Applications of Remote Image Capture System in Agriculture

Protected agriculture is a field in which the use of automatic systems is a key factor. In fact, the automatic harvesting of delicate fruit has not yet been perfected. This issue has received a great deal of attention over the last forty years, although no commercial harvesting robots are available at present, mainly due to the complexity and variability of the working environments. In this work we developed a computer vision system (CVS) to automate the detection and localization of fruit in a tomato crop in a typical Mediterranean greenhouse. The tasks to be performed by the system are: (1) the detection of the ripe tomatoes, (2) the location of the ripe tomatoes in the XY coordinates of the image, and (3) the location of the ripe tomatoes’ peduncles in the XY coordinates of the image. Tasks 1 and 2 were performed using a large set of digital image processing tools (enhancement, edge detection, segmentation, and the feature’s description of the tomatoes). Task 3 was carried out using basic trigonometry and numerical and geometrical descriptors. The results are very promising for beef and cluster tomatoes, with the system being able to classify 80.8% and 87.5%, respectively, of fruit with visible peduncles as “collectible”. The average processing time per image for visible ripe and harvested tomatoes was less than 30 ms. Full article

Surface soil moisture is an important hydrological parameter in agricultural areas. Periodic measurements in tropical mountain environments are poorly representative of larger areas, while satellite resolution is too coarse to be effective in these topographically varied landscapes, making spatial resolution an important parameter to consider. The Las Palmas catchment area near Medellin in Colombia is a vital water reservoir that stores considerable amounts of water in its andosol. In this tropical Andean setting, we use an unmanned aerial vehicle (UAV) with multispectral (visible, near infrared) sensors to determine the correlation of three agricultural land uses (potatoes, bare soil, and pasture) with surface soil moisture. Four vegetation indices (the perpendicular drought index, PDI; the normalized difference vegetation index, NDVI; the normalized difference water index, NDWI, and the soil-adjusted vegetation index, SAVI) were applied to UAV imagery and a 3 m resolution to estimate surface soil moisture through calibration with in situ field measurements. The results showed that on bare soil, the indices that best fit the soil moisture results are NDVI, NDWI and PDI on a detailed scale, whereas on potatoes crops, the NDWI is the index that correlates significantly with soil moisture, irrespective of the scale. Multispectral images and vegetation indices provide good soil moisture understanding in tropical mountain environments, with 3 m remote sensing images which are shown to be a good alternative to soil moisture analysis on pastures using the NDVI and UAV images for bare soil and potatoes. Full article

Infrared thermography has been introduced as an affordable tool for plant water status monitoring, especially in regions where water availability is the main limiting factor in agricultural production. This paper outlines the potential applications of low-cost thermal imaging devices to evaluate the water status of young and mature sweet cherry trees (Prunus avium L.) submitted to water stress. Two treatments per plot were assayed: (i) a control treatment irrigated to ensure non-limiting soil water conditions; and (ii) a water-stress treatment. The seasonal evolution of the temperature of the canopy (Tc) and the difference between Tc and air temperature (ΔT) were compared and three thermal indices were calculated: crop water stress index (CWSI), degrees above control treatment (DAC) and degrees above non-water-stressed baseline (DANS). Midday stem water potential (Ψstem) was used as the reference indicator of water stress and linear relationships of Tc, ΔT, CWSI, DAC and DANS with Ψstem were discussed in order to assess their sensitivity to quantify water stress. CWSI and DANS exhibited strong relationships with Ψstem and two regression lines to young and mature trees were found. The promising results obtained highlight that using low-cost infrared thermal devices can be used to determine the plant water status in sweet cherry trees. Full article

