Application of Artificial intelligence in Agriculture

Svetskiy Arseniy Vladimirovich ORCID: 0000-0002-0678-4841 Junior Researcher of the Environmental, Land and Agrarian Law Sector of the Institute of State and Law of the Russian Academy of Sciences 119019, Russia, Moscow, Znamenka str., 10 arseniy1107@gmail.com Other publications by this author

Currently, it is difficult to imagine any sphere of human activity without the use of modern technologies. The process of digitalization, which is intensively going on in all areas of human activity in modern society, has also affected the agricultural sector. As of 2019, the artificial intelligence (AI) market in agriculture was dominated by North America due to increased investment in research, development and widespread introduction of new technologies. Automated systems with the use of AI are increasingly being used.

Agriculture is the most important branch of the economy of any state. As of November 2022, the world's population is just over 8 billion people (URL: https://countrymeters.info/ru/World / (date of request: 17.11.2022)). In addition, according to UN forecasts, by 2050 the world's population should increase to 9.7 billion. At the same time, the area of land cultivated by the agricultural sector can only be increased by 4% by this point, and in order to feed the entire population of the planet, the amount of food must increase by 60% (URL: https://www.un.org/ru/un75/shifting-demographics (date of application: 17.11.2022)).This means that farmers will have to increase productivity and at the same time reduce production costs to achieve this goal. Setting such a goal has become a strong engine of progress in the agricultural sector, because traditional methods are becoming insufficient.

Intensification of agriculture may be the key to solving the problem of food shortages in the near future.That is why currently technologies related to the use of artificial intelligence are increasingly being introduced into agricultural production, and new directions of its application are also emerging and developing. This process can rightfully be considered part of the technological revolution in agriculture.

Artificial intelligence in the general sense is a human–created program code that uses algorithms that are able to learn and develop independently. Therefore, when mentioning artificial intelligence, one should always remember that this is the process of teaching a machine to analyze and collect a huge amount of data in the shortest possible time. AI is used both in pure software, that is, in software, and in the form of robotic systems that use various algorithms. According to forecasts, in 2025, the costs of states around the world for intelligent agriculture will increase by 3 times (URL: https://www.forbes.com/sites/louiscolumbus/2021/02/17/10-ways-ai-has-the-potential-to-improve-agriculture-in-2021/?sh=67 a 5 d 5 f 87 f 3 b/ (accessed: 11/18/2022)).Technologies based on artificial intelligence increase the efficiency of activities in all areas, as well as solve the problems faced by various industries, including the agricultural sector.

 Agricultural robots are created in order to ensure the effective use of AI in this sector. With the rapid growth of the world's population, the agricultural sector is facing a crisis in which the use of AI can play a decisive role. The use of AI-based technologies makes it possible to produce more products at lower costs and even improve the quality of products, as well as to ensure a faster release of manufactured products to the market. It is expected that the automated system at only one agricultural event by 2050 will process an average of 4.1 million discrete data units every day ^{[5, p. 2-3]}.

In scientific research in the field of agriculture, different authors distinguish different stages and stages of the agricultural production process, for example, forecasting, sowing, harvesting, etc. In turn, agriculture is divided into many different industries, such as forage production, crop production, mushroom farming, animal husbandry, fish farming, etc. These industries, in turn, are already divided into sub-sectors, which complicates the process of unification of equipment used at various stages of production. For example, in crop production, AI can be used in various processes: soil preparation, sowing seeds, adding fertilizers, irrigation, weed protection, harvesting, storage ^{[1, p. 2-3]}.

According to scientists, the global AI market in agriculture is currently developing in three most promising areas: computer vision, machine learning and predictive analytics.  It is expected that, in terms of technology segmentation, the computer vision category will occupy the largest market share in the global AI market in agriculture. Computer vision technology helps the farmer to detect the lack of nutrients in plants and monitor the health of the crop (URL: https://www.globenewswire.com/en/news-release/2021/02/02/2168016/0/en/Market-Size-of-AI-in-Agriculture-is-Projected-to-Reach-USD-2-400-Million-by-2026-According-to-Facts-Factors.html (accessed 21.11.2022)).Predictive analytics is a class of methods for data analysis that are aimed at predicting the future behavior of objects and subjects in order to make an optimal decision.

This class of methods is used in AI technology.

