Search

Irrigation has a strong impact in terms of yield regulation and grape and wine quality, so the implementation of precision watering systems would facilitate the decision-making process about the water use efficiency and the irrigation scheduling in viticulture. The objectives of this work were two-fold. On one hand, to compare and assess grapevine water status using two different spectral devices assembled in a mobile platform and to evaluate their capability to map the spatial variability of the plant water status in two commercial vineyards from July to early October in season 2021, and secondly to develop an algorithm capable of automate the spectral acquisition process using one of the two spectral sensors previously tested. Contemporarily to the spectral measurements collected from the ground vehicle at solar noon, stem water potential (Ψs) was used as the reference method to evaluate the grapevine water status. Calibration and prediction models for grapevine water status assessment were performed using the Partial least squares (PLS) regression and the Variable Importance in the Projection (VIP) method. The best regression models returned a determination coefficient for cross validation (R2cv) and external validation (R2p) of 0.70 and 0.75 respectively, and the standard error of cross validation (RMSECV) values were lower than 0.105 MPa and 0.128 MPa for Tempranillo and Graciano varieties using a more expensive and heavier near-infrared (NIR) spectrometer (spectral range 1200–2100 nm). Remarkable models were also built with the miniaturized, low-cost spectral sensor (operating between 900–1860 nm) ranging from 0.69 to 0.71 for R2cv, around 0.74 in both varieties for R2p and the RMSECV values were below 0.157 MPa, while the RMSEP values did not exceed 0.151 MPa in both commercial vineyards. This work also includes the development of a software which automates data acquisition and allows faster (up to 40% of time saving in the field) and more efficient deployment of the developed algorithm. The encouraging results presented in this work demonstrate the great potential of this methodology to assess the water status of the vineyard and estimate its spatial variability in different commercial vineyards, providing useful information for better irrigation scheduling. [ABSTRACT FROM AUTHOR]

Irrigation has a strong impact in terms of yield regulation and grape and wine quality, so the implementation of precision watering systems would facilitate the decision-making process about the water use efficiency and the irrigation scheduling in viticulture. The objectives of this work were two-fold. On one hand, to compare and assess grapevine water status using two different spectral devices assembled in a mobile platform and to evaluate their capability to map the spatial variability of the plant water status in two commercial vineyards from July to early October in season 2021, and secondly to develop an algorithm capable of automate the spectral acquisition process using one of the two spectral sensors previously tested. Contemporarily to the spectral measurements collected from the ground vehicle at solar noon, stem water potential (Ψs) was used as the reference method to evaluate the grapevine water status. Calibration and prediction models for grapevine water status assessment were performed using the Partial least squares (PLS) regression and the Variable Importance in the Projection (VIP) method. The best regression models returned a determination coefficient for cross validation (R2cv) and external validation (R2p) of 0.70 and 0.75 respectively, and the standard error of cross validation (RMSECV) values were lower than 0.105 MPa and 0.128 MPa for Tempranillo and Graciano varieties using a more expensive and heavier near-infrared (NIR) spectrometer (spectral range 1200–2100 nm). Remarkable models were also built with the miniaturized, low-cost spectral sensor (operating between 900–1860 nm) ranging from 0.69 to 0.71 for R2cv, around 0.74 in both varieties for R2p and the RMSECV values were below 0.157 MPa, while the RMSEP values did not exceed 0.151 MPa in both commercial vineyards. This work also includes the development of a software which automates data acquisition and allows faster (up to 40% of time saving in the field) and more efficient deployment of the developed algorithm. The encouraging results presented in this work demonstrate the great potential of this methodology to assess the water status of the vineyard and estimate its spatial variability in different commercial vineyards, providing useful information for better irrigation scheduling. [ABSTRACT FROM AUTHOR]