Machine learning algorithms are crucial for crop identification and mapping. However, many works only focus on the identification results of these algorithms, but pay less attention to their classification performance and mechanism. In this paper, based on Google Earth Engine (GEE), Sentinel-2 10 m resolution images during a specific phenological period of winter wheat were obtained. Then, support vector machine (SVM), random forest (RF), and classification and regression tree (CART) machine learning algorithms were employed to identify and map winter wheat in a large-scale area. The hyperparameters of the three machine learning algorithms were tuned by grid search and the 5-fold cross-validation method. The classification performance of the three machine learning algorithms were compared, the results of which demonstrate that SVM achieves best performance in identifying winter wheat, and its overall accuracy (OA), user’s accuracy (UA), producer’s accuracy (PA), and kappa coefficient (Kappa) are 0.94, 0.95, 0.95, and 0.92, respectively. Moreover, 50 various combinations of training and validation sets were used to analyze the generalization ability of the algorithms, and the results show that the average OA of SVM, RF, and CART are 0.93, 0.92, and 0.88, respectively, thus indicating that SVM and RF are more robust than CART. To further explore the sensitivity of SVM, RF, and CART to variations of the algorithm parameters—namely, (C and gamma), (tree and split), and (maxD and minSP)—we employed the grid search method to iterate these parameters, respectively, and to analyze the effect of these parameters on the accuracy scores and classification residuals. It was found that with the change of (C and gamma) in (0.01~1000), SVM’s maximum variation of accuracy score is up to 0.63, and the maximum variation of residuals is 76,215 km². We concluded that SVM is sensitive to the parameters (C and gamma) and presents a positive correlation. When the parameters (tree and split) change between (100~600) and (1~6), respectively, the RF’s maximum variation of accuracy score is 0.08, and the maximum variation of residuals is 1157 km², indicating that RF is low in sensitivity toward the parameters (tree and split). When the parameters (maxD and minSP) are between (10~60), the maximum accuracy change value is 0.06, and the maximum variation of residuals is 6943 km². Therefore, compared to RF, CART is sensitive to the parameters (maxD and minSP) and has poor robustness. In general, under the conditions of the hyperparameters, SVM and RF exhibit optimal classification performance, while CART has relatively inferior performance. Meanwhile, SVM, RF, and CART have different sensitivities toward the algorithm parameters; that is, SVM and CART are more sensitive to the algorithm parameters, while RF has low sensitivity toward changes in the algorithm parameters. The different parameters cause great changes in the accuracy scores and residuals, so it is necessary to determine the algorithm hyperparameters. Generally, default parameters can be used to achieve crop classification, but we recommend the enumeration method, similar to grid search, as a practical way to improve the classification performance of the algorithm if the best classification effect is expected. Full article

We introduce here the Grape Berries Counting Net (GBCNet), a tool for accurate fruit yield estimation from smartphone cameras, by adapting Deep Learning algorithms originally developed for crowd counting. We test GBCNet using cross-validation procedure on two original datasets CR1 and CR2 of grape pictures taken in-field before veraison. A total of 35,668 berries have been manually annotated for the task. GBCNet achieves good performances on both the seven grape varieties dataset CR1, although with a different accuracy level depending on the variety, and on the single variety dataset CR2: in particular Mean Average Error (MAE) ranges from 0.85% for Pinot Gris to 11.73% for Marzemino on CR1 and reaches 7.24% on the Teroldego CR2 dataset. Full article

Pasture quality monitoring is a key element in the decision making process of a farm manager. Laboratory reference methods for assessing quality parameters such as crude protein (CP) or fibers (neutral detergent fiber: NDF) require collection and analytical procedures involving technicians, time, and reagents, making them laborious and expensive. The objective of this work was to evaluate two technological and expeditious approaches for estimating and monitoring the evolution of the quality parameters in biodiverse Mediterranean pastures: (i) near infrared spectroscopy (NIRS) combined with multivariate data analysis and (ii) remote sensing (RS) based on Sentinel-2 imagery to calculate the normalized difference vegetation index (NDVI) and the normalized difference water index (NDWI). Between February 2018 and March 2019, 21 sampling processes were carried out in nine fields, totaling 398 pasture samples, of which 315 were used during the calibration phase and 83 were used during the validation phase of the NIRS approach. The average reference values of pasture moisture content (PMC), CP, and NDF, obtained in 24 tests carried out between January and May 2019 in eight fields, were used to evaluate the RS accuracy. The results of this study showed significant correlation between NIRS calibration models or spectral indices obtained by remote sensing (NDVI_RS and NDWI_RS) and reference methods for quantifying pasture quality parameters, both of which open up good prospects for technological-based service providers to develop applications that enable the dynamic management of animal grazing. Full article

