19.1 Introduction

Population and agriculture both are dependent on each other. With every addition in population, the demand for food increases, and simultaneously the pressure on the agriculture sector is increased too. As per the World Bank report, around 40% of the population is still dependent on agriculture as a source of livelihood. Its contribution to GDP is up to 30% in the low-income countries, whereas around one third of the land is used for the purpose of agriculture [1]. The global population is expected to increase by 10 billion at the end of the year 2050 [2] but in the meantime the agriculture land is not going to increase but will simultaneously decrease with the increase in the population. The World Bank report of 2020 says that around 690 million people around the world are hungry or do not have adequate food arrangements. It means more pressure will pile up on the agriculture sector to feed the increased population. It is anticipated that the global food production should be enhance by 60–110% to meet the food requirement of 9–10 billion people by the end of 2050 [3]. How can this pressure be reduced is a major concern for governments globally and majorly for the growing economies is a big question? Lot of research has been conducted in this context to provide a way for sustainable food system [4]. Some researchers have focused on the more technological application in this sector to enhance productivity [5], while many have talked about the organic farming to enhance the quality of agriculture [6]. Many researchers are of the view that agriculture revolution is required [7, 8] unlike the industrial revolution, while many researchers believe that no such revolution is required, but the existing farming techniques should be further improved to raise the level of the current practices [9].

The real question arises that how we can enhance the productivity of our agriculture farms [10, 11]. What can be done to reduce the cost of agriculture and increase the productivity is the major cause of concern [12] for the developing economies and majorly the populous economies. Here comes the application of technology in this sector where the developing economies are lagging behind [13]. By efficiently applying the technology of Artificial Intelligence and machine learning [14], the rule of the game can be changed. The developed nations have already voyaged a long journey in the application of AI and Machine learning in the agriculture sector [15] but the developing nations and poor countries have a long road ahead to follow.

The evolution of new technologies has brought a drastic change in the industrial sector globally [16]. The use of AI and machine learning has transformed the manufacturing sector, healthcare services, banking operations, transportation services to name the few. It has also started playing an important role in our daily activities and the environment around us [17]. If we analyze the adoption of AI and machine learning in the agriculture sector, we find the momentum is slow as far as development and commercialization of farming technologies [18] is concerned mainly in the developing countries. Figure 19.1 reflects the slow growth of various crop yields over the years around the world. The AI and Machine learning have got huge scope for application in the agriculture sector. It is creating new opportunities and providing operational environment in the agriculture sector along with the big data technologies. The recent studies are providing evidence of how Machine learning and big data can improve the management of farms [19, 20]. The Agriculture technology and precision farming are also known as digital agriculture, new scientific areas are evolving where application of data is made to enhance productivity in agriculture and reduce the impact on environment. The smart farming techniques can be applied where sensor technologies are being used for the collection of data from the soil, crop or other allied areas of agriculture. Using the sensor data and application of Internet of Things and machine learning techniques, a farmer can easily predict the weather conditions, monitor its crops, predict the future yield and protect the crops from diseases through early detection [21].

The various activities related to adoption of AI and machine learning has already started from a couple of years in the various countries of the world. The Governments of China, USA, Japan and France have already released their policy paper related to Artificial Intelligence. Artificial revolution in India may mark its presence very soon in the form of the adoption and creation. India cannot be parted from the digital inclusion and will depict undulating effect on economy. Considering its importance and to become a global leader the Government of India has also brought a discussion paper on the National Strategy for Artificial Intelligence in June 2018 [22]. From the application point of view, the approach is to find out the sectors where there is a huge potential for the application of AI and Machine learning. Agriculture sector which forms the backbone of Indian economy has been identified as one of the areas where there is enormous scope for the application of AI and ML.

