSS, GIS, sensor web, mapping, 3D modelling, multivariate and geo-statistics, programming, prototyping

Education: 
2012-2016: Dr. sc. agr. (Ph.D.) in Agricultural Sciences at the Institute of Crop Science, Department of Agronomy, University of Hohenheim, Germany

2009-2012: M.Sc. in Geoinformatics at the Institute for Geoinformatics, University of Münster, Germany

2005-2009: Dipl.-Ing. (FH) in Surveying Engineering and Geoinformatics at the University of Applied Sciences Würzburg-Schweinfurt, Germany

Read more
To document

Abstract

The ageing population, climate change, and labour shortages in the agricultural sector are driving the need to reevaluate current farming practices. To address these challenges, the deployment of robot systems can help reduce environmental footprints and increase productivity. However, convincing farmers to adopt new technologies poses difficulties, considering economic viability and ease of use. In this paper, we introduce a management system based on the Robot Operating System (ROS) that integrates heterogeneous vehicles (conventional tractors and mobile robots). The goal of the proposed work is to ease the adoption of mobile robots in an agricultural context by providing to the farmer the initial tools needed to include them alongside the conventional machinery. We provide a comprehensive overview of the system’s architecture, the control laws implemented for fleet navigation within the field, the development of a user-friendly Graphical User Interface, and the charging infrastructure for the deployed vehicles. Additionally, field tests are conducted to demonstrate the effectiveness of the proposed framework.

Abstract

A process-based model was developed to predict dry matter yields and amounts of harvested nitrogen in conventionally cropped grassland fields, accounting for within-field variation by a node network design and utilizing remotely sensed information from a drone-borne system for increased accuracy. The model, named NORNE, was kept as simple as possible regarding required input variables, but with sufficient complexity to handle central processes and minimize prediction errors. The inputs comprised weather data, soil information, management data related to fertilization, and a visual estimate of clover proportion in the aboveground biomass. A sensitivity analysis was included to apportioning variation in dry matter yield outputs to variation in model parameter settings. Using default parameter values from the literature, the model was evaluated on data from a two-year study (2016–2017, 264 research plots in total each year) conducted at two locations in Norway (i.e. in South-East and in Central Norway) with contrasting climatic conditions and with internal variation in soil characteristics. The results showed that the model could estimate dry matter yields with a relatively high accuracy without any corrections based on remote sensing, compared with published results from comparable model studies. To further improve the results, the model was calibrated shortly before harvest, using predictions of above ground dry matter biomass obtained from a drone-borne remote sensing system. The only parameters which were hereby adjusted in the NORNE model were the starting values of nitrogen content in soil (first cut) and the plant available water capacity (second cut). The calibration based on the remotely sensed information improved the predictive performance of the model significantly. At first cut, the root mean square error (RMSE) of dry matter yield prediction was reduced by 20% to a mean value of 58 g m−2, corresponding to a relative value (rRMSE) of 0.12. For the second cut, the RMSE decreased by 13% to 66 g m−2 (rRMSE: 0.18). The model was also evaluated in terms of the predictions of amounts of nitrogen in the harvested crop. Here, the calibration reduced the RMSE of the first cut by 38%, obtaining a mean RMSE value of 2.1 g N m−2 (rRMSE: 0.28). For the second cut, the RMSE reduction for simulated harvested N was 16%, corresponding to a mean RMSE value of 2.3 g N m−2 (rRMSE: 0.33). The large improvements in model accuracy for simulated dry matter and nitrogen yields obtained through calibration by utilizing remotely sensed information, indicate the importance of considering spatial variability when applying models under Nordic conditions, both for yield predictions and for decision support for nitrogen application.

