Search

The agro-food industry is turning to robots to address the challenge of labour shortage. However, agro-food environments pose difficulties for robots due to high variation and occlusions. In the presence of these challenges, accurate world models, with information about object location, shape, and properties, are crucial for robots to perform tasks accurately. Building such models is challenging due to the complex and unique nature of agro-food environments, and errors in the model can lead to task execution issues. In this paper, MinkSORT, a novel method for generating tracking features using a 3D sparse convolutional network in a deepSORT-like approach, is proposed to improve the accuracy of world models in agro-food environments. MinkSORT was evaluated using real-world data collected in a tomato greenhouse, where it significantly improved the performance of a baseline model that tracks tomato positions in 3D using a Kalman filter and Mahalanobis distance. MinkSORT improved the HOTA from 42.8% to 44.77%, the association accuracy from 32.55% to 35.55%, and the MOTA from 57.63% to 58.81%. Different contrastive loss functions for training MinkSORT were also evaluated, and it was demonstrated that it leads to improved performance in terms of three separate precision and recall detection outcomes. The proposed method improves world model accuracy, enabling robots to perform tasks such as harvesting and plant maintenance with greater efficiency and accuracy, which is essential for meeting the growing demand for food in a sustainable manner. • MinkSORT: A novel method of using 3D sparse convolutions for multi-object tracking. • A novel implementation of the concept of DeepSORT for 3D data like point clouds. • Real world experiments in a challenging tomato greenhouse environment. • An evaluation of MinkSORT showing a tracking accuracy increase up to 7.72 points. [ABSTRACT FROM AUTHOR]

Computer-vision based methods contribute to the availability of high-quality phenotypic datasets. Most computer-vision based methods for plant phenotyping are based on analysis of 2D images. However, previous research showed that for traits related to plant architecture, like internode length, a main limitation of 2D methods was that plants with a curved growing pattern could not be accurately measured. In this work, it was hypothesised that methods based on 3D data can overcome this limitation, while increasing the overall accuracy of the internode length measurements. To test the hypothesis, a method was proposed to estimate internode lengths from 3D point clouds of cucumber plants. First, a deep neural network based on PointNet++ was trained to segment the point clouds into plant parts. The points that were predicted as 'node' were then selected and a clustering algorithm was used to group points belonging to the same node. The Euclidean distance between the detected nodes was used as an estimate of the internode length. The results were compared to the results of a previously published method based on 2D images. The results of the 3D method were significantly more accurate than the results of the 2D method. Moreover, in contrast to the 2D method, the internode length estimates of the 3D method were equally accurate for curved plants as well as for straight plants. The results clearly demonstrated that computer-vision based methods to measure plant architecture in general, and more specifically to measure internode length, greatly benefit from the availability of 3D data. • Digital plant phenotyping greatly benefits from the availability of 3D data. • Internode lengths estimated from 3D point clouds and 2D images were compared. • The accuracy of the estimated internode lengths was higher when using 3D point clouds. • The node detection rate was higher when using 2D images. • Combining 2D images and 3D point clouds could provide the best of both worlds. [ABSTRACT FROM AUTHOR]

The ability to accurately represent and localise relevant objects is essential for robots to carry out tasks effectively. Traditional approaches, where robots simply capture an image, process that image to take an action, and then forget the information, have proven to struggle in the presence of occlusions. Methods using multi-view perception, which have the potential to address some of these problems, require a world model that guides the collection, integration and extraction of information from multiple viewpoints. Furthermore, constructing a generic representation that can be applied in various environments and tasks is a difficult challenge. In this paper, a novel approach for building generic representations in occluded agro-food environments using multi-view perception and 3D multi-object tracking is introduced. The method is based on a detection algorithm that generates partial point clouds for each detected object, followed by a 3D multi-object tracking algorithm that updates the representation over time. The accuracy of the representation was evaluated in a real-world environment, where successful representation and localisation of tomatoes in tomato plants were achieved, despite high levels of occlusion, with the total count of tomatoes estimated with a maximum error of 5.08% and the tomatoes tracked with an accuracy up to 71.47%. Novel tracking metrics were introduced, demonstrating that valuable insight into the errors in localising and representing the fruits can be provided by their use. This approach presents a novel solution for building representations in occluded agro-food environments, demonstrating potential to enable robots to perform tasks effectively in these challenging environments. • A novel method to build world models for robots based on 3D object tracking. • A real world evaluation of our method in highly occluded scenarios. • Detection and association performance decreases with occlusions. • 3D object tracking yields an accuracy up to 71%. • Tracking metrics give more insight than the total count of objects. [ABSTRACT FROM AUTHOR]

Plant scientists require high quality phenotypic datasets. Computer-vision based methods can improve the objectiveness and the accuracy of phenotypic measurements. In this paper, we focus on 3D point clouds for measuring plant architecture of cucumber plants, using spectral data and deep learning (DL). More specifically, the focus of this paper is on the segmentation of the point clouds, such that for each point it is known to which plant part (e.g. leaf or stem) it belongs. It was shown that the availability of spectral data can improve the segmentation, with the mean intersection-over-union rising from 0.90 to 0.95. Furthermore, we analysed the effect of uncertainty in the collection of ground truth data. For this purpose, we hand-labelled 264 point clouds of cucumber plants twice and show that the intra-observer variability between those two annotation sets can be as low as 0.49 for difficult classes, while it was 0.99 for the class with the least uncertainty. Adding the second set of hand-labelled data to the training of the network improved the segmentation performance slightly. Finally, we show the improved performance of a 4-class segmentation over an 8-class segmentation, emphasizing the need for a careful design of plant phenotyping experiments. The results presented in this paper contribute to further development of automated phenotyping methods for complex plant traits. • Adding spectral data boosts 3D point cloud segmentation for plant phenotyping. • High intra-observer variability shows that ground-truth collection is challenging. • Labelling the same dataset multiple times can improve the segmentation performance. • Careful design of phenotyping experiments makes a computer scientist's life easier. [ABSTRACT FROM AUTHOR]