19.2 Overview of AI and Machine Learning

The computers were first used in the field of agriculture in 1983 [23]. Then onwards different approaches have been suggested by various researchers to find the solutions to the existing problems in agriculture. Though the term AI was coined long back in a conference held at Dartmouth in 1956 by John McCarthy, its first use in crop management was proposed in 1985 by McKinion and Lemmon in their paper “Expert systems in Agriculture” [24]. Using Artificial intelligence machines acquire the capability to learn from experience, adjust themselves to new inputs and execute tasks in the way the human intelligence does. The AI techniques can apprehend the multifaceted details of every problem and suggest the best suitable response for that specific situation. Applications using AI prove to be exceptional performers in terms of precision and robustness by progressively learning about the data they are processing.

Machine learning inherits from AI, the mechanism of acquiring the input, process and generate output both in the training and operation phases [25, 26]. The relationship between AI, ML and deep learning is represented in Figure 19.2 given below.

AI is a broader domain comprising of machine learning as a subset, whereas machine learning is considered as a superset of deep learning. Machine learning is used as a tool to explore, understand and recognize patterns in the input dataset. This technique can be used to precisely program the computers to automate the tasks that would rather be impossible for humans to complete. Once the learning is over, the machines can predict values and take decisions with minimal human interventions.

Machine learning algorithms can be broadly divided into four categories: Supervised, unsupervised, semi supervised and reinforcement learning. Supervised machine learning is trained to make decisions in terms of classification or prediction based on example input-output pairs. Such pairs comprise of an input object and a desired output value. The calculated outcomes are compared with desired output values and an acceptable level of accuracy is achieved by tuning the parameters of the internal mapping function through iterative learning. Random forest, Support Vector machines, K nearest neighbour, Naïve Bayes, artificial neural networks are some of the popular supervised machine learning algorithms.

Unsupervised machine learning intends to capture the correlation and draw inferences from datasets where no information about desired outputs is provided. Since there are no labelled responses to which the input data could be related, unsupervised algorithms recognize commonalities among the elements in the input dataset and classify the elements to gain meaningful insights. Hierarchical clustering, principal component analysis, k-means and association rules are examples of unsupervised machine learning algorithms.

Semi-supervised machine learning combines both supervised and unsupervised machine learning methods, where semi-supervised algorithm learns from both labeled and unlabeled data. In situations where for example only a small labeled dataset is available, supervised learning algorithms can determine classification rules from labeled data and then this data can be utilized to label the residual data by using unsupervised machine learning algorithms. Speech analysis and Internet content classification are some of the useful application areas of semi-supervised machine learning.

Reinforcement learning aims at training models to take decisions based on environmental feedback and reward mechanisms. The software agent practices trial and error method to find a solution to any problem. To train the machines to do what is desired, the AI is rewarded or penalized for the actions it performs. Although the reward policy is set, the model must figure out how to complete the task using random trails and at the same time maximizing the rewards. Policy-gradient algorithms, temporal-difference learning, and Q-learning are some of the popular reinforcement machine learning algorithms.

Deep learning, which is a subset of machine learning imitates the functioning of the human brain in processing data for recognizing speech, object detection, language translation, etc. Deep learning takes an edge over previous machine learning algorithms as the former uses a composite structure with many nodes, hidden units and learning algorithms which thereby decreases training time and improves quality of learning. This has been realized by the advancement in computer processors and storage devices which are necessary to accommodate the gigantic, parallel neural networks essential to perform machine learning. Fuzzy logic and Artificial Neural Network (ANN) are grounded on the idea of deep learning.

19.3 Review of Literature

The agriculture sector is facing a drastic change and is moving from the traditional techniques to technological procedures. The application of AI and machine learning with the use of various algos had evolved new technologies which are extremely useful in increasing the productivity of the crops, decreasing cost of produce and saving the environment from the excessive use of hazardous chemicals. Table 19.1 displays the research contributions using various algorithms, the technologies evolved with their use and the benefits to the farmers from such evolution.

