Remote Machine Research Articles

Assessing the Yield of Wheat Using Satellite Remote Sensing-Based Machine Learning Algorithms and Simulation Modeling

Globally, estimating crop acreage and yield is one of the most critical issues that policy and decision makers need for assessing annual crop productivity and food supply. Nowadays, satellite remote sensing and geographic information system (GIS) can enable the estimation of these crop production parameters over large geographic areas. The present work aims to estimate the wheat (Triticum aestivum) acreage and yield of Maharajganj, Uttar Pradesh, India, using satellite-based data products and the Carnegie-Ames-Stanford Approach (CASA) model. Uttar Pradesh is the largest wheat-producing state in India, and this district is well known for its quality organic wheat. India is the leader in wheat grain export, and, hence, its monitoring of growth and yield is one of the top economic priorities of the country. For the calculation of wheat acreage, we performed supervised classification using the Random Forest (RF) and Support Vector Machine classifiers and compared their classification accuracy based on ground-truthing. We found that RF performed a significantly accurate acreage assessment (kappa coefficient 0.84) compared to SVM (0.68). The CASA model was then used to calculate the winter crop (Rabi, winter-sown, and summer harvested) wheat net primary productivity (NPP) in the study area for the 2020–2021 growth season using the RF-based acreage product. The model used for wheat NPP-yield conversion (CASA) showed 3100.27 to 5000.44 kg/ha over 148,866 ha of the total wheat area. The results showed that in the 2020–2021 growing season, all the districts of Uttar Pradesh had similar wheat growth trends. A total of 30 observational data points were used to verify the CASA model-based estimates of wheat yield. Field-based verification shows that the estimated yield correlates well with the observed yield (R2 = 0.554, RMSE = 3.36 Q/ha, MAE −0.56 t ha−1, and MRE = −4.61%). Such an accuracy for assessing regional wheat yield can prove to be one of the promising methods for calculating the whole region’s agricultural yield. The study concludes that RF classifier-based yield estimation has shown more accurate results and can meet the requirements of a regional-scale wheat grain yield estimation and, thus, can prove highly beneficial in policy and decision making.

Improving Estimation of Winter Wheat Nitrogen Status Using Random Forest by Integrating Multi-Source Data Across Different Agro-Ecological Zones.

Timely and accurate estimation of plant nitrogen (N) status is crucial to the successful implementation of precision N management. It has been a great challenge to non-destructively estimate plant N status across different agro-ecological zones (AZs). The objective of this study was to use random forest regression (RFR) models together with multi-source data to improve the estimation of winter wheat (Triticum aestivum L.) N status across two AZs. Fifteen site-year plot and farmers' field experiments involving different N rates and 19 cultivars were conducted in two AZs from 2015 to 2020. The results indicated that RFR models integrating climatic and management factors with vegetation index (R2 = 0.72–0.86) outperformed the models by only using the vegetation index (R2 = 0.36–0.68) and performed well across AZs. The Pearson correlation coefficient-based variables selection strategy worked well to select 6–7 key variables for developing RFR models that could achieve similar performance as models using full variables. The contributions of climatic and management factors to N status estimation varied with AZs and N status indicators. In higher-latitude areas, climatic factors were more important to N status estimation, especially water-related factors. The addition of climatic factors significantly improved the performance of the RFR models for N nutrition index estimation. Climatic factors were important for the estimation of the aboveground biomass, while management variables were more important to N status estimation in lower-latitude areas. It is concluded that integrating multi-source data using RFR models can significantly improve the estimation of winter wheat N status indicators across AZs compared to models only using one vegetation index. However, more studies are needed to develop unmanned aerial vehicles and satellite remote sensing-based machine learning models incorporating multi-source data for more efficient monitoring of crop N status under more diverse soil, climatic, and management conditions across large regions.