Nowadays, the combination of new technologies and the use of mobile devices opens up a new range of teaching–learning strategies in different agricultural engineering degrees. This article presents an augmented reality tool that allows for improved spatial viewing for students who have certain difficulties with viewing graphic representations of agronomic systems and devices. This tool is known as ARTID (Augmented Reality for Teaching, Innovation and Design) and consists in a free-access mobile application for devices using the Android operating system. The proposed method provides each exploded drawing or overall drawing with a QR code that can be used by students to view their 3D models by augmented reality in their own mobile devices. An evaluation experience was carried out to assess the validity of the tool on different devices and the acceptance and satisfaction level of this kind of resources in subjects of graphic expression in engineering. Finally, an example of application in the agronomic domain is provided by the 3D virtual model of portable ferticontrol equipment that comprises the different structures and tanks, which, if viewed by conventional graphical representations, may entail a certain level of difficulty. Thanks to this tool, reality can be merged with the virtual world to help favour the understanding of certain concepts and to increase student motivation in agronomy studies. Full article

Bananas are the world’s most popular fruit and an important staple food source. Recent outbreaks of Panama TR4 disease are threatening the global banana industry, which is worth an estimated $8 billion. Current methods to map land uses are time- and resource-intensive and result in delays in the timely release of data. We have used existing land use mapping to train a U-Net neural network to detect banana plantations in the Wet Tropics of Queensland, Australia, using high-resolution aerial photography. Accuracy assessments, based on a stratified random sample of points, revealed the classification achieves a user’s accuracy of 98% and a producer’s accuracy of 96%. This is more accurate compared to existing (manual) methods, which achieved a user’s and producer’s accuracy of 86% and 92% respectively. Using a neural network is substantially more efficient than manual methods and can inform a more rapid respond to existing and new biosecurity threats. The method is robust and repeatable and has potential for mapping other commodities and land uses which is the focus of future work. Full article

One of the most important applications of remote imaging systems in agriculture, with the greatest impact on global sustainability, is the determination of optimal crop irrigation. The methodology proposed by the Food and Agriculture Organization (FAO) is based on estimating crop evapotranspiration (ETc), which is done by computing the reference crop evapotranspiration (ETo) multiplied by a crop coefficient (Kc). Some previous works proposed methods to compute Kc using remote crop images. The present research aims at complementing these systems, estimating ETo with the use of soil moisture sensors. A crop of kikuyu grass (Pennisetum clandestinum) was used as the reference crop. Four frequency-domain reflectometry sensors were installed, gathering moisture information during the study period from May 2015 to September 2016. Different machine learning regression algorithms were analyzed for the estimation of ETo using moisture and climatic data. The values were compared with respect to the ETo computed in an agroclimatic station using the Penman–Monteith method. The best method was the randomizable filtered classifier technique, based on the K* algorithm. This model achieved a correlation coefficient, R, of 0.9936, with a root-mean-squared error of 0.183 mm/day and 6.52% mean relative error; the second-best model used artificial neural networks, with an R of 0.9470 and 11% relative error. Thus, this new methodology allows obtaining accurate and cost-efficient prediction models for ETo, as well as for the water balance of the crops. Full article