The growth and the harvestability of a broccoli crop is monitored by the size of the broccoli head. This size estimation is currently done by humans, and this is inconsistent and expensive. The goal of our work was to develop a software algorithm that can estimate the size of field-grown broccoli heads based on RGB-Depth (RGB-D) images. For the algorithm to be successful, the problem of occlusion must be solved, which is the partial visibility of the broccoli head due to overlapping leaves. This partial visibility causes sizing errors. In this research, we studied the use of deep-learning algorithms to deal with occlusions. We specifically applied the Occlusion Region-based Convolutional Neural Network (ORCNN) that segmented both the visible and the amodal region of the broccoli head (which is the visible and the occluded region combined). We hypothesised that ORCNN, with its amodal segmentation, can improve the size estimation of occluded broccoli heads. The ORCNN sizing method was compared with a Mask R–CNN sizing method that only used the visible broccoli region to estimate the size. The sizing performance of both methods was evaluated on a test set of 487 broccoli images with systematic levels of leaf occlusion. With a mean sizing error of 6.4 mm, ORCNN outperformed Mask R–CNN, which had a mean sizing error of 10.7 mm. Furthermore, ORCNN had a significantly lower absolute sizing error on 161 heavily occluded broccoli heads with an occlusion rate between 50% and 90%. Our software and data set are available on https://git.wur.nl/blok012/sizecnn. • The image-based size estimation of broccoli heads was studied. • A data set of 2560 broccoli images was acquired with systematic levels of occlusion. • Two deep-learning methods were tested on their ability to size occluded broccolis. • One deep-learning method used a dual segmentation to better deal with the occlusion. • The software and the data set have been made available. [ABSTRACT FROM AUTHOR]

• The YOLOv3 model was trained on data acquired in 20 arable fields. • The generalization error on 5 different arable fields was evaluated. • The effect of the number of training images was investigated. • The effect of the number of training fields was investigated. • The generalization error was mitigated using incremental training. Lack of generalization in plant-detection models is one of the main challenges preventing the realization of autonomous weed-control systems. This paper investigates the effect of the train and test dataset distribution on the generalization error of a plant-detection model and uses incremental training to mitigate the said error. In this paper, we use the YOLOv3 object detector as plant-detection model. To train the model and test its generalization properties we used a broad dataset, consisting of 25 sub-datasets, sampled from multiple different geographic areas, soil types, cultivation conditions, containing variation in weeds, background vegetation, camera quality and variations in illumination. Using this dataset we evaluated the generalization error of a plant-detection model, assessed the effect of sampling training images from multiple arable fields on the generalization of our plant-detection model, we investigated the relation between the number of training images and the generalization of the plant-detection model and we applied incremental training to mitigate the generalization error of our plant-detection model on new arable fields. It was found that the average generalization error of our plant-detection model was 0.06 mAP. Increasing the number of sub-datasets for training, while keeping the total number of training images constant, increased the variation covered by the training set and improved the generalization of our plant-detection model. Adding more training images sampled from the same datasets increased the generalization further. However, this effect is limited and only holds when the new images cover new variation. Naively adding more images does not prepare the model for specific scenarios outside the training distribution. Using incremental training the model can be adapted to such scenarios and the generalization error can be mitigated. Depending on the discrepancy between the training set and the new field, finetuning on as little as 25 images can already mitigate the generalization error. [ABSTRACT FROM AUTHOR]

High-tech greenhouses are characterised by high yields and high energy consumption. Current trends towards expanded use of supplemental lighting further increase the intensity of crop production and energy use, and it is expected that the availability of light-emitting diodes (LEDs) will accelerate this trend. At the same time, an increase in greenhouse lighting reduces the heating energy needed from the heating system. This study presents a novel concept for greenhouses, where both lighting and heating are derived exclusively from lamps. Such greenhouses can be highly efficient, as light is used both for crop growth and for heating. If the electric grid is based on renewable sources, such greenhouses can also be carbon-neutral and fossil-free. By using model simulations for a tomato greenhouse in the Netherlands, it was found that such greenhouses could be realised by employing a heat storage system with a heat storage capacity of 2 MJ m−2 and LEDs with a power capacity of 150 W m−2 and a photosynthetic photon flux density (PPFD) of 450 μmol m−2 s−1. The greenhouse heated by light was predicted to have 44% higher yields and 60% higher energy inputs than a reference greenhouse, equipped with a boiler and LEDs with a PPFD of 200 μmol m−2 s−1. This result was part of a general trade-off that was found between yield and energy efficiency. This exploration helped identify avenues for further improvement of the energy efficiency of greenhouses heated by lamps, highlighting their promise as a potential new direction in greenhouse intensification. • A novel approach for greenhouse climate control, using only lamps and no boiler. • Model simulations uncover the attributes required to make such systems feasible. • Increasing lamp intensity alone increases yields but reduces energy efficiency. • Avenues for improving the energy efficiency of heating by light are proposed. [ABSTRACT FROM AUTHOR]

For many robotic harvesting applications, position and angle between plant parts is required to optimally position the end-effector before attempting to approach, grasp and cut the product. A method for estimating the angle between plant parts, e.g. stem and fruit, is presented to support the optimisation of grasp pose for harvest robots. The hypothesis is that from colour images, this angle in the horizontal plane can be accurately derived under unmodified greenhouse conditions. It was hypothesised that the location of a fruit and stem could be inferred in the image plane from sparse semantic segmentations. The paper focussed on 4 sub-tasks for a sweet-pepper harvesting robot. Each task was evaluated under 3 conditions: laboratory, simplified greenhouse and unmodified greenhouse. The requirements for each task were based on the end-effector design that required a 25° positioning accuracy. In Task I, colour image segmentation for classes back-ground, fruit and stem plus wire was performed, meeting the requirement of an intersection-over-union > 0.58. In Task II, the stem pose was estimated from the segmentations. In Task III, centres of the fruit and stem were estimated from the output of previous tasks. Both centre estimations In Tasks II and III met the requirement of 25 pixel accuracy on average. In Task IV, the centres were used to estimate the angle between the fruit and stem, meeting the accuracy requirement of 25° for 73% of the cases. The work impacted on the harvest performance by increasing its success rate from 14% theoretically to 52% in practice under unmodified conditions. • Relationship between NH 3 mass transfer coefficient and ventilation rate developed. • Models of ammonia mass transfer coefficient under two fan locations proposed. • New models of NH 3 transfer coefficient are compared models in literature. [ABSTRACT FROM AUTHOR]