In agriculture, crop monitoring and plant phenotyping are mainly manually measured. However, this practice gathers phenotyping information at a lower rate than genotyping evolves, thus producing bottleneck. This paper presents vitisBerry, a smartphone application for assessing in the vineyard, using computer vision, the berry number in clusters at phenological stages between berry-set and cluster-closure. The implemented image analysis algorithm is an evolution of a previous development, providing 1.63% and 7.57% of Recall and Precision improvement, respectively. The application was evaluated using two devices, taking and analysing 144 images from 12 different grapevine varieties. The Recall and Precision results ranged between 0.8762 and 0.9082 and 0.9392–0.9508, depending on the device. The average computational time required to analyse the 144 images varied from 3.14 to 8.40 s. According to these results, vitisBerry constitutes a tool for viticulturists to acquire phenotyping information from their vineyards in an easy and practical way. [ABSTRACT FROM AUTHOR]

Abstract: A comparative study is carried out in the problem of selecting a subset of basis functions in regression tasks. The emphasis is put on practical requirements, such as the sparsity of the solution or the computational effort. A distinction is made according to the implicit or explicit nature of the selection process. In explicit selection methods the basis functions are selected from a set of candidates with a search process. In implicit methods a model with all the basis functions is considered and the model parameters are computed in such a way that several of them become zero. The former methods have the advantage that both the sparsity and the computational effort can be controlled. We build on earlier work on Bayesian interpolation to design efficient methods for explicit selection guided by model evidence, since there is strong indication that the evidence prefers simple models that generalize fairly well. Our experimental results indicate that very similar results between implicit and explicit methods can be obtained regarding generalization performance. However, they make use of different numbers of basis functions and are obtained at very different computational costs. It is also reported that the models with the highest evidence are not necessarily those with the best generalization performance. [Copyright &y& Elsevier]

Hyperspectral imaging offers enormous potential for measuring grape composition with a high degree of representativity, allowing all exposed grapes from the cluster to be examined non-destructively. On-the-go hyperspectral images were acquired using a push broom hyperspectral camera (400–100 nm) that was mounted in the front part of a motorized platform moving at 5 km/h in a commercial Tempranillo vineyard in La Rioja, Spain. Measurements were collected on three dates during grape ripening in 2018 on the east side of the canopy, which was defoliated in the basal fruiting zone. A total of 144 grape clusters were measured for Total soluble solids (TSS), Titratable acidity (TA), pH, Tartaric and Malic acid, Anthocyanins and Total polyphenols, using standard wet chemistry reference methods, throughout the entire experiment. Partial Least Squares (PLS) regression was used to build calibration, cross validation and prediction models for the grape composition parameters. The best performances returned determination coefficients values of external validation (R2p) of 0.82 for TSS, 0.81 for Titratable acidity, 0.61 for pH, 0.62 for Tartaric acid, 0.84 for Malic acid, 0.88 for Anthocyanins and 0.55 for Total polyphenols. The promising results exposed in this work disclosed a notable methodology on-the-go for the non-destructive, in-field assessment of grape quality composition parameters along the ripening period. [ABSTRACT FROM AUTHOR]

• Automated disease detection under natural conditions and using a mobile platform. • Convolutional neural networks and vision transformers used for disease detection. • Deep learning models interpreted with Explainable Artificial Intelligence methods. • Symptom localisation and quantification with straightforward annotation. Diseases and pests in agriculture significantly impact crop yield and quality. Downy mildew (Plasmopara viticola) is a particular noteworthy example in grapevines. Traditional detection methods are laborious, subjective and time-consuming. Consequently, a technological solution based on artificial intelligence, would provide higher levels of reproducibility and sampling. The aim of this work was to develop an interpretable, automated method for detection and localisation of plant disease symptoms under field conditions. Images of the grapevine canopy were taken in 14 commercial vineyard plots under a range of lightning conditions, including both static and on-the-go settings. The images were processed using a sliding window, classifying sub-images into areas with and without downy mildew. Transfer learning, fine-tuning and data augmentation were employed to automate the classification, comparing convolutional neural networks (CNNs) and vision transformers (ViT). Subsequently, the trained model was integrated into the sliding window to localise regions within the canopy images exhibiting symptoms of downy mildew. Model predictions were interpreted using explainable artificial intelligence (XAI) methods. The EfficientNetV2S model achieved an accuracy of 91 % and an F1-score of 0.92 when classifying image areas and an Intersection over Union (IoU) of 0.83 when locating symptomatic areas. This method showed promising results, enabling automatic and explainable detection and localisation of plant diseases in complex conditions. The straightforward labelling process facilitated adaptation to new conditions, making it suitable for different crops and diseases. Integration into mobile platforms could enhance disease management and reduce the spread of pathogens, making a significant advance in agricultural technology. [ABSTRACT FROM AUTHOR]

