Search

Weed management is one of the major challenges in viticulture, as long as weeds can cause significant yield losses and severe competition to the cultivations. In this direction, the development of an automated procedure for weed monitoring will provide useful data for understanding their management practices. In this work, a new image-based technique was developed in order to provide maps based on weeds’ height at the inter-row path of the vineyards. The developed algorithms were tested in many datasets from vineyards with different levels of weed development. The results show that the proposed technique gives promising results in various field conditions.

Agricultural activities such as fertilization and other crop management techniques contribute to greenhouse gas emissions and pollution; the cost of such activities is considerably high, with nitrogen pollution costing the EU up to 320 billion euros annually. In the present study, an integrated system has been proposed which utilizes smart farming tools and smart processing methodologies following the concept of a circular economy to reduce the impact of agricultural activity on climate change. Circular agriculture and precision farming together with the use of appropriate crop management tools may contribute to better resource use efficiency and sustainable agriculture.

To establish a safe human–robot interaction in collaborative agricultural environments, a field experiment was performed, acquiring data from wearable sensors placed at five different body locations on 20 participants. The human–robot collaborative task presented in this study involved six well-defined continuous sub-activities, which were executed under several variants to capture, as much as possible, the different ways in which someone can carry out certain synergistic actions in the field. The obtained dataset was made publicly accessible, thus enabling future meta-studies for machine learning models focusing on human activity recognition, and ergonomics aiming to identify the main risk factors for possible injuries.

The digital transformation of agriculture has evolved various aspects of management into artificial intelligent systems for the sake of making value from the ever-increasing data originated from numerous sources. A subset of artificial intelligence, namely machine learning, has a considerable potential to handle numerous challenges in the establishment of knowledge-based farming systems. The present study aims at shedding light on machine learning in agriculture by thoroughly reviewing the recent scholarly literature based on keywords’ combinations of “machine learning” along with “crop management”, “water management”, “soil management”, and “livestock management”, and in accordance with PRISMA guidelines. Only journal papers were considered eligible that were published within 2018–2020. The results indicated that this topic pertains to different disciplines that favour convergence research at the international level. Furthermore, crop management was observed to be at the centre of attention. A plethora of machine learning algorithms were used, with those belonging to Artificial Neural Networks being more efficient. In addition, maize and wheat as well as cattle and sheep were the most investigated crops and animals, respectively. Finally, a variety of sensors, attached on satellites and unmanned ground and aerial vehicles, have been utilized as a means of getting reliable input data for the data analyses. It is anticipated that this study will constitute a beneficial guide to all stakeholders towards enhancing awareness of the potential advantages of using machine learning in agriculture and contributing to a more systematic research on this topic.

Crop sensing is a promising approach for predicting corn (Zea mays L.) yield. Yield prediction is the first step in development of algorithms for sensor-based N management. Here, we evaluated the impact of (i) timing of sensing (growth stage), and (ii) method of reporting sensor data on estimations of corn grain and silage yield in New York. Sensor data were reported as the normalized difference vegetation index (NDVI), in-season estimated yield (INSEY) expressed as NDVI divided by days after planting (DAP; INSEYDAP), growing degree days (GGD; INSEYGGD), and the inverse simple ratio (ISR; [1–NDVI]/[1+NDVI]) divided by DAP (INSEYISR). To evaluate timing of sensing, corn of six fertility trials was scanned at every growth stage between V4 and V11. The replicated trials had up to six N rates (0, 56, 112, 168, 224, and 336 kg ha–¹). The V7 sensor and yield data from zero-N plots of nine additional on-farm trials (varying histories) were added to derive yield algorithms for New York. Drought at three sites in 2016 negatively impacted the accuracy of sensor-based grain yield estimates (R² 0.70 (grain) and >0.77 (silage). We conclude that INSEY data obtained at V7 can be used to accurately predict corn grain and silage yields in non-drought conditions in New York.

Vine vigour assessment has been a major concern of precision viticulture studies in order to identify areas of uniform vine performance within vineyards. Moreover, the counting and weighing of winter dormant canes is considered as the most informative measurement to indicate vine balance and is commonly performed manually by grape growers for management purposes. The main concern of this measurement is that it is time consuming and laborious and it cannot accommodate detailed sampling density. In the present study, the potential of using laser scanner technology as an automated, easy and rapid way to perform mapping of the winter pruning wood across the vineyard was investigated. The study was conducted during 2010 and 2011, in a one hectare commercial vineyard in central Greece, planted with cv. Agiorgitiko, a traditional Greek variety for the production of red wine. Parameters of topography, soil depth, soil texture, canopy properties (NDVI), yield, and grape quality were mapped and analysed in conjunction to winter canes weighing at pruning time. The mapping of the dormant canes was carried out using a 2D laser scanner sensor prior to pruning and manually measuring the pruning weight on a 10 × 20 m grid. Laser scanner measurements showed significant relationship in both 2010 and 2011 with pruning weight (r = 0.809 and r = 0.829 respectively, p