A vision-based weed control robot for agricultural field application requires robust vegetation segmentation. The output of vegetation segmentation is the fundamental element in the subsequent process of weed and crop discrimination as well as weed control. There are two challenging issues for robust vegetation segmentation under agricultural field conditions: (1) to overcome strongly varying natural illumination; (2) to avoid the influence of shadows under direct sunlight conditions. A way to resolve the issue of varying natural illumination is to use high dynamic range (HDR) camera technology. HDR cameras, however, do not resolve the shadow issue. In many cases, shadows tend to be classified during the segmentation as part of the foreground, i.e., vegetation regions. This study proposes an algorithm for ground shadow detection and removal, which is based on color space conversion and a multilevel threshold, and assesses the advantage of using this algorithm in vegetation segmentation under natural illumination conditions in an agricultural field. Applying shadow removal improved the performance of vegetation segmentation with an average improvement of 20, 4.4, and 13.5% in precision, specificity and modified accuracy, respectively. The average processing time for vegetation segmentation with shadow removal was 0.46 s, which is acceptable for real-time application (\1 s required). The proposed ground shadow detection and removal method enhances the performance of vegetation segmentation under natural illumination conditions in the field and is feasible for real-time field applications. [ABSTRACT FROM AUTHOR]

Currently the Dutch greenhouse horticultural sector has a high energy demand. The present use of weather forecasts can exacerbate this high energy consumption by contributing to suboptimal prediction and trading of the greenhouse's power demand. This study investigates the role of weather forecast errors on energy prediction power and trading uncertainty in greenhouse horticulture. This was done using an uncertainty analysis and computer model of a tomato producing Venlo style greenhouse in Bleiswijk, The Netherlands. This greenhouse model was used to predict the greenhouse's gas and electrical power demand. The study concluded that errors in the weather forecast of outdoor radiation, temperature and wind speed caused an overestimation of greenhouse energy demand. A sensitivity analysis showed that the radiation forecast error had the greatest impact on predicted greenhouse power demand errors with a mean relative error of 6.1 %. Predicted gas demand errors were most dependent on the outside wind speed forecast mean relative error (18.0 %) and temperature forecast error (17.2 %). A power trading uncertainty analysis was done to investigate the impact of predicted energy demand errors on the cost of buying power on the Dutch imbalance and Amsterdam Power Exchange day-ahead market. This cost analysis found that the volume of initial power trading was greater than corrective trading. Additionally, the higher volatility in short term power prices resulted in higher corrective power costs per unit of power than if the power demand had been initially predicted with more accuracy. • The study shows that weather forecasts create energy demand uncertainty. • A data-based sensitivity analysis was done for input data weather variables. • The greenhouse power costs are calculated using imbalance and day-ahead markets. • As the weather forecasts lengthen, the energy prediction error and power costs increase. • The radiation forecast has the greatest impact on the predicted power error. [ABSTRACT FROM AUTHOR]

A modular software framework design that allows flexible implementation of eye-in-hand sensing and motion control for agricultural robotics in dense vegetation is reported. Harvesting robots in cultivars with dense vegetation require multiple viewpoints and on-line trajectory adjustments in order to reduce the amount of false negatives and correct for fruit movement. In contrast to specialised software, the framework proposed aims to support a wide variety of agricultural use cases, hardware and extensions. A set of Robotic Operating System (ROS) nodes was created to ensure modularity and separation of concerns, implementing functionalities for application control, robot motion control, image acquisition, fruit detection, visual servo control and simultaneous localisation and mapping (SLAM) for monocular relative depth estimation and scene reconstruction. Coordination functionality was implemented by the application control node with a finite state machine. In order to provide visual servo control and simultaneous localisation and mapping functionalities, off-the-shelf libraries Visual Servoing Platform library (ViSP) and Large Scale Direct SLAM (LSD-SLAM) were wrapped in ROS nodes. The capabilities of the framework are demonstrated by an example implementation for use with a sweet-pepper crop, combined with hardware consisting of a Baxter robot and a colour camera placed on its end-effector. Qualitative tests were performed under laboratory conditions using an artificial dense vegetation sweet-pepper crop. Results indicated the framework can be implemented for sensing and robot motion control in sweet-pepper using visual information from the end-effector. Future research to apply the framework to other use-cases and validate the performance of its components in servo applications under real greenhouse conditions is suggested. [ABSTRACT FROM AUTHOR]

One of the problems in loose housing systems for laying hens is the laying of eggs on the floor, which need to be collected manually. In previous work, PoultryBot was presented to assist in this and other tasks. Here, probabilistic localisation with a particle filter is evaluated for use inside poultry houses. A poultry house is a challenging environment, because it is dense, with narrow static objects and many moving animals. Several methods and options were implemented and tested on data obtained with PoultryBot in a commercial poultry house. Although no animals were present, the localisation problem is still challenging here because of the repetitive nature of the poultry house interior, with its many narrow obstacles. Different parameter configurations were systematically evaluated, based on accuracy and applicability of the results. Estimated paths were quantitatively evaluated based on the Euclidian distance to a ground-truth determined with help of a total station. The presented system reached an accuracy of 0.37 m for 95% of the time, with a mean error of 0.2 m, making it suitable for localising PoultryBot in its future application. [ABSTRACT FROM AUTHOR]

Worldwide competitive challenges urge growers to further improve operational performance. In this paper, the objective ‘model-based analysis and improvement of the operation of horticultural production systems’ was narrowed to ranking simulated labour management scenarios in a multi-operations and multi-worker static cut-rose cultivation system. Eight scenarios with worker skill as a central theme were simulated including a practical labour management scenario applied by a Dutch cut-rose grower. The GWorkS-model was prepared for simulation of disbudding and bending in addition to harvest, three crop operations representing over 90% of crop-bound labour time, as well as for full scale simulation of the greenhouse using all workers and equipment. The sub-models on disbudding and bending were verified using data acquired in practice. Both processes were reproduced accurately. The model study on work scenarios showed that labour organisation choices might yield up to 5 s per harvested rose difference in total labour time for harvest, bending and disbudding between the best and worst scenario, which is equivalent to 7.1 € m −2 labour costs difference per year. Scenarios pointed out that working with low skilled, low paid workers is not effective. Specialised workers were most time effective, −17.5% compared to the reference, but overall a permanent team of skilled generalists ranked best in a multi-factorial assessment. Reduced crop operation diversity per day improved labour organisational outputs but ranked almost the same as the reference. The reference scenario was outranked by 5 scenarios. Overall, the GWorkS-model provided clear answers to research questions using the full complexity of crop operations. [ABSTRACT FROM AUTHOR]