19.4 Application of AI & Machine Learning in Agriculture

Over the past few years various Agri-tech business models have incorporated the technologies in the form of Artificial Intelligence, machine learning, robotics, computer vision and natural language processing [35] Table 19.2 shows how the countries are being segmented on the geographical basis as regards to the application of Artificial Intelligence. The application of AI and Machine learning is turning out to be commercially beneficial for agriculture in recent years due to advancement in big data analytics, cloud-based storage, etc. which helps in reducing the cost in satellite imagery and remote sensors. The application of AI and Machine learning is becoming useful in Soil management, water management, Yield prediction, crop quality management, disease detection, weed detection, agriculture supply chain management, livestock management, Species breeding, species recognition and so on [36].

Figure 19.3 clearly depicts the growing application of AI in agriculture, where the major chunk is with the USA followed by Europe. Both have been a front runner since the beginning of application of AI and Machine learning in agriculture.

The companies like Deere & Company (John Deere) (US), Microsoft Corporation, Descartes Labs, Inc. (Descartes Labs) (US), ec2ce (ec2ce) (Spain), etc. are playing the major role in agriculture technology. They are consistently finding the scope and sprouting the use of AI and Machine Learning in agriculture.

1. 1. Deere & Company (John Deere) (US): The Company has made some big moves in agriculture using AI and Machine learning. The company has become one of the biggest users of cloud computing services in the world. The company is capable of collecting the per second measurement between 5 and 15 million from the machines connected in the whole world. The company has more than 150 million acres in its database. It uses petabytes and petabytes for storage of data. In 2017, it acquired Blue River Technologies by paying $305 million for the control of weeds. This technology makes the use of computer vision and machine learning and can reduce the use of herbicides by 95% and also improves the yield [38]. The herbicides are sprayed only at places where weeds are present, not to the whole crops. This helps in optimization of inputs used in farming, which is a basic philosophy behind precise farming.

   

2. 2. Microsoft Corporation: It is one of the world’s biggest software companies marking important contribution in digital agriculture. In India, the farmers are using AI for increasing the yields of their crops. It is providing technologies like Artificial Intelligence, Satellite Imagery and Advanced Analytics and Cloud Machine Learning for serving the small farmers in increasing their crop yields. In collaboration with the International Crop Research Institute for the Semi-Arid Tropics (ICRISTAT) Microsoft Corporation has technologically advanced an Artificial Intelligence Sowing Application (AISA). AISA is useful in facilitating information to the farmers the exact date to sow the seeds. AISA is a low-cost service provider on a simple phone, which has the capacity to receive text messages as visible from Figure 19.4. This small timing advisory helps in surging the yields up to 30% [39].

   The company has also developed the Pest Risk Prediction Application using AI and Machine Learning in collaboration with United Phosphorous Ltd (UPL). This application is useful in providing information about the possible common pest (Thrips, Aphids, Jassids, Whitefly, etc.) attacks to the famers well in advance, for preventive action.

   Microsoft has made a multiyear strategic alliance with Land O’Lakes for new innovations in technology in agriculture to boost the sustainability practices in agriculture with major focus on rural America. The trustworthy cloud technology and Artificial Intelligence capabilities of Microsoft will provide solutions to the farmers for sustainable agricultural practices. It has also announced the investment in South Africa for big use of technology in agriculture sector. The technological application in agriculture will be in the form of Unmanned Aerial Vehicles, Internet of Things and Remote Sensing for collection of data from farms. The Sensors will be capable of collecting data about the soils while Unmanned Aerial Vehicles can gather imagery [40]. The application of AI systems and techniques like Computer Vision Systems & Machine Learning will make use of this data to study the patterns and make forecast and help in making decisions. The computer Vision and Machine Learning both are valuable in predicting and measuring crop yield. The information attained from remote sensing along with robotics and machine learning, can be pragmatically used for observing and responding to variability in crops [41].

3. 3. Descartes Labs: The Company just started six year ago in 2014 and started providing data on corn yield estimates. Its accuracy rate was much appropriate as compared to USDA’s estimates, having an average error margin of just 2.5% through using more precise algorithms.

