All interviews were held through digital meeting platforms and all but two were recorded with the permission of each respondent.The codes mentioned in Table 1 are used throughout this article to refer to the respective interview respondents.Remote sensing technology enables detection and monitoring of physical characteristics of the earth’s surface.Remote sensing data is collected from a distance, commonly from satellites and drones.The three most common properties of remote sensing data are spatial, spectral, and temporal resolutions.Spatial resolution is the pixel size of an image, a property that affects the ability to detect objects through imagery.differently, spectral resolution refers to the spectral sampling intervals size and number which affect the ability of the sensors to detect objects in electromagnetic regions.The temporal resolution regards the frequency of acquired data.The availability and economics of using remote sensing data collection is addressed by Khanal et al., which present remote sensing technology alternatives both open-accessed and for some cost.However, the resolution of the data varies, where the trend is that medium-resolution data is free whereas the prices for high resolution and very high resolution data increase in proportion to their increasing quality.Regarding data resolution, Meier et al. opine that site-specific smart farming depends on high resolution, as detection of anomalies are impossible or insipid with too large pixel sizes.Of course, depending on what kind of analysis the data aims to contribute to, the need for resolution varies.For example, predicting the crop yield within a field can accomplish a high accuracy despite a coarse resolution while detection of plant diseases through hyperspectral imaging requires a detailed resolution.Internet-of-Things is a collective concept for objects with incorporated electronics and connections that enable remote control and information sharing.In agriculture, IoT is mainly used for collecting data through different types of sensors.
By further data analysis,flower pot valuable information can be derived as decision support, e.g.for farmers.Kamienski et al. define four main challenges for IoT development in smart farming.First, the IoT system must have a high level of adaptability.Since the needs of farmers often significantly vary, the IoT system must be customizable to local circumstances but still not increase the required work for the farmer.Secondly, the IoT deployment must be efficient.As Kamienski et al. write, “there is no ‘one size fits all’ in IoT systems”.Thus, each system needs to be configured, the Internet connection and farm infrastructure must be reliable, and the farmer must deploy enough human and economic resources into this process.Furthermore, the scalability is affected by the previous factors but also depends on whether the system, and the models learned, are supposed to work for just one farm or entire agricultural consortiums.Lastly, the complexity of the IoT system can be interpreted as a trade-off between making the middleware broker complex and the software application simple, or the reverse.Another aspect to IoT in smart farming is security.Since the data often is valuable for the farmer and is regarded as a business secret, Kleinschmidt et al.describe the need for end-to-end encrypted communication from the sensor to the application.In practice, this means that the IoT sensor network must have a synced security strategy to the cloud database and the potential fog computing network.By ensuring security, the probability that the farmer trusts the IoT system increases.Still, trust in IoT systems does not just depend on security but also on the precision of the sensors.Without ensuring that there are no systematic measurement errors in the sensors, few farmers would trust the learned model or the real-time data.The potential of smart farming in animal husbandry, such as dairy-, beef- and fur production, is largely constituted by increasing productivity and profitability by streamlining and automating tasks and information.Much research consists of ways to monitor and look after the animals automatically or semi-automated.These articles suggest that devices, both wearable and non-wearable, may be incorporated in the animal stable and that these devices can gather data that can give indicators on the health of the animals.The data that may be gathered through these devices vary, but the wearable devices can measure heat, hormone levels, rut etc.The non-wearable devices more typically are 3D cameras for body condition scoring and infrared imaging, sensors that monitor environment and weather as well as automatic weighing scales and gates.In arable farming, an important feature of smart farming is to be able to calculate the vegetation index of fields or areas to be able to monitor when it is time for harvest and other activities.This may be done by both remote sensing and IoT solutions.
Viljanen et al. train a machine learning model aimed to optimize the “balance between the highest possible yield quantity and an adequately high digestibility for feeding”.By using an inexpensive drone system that can get multi-spectral data from an RGB camera and an infrared camera, traditional physical tools for predicting ley yield can be replaced by smart machine-learned models with higher accuracy.Furthermore, the research of predicting yield and quality of silage can also be accomplished through satellite data, as presented by Griffiths et al..The study shows that it is possible to detect mowing events of grasslands, and therefore characterize the land use intensity by looking at satellite imagery.In terms of yield prediction, Feng et al. stress the importance of incorporating biophysical characteristics of the crop in machine learning algorithms.This means that to learn a model with high precision, it is important to simulate the growing process of the crop to ensure that the model learns the crop characteristics in different stages of the growing process.Furthermore, Matos-Moreira et al.uses manual soil samples to further improve their model.By including manual sampling and analysis with a variety of existing data sources one may learn a model to predict the quality of a crop or the concentration of some matter at a given place and time.Another application of precision farming is to detect sickness or pests among crops.Torai et al. study how diseases can be detected in crops by classifying, or labeling, areas in pictures as “healthy”, “infected”, “diseased” or “aged”.Thereafter, methods such as hyperspectral imaging, Bayesian networks, and an analysis through probabilistic latent semantics are applied to detect the diseases .This study is a good example of a remote sensing technology applied to agriculture which needs a very high resolution of data, preferably on a scale of centimeters.One dilemma when applying artificial intelligence to arable challenges is how to use the different types of available data.Kerkow et al. use fuzzy mathematical modeling to solve this problem.This approach allows for mixing machine learned climate models with wind data and expert knowledge of the landscapes to build precise models .The literature review also brings up some interesting aspects regarding the implementation of AI in agriculture.Medvedev and Molodyakov highlight both theoretical and practical knowledge of smart farming as requirements for successful implementation.Unfortunately, seldom farmers have either the economic resources or the time to attend longer educations within the subject.To meet the lack of technical education within smart farming, Medvedev and Molodyakov propose smaller model-based courses that should cover technical, economic and management aspects to smart farming.