In the published simulation studies on greenhouse climate control that employ optimal control, often non-realistic weather forecasts are employed, e.g. the realisation of the weather or artificially created forecasts are used. This research aims to quantify the effect of weather forecast errors on the performance of the controlled greenhouse system measured in terms of operational return. The operational return is defined as the difference between the cost of resources (r e s o u r c e u s e × c o s t) and the income through yield (y i e l d × p r o d u c t p r i c e). A stochastic model of the weather forecast error was identified based on historical weather observations and forecasts from a weather forecasting service. An uncertainty analysis using the stochastic model showed that a considerable number of control inputs are sensitive to the forecast errors. A simulation study involving three 7 d a y -intervals throughout the growing season showed, however, that the performance of the controlled greenhouse system is not significantly affected by the forecast error, a performance decrease of 0.03 e u r o. m − 2 (2 %) was observed with respect to the case in which perfect forecasts were used. The results suggest that an optimal control algorithm which (a) is updated every 15 min with the full state information, (b) uses forecasts published every 6 h and (c) uses published forecasts with a weather forecast error similar to the weather forecasting service used here, is able to mitigate the effect of the weather forecast error on the performance of the greenhouse system. • A stochastic forecast error model is identified using historical Dutch climate data. • Quantification of controlled greenhouse system performance loss for forecast errors. • Performance comparison with controller taking into account forecast uncertainty. [ABSTRACT FROM AUTHOR]

Combination of production of algae and tomato increases efficient use of available resources of greenhouse enterprises, such as controlled environment, water and nutrients, carbon dioxide, greenhouse space and infrastructure and knowledge. No information is available, however, about the potential productivity and related costs of a combined tomato and algae production in Dutch greenhouses. The objective was to determine the algae productivity in tubular photobioreactors (PBRs) and the economic feasibility of combined production of tomato and algae in Dutch greenhouses. A model was developed to predict greenhouse climate from outside climate, to predict tomato and algae biomass production and to analyse scenarios of different locations and dimensions of tubular PBR in the greenhouse with regard to algae productivity and cost price of algae production. The results show that algal productivity is low if PBRs are installed under a tomato crop due to limited light levels. Areal algal productivity was calculated to be 5–6.5 kg DM m−2 if PBRs are installed in a separate greenhouse compartment next to tomato. In this case the minimum cost prices of algae production was calculated to be €11 kg−1 DM algae, which give perspectives for the future. The proposed model is important because it gives insight into the feasibility of algae and tomato production in Dutch greenhouses. This novel model approach and the scenario results provide better knowledge about the potential productivity and related costs and returns of algae production in greenhouses. [ABSTRACT FROM AUTHOR]

Labour is the most dominant cost factor in Dutch cut-rose production. To improve crop production systems and labour management, a generic process modelling approach was developed enabling the impact of different scenarios on labour productivity to be assessed. The crop production system with crop handling processes is defined as a stochastic discrete event system. This paper demonstrates the model flexibility and transferability by adapting an existing model developed for a mobile rose production system to a model for a static growing system for cut roses. The paper describes the adaptation process. The adapted model was validated for the harvest process at a 3.6 ha production site in the Netherlands. Work scenarios were simulated to examine effects of skill, equipment, and harvest management. The model reproduces the harvest process accurately. A seven workday validation for an average skilled harvester showed a relative root mean squared error (RRMSE) under 5% for both labour time and harvest rate. A validation over 96 days for various harvesters showed a higher RRMSE, 15.2% and 13.6% for labour time and harvest rate respectively, mainly caused by the absence of model parameters for individual harvesters. The model was successfully used in scenario studies and indicated that worker skill was an important cost factor, differences associated with harvest trolley type are small, and that an extra harvest cycle per day is only feasible when compensated by product price. Overall, the generic model concept performs well for a static growing system when extended with system specific properties and process elements. [Copyright &y& Elsevier]

This study presents an automated lameness detection system that uses deep-learning image processing techniques to extract multiple locomotion traits associated with lameness. Using the T-LEAP pose estimation model, the motion of nine keypoints was extracted from videos of walking cows. The videos were recorded outdoors, with varying illumination conditions, and T-LEAP extracted 99.6% of correct keypoints. The trajectories of the keypoints were then used to compute six locomotion traits: back posture measurement, head bobbing, tracking distance, stride length, stance duration, and swing duration. The three most important traits were back posture measurement, head bobbing, and tracking distance. For the ground truth, we showed that a thoughtful merging of the scores of the observers could improve intra-observer reliability and agreement. We showed that including multiple locomotion traits improves the classification accuracy from 76.6% with only one trait to 79.9% with the three most important traits and to 80.1% with all six locomotion traits. • T-LEAP was used to track keypoints on the hooves, head and back of the cows. • Six locomotion traits associated with lameness were computed from the keypoints. • Lameness classification improves when including multiple locomotion traits. [ABSTRACT FROM AUTHOR]

Greenhouse crop system design for maximum efficiency and quality of labour is an optimisation problem that benefits from model-based design evaluation. This study focussed on the harvest process of roses in a static system as a step in this direction. The objective was to identify parameters with strong influence on labour performance as well as the effect of uncertainty in input parameters on key performance indicators. Differential sensitivity was analysed and results were tested for model linearity and superposability and verified using the robust Monte Carlo analysis method since in the literature, performance and applicability of differential sensitivity analysis has been questioned for models with internal stochastic behaviour. Greenhouse section length and width, single rose cut time, and yield influence labour performance most, but greenhouse section dimensions and yield also affect the number of harvested stems directly. Throughput, i.e. harvested stems per second, being the preferred metric for labour performance, is most affected by single rose cut time, yield, number of harvest cycles per day, greenhouse length and operator transport velocity. The model is insensitive for σ of lognormal distributed stochastic variables describing the duration of low frequent operations in the harvest process, like loading and unloading rose nets. In uncertainty analysis, the coefficient of variation for the most important outputs, labour time and throughput, is around 5%. Total sensitivity as determined using differential sensitivity analysis and Monte Carlo analysis essentially agreed. The combination of both methods gives full insight into both individual and total sensitivity of key performance indicators. [Copyright &y& Elsevier]

