Machine learning has affected many industries and agriculture is no exception. In this post, we are going to analyze the advantages of its implementation, as well as the factors that discourage farmers from implementing it.
Artificial intelligence creates new possibilities for generating value in agriculture, which are indispensable to ecological evolution. They include reducing inputs and automating mundane tasks.
Digitalization has brought new terms to the vocabulary used in agriculture. Today, it is common to hear the following terms: precision agriculture, sensorics, augmented and virtual reality, internet of things, massive data analysis or Agriculture 4.0. Although there is one that stands out above all: artificial intelligence.
Computer awareness
There are many definitions given to refer to artificial intelligence. It consists of mixing different algorithms to obtain models with the ability to mimic human beings. The concepts emphasize four objectives. These differentiate the type of intelligence based on its rationality and thinking.
In a simplified version, the artificial intelligence is a sister to programming and computers, in order to obtain tools capable of solving problems.
Classification of artificial intelligences
Decision-making systems can be differentiated into different types and subcategories.
In the first place, we distinguish what is known as weak or narrow artificial intelligence (ANI, Artificial Narrow Intelligence). Its functionality is limited to performing specific tasks and is present in most commonly used intelligent tools, such as the virtual assistants Siri or Alexa.
Secondly, strong artificial intelligence, which is composed in turn between general (AGI, Artificial General Intelligence) or superior (ASI, Artificial Super Intelligence).
AGI would have a capacity similar to that of a human being. Its algorithms give it some level of self-awareness that allows it to face problems, acquire knowledge and project the future. ASI would surpass human intellect.
The strong artificial intelligence approach presents a theoretical-experimental character. Research centers and large developers are assembling use cases with this orientation.
Two subcategories of artificial intelligence are also highlighted: machine learning and deep learning. The latter is part of the former and consists of different algorithms of intelligent systems that forge expert models to make projections or classifications.
Artificial intelligence categories
The difference between the two subcategories lies in the way the algorithms learn. Deep intelligence is composed of neural networks with more than three layers, including input and output, and is capable of automating the development of the process.
Deep learning enables the use of larger data sets. In addition, unstructured data can be used. The model is able to provide its hierarchy, regardless of whether the primary data source is labeled or not.
Intelligent agriculture
The auxiliary industry of the agri-food system is developing products and services based on artificial intelligence.
The data generated by the sectors included in this economic activity make it possible to generate major innovations, as reported by the European Union.
Tools for irrigation and fertilization
The Green Revolution brought with it the introduction of agrochemicals in the agricultural field and an expansion of irrigation. This has led to a tripling of the consumption of natural resources in just fifty years.
Artificial intelligence systems have integrated algorithms capable of calculating the water demand of crops, which is determined through the evapotranspiration of the plant species.
At Plataforma Tierra we offer a tool for calculating water and fertilization needs.
The values used can be obtained from pre-established databases, weather stations, dendrometers, soil water sensors or hyperspectral images.
The use of one data source or another depends on the degree of sophistication of the model used. The use of data collected in situ increases the accuracy of the values offered, as they are adapted to the requirements of each agricultural field.
Artificial intelligence systems need the integration of all the technologies generated in this Technological Revolution to increase the accuracy of the values provided.
It is common that these systems also determine the nutritional needs of vegetables. Based on soil nutrients, water and amendments applied to the soil profile, the tools provide farmers with the amount of fertilizer to apply to the crop at each growth stage.
These systems can reduce input demand by up to 65%.
On the other hand, the application of the models has been extrapolated to irrigation networks to provide a solution for the integrated and sustainable management of the water cycle.
Intelligent traceability systems can detect water leaks and reduce water and energy losses by 30%.
Tools for pathology identification
Biotic stresses can cause production losses of up to 40%. Therefore, early detection of pathologies affecting crops is of paramount importance.
Part of the algorithms can use meteorological data from nearby stations (temperature, humidity, rainfall, wind, etc.) to calculate the probability rate of incidence of pests and diseases. The system alerts growers to the possibility of incidence.
The expansion of new categories of artificial intelligence has allowed expanding the functionalities of model detection systems capable of diagnosing, from a photograph, the picture of plant symptoms, together with the alternatives available for their control. They can be integrated into autonomous driving vehicles.
Autonomous robot for continuous pest scouting in greenhouses.
Source: Tekeniker TV.
Also, the models can detect the early incidence of pests and diseases by analyzing high-resolution hyperspectral and thermal images for detection.
Selective applicators
Excessive application of agrochemicals can have major impacts on ecosystems, negatively affecting their biodiversity and human health.
The need to reduce the addition of these compounds has led to the development of selective applicators, both for pesticides and fertilizers, which are controlled by computer systems.
The equipment can detect, through ultrasonic or optical sensors, individual targets with an accuracy of up to 95%. The artificial intelligence system triggers the emission of agrochemicals only on the plants of interest.