One of the most used and promising areas of AI implementation is monitoring the condition of plants and soil. The vital activity of plants directly depends on the chemical elements contained in the soil: macronutrients and trace elements. Their content is a critical factor for the health of crops, the number of plants per unit area and, accordingly, the overall quality of the harvested crop. Monitoring plant growth stages is also important for optimizing production efficiency. To make adjustments to improve the health of crops, it is necessary to understand the relationship between the process of crop growth and the environment, namely, the nature and degree of its impact on this process. The traditional method of determining the health of crops, livestock and checking soil quality is direct human observation and assessment. However, despite the indisputable necessity of its application, its accuracy is not devoid of a share of subjectivity.  Despite this, many agricultural producers resort to the introduction of modern technologies to solve this problem. So, they actively use specialized unmanned aerial vehicles (drones), which use computer vision to take aerial photographs, while capturing a huge array of data. These drones monitor the condition of crops and soil. The AI of visual sensing analyzes and interprets the data obtained by this method. This allows you to remotely monitor the health of plants, make a more accurate prediction of yield, as well as detect a lack of fertilizers or the presence of diseases in crops in large areas.

Due to the huge areas of crops, it is relevant to use AI models that are able to inform farmers about specific problems in order to take the necessary measures as soon as possible.

During the period of grain crop growth, there is also a need for a direct inspection of the condition of the ears, which is a time-consuming process, but extremely necessary for crop quality control and timely detection of diseases. However, this process can also be carried out with the help of AI. The researchers were able to solve this problem by forming a database containing images of wheat ears at different stages of growth. The data collection process took place over 3 years, the images were taken under different lighting conditions so that the AI could take into account more factors.

This computer vision model surpassed human observation in accurately determining the stages of wheat growth, which made it possible to save resources, since there was no need for daily trips to the fields to inspect the crop (URL: https://www.v7labs.com/blog/ai-in-agriculture#h1 (accessed: 19.11.2022)).There are many companies that produce various equipment, software and other solutions to help agricultural producers.

One of these companies is Trace Genomics, founded in 2015. She has developed an artificial intelligence system that allows analyzing the composition of farmland soils. The system is a complete solution in the form of equipment with software for analyzing soil samples. This type of application helps in monitoring the state of soil and crops, which contributes to the cultivation of healthier crops with a higher level of productivity ^{[2, pp. 8-15]}.

The classical method of soil assessment is the excavation of samples and subsequent analysis of their composition in the laboratory, which, in turn, is a rather time-consuming and resource-intensive process. To optimize and automate this procedure, scientists have developed an algorithm that allows using data obtained from an inexpensive portable microscope to train artificial intelligence. Then the AI analyzes the color of the fruit in the field and compares it with the color of a fresh market product according to the specified parameters. This algorithm helps agricultural production to estimate yields remotely, as well as to sort fruits ^{[3, pp. 44-47]}.

In 2016, a study was published on the creation of a computer vision model that should determine the texture of the soil and the organic substances contained in it (SOM). The computer vision system was able to assess the content of sand and soil organic substances in the sample with an accuracy comparable to expensive laboratory processing. The introduction of this technology into production should simplify the work of the farmer, replacing manual labor with crop and soil monitoring using modern digital technologies ^{[4, p. 41]}.

Another area of application of artificial intelligence in agriculture is the determination of tomato maturity using a computer vision model. The researchers managed to create an algorithm that analyzes the color of five different parts of a tomato, and then based on these data assesses its maturity. The algorithm has reached the level of successful detection and classification of 99.31%. Reassessing and evaluating the growth of fruits and their maturity is hard and time–consuming work for humans, but AI proves once again that it is able to do most of this work with ease and accuracy.

AI computer vision can not only detect and analyze crop maturity or soil quality, but also apply image recognition technology for automated detection of plant diseases and pests on them. This technology uses methods of classification, detection and segmentation of images to create computer automated systems that can monitor the health of plants. An example of this in practice is the study of black rot on apples. The researchers trained a neural network using images of black rot of an apple tree, which were annotated by botanists according to the four main stages of severity. An alternative to the use of AI in this case requires a lot of time-consuming search and assessment of the degree of damage directly by a person. The developed AI model in this study was able to determine and diagnose the severity of the disease with an accuracy of 90.4% (URL: https://www.frontiersin.org/articles/10.3389/fpls.2020.00898/full (accessed: 19.11.2022)).In order to minimize human labor in such an agricultural industry as cattle breeding, CattleEye has started using AI technology to monitor the health and behavior of cattle.

This makes it possible for employees of agricultural enterprises not to be in close proximity to animals to assess their health. When a problem is detected, responsible employees immediately receive a notification from the system. This technology can be applied not only to assess the condition of cattle. For example, the algorithm has been adapted for poultry farms to view video data and determine the behavior of chickens: how they eat and move, which allows you to assess their health status (URL: https://www.v 7 labs.com/blog/ai-in-agriculture#h 3 (accessed: 11/20/2022)).

