Method and system for treating plants in an agricultural field

Abstract

The present invention relates to a method and system of treating plants in an agricultural field, the method comprising the steps of providing a multispectral camera, an inertial measurement unit, a global navigation satellite system sensor, map data for the agricultural field and combining the signals from the sensors with the map data and the treatment input in order to define a plant need and operate a plant treatment unit according to a treatment input.

Description

[0001] METHOD AND SYSTEM FOR TREATING PLANTS IN AN AGRICULTURAL FIELD

[0002] Field of the invention

[0003] The present invention relates to a method for automated or optimal treatments of plants in an agricultural field.

[0004] Background

[0005] In the agricultural cultivation of plants in industrial scales it is a problem that a desired treatment of crop usually must be applied to the whole field in a homogenous way. For the treatment of plants, a tractor may be used which is applying a certain amount of treatment to the plants. However not all plants, even when they are of the same species, are in need of the same amount of treatment. For example, plants standing in a valley do not need as much water as plants standing in other places. This makes watering the field without drenching and drying up some parts difficult. However, there are many more factors that affect the need of plants that vary over the field and need to be taken into account. A known solution to this is a manual adjustment of an application rate of a treatment equipment. For example, a farmer may adjust the rate while operating a tractor according to his experience or knowledge. While this may work on some occasions it is bothersome and prone to mistakes, wherein some plants are getting to much treatment and other plants to little treatment..

[0006] Summary

[0007] The present invention facilitates the precise treatment of plants in an agricultural field by providing methods and systems. Thus, the invention relates to a method of treating plants in an agricultural field, the method comprising the steps of: providing a sensor system comprising a multispectral camera sensitive to electromagnetic radiation with a wavelength between 500 nm to 900 nm configured to provide pixel maps with each pixel representing an intensity of electromagnetic radiation, an inertial measurement unit having at least one of an accelerometer, a gyroscope and a magnetometer, a global navigation satellite system (GNSS) sensor, defining a plurality of regions of interest of the agricultural field, providing map data for the agricultural field, the map data comprising at least one of a soil type, a water content, soil moisture, a vegetation state, a vegetation index, and a vegetation type, obtaining a pixel map from the multispectral camera, identifying a first region of interest in the pixel map based on positioning data from the inertial measurement unit and the GNSS sensor, defining a vegetation index for the first region of interest based on the pixel map of the first region of interest or the map data, providing a treatment input for the agricultural field selected from at least one of a crop type, a crop stage, an operation type, and an operation strategy, defining a plant treatment need of a plant in the first region of interest based on at least the vegetation index, the treatment input, and the map data, and actuating a plant treatment unit in accordance with the plant treatment need.

[0008] In another aspect, the invention relates to a system for the treatment of plants. The system comprises: a memory unit configured to store map data, a multispectral camera sensitive to electromagnetic radiation with a wavelength between 500 nm and 900 nm, an inertial measurement unit having at least one of an accelerometer, a gyroscope and a magnetometer, a GNSS sensor, a plant treatment unit, a treatment input unit configured to receive a treatment input from a user, a processing unit configured to, receive an image signal from the multispectral camera, receive a location signal from at least one of the GNSS sensor and the inertial measurement unit, receive a map data signal from the memory unit, receive a treatment input from a user, and send an instruction signal to the plant treatment unit based on the received signals.

[0009] The system may in particular be configured to perform the method of treating plants in an agricultural field of the invention. The overall inventive idea is to determine the need of plants for relatively small regions of interest, and then treating the plants in the region of interest according to the determined needs. These regions of interest are chosen such that they best match the location and effective radius of a plant treatment unit.

[0010] In another aspect, the invention relates to a method of producing a treatment prescription map of plants in an agricultural field, the method comprising the steps of: providing a sensor system comprising a multispectral camera configured to provide two-dimensional pixel maps with each pixel representing an intensity of light of a wavelength in the range of 500 nm to 900 nm, an inertial measurement unit having at least one of an accelerometer, a gyroscope and a magnetometer, a GNSS sensor, defining a plurality of regions of interest of the agricultural field and providing each region of interest with an identifier, providing map data for the agricultural field, the map data comprising at least one of a soil type, a water content, soil moisture, a vegetation state and a vegetation type, obtaining a two-dimensional pixel map from the multispectral camera, identifying a region of interest in the two-dimensional pixel map based on positioning data from the inertial measurement unit and the GNSS sensor and linking the two-dimensional pixel map with the identifier of the region of interest, defining a vegetation index for the region of interest based on the two- dimensional pixel map of the region of interest and the map data for the agricultural field, providing a treatment input for the agricultural field selected from at least one of a crop type, a crop stage, an operation type, and an operation strategy, defining a plant treatment need of a plant in the region of interest based on at least the vegetation index and the treatment input, and linking the plant treatment need with the identifier for the region of interest to produce the treatment prescription map.