   It uses machine learning techniques and watches tons of pixels through the satellite and informs what is being grown. Figure 19.5 displays the satellite being used by the company for obtaining information and pictures about every aspect of crops being grown in the farm. It uses the technique spectral information and measures the level of chlorophyll which cannot be visible from our eyes. It analyzes the data received from the satellite of every farm in the US on day to day basis and updates the prediction of corn yields in every two days where as USDA updates this data once in a month. It gives the real time geospatial data catalog & flexible modeling by using the cloud based geospatial analytics platform.

4. 4. Ec2ce Spain: The company has developed a tech tool which uses Artificial Intelligence for the agriculture sector, which provides forecast on the various areas like yield parameters, management of fertilizers, spreading of pests, irrigation etc. on the basis of agricultural data obtained from different sources, which assist the farmers in decision making. These technological upgraded models give forecasts and optimizations using smart tools for processing of data and make available smart solutions for the purpose of making decisions in the agriculture sector. The use of predictive models for the purpose of agriculture management increases the productivity of the farms and increases profitability. AI and machine learning when applied for the purpose of integrated pest management, predict the spread of pests thereby reducing the consumption of pesticides and protects the environment.

There are many more companies making serious contributions for the agriculture sector globally using AI and Machine learning techniques.

19.5 Current Scenario and Emerging Trends of AI and ML in Indian Agriculture Sector

India is the second most populous country in the world with a population of 1.3 billion. The growing population required food to feed them. In terms of the land area India is the seventh largest country. Approximately 67% of the Indian population depends on agriculture [43]. The sudden outbreak of COVID-19 has created the regional imbalance as large population has shifted to their native places, thus putting huge pressure on agriculture at some places as well as shortage of labor at some places. Although the Indian agriculture sector has made sufficient progress and there has been continuous focus of the various government in the promotion of this sector but still it relies on unpredictable variables like inadequate supply chain and low production in comparison to the developed nations (Table 19.3).

There is a continuous degradation of land, the fertility of soil is decreasing, there is increased dependency on chemical-based fertilizers for improved production and yield, decreasing level of ground water and pesticides resistance are among the important barriers in the rapid growth of the sector. The traditional techniques of agriculture whether irrigation, sowing of crops, use of pesticides, harvesting, etc. require to be supported technically and necessitates upgradation of agriculture methods to fulfill the growing demand of food in the country [44].

Across the world, digitalization and AI technology are proving helpful in solving the issues related to value chain. India had approx. Approximately 30 million farmers who used smart phones, which was expected to increase three times by the end of 2020. As per the study conducted by Accenture, digital farming and other related farm services will affect 70 million Indian farmers, which are expected to add more than $9 billion towards farmers’ income in 2020. The Government of India has identified the relevance of AI and Machine learning for agriculture, and NitiAyog has come up with the national strategy on it [22]. The latest technological development in the field of agriculture using AI and Machine learning has abundance of opportunities for this vulnerable sector as depicted from Figure 19.6.

1. Monitoring of Soil Health: Health of soil is the critical element for the production of crops. Soil health monitoring can be done by means of image recognition and deep learning techniques without the requirement of laboratory testing infrastructure. The remote satellites are used for the purpose where Artificial intelligence solutions are combined with data signals. The local images of the agriculture farms are captured, which helps farmers in taking immediate actions for restoring the health of soil. The Berlin-founded agricultural tech company has started PEAT in India which has developed a deep learning application known as Plantix, which is capable of finding out the possible defects and deficiencies in the nutrition of the soil. The analysis can be done by using software algo, which relate the specific foliage trends with specific soil defects, pests and diseases in plants. The image recognition application finds out the likely defects from the images taken from user’s smartphone camera. The users are provided the soil renewal techniques, guidance, and other necessary solutions.
2. Monitoring of Healthy Crops and Advisory for Timely Action: The agriculture sector of India is susceptible to climate change due to huge dependence on rain. There is a variation in the pattern of weather like increase of temperature, variation in the levels of precipitation and density of ground water, which distress the famers mainly in the rainfed areas. AI can successfully be applied for issued advisories regarding sowing, control of pests and input control for increasing the income of the farmers. By using the remote sensing technique, farmers can monitor the health of vegetation and moisture of soil for individual farms. The monitoring of crops and supplementary information to the farmers about their farms can be provided using remote sensing data, weather data through Artificial Intelligence technology and other platforms of AI. Microsoft has developed an AI Sowing app which issues advisories to farmers about the best possible sowing date, use of fertilizers based on soil requirement, treatment of seeds, the depth to which seeds can be shown and many things more [46]. The best part of this app is that it can be used in a simple mobile phone. AI is being used for the optimization of the herbicides. For this purpose, Blue River Technology has prepared a unified computer vision and machine learning technology which helps the farmers to lower the use of herbicides by spraying at places where weeds exist.
3. Crop Insurance: As agriculture sector is adversely affected due to natural calamities and farmers are the only risk bearers. The insurance of crops will cover the risk of farmers from such natural calamities. In India Pradhan Mantri Fasal BimaYojna has been initiated using the AI technology. In this pilot studies are being carried out in various states to know the optimization of crop cutting experiments (CCEs). The study uses the data received from the satellites and modeling tools in order to reduce the number of CCEs needed for insurance unit level for estimation of yields.