• Development of four models using YOLOv4 for accurate detection of grape bunches under varied occlusion. • Training and testing were conducted on over 1700 images from 11 distinct vineyards. • The ComprehensiveModel achieved high accuracy with R2 values of 0.83 in validation and 0.81 in the testing phases. • The most effective models necessitated images depicting various occlusion conditions. Accurately and automatically estimating vineyard yield is a significant challenge. This study focuses on grape bunch counting in commercial vineyards using advanced deep learning techniques and object detection algorithms. The aim is to overcome the limitations of conventional yield estimation techniques, which are labour intensive, costly, and often inaccurate due to the spatial and temporal variability of the vineyard. This research proposes a non-invasive methodology for identifying grape bunches under different occlusion conditions using RGB cameras and deep learning models. The methodology is based on the collection of RGB images captured under field conditions, coupled with the implementation of the YOLOv4 architecture for data processing and analysis. Statistical indicators were used to evaluate the performance of the developed models. The comprehensive model produced a favourable outcome during validation, with an error rate of 1.12 bunches (R2 = 0.83). In the test dataset, the model achieved an error rate of 1.12 (R2 = 0.81). The results highlight the potential of emerging technologies to significantly improve vineyard yield estimation. This approach has the potential to assist vineyard management practices, enabling more informed and efficient decisions that could increase both the quantity and quality of grape production intended for winemaking. [ABSTRACT FROM AUTHOR]

Efficient irrigation in viticulture requires objective and representative monitoring of the vineyard water status variability. In this work the combination of multispectral, environmental and thermal data (using an infrared radiometer) acquired simultaneously on-the-go (at midday), from a ground moving vehicle (moving at 3 km/h) was tested to assess the vineyard stem water potential (Ψ s) and its spatial variability (three different irrigation treatments were imposed) over two seasons in north Spain. Partial least squares (PLS) cross-validation regression models involving the canopy temperature (T c), environmental and spectral variables yielded determination coefficients (R2 cv) of ~ 0.63 and root mean square error of cross-validation (RMSECV) between 0.124 MPa and 0.206 MPa in the two seasons. Linear discriminant analysis (LDA) involving only the variables used to build the regression models was run to distinguish among low, medium and high water stressed vines, yielding an average percentage of correct classification samples of 74%. The satisfactory performance of the multivariate models involving thermal, environmental and spectral data to either estimate or classify the plant water status within a vineyard supports the approach towards the combination of different data source to improve the capabilities of thermography itself. The inclusion of vegetative spectral indices in the regression and classification models of grapevine water status may provide real-time feedback on grapevine water use as influenced by actual vegetative growth, abiotic and/or biotic stress patterns. This combined approach can be seen as an advancement from existing solutions to assess plant water status variability given the simplicity and potential to automation of the thermal sensor employed and the integration of environmental and canopy vigour data into the model. • On-the-go monitoring of the grapevine water status in commercial vineyards. • Non-destructive technologies on a ground mobile platform to assess water status. • Thermography, multispectral imaging and environmental information were combined. [ABSTRACT FROM AUTHOR]