Investigation on the deformation mechanism of sandstone is crucial to understanding the life cycle patterns of pertinent infrastructure systems considering the extensive adoption of sandstone in infrastructure construction of various engineering systems, e.g., agricultural engineering systems. In this study, the state-of-the-art digital image correlation (DIC) method, which uses classical digital photography, is employed to explore the detailed failure course of sandstone with physical uniaxial compression tests. Four typical points are specifically selected to characterize the global strain field by plotting their corresponding strain–time relationship curves. Thus, the targeted failure thresholds are identified. The Hill–Tsai failure criterion and finite element simulation are then used for the cross-check process of DIC predictions. The results show that, though errors exist between the experimental and the theoretical values, overall, they are sufficiently low to be ignored, indicating good agreement. From the results, near-linear relationships between strain and time are detected before failure at the four chosen points and the failure strain thresholds are almost the same; as low as 0.004. Failure thresholds of sandstone are reliably determined according to the strain variation curve, to forecast sandstone damage and failure. Consequently, the proposed technology and associated information generated from this study could be of assistance in the safety and health monitoring processes of relevant infrastructure system applications. Full article

Agricultural activity has always been threatened by the presence of pests and diseases that prevent the proper development of crops and negatively affect the economy of farmers. One of these pests is Coffee Leaf Rust (CLR), which is a fungal epidemic disease that affects coffee trees and causes massive defoliation. As an example, this disease has been affecting coffee trees in Colombia (the third largest producer of coffee worldwide) since the 1980s, leading to devastating losses between 70% and 80% of the harvest. Failure to detect pathogens at an early stage can result in infestations that cause massive destruction of plantations and significantly damage the commercial value of the products. The most common way to detect this disease is by walking through the crop and performing a human visual inspection. As a result of this problem, different research studies have proven that technological methods can help to identify these pathogens. Our contribution is an experiment that includes a CLR development stage diagnostic model in the Coffea arabica, Caturra variety, scale crop through the technological integration of remote sensing (through drone capable multispectral cameras), wireless sensor networks (multisensor approach), and Deep Learning (DL) techniques. Our diagnostic model achieved an F₁-score of 0.775. The analysis of the results revealed a p-value of 0.231, which indicated that the difference between the disease diagnosis made employing a visual inspection and through the proposed technological integration was not statistically significant. The above shows that both methods were significantly similar to diagnose the disease. Full article

The first step in identifying fruits on trees is to develop garden robots for different purposes such as fruit harvesting and spatial specific spraying. Due to the natural conditions of the fruit orchards and the unevenness of the various objects throughout it, usage of the controlled conditions is very difficult. As a result, these operations should be performed in natural conditions, both in light and in the background. Due to the dependency of other garden robot operations on the fruit identification stage, this step must be performed precisely. Therefore, the purpose of this paper was to design an identification algorithm in orchard conditions using a combination of video processing and majority voting based on different hybrid artificial neural networks. The different steps of designing this algorithm were: (1) Recording video of different plum orchards at different light intensities; (2) converting the videos produced into its frames; (3) extracting different color properties from pixels; (4) selecting effective properties from color extraction properties using hybrid artificial neural network-harmony search (ANN-HS); and (5) classification using majority voting based on three classifiers of artificial neural network-bees algorithm (ANN-BA), artificial neural network-biogeography-based optimization (ANN-BBO), and artificial neural network-firefly algorithm (ANN-FA). Most effective features selected by the hybrid ANN-HS consisted of the third channel in hue saturation lightness (HSL) color space, the second channel in lightness chroma hue (LCH) color space, the first channel in L*a*b* color space, and the first channel in hue saturation intensity (HSI). The results showed that the accuracy of the majority voting method in the best execution and in 500 executions was 98.01% and 97.20%, respectively. Based on different performance evaluation criteria of the classifiers, it was found that the majority voting method had a higher performance. Full article

This paper is about the geometric and radiometric consistency of diverse and overlapping datasets acquired with the Parrot Sequoia camera. The multispectral imagery datasets were acquired above agricultural fields in Northern Italy and radiometric calibration images were taken before each flight. Processing was performed with the Pix4Dmapper suite following a single-block approach: images acquired in different flight missions were processed in as many projects, where different block orientation strategies were adopted and compared. Results were assessed in terms of geometric and radiometric consistency in the overlapping areas. The geometric consistency was evaluated in terms of point cloud distance using iterative closest point (ICP), while the radiometric consistency was analyzed by computing the differences between the reflectance maps and vegetation indices produced according to adopted processing strategies. For normalized difference vegetation index (NDVI), a comparison with Sentinel-2 was also made. This paper will present results obtained for two (out of several) overlapped blocks. The geometric consistency is good (root mean square error (RMSE) in the order of 0.1 m), except for when direct georeferencing is considered. Radiometric consistency instead presents larger problems, especially in some bands and in vegetation indices that have differences above 20%. The comparison with Sentinel-2 products shows a general overestimation of Sequoia data but with similar spatial variations (Pearson’s correlation coefficient of about 0.7, p-value < 2.2 × 10⁻¹⁶). Full article

