1.1 Introduction
Agriculture continues to remain fundamental to the global economy, with 60% of the world’s population relying on it for survival. The Food and Agriculture Organization (FAO) of the United Nations has stated that 5 billion hectares of land, which is 38% of the global land surface, is currently employed in agriculture and related activities. Though this figure seems large, each and every aspect of agricultural activities face numerous challenges, such as soil testing, efficient planting, controlling weeds, pesticide control, disease treatment, and lack of proper irrigation (Bannerjee et al., 2018). As such, agricultural industries are on the hunt for novel techniques to improve crop yielding and productivity in order to feed the rising population. Smart technologies such as artificial intelligence (AI), the Internet of Things (IoT), and robotics were incorporated into agriculture a few decades ago. They have led to a period of revolution in agriculture and recently have been paid more attention. Although the integration of this trio of smart technologies can maximize farming efficiency, there are some drawbacks that accompany the implementation and commercialization of such automation technologies (Talaviya et al., 2020). This review aims to revise the numerous desirable applications of AI, the IoT, and robotics in various stages of agriculture and present the major challenges and future recommendations for the successful implementation of advanced farming.
1.2 The Role of Artificial Intelligence in Advanced Farming
Artificial intelligence-based technologies support farming by increasing the efficiency of conventional farming and overcoming the challenges and drawbacks faced by traditional farmers. Artificial intelligence (AI) is the process where humans produce artificial machines similar to the human brain but with an ability to deal with larger amounts of data than the human brain. AI directly falls within the computer science field, but it should surpass this boundary to contribute to agriculture (Jha et al., 2019). Various technical devices and instruments have been developed based on AI that have been tested on agricultural fields and optimized. They have been successful in developing various field-steps of agriculture, such as soil testing, weeding, pesticide control, the treating of diseased crops, lack of proper irrigation to match the needs of crops, post-harvest activities such as storage management, optimising storage parameters, etc. Farmers have attained a high output as well as increased quality of output (Talaviya et al., 2020).
On the other hand, AI can be involved in agriculture to mitigate the environmental concern raised due to unfavourable agricultural activities, such as the heavy usage of pesticides, uncontrolled irrigation resulting in loss of water, and water being polluted with fertilizers. The implementation of AI would help in both these ways (Jha et al., 2019). There have been various AI systems proposed and developed by various scientists for various plantations in the past (Bannerjee et al., 2018).
The foremost objective of utilizing AI-based technologies is to reduce the labour force needed to achieve the required yield. Also, questions unanswered by humans are easily attended to by AI-based devices due to their ability to gather large amounts of data from governmental websites up to the real-time field data and analyse them. They can then provide suggestions to problems that would take a lot of time and high-end skills if they were to be made by humans. AI requires training with the biological skills of the farmer and vice versa; hence, farmers with the required skills will also need to be trained with these AI technologies (Talaviya et al., 2020).
1.2.1 The Fundamentals of AI Technologies Involved in Agriculture
The foremost step in involving AI in any field is machine learning. The data that needs to be processed should be fed in a machine-readable manner, and the processed solution should be delivered in a human language. As the AI-based machine processes the fed data, it should be able to gather information from the directed databases to meet the problem that has arisen. On occasion, real-time data would be needed for the AI to arrive at a conclusion, where the AI should be competent enough to read the real-time parameters. Weather prediction is an important factor needed to make decisions about the cropping season.
Chatbots are devices that virtually assist farmers with less experience of interaction with technologies by engaging them in conversations. Unmanned aerial vehicles (UAVs) are popular among governmental and institutional officers of farming for detecting any potential harm to the fields, such as the spreading of forest fires, pest invasions, pathogen attacks, and many more by geolocalization (Talaviya et al., 2020).
Neuro-fuzzy logic, fuzzy logic, expert systems, and artificial neural networks (ANNs) are four methods designed to solve problems (Jha et al., 2019). ANNs are the most common method utilized when designing AI-based technologies. An ANN simulates the processes within a human brain in a machine. In the brain, electric signals pass through neurons by axons and synapses. Various algorithms, such as delta-bar-delta, Silva, and Almeida, are used. The difference between conventional computer programmes and these algorithms is that this method allows the machine to perform an inbuilt task (Jha et al., 2019). A hardware-software interface should be built for the user-friendly functioning of the machine by farmers and other stakeholders. “Embedded systems” are machines into which software is fed.
1.2.2 AI in Crop or Seed Selection
High vigour, good germination, and the seedling emergence rate of seeds have always ensured emergence even under varying agricultural conditions and have been the key to optimising yields and ensuring uniformity in production (TeKrony & Egli, 1991). Traditionally, farmers have optimised seed choice based on experience, and any laboratory experiments for seed-choice optimisation are laborious and prone to error. The way that individual seed varieties react to different weather conditions and disease resistance, etc., are understood by AI technical devices by analysing the previous data to a greater extent than could be accessed by a general farmer.
SeedGerm is a phenotyping platform developed from automated seed imaging and phenotypic analysis based on machine learning. The core algorithm of SeedGerm has been developed with features such as background remover, feature extraction and germination detection, and measurements of traits. The hardware design of the SeedGerm system consists of a transluc...