Remote Run-Time Failure Detection and Recovery Control For Quadcopters

We propose an adaptive run-time failure recovery control system for quadcopter drones, based on remote real-time processing of measurement data streams. Particularly, the measured RPM values of the quadcopter motors are transmitted to a remote machine which hosts failure detection algorithms and performs recovery procedure. The proposed control system consists of three distinct parts: (1) A set of computationally simple PID controllers locally onboard the drone, (2) a set of computationally more demanding remotely hosted algorithms for real-time drone state detection, and (3) a digital twin co-execution software platform — the ModelConductor-eXtended — for two-way signal data exchange between the former two. The local on-board control system is responsible for maneuvering the drone in all conditions: path tracking under normal operation and safe landing in a failure state. The remote control system, on the other hand, is responsible for detecting the state of the drone and communicating the corresponding control commands and controller parameters to the drone in real time. The proposed control system concept is demonstrated via simulations in which a drone is represented by the widely studied Quad-Sim six degrees-of-freedom Simulink model. Results show that the trained failure detection binary classifier achieves a high level of performance with F1-score of 96.03%. Additionally, time analysis shows that the proposed remote control system, with average execution time of 0.49 milliseconds and total latency of 6.92 milliseconds in two-way data communication link, meets the real-time constraints of the problem. The potential practical applications for the presented approach are in drone operation in complex environments such as factories (indoor) or forests (outdoor).

Identification and Classification of Downy Mildew Severity Stages in Watermelon Utilizing Aerial and Ground Remote Sensing and Machine Learning.

Remote sensing and machine learning (ML) could assist and support growers, stakeholders, and plant pathologists determine plant diseases resulting from viral, bacterial, and fungal infections. Spectral vegetation indices (VIs) have shown to be helpful for the indirect detection of plant diseases. The purpose of this study was to utilize ML models and identify VIs for the detection of downy mildew (DM) disease in watermelon in several disease severity (DS) stages, including low, medium (levels 1 and 2), high, and very high. Hyperspectral images of leaves were collected in the laboratory by a benchtop system (380–1,000 nm) and in the field by a UAV-based imaging system (380–1,000 nm). Two classification methods, multilayer perceptron (MLP) and decision tree (DT), were implemented to distinguish between healthy and DM-affected plants. The best classification rates were recorded by the MLP method; however, only 62.3% accuracy was observed at low disease severity. The classification accuracy increased when the disease severity increased (e.g., 86–90% for the laboratory analysis and 69–91% for the field analysis). The best wavelengths to differentiate between the DS stages were selected in the band of 531 nm, and 700–900 nm. The most significant VIs for DS detection were the chlorophyll green (Cl green), photochemical reflectance index (PRI), normalized phaeophytinization index (NPQI) for laboratory analysis, and the ratio analysis of reflectance spectral chlorophyll-a, b, and c (RARSa, RASRb, and RARSc) and the Cl green in the field analysis. Spectral VIs and ML could enhance disease detection and monitoring for precision agriculture applications.

Increase in rock drill penetration depth

The application range of remote drilling machines is described. The common shortage of these machines is the dependence of the blow energy transfer to the bottomhole on the bottomhole depth. Aimed to reduce the influence of this factor, it is proposed to use the mixed-type method of drilling with a downhole rock drill to be placed directly at the bottomhole. The key element of this method is a small-size DTH rock drill. The layout of the rock drill and its prototype testing results are presented.

The Geothermal Artificial Intelligence for geothermal exploration

Exploration of geothermal resources involves analysis and management of a large number of uncertainties, which makes investment and operations decisions challenging. Remote Sensing (RS), Machine Learning (ML) and Artificial Intelligence (AI) have potential in managing the challenges of geothermal exploration. In this paper, we present a methodology that integrates RS, ML and AI to create an initial assessment of geothermal potential, by resorting to known indicators of geothermal areas namely mineral markers, surface temperature, faults and deformation. We demonstrated the implementation of the method in two sites (Brady and Desert Peak geothermal sites) that are close to each other but have different characteristics (Brady having clear surface manifestations and Desert Peak being a blind site). We processed various satellite images and geospatial data for mineral markers, temperature, faults and deformation and then implemented ML methods to obtain pattern of surface manifestation of geothermal sites. We developed an AI that uses patterns from surface manifestations to predict geothermal potential of each pixel. We tested the Geothermal AI using independent data sets obtaining accuracy of 92–95%; also tested the Geothermal AI trained on one site by executing it for the other site to predict the geothermal/non-geothermal delineation, the Geothermal AI performed quite well in prediction with 72–76% accuracy.