Monitoring the cotton budding rate is important for growers so that they can replant cotton in a timely fashion at locations at which cotton density is sparse. In this study, a true-color camera was mounted on an unmanned aerial vehicle (UAV) and used to collect images of young cotton plants to estimate the germination of cotton plants. The collected images were preprocessed by stitching them together to obtain the single orthomosaic image. The support-vector machine method and maximum likelihood classification method were conducted to identify the cotton plants in the image. The accuracy evaluation indicated the overall accuracy of the classification for SVM is 96.65% with the Kappa coefficient of 93.99%, while for maximum likelihood classification, the accuracy is 87.85% with a Kappa coefficient of 80.67%. A method based on the morphological characteristics of cotton plants was proposed to identify and count the overlapping cotton plants in this study. The analysis showed that the method can improve the detection accuracy by 6.3% when compared to without it. The validation based on visual interpretation indicated that the method presented an accuracy of 91.13%. The study showed that the minimal resolution of no less than 1.2 cm/pixel in practice for image collection is necessary in order to recognize cotton plants accurately. Full article

The area of remote sensing techniques in agriculture has reached a significant degree of development and maturity, with numerous journals, conferences, and organizations specialized in it. Moreover, many review papers are available in the literature. The present work describes a literature review that adopts the form of a systematic mapping study, following a formal methodology. Eight mapping questions were defined, analyzing the main types of research, techniques, platforms, topics, and spectral information. A predefined search string was applied in the Scopus database, obtaining 1590 candidate papers. Afterwards, the most relevant 106 papers were selected, considering those with more than six citations per year. These are analyzed in more detail, answering the mapping questions for each paper. In this way, the current trends and new opportunities are discovered. As a result, increasing interest in the area has been observed since 2000; the most frequently addressed problems are those related to parameter estimation, growth vigor, and water usage, using classification techniques, that are mostly applied on RGB and hyperspectral images, captured from drones and satellites. A general recommendation that emerges from this study is to build on existing resources, such as agricultural image datasets, public satellite imagery, and deep learning toolkits. Full article

Remote Sensing (RS) allows the estimation of some important vineyard parameters. There are several platforms for obtaining RS information. In this context, Sentinel satellites are a valuable tool for RS since they provide free and regular images of the earth’s surface. However, several problems regarding the low-resolution of the imagery arise when using this technology, such as handling mixed pixels that include vegetation, soil and shadows. Under this condition, the Normalized Difference Vegetation Index (NDVI) value in a particular pixel is an indicator of the amount of vegetation (canopy area) rather than the NDVI from the canopy (as a vigour expression), but its reliability varies depending on several factors, such as the presence of mixed pixels or the effect of missing vines (a vineyard, once established, generally loses grapevines each year due to diseases, abiotic stress, etc.). In this study, a vine removal simulation (greenhouse experiment) and an actual vine removal (field experiment) were carried out. In the field experiment, the position of the Sentinel-2 pixels was marked using high-precision GPS. Controlled removal of vines from a block of cv. Cabernet Sauvignon was done in four steps. The removal of the vines was done during the summer of 2019, matching with the start of the maximum vegetative growth. The Total Leaf Area (TLA) of each pixel was calculated using destructive field measurements. The operations were planned to have two satellite images available between each removal step. As a result, a strong linear relationship (R² = 0.986 and R² = 0.72) was obtained between the TLA and NDVI reductions, which quantitatively indicates the effect of the missing vines on the NDVI values. Full article