A crucial part of the education is that the courses are on-demand, so that busy farmers can access it whenever it suits them.Both business cases and clear driving forces are named as critical components to spreading the use of smart farming technologies in society.Barriers that hinder the drive towards smart farming are categorized as economic, institutional behavioral, and organizational as well as market.Furthermore, they identify social and moral drivers to play a key role in terms of creating a societal demand for smart farming.Without the support from society at large, innovations will not be adopted by key actors, they conclude.Other research aims to map the barriers to implementing and diffusing smart farming technologies.Kernecker et al., describe that farmers approach smart farming technologies differently given how much smart farming technologies the farmers have already adopted.The so-called adopters perceive the barriers to adopt smart farming technology as high investment costs, a difficulty in interpreting data, a lack of interoperability or precision in devices, that farmers cannot see the added value of the new technology or the relative advantage of the system, as well as a lack of neutral advice from advisors and other actors.The non-adopters also perceive high investment costs and unclear added value as barriers.Additionally, they regard too demanding complexity of use, that the technology is not appropriate for their context or farm size, as well as a lack of access to proof of concept from a neutral point of view, as obstacles.Finally, the literature review highlights the importance of data presentation and visualization, both in arable farming and livestock farming.Beside identifying possible applications of the technology in agriculture, several research groups argue that methods within machine learning and AI require decision support tools that visualize the data in comprehensive ways.One pattern, stated by a farmer respondent,berry pots is that farmers of different agricultural sectors almost always believe that the implementation of smart farming technologies has come further in other sectors than in their own.The agricultural sector that most farmers highlight as currently the most technologically advanced is the milk production.Milk robots were introduced to the commercial market decades ago, and with the milk robots the fodder of an individual cow can be customized, increasing its health status and production capacity.Due to the milk robots, the dairy industry is regarded notably data driven.One important aspect to consider when evaluating the success of the milk robots is the short feedback loop.Since cows are both fed and milked daily, the machines can adjust quickly depending on the latest input.Furthermore, Swedish dairy farmers have a long history of collecting data by being part of the so-called Kokontrollen, a cow data collection application owned by Växa Sverige.Even if Kokontrollen today is web-based, Swedish dairy farmers have been reporting to it for more than 100 years.Previously, all data was collected manually but today almost all data connected to milk production is automatically gathered by the milking robots.Contrasting to milk production, arable farming is diverse with different crops requiring distinct machines and technologies.Hence, a single successful machine is difficult to implement for the entire arable farming sector, making its technological development more complex.
However, it is possible to create effective technology for specific crops.As a rule of thumb, crops with high manual work, such as vegetables, use lots of technology since they operate on small, more controlled areas.In such environments, such as green houses, the feedback loop is faster and there are less uncontrollable factors, such as weather or wild hogs, which makes the application of new technology and AI easier.Of the three agricultural sectors compared in this study, beef production is considered by the respondents as the least technologically developed.Nevertheless, one respondent at a major company believes that meat production will have a central role in the development of the Swedish primary food production.The list of possible innovations includes making the value chain digital by automatically transferring information to the slaughterhouses regarding characteristics of the animals they will receive.By mandatory RFID tags for all cattle, the respondent argues there is an enormous potential, since the development of the animals could be followed in real time throughout the value chain.With such a system, the slaughterhouse could plan far in advance for incoming meat quality and volume.Simultaneously, a grocery store could send data to the farmers regarding the current popularity of different kinds of meat, enabling the farmers to adjust their production to the current consumer behavior.Furthermore, if one could autonomously and automatically weigh the cattle, their growth curves can be predicted which would enable optimization of the timing for sending animals to slaughter.By this optimization, one could avoid having full-grown animals that both drain economical resources and emit environmentally damaging methane gas.Regarding data and the activity of collecting data, the responses from the interviews reflects different realities within the agricultural sector.On the one hand, some respondents say that farmers generally are positive towards gathering data on their farm.On the other hand, some responding farmers state that they collect almost no data on their farms, although they say that they understand that data could add value to them.In-between is a spectrum of attitudes towards data gathering and implementation of technology in the farms.