Predicting Terrorism in Europe with Remote Sensing, Spatial Statistics, and Machine Learning

This study predicts the presence or absence of terrorism in Europe on a previously unexplored spatial scale. Dependent variables consist of satellite imagery and socio-environmental data. Five machine learning models were evaluated over the following binary classification problem: the presence or absence of historical attacks within hexagonal-grid cells of 25 square kilometers. Four spatial statistics were conducted to assess the validity of the results and improve our inferential understanding of spatial processes among terror attacks. This analysis resulted in a Random Forest model that achieves 0.99 accuracy in predicting the presence or absence of terrorism at a spatial resolution of approximately 5 km. The results were validated by robust F1 and average precision scores of 0.96 and 0.97, respectively. Additionally, statistical analysis revealed spatial differences between separatists and all other terrorist types. This work concludes that remote sensing, machine learning, and spatial techniques are important and valuable methods for providing insight into terrorist activity and behavior.

Remote Desktop Software as a forensic resource

ABSTRACT Remote Desktop Software (RDS) enables the controlling of a computer system without the need for physical access. Operations are sent to the remote machine and executed as if performed by a local user. With an unprecedented shift to remote working due to the COVID-19 Pandemic, more people are working on home devices without enterprise IT support and therefore reliant upon this software to collaborate and keep their systems available and secure. RDS complicates a Forensic Investigation as any person with remote access privileges or knowledge of bypassing them could be responsible for an action. Despite its importance and prevalence, forensic research into RDS is minimal. As a market-leading solution for Windows, TeamViewer is an impactful starting point to demonstrate that such software is forensically-valuable to explore. This paper shows that with suitable evidence, an Investigator can identify which machines have performed remote control or been controlled, transferred files and have been remotely rebooted, among other events. We also highlight a potential privacy concern due to inadequate uninstallation processes. To illustrate the value of our findings we publish a Python module for Autopsy that automatically locates, processes and visualises key TeamViewer artefacts for an Investigator.

AI Back-End as a Service for Learning Switching of Mobile Apps Between the Fog and the Cloud

Given that cloud servers are usually remotely located from the devices of mobile apps, the end-users of the apps can face delays. The Fog has been introduced to augment the apps with machines located at the network edge close to the end-users. However, edge machines are usually resource constrained. Thus, the execution of online data-analytics on edge machines may not be feasible if the time complexity of the data-analytics algorithm is high. To overcome this, multiple instances of the back-end should be deployed on edge and remote machines. In this case, the research question is how the switching of the app among the instances of the back-end can be dynamically decided based on the response time of the service instances. To answer this, we contribute an AI approach that trains machine-learning models of the response time of service instances. Our approach extends a back-end as a service into an AI self-back-end as a service that self-decides at runtime the right edge/remote instance that achieves the lowest response-time. We evaluate the accuracy and the efficiency of our approach by using real-word machine-learning datasets on an existing auction app.

Integrating Remote Sensing and Machine Learning for Regional-Scale Habitat Mapping: Advances and future challenges for desert locust monitoring

Increased access to reliable data and computationally efficient systems has created more collaborative potential between remote sensing and machine learning for species habitat prediction. Exploiting the integrative opportunities will require a deeper understanding of methods remote sensing instruments use to capture biophysical variables and the extent that data science can model ecological relationships. In this article, we provide the first systemic review of the integration of remote sensing and machine learning to predict habitat for the highly destructive desert locust and explore deep learning as a new method for increased classification. We evaluated the performance of six machine learning algorithms in two study regions (Niger and Sudan), using locust observations, multiple pseudoabsence data sets, and remotely sensed habitat data. In both regions, the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">k</i> -nearest neighbor ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">k</i> NN) and a deep neural network (DNN) were the best-preforming models. In Niger, the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">k</i> NN average accuracy score was 88%, and the F-1 score was 89% for the Present <xref ref-type="disp-formula" rid="deqn1" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">(1)</xref> class. The DNN average accuracy score was 88%, and the F-1 score was 89%. In Sudan, the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">k</i> NN average accuracy score was 88%, and the F-1 score was 88% for the Present <xref ref-type="disp-formula" rid="deqn1" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">(1)</xref> class. The DNN average accuracy score was 88%, and the F-1 score was 89%. Additionally, we outline a process for robustly modeling habitat through remote sensing data and highlight important limitations. We propose that the DNN model has the best potential for constructing a transferable representation and discuss novel methods to meet future challenges in desert locust management in a sustainable way.

