In recent years, artificial intelligence (AI) techniques have been adopted extensively in farming to manage pests, monitor growing and soil conditions, and cultivate healthier crops. AI is also assisting farmers in selecting the optimum seed for a specific weather scenario and by providing data on weather forecasts. This article discusses the increasing role and applications of AI techniques in agriculture.
Image credit: Suwin/Shutterstock
Importance of AI in Agriculture
AI has received significant attention in agriculture owing to its robust applicability in problems that cannot be solved effectively by humans and conventional computing structures. Currently, several issues adversely affect the agricultural sector, including disease and pest infestation, inadequate chemical applications, inaccurate yield prediction and improper irrigation, weed control, and drainage.
In dynamic domains such as agriculture, situations cannot be generalized to obtain a common solution. Thus, AI systems with excellent robustness and accuracy must be used to resolve the existing issues in agriculture. AI techniques can capture the complex details of every situation and offer the most feasible solution for that specific issue. Different AI techniques, such as fuzzy systems, expert systems, and artificial neural networks (ANNs), are increasingly used to solve complex issues in agriculture.
Applications of AI-based Tools in Agriculture
General Crop Management
Crop management systems primarily provide an interface for overall crop management covering all aspects of farming. COTton Farm-Level EXpert (COTFLEX) and COtton MAnagement eXpert (COMAX) for cotton crop management and POMME expert system for apple plantation management were the first AI techniques designed for crop management.
PROgramming in LOGic (PROLOG) and CALEX can be used to remove less utilized farm tools and formulate scheduling guidelines for crop management activities, respectively. A multi-layered feed-forward ANN-based system was developed for citrus crop protection from frost damage, with the best model in the system possessing 94% accuracy with six inputs and two output classes.
ANN-based systems can also be used to precisely detect the crop nutrition disorder and predict the crop response to soil salinity and moisture. ANN and fuzzy logic were used to reduce insect attacks on crops. Similarly, an image-based AI technique was formulated for wheat crops using a pixel labeling algorithm, with the best network having five hidden layers trained up to 300000 iterations and an 85.9% average accuracy. A soybean crop management system based on fuzzy logic was also developed that can advise about crop selection and fertilizer application.
Pest and Disease Management
Pest manifestation and crop diseases lead to extensive crop damage and significant economic losses. Several rule-based expert systems were initially developed for pest management. However, using rule-based systems led to uncertainty as the knowledge involved in pest management is imprecise, vague, and imperfect. Thus, multiple fuzzy logic-based expert systems were later developed for more effective pest management. For instance, FuzzyXpest was developed to provide pest information with high accuracy to farmers.
Rule-based and fuzzy logic-based systems were also developed initially for disease management. For instance, a fuzzy logic-based model was designed to forecast diseases based on leaf wetness duration. Similarly, a fuzzy logic approach combined with image processing was implemented to detect the infection percentage in the leaf.
Rule-based expert systems can be used for faster diagnosis and treatment of disease cost-efficiently. Additionally, different ANN-based models and hybrid systems were designed for disease control in different crops. For instance, an image processing model coupled with an ANN model was developed for phalaenopsis seedling disease classification.
Similarly, an approach based on ANN and geographic information system (GIS) was formulated to detect crop diseases with 95% accuracy. Computer vision systems (CVS) and genetic algorithms (GAs) can also be used to quickly detect diseases. Web-based intelligent disease diagnosis systems (WIDDS) can swiftly respond to the nature of crop diseases with high accuracy.
Irrigation and Soil Management
Irrigation and soil management are necessary to prevent degradation in soil quality and crop loss, which can be achieved using AI techniques. For instance, a rule-based expert system was developed to evaluate the performance and design of microirrigation systems. A fuzzy logic-based soil risk characterization decision support system (SRCDSS) can be used for soil classification based on associated risks.
Similarly, the knowledge of farmers was used to model a fuzzy-based system that can recommend crops based on the fuzzy system-generated land suitability maps, while a Takagi Sugeno Kang fuzzy inference system was employed to assess the stem water potential of a plant based on soil water content and meteorological data.
ANN models were also developed to predict soil enzyme activity, classify and predict soil structure, estimate soil moisture, predict monthly mean soil temperature and texture, and estimate soil nutrients after erosion. Management-oriented modeling was used to maximize production and minimize nitrate leaching.