Abstract

Weeds affect crop yield and quality due to competition for resources. In order to reduce the risk of yield losses due to weeds, herbicides or non-chemical measures are applied. Weeds, especially creeping perennial species, are generally distributed in patches within arable fields. Hence, instead of applying control measures uniformly, precision weeding or site-specific weed management (SSWM) is highly recommended. Unmanned aerial vehicle (UAV) imaging is known for wide area coverage and flexible operation frequency, making it a potential solution to generate weed maps at a reasonable cost. Efficient weed mapping algorithms need to be developed together with UAV imagery to facilitate SSWM. Different machine learning (ML) approaches have been developed for image-based weed mapping, either classical ML models or the more up-to-date deep learning (DL) models taking full advantage of parallel computation on a GPU (graphics processing unit). Attention-based transformer DL models, which have seen a recent boom, are expected to overtake classical convolutional neural network (CNN) DL models. This inspired us to develop a transformer DL model for segmenting weeds, cereal crops, and ‘other’ in low-resolution RGB UAV imagery (about 33 mm ground sampling distance, g.s.d.) captured after the cereal crop had turned yellow. Images were acquired during three years in 15 fields with three cereal species (Triticum aestivum, Hordeum vulgare, and Avena sativa) and various weed flora dominated by creeping perennials (mainly Cirsium arvense and Elymus repens). The performance of our transformer model, 1Dtransformer, was evaluated through comparison with a classical DL model, 1DCNN, and two classical ML methods, i.e., random forest (RF) and k-nearest neighbor (KNN). The transformer model showed the best performance with an overall accuracy of 98.694% on pixels set aside for validation. It also agreed best and relatively well with ground reference data on total weed coverage, R2 = 0.598. In this study, we showed the outstanding performance and robustness of a 1Dtransformer model for weed mapping based on UAV imagery for the first time. The model can be used to obtain weed maps in cereals fields known to be infested by perennial weeds. These maps can be used as basis for the generation of prescription maps for SSWM, either pre-harvest, post-harvest, or in the next crop, by applying herbicides or non-chemical measures.

Abstract

The remote sensing of the biophysical and biochemical parameters of crops facilitates the preparation of application maps for variable-rate nitrogen fertilization. According to comparative studies of machine learning algorithms, Gaussian process regression (GPR) can outperform more popular methods in the prediction of crop status from hyperspectral data. The present study evaluates GPR model accuracy in the context of spring wheat dry matter, nitrogen content, and nitrogen uptake estimation. Models with the squared exponential covariance function were trained on images from two hyperspectral cameras (a frenchFabry–Pérot interferometer camera and a push-broom scanner). The most accurate predictions were obtained for nitrogen uptake (R2=0.75–0.85, RPDP=2.0–2.6). Modifications of the basic workflow were then evaluated: the removal of soil pixels from the images prior to the training, data fusion with apparent soil electrical conductivity measurements, and replacing the Euclidean distance in the GPR covariance function with the spectral angle distance. Of these, the data fusion improved the performance while predicting nitrogen uptake and nitrogen content. The estimation accuracy of the latter parameter varied considerably across the two hyperspectral cameras. Satisfactory nitrogen content predictions (R2>0.8, RPDP>2.4) were obtained only in the data-fusion scenario, and only with a high spectral resolution push-broom device capable of capturing longer wavelengths, up to 1000 nm, while the full-frame camera spectral limit was 790 nm. The prediction performance and uncertainty metrics indicated the suitability of the models for precision agriculture applications. Moreover, the spatial patterns that emerged in the generated crop parameter maps accurately reflected the fertilization levels applied across the experimental area as well as the background variation of the abiotic growth conditions, further corroborating this conclusion.

Abstract

This paper describes a tool that enables farmers to time harvests and target nitrogen (N) inputs in their forage production, according to the prevailing yield potential. Based on an existing grass growth model for forage yield estimation, a more detailed process-based model was developed, including a new nitrogen module. The model was tested using data from an experiment conducted in a grassland-rich region in central Norway and showed promising accuracy with estimated root mean square error (RMSE) of 50 and 130 g m-2 for dry matter yield in the trial. Three parameters were detected as highly sensitive to model output: initial value of organic N in the soil, fraction of humus in the initial organic N in the soil, and fraction of decomposed N mineralized. By varying these parameters within a range from 0.5 to 1.5 of their respective initial value, most of the within-field variation was captured. In a future step, remotely sensed information on model output will be included, and in-season model correction will be performed through re-calibration of the highly sensitive parameters.