   

4. Smart Irrigation: The proportion of Indian population is 17% but the holdings of the fresh water reserves are just 4% [47]. This water is also not evenly distributed in the country. The Indian agriculture sector accounts for 90% of water use due to improper irrigation system. According to the OECD environment outlook 2050, India will face severe water crises by the end of 2050, the signs of which can be seen in various districts during summer season [48]. The inappropriate irrigation contributes more to poor quality of crops. So, a smart management system is needed to enhance the productivity of crops.

   The smart irrigation system uses devices based on Internet of Things, which can mechanize the process of irrigation by investigating the moisture in soil and the climatic conditions. The use of water for agriculture determination (Figure 19.7) can be improved by using thermal imaging cameras which regularly monitor that crops get enough water. KisanRaja provides cloud based IOT solutions using wireless sensors, wireless valve controllers and mobile pump controllers for intelligent automatic irrigation system. Using this system farmers would not only save huge quantity of water, whereas 25% saving would be thru reduced power consumption and 30% reduction in maintenance cost of pumps and will increase their farm productivity up to 15%.

5. Smart Farm Management: The smart farm administration identifies the farms for weather and information about field, inspects the possible risk, and encapsulates precise location and size of the farm, details of farmers and crops, since the pre-harvest stage. SmartRisk is an Artificial Intelligence and Machine Learning based foretelling solution, which collects the historical data of the crops in a farm for the duration of growth cycle. Both SmartRisk and SmartFarm make sure that the farm is monitored from the starting of the cultivation cycle. In India CropIn is providing such services and is being able to transform farming by using technology in the day to day operations in the field. The most important part of it is that solutions are provided in the local languages, which benefits the farmers to comprehend the crops types, yields and the diseases which may harm them.

19.6 Opportunities for Agricultural Operations in India

The share of agriculture and its related sector contribution is less than 17% in India’s $3 trillion economy. The average annual income of the farmers is around Rs. 99,976 [49]. The output to input ratio is quite low in the country as the farmers are having tough time due to increasing input costs, decreasing productivity, changes in the climate, shortage of water, poor access to the market and lower use of technology. Such issues can be successfully addressed by the augmented use of technology in the agriculture sector. The momentum for using AI and Machine learning in Indian agriculture sector has already begun with the establishment of INDIAai by the Government of India. Both the government and private sector necessitate developing an AI & machine learning based ecosystem in India.

The application of AI for monitoring health of soil can lead to reduction in the cultivating costs and improved crop productivity resulting more revenue for farmers. The use of AI based soil monitoring system can reduce the consumption of chemical fertilizer by 40% [50]. The spatial assessment abilities of geographic information system technology are advantageous in efficient management of water for irrigation purpose. This technology has an ability to reduce the consumption of water by 20% and increase productivity up to 37.50% per acre [51]. In India KisanRaja has already been put into practice for such drive. There is enormous scope for the application of these technologies in India where water crises are taking tiny forthcoming disparaging steps.