Usually crop models are run as point-based at field level. However, various soil properties may cause crop growth and yield to vary significantly at a smaller spatial scale than the field. Thus the objective of this study was to determine whether within-field variation in yield can be simulated when appropriate input data are available. A study was performed on a 64-ha maize field located in Vojvodina region (northern Serbia). The soil was characterized as Chernozem. The field was managed by the farmer at a sub-field level in 2017. Apparent electrical conductivity zones were used for targeted soil sampling and final yield was recorded by yield monitors installed on the two harvesters used to harvest the field. According to the results, field slope, water flow direction and accumulation were important yield driving factors. Spatially variable soil properties were introduced into the WOFOST crop model by estimating available water within the field, based on calculated water flow accumulation. Points were selected within management zones. Yield predicted by the model was correlated with the yield measured by the yield monitors.

Plant stress monitoring is of crucial importance to understand the plant response to environmental conditions, and has been widely applied in various fields including biology, agronomy, botany, and horticulture. A number of methods and instruments for plant stress monitoring have been proposed, most of which show significant disadvantages such as high cost, poor accuracy or complex operation that limit the use of such devices. In this paper, we propose a low cost, portable active multispectral optical device for precise plant stress detection and field mapping named Plant-O-Meter. The device has an integrated multispectral source that comprises light sources of the four most indicative wavelengths (850, 630, 535 and 465 nm), and enables simultaneous illumination of the whole plant. Sequential illumination and detection provide rapid reflectance measurements, which are wirelessly transmitted to android operated devices for processing and data storing. The device was tested in laboratory conditions comparing the Plant-O-Meter measurements with the image results from a SPECIM hyperspectral camera and a GreenSeeker handheld device and under field conditions with the GreenSeeker. The comparison revealed comparable performance, showing strong correlation with both the hyperspectral (R2=0.997) and the GreenSeeker handheld (R2=0.954 from the laboratory measurements and R2=0.886 from the field experiments), indicating that the device exhibits strong potential for accurate plant stress measurements. Moreover, owing to the very simple operating principle, the device represents a compact, cost-effective, and easy-to-operate solution.

The present study deals with human awareness, which is a very important aspect of human–robot interaction. This feature is particularly essential in agricultural environments, owing to the information-rich setup that they provide. The objective of this investigation was to recognize human activities associated with an envisioned synergistic task. In order to attain this goal, a data collection field experiment was designed that derived data from twenty healthy participants using five wearable sensors (embedded with tri-axial accelerometers, gyroscopes, and magnetometers) attached to them. The above task involved several sub-activities, which were carried out by agricultural workers in real field conditions, concerning load lifting and carrying. Subsequently, the obtained signals from on-body sensors were processed for noise-removal purposes and fed into a Long Short-Term Memory neural network, which is widely used in deep learning for feature recognition in time-dependent data sequences. The proposed methodology demonstrated considerable efficacy in predicting the defined sub-activities with an average accuracy of 85.6%. Moreover, the trained model properly classified the defined sub-activities in a range of 74.1–90.4% for precision and 71.0–96.9% for recall. It can be inferred that the combination of all sensors can achieve the highest accuracy in human activity recognition, as concluded from a comparative analysis for each sensor’s impact on the model’s performance. These results confirm the applicability of the proposed methodology for human awareness purposes in agricultural environments, while the dataset was made publicly available for future research.

Fuzzy inference systems (FIS) are particularly suited for aggregating multiple data to feed multi-variables decision support systems. Moreover, grape quality is a complex concept that refers to the simultaneous achievement of optimal levels in many parameters, thus single berry attributes spatial data are not adequate to define grape suitability for a specific end use. The aim of the present study was to develop and validate a FIS to classify grape quality based on selected grape attributes in a commercial vineyard in Central Greece planted with Vitis vinifera cv. Agiorgitiko, during 2010, 2011 and 2012. The vineyard was sectioned in 48 cells sized 10 × 20 m; total soluble solids, titratable acidity, total skin anthocyanins and berry fresh weight were measured at harvest on the same grid and were used in the FIS as inputs to build linguistic rules based on expert knowledge. The result of the FIS was a numerical value (Grape Total Quality, GTQ) which corresponded to a fuzzy set of grape quality classes (very poor, poor, average, good, and excellent). The validation process for the proposed FIS consisted of two parts: a comparison of GTQ with an independent set of data by viticulture experts and a comparison with soil and grapevine properties to verify its spatial relevancy. The evaluation process showed high general agreement between GTQ and expert evaluation suggesting that the FIS was able to model expert knowledge successfully. Moreover, GTQ exhibited higher variability than the individual grape quality attributes in all years. Among individual grape components, anthocyanins and berry weight seemed to be more important in determining GTQ than total soluble solids and titratable acidity. According to the results, FIS could allow the aggregation of grape quality parameters into a single index providing grape growers with a valuable tool for classifying grape quality at harvest.