Fine-scale spatial information on soil properties is needed to successfully implement precision agriculture. Proximal gamma-ray spectroscopy has recently emerged as a promising tool to collect fine-scale soil information. The objective of this study was to evaluate a proximal gamma-ray spectrometer to predict several soil properties using energy-windows and full-spectrum analysis methods in two differently managed sandy loam fields: conventional and organic. In the conventional field, both methods predicted clay, pH and total nitrogen with a good accuracy (R² ≥ 0.56) in the top 0-15 cm soil depth, whereas in the organic field, only clay content was predicted with such accuracy. The highest prediction accuracy was found for total nitrogen (R² = 0.75) in the conventional field in the energy-windows method. Predictions were better in the top 0-15 cm soil depths than in the 15-30 cm soil depths for individual and combined fields. This implies that gamma-ray spectroscopy can generally benefit soil characterisation for annual crops where the condition of the seedbed is important. Small differences in soil structure (conventional vs. organic) cannot be determined. As for the methodology, we conclude that the energy-windows method can establish relations between radionuclide data and soil properties as accurate as the full-spectrum analysis method. [ABSTRACT FROM AUTHOR]

Measurement of volatile organic compounds (VOCs) emitted by plants allows us to monitor plant health status without touching the plant. To bring this technique a step further towards a practical plant diagnosis technique for greenhouse crop production, we have defined a numerical index named “Emission index” to evaluate VOC emissions from tomato plants isolated from ambient greenhouse air using an open-bottom chamber. The emission index of a VOC is a ratio of the concentration of the VOC in the air inside the chamber to that in the ambient greenhouse air and the index should be larger than one if the VOC is emitted by the plants in the chamber. Measurement of the emission indices of representative tomato VOCs proved that non-stressed tomato plants in a greenhouse emit n-hexanal, 2-carene, β-phellandrene, α-copaene, β-caryophyllene and (3E,7E)-4,8,12-trimethyl-1,3,7,11-tridecatetraene (TMTT), do not emit (Z)-3-hexenol, and might absorb and/or adsorb n-hexanol. Another experiment proved that a routine crop maintenance operation, i.e. removal of old leaves from the lower part of plants and removal of side shoots, enhances the emissions of n-hexanal, (E)-2-hexenal, (Z)-3-hexenol, but does not enhance the emissions of α-copaene, methyl salicylate (MeSA) and TMTT. The results suggest that the measurement of emission index with the open-bottom chamber is a useful technique for monitoring VOC emissions by plants in a greenhouse under practical conditions. [ABSTRACT FROM AUTHOR]

Mobile rose systems are designed to increase labour efficiency. However, many questions remain on best design and settings of operational parameters for best performance. The ultimate goal of this research is an assessment of re-designed horticultural crop production systems and work scenarios on labour and machine performance before implementation. To attain this goal, a queueing network model, GWorkS 1 [1] GWorkS is an acronym for Greenhouse Work Simulation, and is a discrete event simulation model on the crop handling processes and logistics inside a greenhouse production site. -rose, is presented for simulation of labour processes in a greenhouse with a mobile rose cultivation system. The objective for modelling is to quantify effects of production system changes by means of a flexible and generic model approach. A state-of-the-art mobile rose production system was used to validate and test the GWorkS-rose model. Data from the labour registration system and from video recordings were used for validation. System performance was simulated and compared to the performance measured in the real situation. Results of a single day validation show that the model estimates harvest labour time with an accuracy of 94%. For a one month validation an accuracy of 92% and RRMSE of 18% resulted. The value of RRMSE was caused by missing data on the number of workers at the loop and on the actual gutter speed level. The model can determine best system settings as is illustrated for operator and mobile gutter speed settings at given rose yield levels. It is concluded that the model can be used for studies on design and management of this kind of production systems. [Copyright &y& Elsevier]

An optimisation algorithm, as an essential part of a model-based method to design greenhouses for a broad range of climatic and economic conditions, was described. This algorithm – a modified controlled random search using parallel computing – maximised the annual Net Financial Return (NFR) for a tomato grower by selecting the best alternative to fulfil eight design elements: type of greenhouse structure, material of the cover, outdoor shade screen, whitewash properties, thermal screen, heating system, cooling system and CO2 enrichment system. As an example, the algorithm was applied to two locations with different climatic and economic conditions, Almeria and The Netherlands. Due to the warm climate with high radiation levels in Almeria, a greenhouse with a relatively large specific ventilation area (20% compared to 14% for Dutch conditions), seasonal whitewash and a low-capacity direct air heater (50 W m−2 compared to 200 W m−2 for Dutch conditions) was selected. In contrast, for the relatively cold climate with low radiation levels of the Netherlands, a 100% aluminium thermal screen and no whitewash would give the best result. The design method produced realistic greenhouses and related annual NFR, indicating that the method performs well. An analysis of the close-to-best greenhouses showed that, for both locations, a structure with high light transmissivity considerably enhanced the greenhouse performance whereas an outdoor shade screen, geothermal heating and mechanical cooling would be not economical. These results demonstrate the feasibility of a model-based design approach that produces suitable greenhouse designs for given climatic and economic conditions. [ABSTRACT FROM AUTHOR]

This paper presents a receding horizon optimal control (RHOC) method with an economic objective function for balancing the cost of resources (resource use × cost) with income through yield (yield × product price). This paper considers the two elements that determine the income through yield. The first element is yield and associated fruit development. A new, computationally viable, approach to model the income through yield is proposed and its prediction accuracy with respect to the original model is evaluated. The new approach employs a model that predicts at each time step, the future income through yield based on the assimilates partitioned to the fruits at the current time step. Simulations suggest that the assumptions made to arrive at the model for the new approach, do not significantly affect the accuracy of the predictions. The second element considered in this paper is the product price and the uncertainty inherent in its forecasts. Historical product price data are used to generate artificial product price forecasts. An uncertainty analysis, in combination with the artificial product price forecasts, showed that the product price forecast error does not considerably affect the optimised control strategy. Season-wide simulations with RHOC suggest that the product price forecast error does not considerably affect the value of the economic objective function. • A computationally viable approach to model the income through yield is proposed. • The new approach to model income through yield allows for RHOC with a short horizon. • Historical product price data are used to create artificial product price forecasts. • Product price forecast errors do not considerably affect the control strategy. [ABSTRACT FROM AUTHOR]

