1
Computer Vision-Based Innovations for Smart Agriculture and Crop Surveillance: Evolution, Trends, and Future Challenges

M. Nalini* and B. Yoga Bhuvaneswari

Department of Electronics and Instrumentation Engineering, Sri Sairam Engineering College, Chennai, India

Abstract

The economic prosperity of a nation relies heavily on its agricultural industry. Due to population expansion, frequent climate change, and material restriction, it has become harder to supply the food needs of the present population. Precision agriculture, also frequently referred to as smart agriculture, is a cutting-edge method for addressing the challenges associated with agricultural sustainability. This cutting-edge technology is powered by a system based on computer vision (CV) and artificial intelligence (AI). These AI and CV techniques have made remarkable contributions to the agricultural industry in the areas of plant health detection and monitoring, planting, weeding, harvesting, and modern weather forecast analysis. So many use cases of smart farming have an influence on the whole food supply chain by offering insightful data on the entire agricultural process, easing operational decision-making in real-time, and improving farming methods by adding smart sensors and equipment to the field. The use of CV in agriculture has recently increased. Computer vision has enormous potential to improve the whole functioning of the agricultural industry, from lowering production costs through intelligent automation to increasing output.

Artificial intelligence includes the field of CV. Because of breakthrough technology, machines can now understand and perceive the visual environment in a manner that is comparable to that of humans. Computer vision techniques combined with image capturing from remote cameras provide agriculture-specific, non-contact, and scalable sensing solutions. The CV–AI models have made several contributions to the agricultural business in areas including harvesting, enhanced weather analysis, weeding, planting, and plant health detection and monitoring. This chapter covers the evolution, trends, and future challenges of smart farming using CV.

Keywords: Computer vision, smart farming, precision agriculture, artificial intelligence, intelligent farming, internet of things-based smart farming, machine learning, precision farming

1.1 Introduction

Precision agriculture is an agricultural management idea built on surveilling, measuring, and reacting to crop variances. By maximizing input returns while protecting resources, precision agriculture research tries to establish a decision-making system for farm management. This type of agriculture has benefited greatly from machine vision, which makes automated alternatives to chores that are often done by hand. Manual processes are usually laborious and vulnerable to mistakes. Accurate, precise, and effective solutions may be provided by machine vision to help agricultural activities. Additionally, machine learning (ML) techniques make it possible to analyze enormous amounts of data swiftly and reliably, opening the door for the adoption of machine vision applications in agriculture. For agricultural production to meet the issues of productivity, environmental impact, food safety, and sustainability, smart farming is crucial. Due to the continually growing world population, a significant increase in food production must be made while simultaneously maintaining high nutritional quality and availability everywhere and safeguarding the natural ecosystems by employing sustainable agricultural practices. Advanced computer vision (CV) algorithms and ML-based artificial intelligence (AI) can continually analyze the data gathered from multiple sources. Artificial intelligence, deep computing, and computational intelligence are all related technologies created for intelligent systems. A larger definition of AI is the development of intelligent computers that can mimic human thought and behavior. Simply expressed, the basic goal of AI is to develop computer systems that are intelligent enough to resemble human intellect. These AI systems need knowledge engineering to function properly.

Machine learning, on the other hand, is a branch of AI that enables computers to learn from collected data without explicit programming. Machine learning makes accurate predictions on new data by using sophisticated algorithms that repeatedly cycle over the massive data set. One dynamic and fascinating area of AI is CV, which replicates the intricate human visual system. The basic objective of CV is to enable computer systems to detect and locate objects in images and videos in a similar way to how people do. Two examples of cutting-edge methods that have considerably enhanced CV are deep learning and neural networks. Deep learning algorithms like the convolutional neural networks (CNNs) and their derivatives are frequently employed in in-plant phenology investigations. For classifying images, CNNs are frequently employed. The main building blocks of CNNs are convolutional layers, and to create a feature map, a window (filter) is utilized to scan the pictures and seek for certain features. Convolution, pooling, and fully connected layers are utilized to construct an entire CNN, which produces precise feature recognition and accurate picture categorization. With little information loss, pooling layers attempt to minimize the dimensionality of the feature map produced by the convolutional layers [12].

1.2 Artificial Intelligence in Agriculture

Artificial intelligence techniques are frequently employed in agricultural industries to solve a unique set of issues and improve production and operating methods. Agriculture can quickly adapt to using AI and ML when reading each sentence of agricultural products and agricultural techniques in the topic. Because it can comprehend, learn, and react to a one-of-a-kind (mostly based on learning) rising efficiency, of particular importance, cognitive computing is positioned as the catalyst in the next great agricultural revolution. With the use of AI, farmers may gain valuable information about their farm’s weather, temperature, water consumption, and soil monitoring, which will help them increase their earnings. By discovering which crops they can grow, creating high-quality hybrid seeds, and maximizing resource efficiency, AI technologies aid farmers. Both the quality of harvested products and the accuracy with which they are harvested may be improved with AI [11].

Precision agriculture uses AI technology to assist in the early diagnosis of diseased plants, insect infestations, and nutritional issues in agricultural areas. Artificial intelligence-powered sensors can quickly identify weeds and advise users to apply pesticides. Utilizing this method can save cultivation expenses since herbicides were made more widely known and sprayed throughout the entire field. Farms of any size across the world are using AI technology to operate more effectively and meet global food demand. With the use of AI, forecasting models may be more accurately and effectively learned. These models can predict weather accurately months in advance. The most efficient method for assisting small farmers is seasonal forecasting. Applications of AI are shown in Figure 1.1.

Artificial intelligence is positively transforming Indian agriculture in several ways. The value of AI applications in agriculture is predicted to increase from $852.2 million in 2019 to $238.38 billion in 2030, or a growth of about 25%. Access to markets, inputs, loans, and crop insurance are all improved by technology. A supply chain that is driven by demand may be built with the use of appropriate timing and precise data. Many AI models become inexpensive and accessible by utilizing sensors, phone photos, drones, agricultural weather data, and data on the condition of the soil. Various challenges, such as climate change, population increase, and job issues, are being addressed by AI, which is functioning as a catalyst for improved yield. Nowadays, relatively few individuals are interested in farming, which causes a labor shortage on many farms. In traditional farming, numerous laborers are needed to seed the crops and make the fields profitable. The use of AI farm bots is a way to deal with the labor scarcity. The AI bots are used in a variety of ways to supplement human labor. Bots can harvest crops more quickly than human laborers can, and they can readily spot and get rid of weeds, which can lower expenses [10].

Figure 1.1 Applications of AI in the agricultural sector.

Thanks to a multitude of technologies, such as wireless, wired, and cellular connections (5G or beyond) for the Internet of Things (IoT) and robotics (agricultural drones and agribots), smart agriculture (SA) has started to be deployed. Specialized hardware and expertise in wireless communication are required for AI in 5G and beyond. Deep learning (DL) and machine learning (ML) are frequently employed for communications that require a massive number of inputs, outputs, and beamforming to support 5G and beyond. The DL, reinforcement learning, and CNN are utilized in channel coding for effectively using the air interface for 5G communications. Long short-term memory-type algorithms are used to foresee resource requirements for 5G slicing, enabling the operator to deliver diverse services over a single infrastructure.

1.3 Evolution of Smart Agriculture

Around the turn of the 20th century, the first real-world uses of so-called smart agricultural technology have been concentrated on the...