Measuring soil compaction is a factor of interest to monitor soil fertility, which plays an important role in crop production cycle. Soil penetration resistance is the most commonly used method to measure soil compaction. It is fast and simple although it presents important limitations due to its close relationship with soil water content and the existence of high variability in the field, which requires increased number of samples that is effort demanding and time consuming. In this work, a fast and robust 5-point penetration resistance system was developed attached on the tractor three point hitch using load cells and combined with a GPS receiver. An ultrasonic sensor to monitor the penetrating depth was also attached. A software program using Microsoft Visual Basic was developed for data acquisition. Security pins and software alarm was added to secure the equipment safety when stones or soil harder than a limit was encountered. The system was successfully tested in an experimental field, where five tillage methods were studied, including no tillage. The results indicated the system's ability to recognize compacted soil layers and depict the spatial variability.

Precision agriculture was applied in handpicked fruit orchards in Greece in the last few years. The results of applications in apples and vines are presented in this paper. Yield, quality and NDVI maps were created for three years. The results for both crops showed that there was high variability in the measured parameters. Comparison of the maps indicated that fruit quality was inferior in areas where yield was high. NDVI maps in apples at the beginning of June were correlated to the final yield variability. This result can be used as an early warning of yield variability and could be used to manage the predicted variability. Similar results were indicated for vines.

Precision viticulture aims at managing vineyards at a sub-field scale according to the real needs of each part of the field. The current study focused on delineating management zones using fuzzy clustering techniques and developing a simplified approach for the comparison of zone maps. The study was carried out in a 1.0 ha commercial vineyard in Central Greece during 2009 and 2010. Variation of soil properties across the field was initially measured by means of electrical conductivity, soil depth and topography. To estimate grapevine canopy properties, NDVI was measured at different stages during the vine growth cycle. Yield and grape composition (must sugar content and total acidity) mapping was carried out at harvest. Soil properties, yield and grape composition parameters showed high spatial variability. All measured data were transformed on a 48-cell grid (10 × 20 m) and maps of two management zones were produced using the MZA software. Pixel-by-pixel comparison between maps of electrical conductivity, elevation, slope, soil depth and NDVI with yield and grape composition maps, set as reference parameters, allowed for the calculation of the degree of agreement, i.e. the percentage of pixels belonging to the same zone. The degree of agreement was used to select the best-suited parameters for final management zones delineation. For the year 2009 soil depth, early and mid season NDVI were used for yield-based management zones while for quality-based management zones ECa, early and mid season NDVI were utilized. For the year 2010 ECa, elevation and NDVI acquired during flowering and veraison were used for the delineation of yield-based management zones while for quality-based management zones ECa and NDVI acquired during flowering and harvest were utilized. Results presented here could be the basis for simple management zone delineation and subsequent improved vineyard management.

Irregular or insufficient rainfall can be a serious limitation to the final yield, causing low yields and even crop failure. Moreover the total amount of water is assessed during the growing stages by the plants¢ characteristics. It is possible, however, to find out how much soil moisture by using sensors This can help farmers in decision making Furthermore the combination of soil moisture sensors with management zones may lead to increase of the final yield. The aim of this research was to calibrate and install the WATERMARK sensors in a commercial vineyard according to the Precision Agriculture practices. The results of the research proved that the WATERMARK sensors can assess the soil moisture with high accuracy (R2 = 0.85).

With an increasing interest of the agricultural community in precision agriculture, this paper aims to compare two novel sensing approaches for crop monitoring. The recently developed multispectral proximal sensor named Plant-O-Meter and Sentinel-2 satellite, which carries a multispectral optical instrument, are two sensors suitable for agricultural applications. Each of them has pros and cons regarding spatial, spectral and temporal resolutions and their complementary use will surely bring added value compared to information retrieved by a single sensor. In order to correctly address the problem of data fusion, compatibility studies between the two sensors are necessary. In this study, a maize field was sensed on several dates in 2018 growing season using both sensors. Numerous vegetation indices based on different spectral channel combinations were calculated and the results were compared using linear regression analysis. First results showed good positive correlations between the indices obtained by the two sensors.