New Trends and Challenges in Precision and Digital Agriculture

1. Introduction

Real change is needed in the agricultural sector to meet the challenges of the 21st century in terms of humanity’s food needs. Activities related to the modernization of agricultural production must be safe for the intensively exploited environment. In this context, precision agriculture, which is closely linked to the digitization of agriculture, is becoming increasingly important. The role of precision agriculture is to increase yields and support management decision making on farms using modern, advanced technologies such as sensors and measurement and analytical tools. Accordingly, production carried out in accordance with the principles of precision and digital agriculture uses a great deal of data and a variety of information. All this is carried out to effectively reduce labor time, efficiently manage agricultural resources, and rationally use inputs [1,2,3,4,5]. More and more producers are successfully applying information technology to their farms, so the concept of digital agriculture is gaining importance. This solution focuses the automation of machinery and processes, involving the latest developments in artificial intelligence: classical and convolutional neural networks [1,6,7,8,9,10]; analysis of diverse images [8,11,12,13,14]; cloud computing and unmanned aerial vehicles [15,16,17,18,19], etc. Digital technologies in agriculture enable a better understanding of the interdependence of factors that determine various aspects of the business. This is because the collection of data, analysis, and subsequent control of results can be carried out within a single system, and they are compatible with multiple machines and equipment. It is noteworthy that, with appropriate programming, the equipment has the ability to take various simple actions, for example, based on neural models [20,21,22,23,24,25]. The continuous development and improvement of precision and digital farming elements are providing innovative solutions in many aspects. More and more tools for detecting/identifying various undesirable objects in crop conditions, such as weeds or pests, are being researched. Dynamic advances in image analysis in agriculture make it possible to define disease symptoms observed in plants during the growing season. Typically, such identification is subject to fewer errors [2,12,26]. Proper analyses of images from drones or satellites provide a chance to assess the condition of plants in the field at each stage of their growth to identify potential abiotic threats and stress conditions [27,28].

2. Methods Used in Machine Learning

Implementing machine learning (ML) into precision agriculture has the real potential to bring many benefits. However, it requires the right technological infrastructure, properly processed data, and adequate training for farmers to take full advantage of the potential of machine learning. To be able to implement products based on machine learning for agricultural purposes, it is necessary to involve the scientific community. The development of suitable ML models not only requires a large amount of learning data, but a large dataset for model testing is also necessary. Only the most accurate models can find applications in precision agriculture. In this regard, deep learning methods [29] such as convolutional neural networks (CNNs) are widely used. CNNs are used, among others, in the popular image detection algorithm—YOLO (You Only Look Once) [30,31,32,33]. In addition, CNNs can be successfully used to map various crops [34] or predict soil moisture in vegetated areas [35]. Artificial neural networks (ANNs) and random forests (RFs) are appropriate tools for yield prediction [18,28,36,37,38,39,40,41]. In order to obtain more precise and accurate results, efforts are being made to improve existing machine learning methods. The ResNet50 model, Faster RCNN, and Focal Voxel R-CNN have been used to diagnose rice blight, detect defects in groundnut crops, and detect obstacles by automated agricultural machinery, respectively [42,43,44]. Diagnosing agricultural crop diseases or detecting obstacles in real time is not possible without advanced visual technologies. Image analysis combined with ML improves agricultural production and minimizes losses. With advanced technologies and tools, farmers and gardeners can regularly and systematically collect data on plant health, [45,46,47], enabling faster responses and reducing losses [48]. The introduction of smart irrigation systems [49,50,51] and the accurate dosing of fertilizers [52] or crop protection products [53] allows the better use of water resources and minimizes environmental pollution.