Assessing the Efficiency of Remote Sensing and Machine Learning Algorithms to Quantify Wheat Characteristics in the Nile Delta Region of Egypt

Monitoring strategic agricultural crops in terms of crop growth performance, by accurate cost-effective and quick tools is crucially important in site-specific management to avoid crop reductions. The availability of commercial high resolution satellite images with high resolution (spatial and spectral) as well as in situ spectra measurements can help decision takers to have deep insight on crop stress in a certain region. The research attempts to examine remote sensing dataset for forecasting wheat crop (Sakha 61) characteristics including the leaf area index (LAI), plant height (plant-h), above ground biomass (AGB) and Soil Plant Analysis Development (SPAD) value of wheat across non-stress, drought and salinity-induced stress in the Nile Delta region. In this context, the ability of in situ spectroradiometry measurements and QuickBird high resolution images was evaluated in our research. The efficiency of Random Forest (RF) and Artificial Neural Network (ANN), mathematical models was assessed to estimate the four measured wheat characteristics based on vegetation spectral reflectance indices (V-SRIs) extracted from both approaches and their interactions. Field surveys were carried out to collect in situ spectroradiometry measurements concomitant with the acquisition of QuickBird imagery. The results demonstrated that several V-SRIs extracted from in situ spectroradiometry data and the QuickBird image correlated with the LAI, plant-h, AGB, and SPAD value of wheat crop across the study site. The determination coefficient (R2) values of the association between V-SRIs of in situ spectroradiometry data and various determined wheat characteristics varied from 0.26 to 0.85. The ANN-GSIs-3 was found to be the optimum predictive model, demonstrating a greater relationship between the advanced features and LAI. The three features of V-SRIs comprised in this model were strongly significant for the prediction of LAI. The attained results indicated high R2 values of 0.94 and 0.86 for the training and validation phases. The ANN-GSIs-3 model constructed for the determination of chlorophyll in the plant which had higher performance expectations (R2 = 0.96 and 0.92 for training and validation datasets, respectively). In conclusion, the results of our study revealed that high resolution remote sensing images such as QuickBird or similar imagery, and in situ spectroradiometry measurements have the feasibility of providing necessary crop monitoring data across non-stressed and stressed (drought and salinity) conditions when integrating V-SRIs with ANN and RF algorithms.

Use Remote Sensing and Machine Learning to Study the Changes of Broad-Leaved Forest Biomass and Their Climate Driving Forces in Nature Reserves of Northern Subtropics

Forest is the largest vegetation carbon pool in the global terrestrial ecosystem. The spatial distribution and change of forest biomass are of importance to reveal the surface spatial variation and driving factors, to analyze and evaluate forest productivity, and to evaluate ecological function of forest. In this study, broad-leaved forests located in a typical state nature reserve in northern subtropics were selected as the study area. Based on ground survey data and high-resolution remote sensing images, three machine learning models were used to identify the best remote sensing quantitative inversion model of forest biomass. The biomass of broad-leaved forest with 30-m resolution in the study area from 1998 to 2016 was estimated by using the best model about every two years. With the estimated biomass, multiple leading factors to cause biomass temporal change were then identified from dozens of remote sensing factors by investigating their nonlinear correlations. Our results showed that the artificial neural network (ANN) model was the best (R2 = 0.8742) among the three, and its accuracy was also much higher than that of the traditional linear or nonlinear models. The mean biomass of the broad-leaved forest in the study area from 1998 to 2016 ranged from 90 to 145 Mg ha−1, showing an obvious temporal variation. Instead of biomass, biomass change (BC) was studied further in this research. Significant correlations were found between BC in broad-leaved forest and three climate factors, including average daily maximum surface temperature, maximum precipitation, and maximum mean temperature. It was also found that BC has a strong correlation with the biomass at the previous time (i.e., two years ago). Those quantitative correlations were used to construct a linear model of BC with high accuracy (R2 = 0.8873), providing a new way to estimate the biomass change of two years later based on the observations of current biomass and the three climate factors.