Another area of AI use in agriculture is its application in various technologies for weed control. The spraying of pesticides or the distribution of fertilizers over the entire field area can be automated through the use of drones equipped with artificial intelligence. Computer vision on drones allows you to recognize a specific target area for spraying pesticides or distributing fertilizers in real time. This reduces the risk of contamination of the territory with substances harmful to the environment. Although the potential here is great, there are still some challenges at present. For example, spraying a large area field is much more efficient using multiple UAVs, but assigning specific task sequences and flight paths for individual vehicles can be a difficult task. So, complex management of several devices at once is quite a difficult task.

An example of the use of artificial intelligence is the use of "smart" spraying systems. Researchers from the Virginia Institute of Technology have developed an intelligent spraying system based on the use of servo-motor sprayers that use computer vision to detect weeds. The camera mounted on the sprayer, when a weed plant is detected, records its geolocation, analyzes its size and color in order to calculate and enter the exact amount of herbicides necessary for its destruction. Thus, the destruction of weeds is carried out more efficiently and without causing damage to cultivated plants (URL: https://www.researchgate.net/publication/324174493_Computer_vision_A_promising_tool_for_weed_management (accessed: 20.11.2022)).Sprayers with AI are not the only way to eliminate weeds.

Modern agricultural enterprises use AI systems that allow weeding without using herbicides, which, in turn, makes this approach more environmentally friendly. The ability to physically remove weeds not only saves agricultural employees a lot of effort and time, but also reduces the need for herbicides and, thus, makes all agricultural activities less harmful to the environment.

It should be noted that machine learning is, in fact, a class of artificial intelligence methods. Machine learning is based on algorithms that use the provided data for training and forecasting. Machine learning is used in cases where the use of precise algorithms will not give sufficient flexibility in performing the task. In the process of weeding, the ability to detect objects and the ability of AI to identify weeds and distinguish them from crops of cultivated plants is important. Computer vision algorithms are used together with machine learning to create robots that perform automatic weeding.

The Deepfield Robotics startup of Bosch, Amazonen Werke, together with the Osnabr?ck Technical Institute, were able to design a BoniRob robot. It is an autonomous field robot for experiments on the processing of individual plants, equipped with an independent navigation system, as well as capable of mapping the work carried out, preparing documentation, including the creation of a database of statistical data.

The robot learns to distinguish between weeds and crops by adding images of leaves of various sizes, shapes and colors to its database. By analyzing the data embedded in it and comparing it with what it sees in real time, BoniRob can move around the field, destroying unwanted plants. Experimental results of the application of this system show that its classification of plants and the weeding rate have a high success rate, which is at or above 90% (URL: https://robotrends.ru/robopedia/bonirob (accessed: 20.11.2022)).AI is able to use images not only from drones, but also from satellites, which, in turn, helps to control both crops in crop production and animals in cattle breeding.

Due to the implementation of round-the-clock monitoring with the help of artificial intelligence, farmers receive timely information about the slightest deviations, without wasting time and other resources for monitoring. The accuracy and efficiency in the use of pesticides is ensured by aerial photography. This allows you to save finances and preserve the environment, since pesticides are sprayed more accurately.

AI functions are not limited to those described above. Visualization algorithms that identify defects, diseases and pests can be used for qualitative assessment and sorting of the harvested crop. To do this, the program compares the shape, color and volume of fruits and vegetables with the standard indicators embedded in it. At the same time, the automated evaluation and sorting process is much more accurate. Thus, AI is able to completely replace manual labor in sorting vegetables, while sorting according to the specified criteria (URL: https://elibrary.asabe.org/abstract.asp?aid=47686 (accessed: 20.11.2022)).

Another application of AI in the agricultural sector is the use of surveillance systems based on artificial intelligence and machine learning to monitor the terrain by video surveillance in each field in real time. This reveals violations such as illegal access to the territory of the agricultural enterprise. The system immediately sends a warning, which helps to identify illegal actions online.  Given the rapid development of video analytics, fueled by artificial intelligence and machine learning algorithms, anyone engaged in agriculture can more effectively protect their fields and building perimeters from illegal entry. Video surveillance systems with artificial intelligence and machine learning are also easily scaled for both a large agricultural enterprise and a separate farm. Surveillance systems based on machine learning can be programmed or trained over time to distinguish objects (people, vehicles) whose access to the territory is authorized from unauthorized ones. Twenty20 Solutions is one of the leading companies in the field of video surveillance equipped with AI. The company shows its effectiveness in ensuring the security of remote facilities, deterring violators through the use of machine learning to identify authorized and unauthorized facilities on the territory of an agricultural enterprise.