• The performance of 40 combinations for vegetation segmentation was assessed. • A clear difference in performance was observed among various combinations under the given conditions. • CIVE + Kapur showed the best performance; while VEG + Kapur showed the worst. • When adapting the combination to the given conditions yielded a slightly higher performance. • However, the expected advantage of adapting the combination to the field condition is not large. Robust vegetation segmentation is required for a vision-based weed control robot in an agricultural field operation. The output of vegetation segmentation is a fundamental element in the subsequent process of weed/crop discrimination as well as weed control actuation. Given the abundance of colour indices and thresholding techniques, it is still far from clear how to choose a proper threshold technique in combination with a colour index for vegetation segmentation under agricultural field conditions. In this research, the performance of 40 combinations of eight colour indices and five thresholding techniques found in the literature was assessed to identify which combination works the best given varying field conditions in terms of illumination intensity, shadow presence and plant size. It was also assessed whether it was better to use one specific combination at all times or whether the combination should be adapted to the field conditions at hand. A clear difference in performance, represented in terms of MA (Modified Accuracy) which indicates the harmonic mean of relative vegetation area error and balanced accuracy, was observed among various combinations under the given conditions. On the image dataset that was used in this study, CIVE+Kapur (Colour Index of Vegetation Extraction+Max Entropy threshold) showed the best performance while VEG+Kapur (Vegetative Index+Max Entropy threshold) showed the worst. Adapting the combination to the given conditions yielded a slightly higher performance than when using a single combination for all (in this case CIVE+Kapur). Consistent results were obtained when validated on a different independent image dataset. Although a slightly higher performance was achieved when adapting the combination to the field conditions, this slight improvement seems not to outweigh the potential investment in sensor technology and software that are needed in practice to accurately determine the different conditions in the field. Therefore, the expected advantage of adapting the combination to the field condition is not large. [ABSTRACT FROM AUTHOR]

Greenhouse models are important tools for the analysis and design of greenhouse systems and for offering decision support to growers. While many models are available, relatively few include the influence of supplementary lighting on the greenhouse climate and crop. This study presents GreenLight, a model for greenhouses with supplemental lighting. GreenLight extends state of the art models by describing the qualitative difference between the common lighting system of high-pressure sodium (HPS) lamps, and the newest technology for horticultural lighting - the light-emitting diodes (LEDs). LEDs differ from HPS lamps in that they operate at lower temperatures, emit mostly convective heat and relatively little radiative heat, and can be more efficient in converting electricity to photosynthetically active radiation (PAR). These differences can have major implications on the greenhouse climate and operation, and on the amount of heat that must be supplied from the greenhouse heating system. Model predictions have been evaluated against data collected in greenhouse compartments equipped with HPS and LED lamps. The model predicted the greenhouse's heating needs with an error of 8–51 W m−2, representing 1–12% of the measured values; the RMSE for indoor temperature was 1.74–2.04 °C; and the RMSE for relative humidity was 5.52–8.5%. The model is freely available as open source MATLAB software at https://github.com/davkat1/GreenLight. It is hoped that it may be further evaluated and used by researchers worldwide to analyse the influence of the most recent lighting technologies on greenhouse climate control. Image 1 • Model for lighting in greenhouses, including HPS and LED lamps, was developed. • Model was evaluated using data from a greenhouse with HPS and LED top-lights. • Model evaluates energy requirements of the greenhouse with an accuracy of 1–12%. • Model predicts indoor temperature and humidity with an RRMSE of 6–11%. • Model is publicly available in free open source format. [ABSTRACT FROM AUTHOR]

Crop modelling is an essential part of biosystems engineering; selecting or developing a crop model for a specific application, having its requirements and desires, is difficult if not impossible without the required domain knowledge. This paper presents a fundamentally different model selection approach based on biological functionalities. This is enabled by a common structure, which allows for a combining of components, yielding new models. This increased design space allows the development of models which are better suited to the application than the original models. The use of a common structure, and its potential, are demonstrated by a use-case involving the selection of a tomato crop model for a model-based control application, but the rationales and methodologies can apply to other crops and applications as well. In this paper, 27 valid model combinations have been created from 4 models. In the use-case presented, the models are compared to data originating from a real system. The predictive performance of a model is quantified by the Root-Mean-Squared-Error (RSME) between the predictions and data. One trade-off is model accuracy versus computational speed. With the model set used, a 13% decrease in RSME was obtained by allowing a 7.5% increase in model computation time compared to one of the original models. • A fundamentally different model selection approach using biological functionalities. • 27 valid model combinations have been created from 4 models. • A model combination has been found which predicts better than the original models. [ABSTRACT FROM AUTHOR]

The possibility of improving the accuracy of climate monitoring in greenhouse cultivation by way of model based filtering was explored. The focus was on estimating the average climate inside a greenhouse compartment. Starting point was employing an extended Kalman filter (EKF), combined with a greenhouse climate differential equation model. In two different greenhouses (A and B), temperature and humidity were monitored with a 5-min sampling resolution with a sensor grid. The available data sets spanned 1 and 0.5 years. With the average over all sensors as reference signal, the root mean squared errors (RMSEs) of the unfiltered signals (coming from single sensors) were 0.43 °C and 0.48 g m−3 for greenhouse A, and 0.80 °C and 0.64 g m−3 for greenhouse B. The filter was compared with a moving average (MA) filter, and an unscented Kalman filter (UKF). Overall, monitoring accuracy was not improved by any of the filters, and in most cases it deteriorated. Performance was strongly linked to the choice of filter, where the EKF outperformed the other filters by a considerable difference. The violations on the assumptions of whiteness and normality of the noise were severe but had a moderate effect on the RMSEs (0.11 °C and 0.10 g m−3 for greenhouse A). A clear link was found between model accuracy and monitoring accuracy. A 10–15 fold decrease of state errors was associated with an RMSE reduction down to 0.1 °C and 0.1 g m−3, the expected equivalent of increasing the number of climate sensors from 1 to 25. • A novel methodology for assessing data filtering methods was developed. • Three data filters were tested in two different greenhouses for in total 1.5 years. • None of the filters improved climate monitoring accuracy. • Bias, non-Gaussianity, and auto-correlation had a moderate effect on performance. • Increasing model accuracy has the same effect as increasing the number of sensors. [ABSTRACT FROM AUTHOR]