Surface Water Salinity Evaluation and Identification for Using Remote Sensing Data and Machine Learning Approach

Knowledge of the spatio-temporal distribution of salinity provides valuable information for understanding different processes between biota and environment, especially in hypersaline lakes. Remote sensing techniques have been used for monitoring different components of the environment. Currently, one of the biggest challenges is the spatio-temporal monitoring of the salinity level in water bodies. Due to some limitations, such as the inability to be located there permanently, it is difficult to obtain these data directly. In this study, machine learning techniques were used to evaluate the salinity level in hypersaline East Sivash Bay. In total, 93 in situ data samples and 6 Sentinel-2 datasets were used, according to field measurements. Using linear regression, random forest and AdaBoost models, eight water salinity evaluation models were built (six with simple, one with random forest and one with AdaBoost). The accuracy of the best-fitted simple linear regression model was 0.8797; for random forest, it was equal, at 0.808, and for AdaBoost, it was −0.72. Furthermore, it was found that with an increase in salinity, the absorbing light shifts from the ultraviolet part of the spectrum to the infrared and short-wave infrared parts, which makes it possible to produce continuous monitoring of hypersaline water bodies using remote sensing data.

How to Predict Parturition in Cattle? A Literature Review of Automatic Devices and Technologies for Remote Monitoring and Calving Prediction.

Simple SummaryCattle farming is progressively facing an increase in the number of animals that farmers must care for, resulting in decreasing time for observation of the single cow. A large amount of the scientific literature has been published concerning remote automatic devices and machine learning technologies for continuous monitoring of animal behavior and health status, including sensors for calving prediction This review summarizes the current status of the art concerning available automatic devices for the identification of the beginning of calving.Cattle farming is facing an increase in number of animals that farmers must care for, together with decreasing time for observation of the single animal. Remote monitoring systems are needed in order to optimize workload and animal welfare. Where the presence of personnel is constant, for example in dairy farms with great number of lactating cows or with three milking/day, calving monitoring systems which send alerts during the prodromal stage of labor (stage I) could be beneficial. On the contrary, where the presence of farm personnel is not guaranteed, for example in smaller farms, systems which alert at the beginning of labor (stage II) could be preferred. In this case, time spent observing periparturient animals is reduced. The reliability of each calving alarm should also be considered: automatic sensors for body temperature and activity are characterized by a time interval of 6–12 h between the alarm and calving. Promising results have been shown by devices which could be placed within the vaginal canal, thus identifying the beginning of fetal expulsion and optimizing the timing of calving assistance. However, some cases of non-optimal local tolerability and cow welfare issues are reported. Future research should be aimed to improve Sensitivity (Se), Specificity (Sp) and Positive Predictive Value (PPV) of calving alert devices in order to decrease the number of false positive alarms and focusing on easy-to-apply, re-usable and well tolerated products.

A simple yet effective approach for log based critical errors prediction

With the advent of Industry 4.0, failure anticipation is becoming one of the key objectives in industrial research. In this context, predictive maintenance is an active research area for various applications. This paper presents an approach to predict high importance errors using log data emitted by machine tools. It uses the concept of bag to summarize events (or errors) provided by remote machines, available within log files. The idea of bag is inspired by the Multiple Instance Learning paradigm introduced by Dietterich et al. However, our proposal follows a different strategy to label bags, that we wanted as simple as possible. Three main setting parameters are defined to build the training set allowing the model to fine-tune the trade-off between early warning, historical data informativeness and time accuracy. The effectiveness of the approach is demonstrated using a real industrial application where failures can be predicted up to seven days in advance thanks to a classification model.

Psycholinguistics of AI, Psycholinguistics versus Machine code

Abstract: I will examine the background of the language phylogeny in emerging Homo Sapiens as a fast, bipedal, long-distance runner in Black Africa;, followed by language psychogenesis in children from their gestation twenty-fourth week onward. I will concentrate on the audio-visual machiness impact, Lacan’s mirror stage, AV machines, the discontinuity between real and virtual realities, the remote control and AI machines as smart speakers and smart homes. In addition, I will discuss the following questions: Is the Machine beyond human intelligence? Is the human individual beyond Homo Sapiens? Is the human community beyond social contract? My working hypotheses on education within phylogenetic psycholinguistics are built on the following topics: Tomorrow’s AI class (unit and room); guided self-learning and who is the guide; can transference and countertransference take place in AI-guided self-learning? Can a human subject develop such transference/countertransference with a machine? Can a machine “play the game”? In conclusion, I will debate “The utopian vision of an improved human being versus the dystopic vision of human beings and human communities totally enslaved to AI machines. Keywords: AI; linguistic phylogeny; guided self-learning; (counter)-transference; enslaved-to-machines

