Search

Agricultural production is a key economic activity providing a growing world population withresources like food, fibers, fuels, andchemicalbasic components. Yet, agriculture also plays acrucial role in the depletion of resources likefresh water, fossil fuels, and nutrients (e.g., phosphate) and has a significant impact on our living environmentthrough emissions and the use of chemicals. Despite a growing world population, labor participation in agricultural production is declining. Technology hasbeen instrumental in the very successful development of agriculture throughout the centuries and will play acrucial role in the European Green Deal that addresses the transition to a more sustainable agri-food productionin the years that lay ahead of us. It is expected that technologies like Artificial Intelligence, Data, and Robotics willbe leading in this transition. During this transition, it is critically important to support agri-food production withnovel technologies. Covering the value chain from breeding, through on-farm production, post-harvest storage,logistics, packaging, and food processing, the agri-food production network is an ecosystem providing agri-foodproducts and services. Besides food, other off-farm productions do exist including harvesting of wild food, fisheries, indoor farming, both large scale commercial, and home growing, forestry, and, last but not least, urban landscaping and maintenance where robot technologies can and will be applied as well. The current release of this agenda mainly focuses on the food production part. Being a working document, theauthors wish toexpress the hope that it will serve as a basis for further elaboration to encompass the whole agrifood domain in its full widthanddepth. With an emphasis on robotics and to a lesser extent on AI and Data, thisreport provides a vision and mission on what robotics should contribute and how to facilitate the transition to amore sustainable agri-food production. Key challenges in terms of technology, ecosystem, and market are derived from 12 use-case themes within theagri-food production domain. Use-case themes represent classes of functionality that go beyond specific individual applications and application domains. Despite of aiming to identify high-level cross-domain challenges in robotics, AI, and Data agri-food and offering the opportunity to connect to other technical and application domains,this classification is not considered to be exhaustive. &nbsp

In Kaljaca, Vroegindeweij, and van Henten (2020), there is a change in the author group, name of one of the author, Angelo Mencarelli, was accidentally removed from the author list and needs to be added back. This article was published in Journal of Field Robotics in February 2020 (https://doi.org/10.1002/rob.21917). We apologize for the error. Updated author group is: Dejan Kaljaca, Bastiaan Vroegindeweij, Angelo Mencarelli, and Eldert van Henten with the following affiliation for Angelo Mencarelli: Stichting Wageningen Research, Wageningen University & Research, Wageningen, The Netherlands.

In this research, we investigated the robustness of an end-to-end deep-learning plant detection algorithm with respect to influences from uncontrolled illumination. For this research, we acquired two datasets, one containing images of plants taken under controlled illumination and one dataset with images acquired under uncontrolled illumination. We trained and evaluated the YOLOv3 object detector on both datasets. The object detector scored a mean Average Precision of 0.96 on controlled illumination conditions and 0.90 on the uncontrolled illumination conditions. This difference in performance is significant.

A key challenge in precision farming is complex decision making under variable and uncertain circumstances. A possible solution is offered by mathematical models and algorithms.

Wageningen University and Research developed a machine vision based intra row hoeing machine in collaboration with Steketee, a manufacturer of agricultural machinery. Although the machine is used in agricultural practice for several years already there is still a lack of an experimental method to evaluate the field performance of such a device. Reliable, quantitative information about the performance of the machine in the field is therefore not known and the effect of variations in field conditions and machine settings on the performance were not previously researched. The relative number of crop plants damaged by the machine and the relative amount of soil around the crop plants not hoed were used as criteria to assess the field performance. The performance was measured both manually on the field and with the help of sensors by logging the actuation response of the hoeing elements. In the field trial four treatments were combined: 1) the relative number of lettuce plants, 2) the weed density on the field, 3) the driving speed during hoeing, and 4) the moment of hoeing. Each possible combination of treatments was repeated three times. It was concluded that the developed method was suitable to determine statistically significant differences in the hoeing results for the applied treatments. No difference in the machine performance was found for different numbers of missing crop plants in a plant row. The machine performance decreased in terms of the relative number of damaged crop plants in case of high weed density treatments and decreased in terms of the relative number of not hoed intra row spaces in case of high speed treatments.

A bottleneck of state-of-the-art machine learning methods, e.g. deep learning, for plant part image segmentation in agricultural robotics is the requirement of large manually annotated datasets. As a solution, large synthetic datasets including ground truth can be rendered that realistically reflect the empirical situation. However, a dissimilarity gap can remain between synthetic and empirical data by incomplete manual modelling. This paper contributes to closing this gap by optimising the realism of synthetic agricultural images using unsupervised cycle generative adversarial networks, enabling unpaired image-to-image translation from the synthetic to empirical domain and vice versa. For this purpose, the Capsicum annuum (sweet- or bell pepper) dataset was used, containing 10,500 synthetic and 50 empirical annotated images. Additionally, 225 unlabelled empirical images were used. We hypothesised that the similarity of the synthetic images with the empirical images increases qualitatively and quantitively when translated to the empirical domain and investigated the effect of the translation on the factors color, local texture and morphology. Results showed an increased mean class color distribution correlation with the empirical dataset from 0.62 prior and 0.90 post translation of the synthetic dataset. Qualitatively, synthetic images translate very well in local features such as color,illumination scattering and texture. However, global features like plant morphology appeared not to be translatable.

A plant detection device is provided that includes a robotic arm having gripping element that includes first and second curved grippers with opposing concave surfaces that move between an open and closed states, and the arm moves and vibrates the gripping element, a proximity force sensor that is disposed on the gripper and outputs a measurement signal of a force between the gripping element and an outgrowth from a stem of a plant under test to a computer, a force and frequency sensor that is orthogonal to the proximity force sensor outputs a gripping force measurement and a frequency response measurement of the stem of the plant under test to the computer, where the computer moves the gripping and vibrating arm according to the sensor signal outputs.