There is a growing need to provide support and applicable tools to farmers and the agro-industry in order to move from their traditional water status monitoring and high-water-demand cropping and irrigation practices to modern, more precise, reduced-demand systems and technologies. In precision viticulture, very few approaches with ground robots have served as moving platforms for carrying non-invasive sensors to deliver field maps that help growers in decision making. The goal of this work is to demonstrate the capability of the VineScout (developed in the context of a H2020 EU project), a ground robot designed to assess and map vineyard water status using thermal infrared radiometry in commercial vineyards. The trials were carried out in Douro Superior (Portugal) under different irrigation treatments during seasons 2019 and 2020. Grapevines of Vitis vinifera L. Touriga Nacional were monitored at different timings of the day using leaf water potential (Ψl) as reference indicators of plant water status. Grapevines' canopy temperature (Tc) values, recorded with an infrared radiometer, as well as data acquired with an environmental sensor (Tair, RH, and AP) and NDVI measurements collected with a multispectral sensor were automatically saved in the computer of the autonomous robot to assess and map the spatial variability of a commercial vineyard water status. Calibration and prediction models were performed using Partial Least Squares (PLS) regression. The best prediction models for grapevine water status yielded a determination coefficient of cross-validation (r2cv) of 0.57 in the morning time and a r2cv of 0.42 in the midday. The root mean square error of cross-validation (RMSEcv) was 0.191 MPa and 0.139 MPa at morning and midday, respectively. Spatial–temporal variation maps were developed at two different times of the day to illustrate the capability to monitor the grapevine water status in order to reduce the consumption of water, implementing appropriate irrigation strategies and increase the efficiency in the real time vineyard management. The promising outcomes gathered with the VineScout using different sensors based on thermography, multispectral imaging and environmental data disclose the need for further studies considering new variables related with the plant water status, and more grapevine cultivars, seasons and locations to improve the accuracy, robustness and reliability of the predictive models, in the context of precision and sustainable viticulture. [ABSTRACT FROM AUTHOR]

Yield assessment has been identified as critical topic for grape and wine industry. Computer vision has been applied for assessing yield, but the accuracy was greatly affected by fruit occlusion affected by leaves and other plant organs. The objective of this work was the consistent, continuous evaluation of the impact of leaf occlusions in different commercial vineyard plots at different defoliation stages. RGB (red, green and blue) images from five Tempranillo (Vitis vinifera L.) vineyards were manually acquired using a digital camera under field conditions at three different levels of defoliation: no defoliation, partial defoliation and full defoliation. Computer vision was used for the automatic detection of different canopy features, and for the calibration of regression equations for the prediction of yield computed per vine segment. Leaf occlusion rate (berry occlusion affected by leaves) was computed by machine vision in no defoliated vineyards. As occlusion rate increased, R2 between bunch pixels and yield was gradually reduced, ranging from 0.77 in low occlusion, to 0.63. [ABSTRACT FROM AUTHOR]

• Spider mite and downy mildew can be differentiated automatically. • Computer vision and deep learning allow for fast modelling of disease in grapevine. • The results show that accurate models can be trained without the need of large numbers of images. Diseases and pests cause serious damage in crop production, reducing yield and fruit quality. Their identification is often time-consuming and requires trained personnel. New sensing technologies and artificial intelligence could be used for automatic identification of disease and pest symptoms on grapevine in precision viticulture. The aim of this work was to apply deep learning modelling and computer vision for the detection and differentiation of downy mildew and spider mite symptoms in grapevine leaves under field conditions. RGB images of grapevine canopy leaves with downy mildew symptoms, with spider mite symptoms and without symptoms were taken under field conditions in a commercial vineyard. The images were prepared using computer vision techniques to increase disease visual features. Finally, deep learning was used to train a model capable of differentiating leaf images of the three classes. An accuracy up to 0.94 (F1-score of 0.94) was obtained by classifying leaves with downy mildew, spider mite and without symptoms at the same time, using a hold-out validation. Additionally, accuracies between 0.89 and 0.91 (F1-scores between 0.89 and 0.91) were obtained in the binary classification of the disease and pest, obtaining the best results in differentiating downy mildew from spider mite symptoms. This high accuracy demonstrates the effectiveness of deep learning and computer vision techniques for the classification of grapevine leaf images taken under field conditions, automatically finding complex features capable of differentiating leaves with spider mite symptoms, with downy mildew symptoms and without any. These results prove the potential of these non-invasive techniques in the detection and differentiation of pests and diseases in commercial crop production. [ABSTRACT FROM AUTHOR]