3. Precision Agriculture in Plant Cultivation

Due to its versatility, precision and digital agriculture has found applications in all types and specializations of agricultural and horticultural production. Precision fertilization, irrigation, current assessments of vegetation condition, decision support and management systems, yield prediction, and the classification of disease signs and symptoms are implemented in greenhouse crops, production fields, orchards, vegetation halls, etc. [4,37,54,55,56,57,58]. The above solutions are practiced extensively in cereal and rapeseed crops, cotton, and soybeans [15,59,60]. The most popular implementations involve the use of detailed images to assess the state of vegetation by analyzing vegetation indices; yield predictions in qualitative and quantitative terms; and combining the work of various systems that collect important environmental data—meteorological, soil, and yields—quantitatively and qualitatively. In more demanding and specialized production, i.e., vegetable, fruit, potato, and herb crops, the greatest importance is attributed to image analysis, allowing the identification of diseases, physiological disorders, and yield quality defects [8,61,62,63].

4. Data and Sources

Satellite imaging plays a key role in the development of modern agriculture. With the help of advanced satellite technology, farmers can obtain valuable information about their crops and soil, allowing them to optimize their farming processes. More and more scientific work in the field of precision agriculture is focusing on the use of satellite data in the development of ML models. Data from Sentinel 1 and Sentinel 2 satellites [34,64], and the MODIS optical-mechanical scanner deployed on Terra and Aqua satellites [16,65,66,67] are commonly used. These tools allow the collection of spectral and multispectral data, which are aids for calculating vegetation indices (VIs). VIs as parameters or measures are used to assess plant growth and development activity. Based on indicators such as NDVI and GNDVI, it is possible to identify diseases: for example, in cotton cultivation [68]. NDVI has been shown to be strongly correlated with potassium and phosphorus content in the soil [69]. Plant vegetation indices such as EVI, NDRE, SAVI or SARVI are helpful in detecting chlorophyll content: for example, in corn plants [70]. Data for VI calculations are also obtained using UAVs (unmanned aerial vehicles). Drones are becoming an indispensable part of digital agriculture, the use of which contributes to the further development of farms. The ability to work on cloudy days, higher image resolution, omni-directional analysis, and positioning accuracy make UAVs have a certain advantage over satellites. These features make drones increasingly applicable to agriculture. Assessing the field germination of soybean plants [71]; determining the nutritional status of spring wheat after the earing stage [72]; and monitoring nitrogen concentrations in walnut [73], cotton [74], or phenotyping in corn breeding [75] are just a few examples of what can be achieved using UAVs and spectral cameras in modern agriculture.

5. Increased Interest in the Area of Precision and Digital Agriculture

Precision and digital agriculture encompasses a vast and growing research area. Global issues such as changes in weather patterns during the growing season, the need to quickly and accurately identify diseases and pests, and monitoring the nutritional status of plants in nitrogen and other elements require further research that will increase the number of publications. To be able to measure the scientific community’s interest in these topics, we conducted an analysis of the number of published papers from the Scopus database. Based on keywords such as agriculture, deep learning, image, UAV, and remote sensing, 23,087 papers were obtained.

The number of publications in the area of precision and digital agriculture has been growing for many years (Figure 1A). The number of papers published in 2022 has increased more than 4 times compared to 2014. By the end of June 2023, 2366 articles had already been published. Among these publications, research articles dominate the body of research (55.9%), as can be observed in Figure 1B. Conference proceedings account for 34.6% of the total published papers, and review articles and chapters in books account for 3.5 and 3.4%, respectively.

6. Conclusions

The future of precision and digital agriculture is a highly complex issue. On the one hand, their daily use is expected to contribute to a real increase in farm income levels; these solutions will ensure that higher yields are achieved with reduced inputs. On the other hand, implementations must be carried out in such a way that they do not pose a threat to the environment. Artificial intelligence is forecasted to play an even more important role in the economic performance of farms. Its latest developments, especially in the areas of deep learning and machine learning, are expected to complement certain elements of precision agriculture that have not yet been developed. All data produced by farms, particularly via various IoT sensors, will be able to feed numerous algorithms based on deep learning and machine learning technologies. These algorithms will facilitate and optimize the operations of food producers by automating agricultural robots and monitoring crops and animals in real time while mitigating environmental impacts.