Predictive Framework of Plant Height in Commercial Cotton Fields Using a Remote Sensing and Machine Learning Approach

Predictive Framework of Plant Height in Commercial Cotton Fields Using a Remote Sensing and Machine Learning Approach

A Global Meta-Analysis of Soil Salinity Prediction Integrating Satellite Remote Sensing, Soil Sampling, and Machine Learning

Despite the growing interest among researchers, satellite-based prediction of soil salinity remains highly uncertain. The improvements in prediction accuracy reported in previous studies are usually limited to a single area. We performed a meta-analysis of regional satellite-based soil salinity predictions combined with <i>in situ</i> soil sampling and machine learning. Based on <inline-formula> <tex-math notation="LaTeX">$R^{2}$ </tex-math></inline-formula> and root-mean-square error (RMSE) collected, we evaluated the effects of various features on the model accuracy and established a Bayesian network to evaluate the joint causal effect of multifeatures. Most significant differences were found in soil sampling schemes and characteristics of the study area, including the mean and variability (averaged <inline-formula> <tex-math notation="LaTeX">$R^{2}$ </tex-math></inline-formula> of 0.75 for soil sample sets with lower salinity variation and 0.62 for others) of the salinity, climate type (<inline-formula> <tex-math notation="LaTeX">$R^{2}$ </tex-math></inline-formula> of 0.64 in arid areas and 0.74 in others), soil texture (<inline-formula> <tex-math notation="LaTeX">$R^{2}$ </tex-math></inline-formula> of 0.66 in sandy areas and 0.57 in others), and the interval between sampling date and satellite data acquisition date (<inline-formula> <tex-math notation="LaTeX">$R^{2}$ </tex-math></inline-formula> of 0.53 under the condition of over 15 days and 0.65 in others). Generally, using different satellite data has limited effects on model performance among which Sentinel-2 performed better (<inline-formula> <tex-math notation="LaTeX">$R^{2} $ </tex-math></inline-formula> &#x003D; 0.72) than Landsat (<inline-formula> <tex-math notation="LaTeX">$R^{2} $ </tex-math></inline-formula> &#x003D; 0.66). The sampling of subsamples for each sample should focus on their subpixel-scale spatial heterogeneity across satellite data rather than the number of subsamples. It is also necessary to select appropriate vegetation and salinity indices for different satellite data under different vegetation conditions. Among algorithms, random forests (<inline-formula> <tex-math notation="LaTeX">$R^{2} $ </tex-math></inline-formula> &#x003D; 0.70) and support vector machines (<inline-formula> <tex-math notation="LaTeX">$R^{2} $ </tex-math></inline-formula> &#x003D; 0.71) performed best.

Assessment of Forest Dieback in the Moroccan Central Plateau Using Remote Sensing and Machine Learning

Assessment of Forest Dieback in the Moroccan Central Plateau Using Remote Sensing and Machine Learning

Remote Sensing and Machine Learning Modeling to Support the Identification of Sugarcane Crops

One of the main concerns of agricultural financing institutions is to make sure the loans they grant are used for the stated objective when the loan was requested. Specifically, when Banco Agrario de Colombia grants loans for crop farmers, it schedules verification visits to the cultivation sites to check if the crop stipulated in the loan agreement exists and assess its health. These visits are challenging to make due to the number of visits over vast areas that they need to schedule, lack of trained personnel, and difficulty of access. This article proposes a software tool, based on a machine learning model for processing free satellite imagery, to support the bank’s identification of non-compliant crops with the investment plan before making field visits, minimizing the loss of investment by focusing on those areas to prioritize the visits. Sugarcane along the department of Boyacá, Colombia was chosen as the case of study. Free access satellite imagery through the Colombian Data Cube (CDCol) was used and machine learning models were applied on them to classify the land and predict the presence of the crop, a Random Forest model achieved an overall F1-score of 91% using Landsat-8 imagery and a K-nearest Neighbors model achieved an overall F1-score of 98% using Sentinel-2 imagery.