Assessment of animal status, housing conditions and manually collecting floor eggs are the major daily tasks for poultry farmers. To assist the farmer in these tasks, PoultryBot, an autonomous mobile robot for use in poultry houses has been developed. In earlier research, several components of PoultryBot were discussed in detail. Here, performance of the robot is evaluated under practical conditions. For navigation, different paths were used to assess its navigation performance for various tasks, such as area sweeping and surveying close to walls. PoultryBot proved capable of navigating autonomously more than 3000 m, while avoiding obstacles and dealing with the hens present. The robustness of its navigation performance was tested by confronting PoultryBot with obstacles in different positions with respect to its path and using different settings of the navigation parameters. Both factors clearly influenced the driving behaviour of PoultryBot. For floor egg collection, detection and collection of eggs was assessed at 5 predefined egg positions lateral to the path of the robot. Over 300 eggs were tested; 46% were collected successfully, 37% was not collected successfully, and 16% were missed. The most observed failures occurred when the collection device was just next to the egg. It is thought that this problem can be solved by improving the control algorithm. The results demonstrate the validity of the PoultryBot concept and the possibility of autonomous floor egg collection in commercial poultry houses. Furthermore, they indicate that application of smart autonomous vehicles in dense animal environments is feasible. Highlights • Integration of previous work on PoultryBot, tested in environment with live hens. • Performance evaluation for autonomous navigation and floor egg collection. • Capable of dealing with obstacles and animals using different path types. • Floor egg collection with 46% success, 37% failure and 16% missed from >300 eggs. • Results show that smart autonomous vehicles can be used in poultry houses. [ABSTRACT FROM AUTHOR]

Classification of weeds amongst cash crops is a core procedure in automated weed control. Addressing volunteer potato control in sugar beets, in the EU Smartbot project the aim was to control more than 95% of volunteer potatoes and ensure less than 5% of undesired control of sugar beet plants. A promising way to meet these requirements is deep learning. Training an entire network from scratch, however, requires a large dataset and a substantial amount of time. In this situation, transfer learning can be a promising solution. This study first evaluates a transfer learning procedure with three different implementations of AlexNet and then assesses the performance difference amongst the six network architectures: AlexNet, VGG-19, GoogLeNet, ResNet-50, ResNet-101 and Inception-v3. All nets had been pre-trained on the ImageNet Dataset. These nets were used to classify sugar beet and volunteer potato images taken under ambient varying light conditions in agricultural environments. The highest classification accuracy for different implementations of AlexNet was 98.0%, obtained with an AlexNet architecture modified to generate binary output. Comparing different networks, the highest classification accuracy 98.7%, obtained with VGG-19 modified to generate binary output. Transfer learning proved to be effective and showed robust performance with plant images acquired in different periods of the various years on two types of soils. All scenarios and pre-trained networks were feasible for real-time applications (classification time < 0.1 s). Classification is only one step in weed detection, and a complete pipeline for weed detection may potentially reduce the overall performance. Highlights • Transfer learning provided very promising performance for weed/crop classification. • The highest classification accuracy of 98.7% was obtained with VGG-19. • All scenarios and pre-trained networks were feasible for real-time applications. • Data augmentation may improve the classification accuracy. [ABSTRACT FROM AUTHOR]

To handle surrounding objects, autonomous poultry house robots need to discriminate between various types of object present in the poultry house. A simple and robust method for image pixel classification based on spectral reflectance properties is presented. The four object categories most relevant for the autonomous robot PoultryBot are eggs, hens, housing elements and litter. Spectral reflectance distributions were measured between 400 and 1000 nm and based on these spectral responses the wavelength band with lowest overlap between all object categories was identified. This wavelength band was found around 467 nm with an overlap of 16% for hens vs. eggs, 12% for housing vs. litter, and less for other combinations. Subsequently, images were captured in a commercial poultry house, using a standard monochrome camera and a band pass filter centred around 470 nm. In 87 images, intensity thresholds were applied to classify each pixel into one of four categories. For eggs, the required 80% correctly classified pixels was almost reached with 79.9% of the pixels classified correctly. For hens and litter, 40–50% of the pixels were classified correctly, while housing elements had lower performance (15.6%). Although the imaging setup was designed to function without artificial light, its optical properties influenced image quality and the resulting classification performance. To reduce these undesired effects on the images, and to improve classification performance, artificial lighting and additional processing steps are proposed. The presented results indicate both the simplicity and elegance of applying this method and are a suitable starting point for implementing egg detection with the robot. [ABSTRACT FROM AUTHOR]

One of the most important steps in vision-based weed detection systems is the classification of weeds growing amongst crops. In the EU SmartBot project it was required to effectively control more than 95% of volunteer potatoes and ensure less than 5% of damage of sugar beet. Classification features such as colour, shape and texture have been used individually or in combination for classification studies but they have proved unable to reach the required classification accuracy under natural and varying daylight conditions. A classification algorithm was developed using a Bag-of-Visual-Words (BoVW) model based on Scale-Invariant Feature Transform (SIFT) or Speeded Up Robust Feature (SURF) features with crop row information in the form of the Out-of-Row Regional Index (ORRI). The highest classification accuracy (96.5% with zero false-negatives) was obtained using SIFT and ORRI with Support Vector Machine (SVM) which is considerably better than previously reported research although its 7% false-positives deviated from the requirements. The average classification time of 0.10–0.11 s met the real-time requirements. The SIFT descriptor showed better classification accuracy than the SURF, but classification time did not vary significantly. Adding location information (ORRI) significantly improved overall classification accuracy. SVM showed better classification performance than random forest and neural network. The proposed approach proved its potential under varying natural light conditions, but implementing a practical system, including vegetation segmentation and weed removal may potentially reduce the overall performance and more research is needed. [ABSTRACT FROM AUTHOR]