Weed Management
The application of herbicides in agriculture directly affects the environment and human health. Several modern AI methods have minimized herbicide application through proper weed management. For instance, a rule-based expert system was designed to identify and eliminate weeds in crops such as wheat, triticale, barley, and oats. Similarly, machine vision with a back-propagation-trained neural network was developed to identify five different weed species.
The back-propagation network demonstrated the highest 97% accuracy among the three neural network models, including counter propagation, radial basis function, and back-propagation, evaluated for weed management. Neural network and image analysis were used to formulate a new approach for weed management. Learning vector quantization (LVQ) ANNs can accurately identify weeds within a short processing time. Digital image analysis and global positioning system (GPS) can be used to identify weeds with medium/60% accuracy.
Additionally, the support vector machine (SVM) and ANN can detect crop stress quickly to ensure timely site-specific remedies. Invasive weed optimization (IVO) coupled with ANN offers cost-effective solutions for weed management with improved performance.
Yield Prediction
Accurate crop yield prediction greatly benefits crop cost estimation and marketing strategies. Prediction models can analyze the relevant factors that directly affect the yield. For instance, an ANN model using a back-propagation learning algorithm was designed to predict crop yield from soil parameters.
Similarly, a neural model was built to predict tomato growth and yield, and water use in a greenhouse environment. Neural models were also developed to predict the crop yields of seven crops using fertilizer consumption and atmospheric inputs.
Challenges of Using AI in Agriculture
Although AI has demonstrated significant potential in the agricultural sector, multiple challenges regarding the use of AI in agricultural activities have reduced the positive impact of AI in agriculture compared to other sectors.
For instance, several AI-based expert/intelligent systems used in agriculture have longer response times and/or low accuracy, adversely affecting the user's task strategy selection. Automatic performance, monitoring, and pacing are the three strategies used to maximize accuracy and minimize efforts.
The method employed to look up and train an expert system using big data is often not defined properly, which affects the accuracy and speed of the system. Additionally, most AI systems are based on the Internet, which reduces/limits their usage in rural/remote areas.
A web service enabling device with lower tariffs can be developed that can work with AI systems to assist farmers in adopting AI. The lack of simple solutions that can seamlessly embed and incorporate AI hinders the extensive adoption of AI in agriculture. Many farmers lack the digital skills and time to evaluate these AI solutions.
Thus, the new AI solutions must be integrated into legacy and existing systems and infrastructure used by farmers to address this issue. Moreover, AI systems require a substantial amount of data to make accurate predictions and train machines. Although the spatial data of primarily agricultural lands can be obtained easily, obtaining the temporal data is challenging.
Moreover, crop-specific data can only be obtained yearly while growing crops. Thus, the development of effective machine learning models is a time-consuming process as data infrastructure requires significant time to mature.
Conclusion and The Journey Ahead
In summary, AI-enabled farming operations can improve the overall agricultural output as AI solutions can effectively address farming issues, such as weed and insect infestations, which reduce farm yields.
In the next few decades, agricultural production must be increased substantially to meet the demands of the increasing global population. A leading share of this higher production will be fulfilled by the intensification of current production, which necessitates the use of AI-driven solutions to increase farming efficiency.
AI will assist farmers in using data to optimize yields to individual plant rows. Additionally, robots can be engineered to perform different tasks in farm settings more quickly and thoroughly compared to human workers. Moreover, a greater focus is required to ensure universal access to AI solutions in agriculture. Specifically, all AI-driven solutions must be made accessible to small farms in distant/remote regions to maximize efficiency and resource utilization by resolving labor and resource shortage.
References and Further Reading
Javaid, M., Haleem, A., Khan, I. H., Suman, R. (2023). Understanding the potential applications of Artificial Intelligence in Agriculture Sector. Advanced Agrochem, 2(1), 15-30. https://doi.org/10.1016/j.aac.2022.10.001
Eli-Chukwu, N. C. & Ogwugwam, Ezeagwu. (2019). Applications of Artificial Intelligence in Agriculture: A Review. Engineering, Technology and Applied Science Research, 9, 4377-4383. https://doi.org/10.48084/etasr.2756
Bannerjee, G., Sarkar, U., Das, S., & Ghosh, I. (2018). Artificial intelligence in agriculture: A literature survey. International Journal of Scientific Research in Computer Science Applications and Management Studies, 7(3), 1-6. https://www.researchgate.net/publication/326057794_Artificial_Intelligence_in_Agriculture_A_Literature_Survey