The Indian farmers sow the crops on the basis of time period but for the past several years there has been change in the climatic conditions. AI can be efficaciously advantageous for this purpose. It communicates the farmers the exact time for sowing the seeds, procedure of sowing, usage of fertilizers, etc. Microsoft has taken good initiative and developed the technology and now information can be shared with the farmers from a simple mobile phone. The application of AI for these purposes can increase the farm productivity up to 30% [52].

Weeds and pests’ control have been a severe problem for the Indian farmers. Approximately 90% of the crops production is damaged due to incursion of weeds [53]. The pests are capable of incurring losses up to 19% of the crops [54]. To control the pest difficulties farmers are making the heavy use of fertilizers, which makes the land and ground water poisonous. AI is useful in controlling the weeds and management of pests and can reduce the use of pesticides and weedicides up to 80%. The Blue River technology has ample of scope in India as it can control 90% of the weeds and the use of weedicides is reduced from 20 to 2 gallons per acre [55].

The cost of labor in agriculture is among one of the important components of cost of production. Increased use of various agricultural machines and equipment’s has reduced the use of labor and thereby reducing the cost of producing the crops. As per the study by ICRISAT for the period 2007– 08 to 2014–15, the use of labor in the production of cotton per hectare was decreased by 43% i.e. from 153 to 87 labor days. During the same period, the use of labor to produce soybean per hectare decreased by 58% i.e. from 55 to 23 labor days. Similarly, in the production of pigeon pea labor days decreased by 52% from 48 to 23 labor days. Similarly, the cutting of crops like wheat and rice required large number of labors and time but though use of machines, it can be effortlessly done within less time period [56]. So, we can say that application of AI and Machine learning can directly contribute in the reduction of direct labor costs as well as the time involved for such activities.

In the Global Hunger Index India’s stands in the 103 position among 119 countries of the world, so the wastage of agri-based products is a serious cause of concern. As per the reports 16% of the fruits and vegetables are wasted every year due to poor quality supply chain. Out of the total production of oil seeds, cereals and pulses produced in India, 10% is completely wasted, again because of the same reason [57]. The infrastructure of supply chain management in India requires to be strengthened. The policy makers are still not able to manage aforementioned issues and the results are that farmers throw their produce in the fields or even do not harvest their labor intensive crops from the fields. The initiative like Jivabhumi food print using block chain technology provides a digital market place to connect with the institutional buyers and farmers [58].

19.7 Conclusion

Agriculture contributes an imperative share in the global economy. Due to increasing population there is a continuous burden on the agriculture sector to intensify productivity and cultivate more corps within the same agricultural land. The use of Artificial Intelligence and machine learning is going to be a game changer for the agricultural sector. The world agriculture production increased by 1,225 million tons in between 1961/63 and 2005/2007 and is further expected to increase up to 3 billion tons till the end of 2050 [59]. To meet the food demand of 10 billion population by the end of 2050 globally, more technological innovations are to be made and applied, as the agricultural land is going to reduce with the increase in population. According to the United Nations Report, the population of India is expected to be more than China by the end of 2027. India will add more 273 million people between 2019 and 2050 and will continue to be the most populous country of the word in the current century [60]. The total gross cropped area increased from 131.89 ha in 1950–51 to just 198.36 ha (Provisional) in 2014–15 [61]. The developing country like India has ample of opportunities for enhancement in the productivity of agriculture produces using AI and Machine learning. In India NitiAayog has already issued a discussion paper in 2018, suggesting how the AI solutions can contribute for the agriculture sector. If the Indian farmers are exposed with the best practices, there could be a next revolution in the Indian agriculture sector after green revolution, which can be termed technological revolution. The various sub areas of Artificial Intelligence like Machine Learning, Computer vision, Deep Learning, Cloud Computing, Internet of Things, and Expert systems are among the important tools which have ample of scope in solving the complex problems of farmers like soil management, irrigation management, sowing of crops, smart farm management, control over weeds and pests and the most important supply chain management.