The use of robotic technology helps to solve the problem of shortage of personnel, and also provides an element of security around the perimeter of remote locations. Programming self-propelled robotic equipment for the distribution of fertilizers for each row of crops helps to reduce operating costs and further increase the yield of fields. Every year agricultural robots are trained to perform more and more complex tasks. An example of this is the VineScout robot, controlled by an advanced artificial intelligence system developed by the European Union research Consortium. This robot uses the input signal from a machine with a three-dimensional stereoscopic vision system and ultrasonic sensors to follow the rows without colliding with anything, turn and move when moving from one row to another. In addition, it can work around the clock, while collecting up to 72 thousand bunches of grapes per day. When VineScout moves, it uses an infrared sensor and a multispectral camera to measure the temperature of plant leaves and the amount of water contained in them. This data is displayed in the form of a harvest map, which allows producers to find out whether plants are getting enough water at the same time as assessing the current level of fruit maturity (URL: https://robroy.ru/novyij-i-uluchshennyij-agro-robot-budet-rabotat-v-vinogradnikax.html (accessed: 11/20/2022)).

The agricultural industry is facing various problems, such as the lack of effective irrigation systems, weeds, problems with monitoring plants due to crop height and extreme weather conditions. But with the use of new technologies, it is possible to increase productivity and thereby ensure their effective solution.

Taking into account the possibilities of using artificial intelligence technology in agriculture, it can be concluded that digitalization will: reduce possible risks by predicting climate change and use computer vision to detect diseases of agricultural crops and livestock. Automation of the process of weeding, spraying with pesticides, as well as harvesting is another opportunity to increase the quantity and quality of products that are opened using drones, robots, sensors and scanners, as well as other hardware and software solutions.

Although agriculture is the least digitalized industry, but today agricultural enterprises have begun to use methods of precise weeding using AI. This method avoids the loss of a large amount of crop during the weeding process. Such autonomous robots not only increase efficiency, but also reduce the need for unnecessary pesticides and herbicides. In addition, agricultural producers can effectively spray pesticides and herbicides at their enterprises using drones, as well as monitor the condition of plants and livestock.

Machine learning is constantly evolving, making more complex decisions based on the data it processes. At the same time, there is a clear probability of an unforeseen or unfavorable outcome in the absence of human control. Despite its value, AI technology can cause damage in the event of unauthorized access and modification of data in the database, failure of autonomous systems, etc. For full implementation and optimization of use, it is necessary to make certain amendments to the current legislation. Due to the novelty of the technology, the legal regulation of AI is not sufficiently developed, many relationships arising in the process of using AI do not have a clear regulation. This issue is not sufficiently regulated not only in the national legislation of the Russian Federation, but also in international law and the legislation of foreign countries. The problem is also connected with the lack of a unified approach to determining the characteristics of artificial intelligence. In 2019, the Decree of the President of the Russian Federation No. 490 dated October 10, 2019 "On the development of artificial intelligence in the Russian Federation" was issued, which gives the concept of artificial intelligence. In accordance with this regulatory legal act, artificial intelligence is a set of technological solutions that allow you to perform specific tasks and obtain results comparable to human intellectual activity. This concept is seen as general and incomplete, since AI includes various classes of methods containing both data replenishment for human learning and an independent algorithm using the Internet to expand its database.

In the fall of 2019, UNESCO member countries decided to develop an international document containing recommendations on ethical aspects of the use of artificial intelligence. In November 2021, 193 UN member states signed an agreement on artificial intelligence. This document is the first universal standard in the field of AI application. It defines the common values and principles necessary to ensure the safe development and use of AI. In addition to this recommendation, there is a guide on data protection when using AI, adopted by the Council of Europe on January 25, 2019. This international act establishes a number of recommendations for the government, the AI developer, the manufacturer and the service provider. These recommendations serve to ensure human rights and freedoms, in particular, the right to data protection ^{[7, pp. 69-70]}.

One of the key issues related to the legal regulation of the use of AI is the problem of liability for unauthorized actions of robotic devices endowed with artificial intelligence. The question of who should be responsible for the actions of a robot, especially an autonomous or intelligent one, is both urgent and complex. Approaches to solving this problem have not yet been found ^{[6, pp. 32-35]}. The use of AI technology entails certain risks associated with equipment failure, system errors, hacking of drone control systems, etc. Therefore, it is necessary to have a detailed legal regulation of the use of AI not only in the agricultural sector, but also in all areas of human activity.