To reach a fruit in an obstacle-dense crop environment, robotic fruit harvesting requires a collision-free motion of the manipulator and end-effector. A novel two-part analysis was conducted of a sweet-pepper harvesting robot based on data of fruit (N = 158) and stem locations collected from a greenhouse. The first part of the analysis compared two methods of selecting the azimuth angle of the end-effector. The new ‘constrained-azimuth’ method avoided risky paths and achieved a motion planning success similar to the ‘full-azimuth’ method. In the second part, a sensitivity analysis was conducted for five parameters specifying the crop (stem spacing and fruit location), the robot (end-effector dimensions and robot position) and the planning algorithm, to evaluate their effect on successfully finding a collision-free goal configuration and path. Reducing end-effector dimensions and widening stem spacing are promising research directions because they significantly improved goal configuration success, from 63% to 84%. However, the fruit location at the stem is the strongest influencing parameter and therefore provides an incentive to train or breed plants that develop more fruit at the front side of the plant stem. The two analyses may serve as useful tools to study motion planning problems in a dense obstacle environment. [ABSTRACT FROM AUTHOR]

• MaskAL is software that can reduce the annotation effort for training Mask R-CNN. • MaskAL uses active learning to select images about which Mask R-CNN is uncertain. • On a broccoli dataset, MaskAL performed significantly better than random sampling. • By using MaskAL, 1400 image annotations were saved compared to random sampling. • The software is available on https://github.com/pieterblok/maskal. The generalisation performance of a convolutional neural network (CNN) is influenced by the quantity, quality, and variety of the training images. Training images must be annotated, and this is time consuming and expensive. The goal of our work was to reduce the number of annotated images needed to train a CNN while maintaining its performance. We hypothesised that the performance of a CNN can be improved faster by ensuring that the set of training images contains a large fraction of hard-to-classify images. The objective of our study was to test this hypothesis with an active learning method that can automatically select the hard-to-classify images. We developed an active learning method for Mask Region-based CNN (Mask R-CNN) and named this method MaskAL. MaskAL involved the iterative training of Mask R-CNN, after which the trained model was used to select a set of unlabelled images about which the model was most uncertain. The selected images were then annotated and used to retrain Mask R-CNN, and this was repeated for a number of sampling iterations. In our study, MaskAL was compared to a random sampling method on a broccoli dataset with five visually similar classes. MaskAL performed significantly better than the random sampling. In addition, MaskAL had the same performance after sampling 900 images as the random sampling had after 2300 images. Compared to a Mask R-CNN model that was trained on the entire training set (14,000 images), MaskAL achieved 93.9% of that model's performance with 17.9% of its training data. The random sampling achieved 81.9% of that model's performance with 16.4% of its training data. We conclude that by using MaskAL, the annotation effort can be reduced for training Mask R-CNN on a broccoli dataset with visually similar classes. Our software is available on https://github.com/pieterblok/maskal. [ABSTRACT FROM AUTHOR]

A problem in loose housing systems for laying hens is the laying of eggs on the floor; these eggs need manual collection. This job is heavy and time-consuming and automated collection is desired. For collection using a robot, a collection path is required. A novel path planning algorithm is introduced for non-uniform repetitive area coverage (NURAC) paths and evaluated based on information about floor egg distribution probability. Firstly, a spatial map was developed that describes the potential for floor eggs at each location in a poultry house. Next, paths for floor egg collection are planned with a dynamic programming approach that covers the house floor area and frequently revisits locations with a high potential on floor eggs. These paths are compared with the paths used for floor egg collection by a farmer and evaluated with help of a simulated set of floor eggs. With respect to the average time eggs are present on the floor, paths planned for a robot are compared to two collection rounds of a farmer. With respect to the structure of the path and the number of visits to locations with a high potential, the robot paths outperform the farmer. Although optimality of the path is not guaranteed, the presented results are promising for the use of a robot to collect floor eggs, and will result in a reduction of the demand for manual labour. Extending the floor egg model with feedback information could further improve the results. [Copyright &y& Elsevier]

For robotic harvesting of sweet-pepper fruits in greenhouses a sensor system is required to detect and localize the fruits on the plants. Due to the complex structure of the plant, most fruits are (partially) occluded when an image is taken from one viewpoint only. In this research the effect of multiple camera positions and viewing angles on fruit visibility and detectability was investigated. A recording device was built which allowed to place the camera under different azimuth and zenith angles and to move the camera horizontally along the crop row. Fourteen camera positions were chosen and the fruit visibility in the recorded images was manually determined for each position. For images taken from one position only with the criterion of maximum 50% occlusion per fruit, the fruit detectability (FD) was in no case higher than 69%. The best single positions were the front views and looking with a zenith angle of 60° upwards. The FD increased when a combination was made of multiple viewpoint positions. With a combination of five favourite positions the maximum FD was 90%. [ABSTRACT FROM AUTHOR]

Highlights: [•] We test the accuracy of different vis-NIR models to predict soil properties. [•] We use local, general, spiked and true validation models in five fields. [•] Local models predict tillage properties well, but more effort is needed in sampling. [•] Spiked models yield results comparable to local models with a little lower accuracy. [•] Spiked models of vis-NIR spectroscopy should be used in field level studies. [Copyright &y& Elsevier]

An economic model was developed as a key component of a model-based method to design greenhouses for a broad range of climatic and economic conditions. This economic model was linked to an existing greenhouse climate-crop yield model to calculate the annual Net Financial Result (NFR) of a greenhouse. The aim of this study was to identify – among ten predefined design alternatives – the greenhouse with the highest annual NFR for tomato production under southern Spanish conditions. The basic designs were either the parral greenhouse, or a multi-tunnel, possibly fitted with any combination of heating, fogging and CO2 enrichment. Results demonstrated that the multi-tunnel, fitted with only a fogging system was most profitable, followed by the multi-tunnel with heating, CO2 enrichment and fogging. However, the difference in NFR between such a design and a simple parral was small with respect to the difference in investment. A sensitivity analysis of the NFR of the two technology extremes shows that tomato price, the fraction of marketable yield and the Photosynthetically Active Radiation (PAR) transmission of the cover had the largest bearing on NFR. With increasing technology level, the NFR depended less on outdoor climate and more on tomato price. This indicates that a low-tech greenhouse diminishes the risk of variations among price paths in different years, whereas a high-tech greenhouse covers better the “weather risk”. The best design was also affected by climate management and the joint impact of climate modification techniques. These results demonstrated that a model-based design approach can cope with multi-factorial design aspects. [ABSTRACT FROM AUTHOR]