Abstract

This study investigated the potential of in-season airborne hyperspectral imaging for the calibration of robust forage yield and quality estimation models. An unmanned aerial vehicle (UAV) and a hyperspectral imager were used to capture canopy reflections of a grass-legume mixture in the range of 450 nm to 800 nm. Measurements were performed over two years at two locations in Southeast and Central Norway. All images were subject to radiometric and geometric corrections before being processed to ortho-images, carrying canopy reflectance information. The data (n = 707) was split in two, using half the data for model calibration and the remaining half for validation. Several powered partial least squares regression (PPLSR) models were fitted to the reflectance data to estimate fresh (FM) and dry matter (DM) yields, as well as crude protein (CP), dry matter digestibility (DMD), neutral detergent fibre (NDF), and indigestible neutral detergent fibre (iNDF) content. Prediction performance of these models was compared with the prediction performance of simple linear regression (SLR) models, which were based on selected vegetation indices and plant height. The highest prediction accuracies for general models, based on the pooled data, were achieved by means of PPLSR, with relative root-mean-square errors of validation of 14.2% (2550 kg FM ha−1), 15.2% (555 kg DM ha−1), 11.7% (1.32 g CP 100 g−1 DM), 2.4% (1.71 g DMD 100 g−1 DM), 4.8% (2.72 g NDF 100 g−1 DM), and 12.8% (1.32 g iNDF 100 g−1 DM) for the prediction of FM, DM, CP, DMD, NDF, and iNDF content, respectively. None of the tested SLR models achieved acceptable prediction accuracies.

Abstract

Grassland farmers face ever increasing demands on their production systems and the quality of their grassland yields. Estimating pasture quality using traditional field methods is limited as it is time consuming and costly, and requires some destructive sampling. The field of remote sensing offers alternative tools and techniques to overcome some of the limitations and thereby help farmers to receive spatial continuous and near real-time information about grassland quality parameters. This review gives an overview about recent developments in the remote sensing-based estimation of three aspects of grassland quality: feed quality, biological nitrogen fixation by legumes, and the identification of unwanted plant species.

Abstract

Today’s modern precision agriculture applications have a huge demand for data with high spatial and temporal resolution. This leads to the need of unmanned aerial vehicles (UAV) as sensor platforms providing both, easy use and a high area coverage. This study shows the successful development of a prototype hybrid UAV for practical applications in precision agriculture. The UAV consists of an off-the-shelf fixed-wing fuselage, which has been enhanced with multi-rotor functionality. It was programmed to perform pre-defined waypoint missions completely autonomously, including vertical take-off, horizontal flight, and vertical landing. The UAV was tested for its return-to-home (RTH) accuracy, power consumption and general flight performance at different wind speeds. The RTH accuracy was 43.7 cm in average, with a root-mean-square error of 39.9 cm. The power consumption raised with an increase in wind speed. An extrapolation of the analysed power consumption to conditions without wind resulted in an estimated 40 km travel range, when we assumed a 25 % safety margin of remaining battery capacity. This translates to a maximal area coverage of 300 ha for a scenario with 18 m/s airspeed, 50 minutes flight time, 120 m AGL altitude, and a desired 70 % of image side-lap and 85 % forward-lap. The ground sample distance with an in-built RGB camera was 3.5 cm, which we consider sufficient for farm-scale mapping missions for most precision agriculture applications.

Abstract

In this study, we investigated the potential of airborne imaging spectroscopy for in-season grassland yield estimation. We utilized an unmanned aerial vehicle and a hyperspectral imager to measure radiation, ranging from 455 to 780 nm. Initially, we assessed the spectral signature of five typical grassland species by principal component analysis, and identified a distinct reflectance difference, especially between the erectophil grasses and the planophil clover leaves. Then, we analyzed the reflectance of a typical Norwegian sward composition at different harvest dates. In order to estimate yields (dry matter, DM), several powered partial least squares (PPLS) regression and linear regression (LR) models were fitted to the reflectance data and prediction performance of these models were compared with that of simple LR models, based on selected vegetation indices and plant height. We achieved the highest prediction accuracies by means of PPLS, with relative errors of prediction from 9.1 to 11.8% (329 to 487 kg DM ha−1) for the individual harvest dates and 14.3% (558 kg DM ha−1) for a generalized model.