[0011] Any embodiment for the method of treating plants in an agricultural field is equally relevant for the method of producing a treatment prescription map of plants in an agricultural field, and any specific combination of features described for the method of treating plants in an agricultural field can be used in the same combination for the method of producing a treatment prescription map of plants in an agricultural field.

[0012] In yet at further aspect the invention relates to a system for producing a treatment prescription map of plants in an agricultural field. The system comprises: a memory unit configured to store map data, a multispectral camera sensitive to electromagnetic radiation with a wavelength between 500 nm and 900 nm, an inertial measurement unit having at least one of an accelerometer, a gyroscope and a magnetometer, a global navigation satellite system sensor, a treatment input unit configured to receive a treatment input from a user, a processing unit configured to, receive an image signal from the multispectral camera, receive a location signal from at least one of the global navigation satellite system sensor and the inertial measurement unit, receive a map data signal from the memory unit, and receive a treatment input from a user.

[0013] The system for producing a treatment prescription map of plants in an agricultural field may especially be configured to perform the method of producing a treatment prescription map of plants in an agricultural field of the invention. The overall inventive idea is to determine the need of plants for relatively small regions of interest, and then define a treatment prescription map of plants in an agricultural field that can be used for treating the plants in the region of interest according to the determined needs. These regions of interest are chosen such that they best match the location and effective radius of a plant treatment unit.

[0014] In general terms, any embodiment of the system for the treatment of plants can operate as a system for producing a treatment prescription map of plants in an agricultural field of the disclosure.

[0015] The systems of the disclosure may be configured to be operated by a user, e.g. by receiving input, such as treatment input, from a user, or the system may be configured to operate autonomously from defined input, e.g. predefined treatment input. The systems may be configured to be installed on a tractor and in particular, the system may be configured to operate passively, e.g. be active, when the tractor housing the system is active. It is especially advantageous that the system for producing a treatment prescription map of plants in an agricultural field is active when the tractor housing the system is active. Thereby, the treatment prescription map of plants in the agricultural field is continuously updated. A continuously updated treatment prescription map of plants is especially advantageous for treating plants in the agricultural field for which the treatment prescription map is defined.

[0016] In examples of the disclosure, the systems may operate without requiring manual activation. For example, the systems may be connected electrically to an electric branch circuit of the tractor which is carrying voltage whenever the tractor is in an on state. The systems may autonomously detect the field and type of operation being performed or intended to be performed, continuously map the field, and may upload data to a server for post-analysis. This persistent monitoring may transform the tractor into an intelligent scouting tool, enabling advanced analytics and data-driven decision-making for precision agriculture. Thereby data is automatically collected whenever the tractor is in use. This allows the system to estimate growth rates, by tracking plant biomass over time. The system may identify areas with uneven or slow growth, allowing farmers to adjust inputs and / or optimize field management.

[0017] Moreover, weed monitoring may be achieved by continuous mapping of weed density and regrowth patterns which may enable precise herbicide application adjustments and minimizing herbicide use. Thereby, improved prescription map of plants in an agricultural field, which is correspondingly useful for the method of treating plants in an agricultural field of the disclosure.

[0018] The systems may be configured to receive or measure a yield of crops on a field which may be correlated with other parameters to predict future yields or yields of currently growing plants. An Al model may be used to analyse growth trends and environmental conditions to predict potential yield more accurately, supporting better planning and resource allocation.

[0019] The systems may provide real-time feedback on the effectiveness of the plant treatment used in the method of treating plants in an agricultural field, e.g. application of fertilizer or pesticide, as the system can compare the effect of a treatment with the result of the treatment and thereby ensuring optimal utilization, reducing waste and improving sustainability.

[0020] The systems may further analyse soil degradation indicators, especially compaction zones, erosion risks, and organic matter distribution, allowing to implement regenerative practices to maintain long-term soil health.

[0021] Plants absorb a certain amount of light, e.g. at wavelengths in the visible range, and correspondingly also reflect light. The absorbed light is generally referred to as being in the photosynthetically active radiation region. Plants generally absorb little light in the infrared range, and therefore plants can be detected from light reflected in the range of 500 nm to 900 nm, and the reflection can also be used to classify plants. Thus, the reflected light can be used to define a vegetation index. In general, the vegetation index is calculated by comparing reflected light, e.g. in the range of 500 nm to 900 nm, with the available light in the same region, e.g. from the sun or the sky in general. The available light may also be referred to as background light.

[0022] In general, many procedures for calculating vegetation indices are known to the skilled person, and any known vegetation index calculation may be used in the present methods. It is preferred that the vegetation index in the present methods is calculated from at least two different wavelengths, especially an infrared and another wavelength. The multispectral camera is configured to provide pixel maps with each pixel representing an intensity of electromagnetic radiation at a wavelength in the range of 500 nm to 900 nm. The pixel map may for example represent a specific wavelength. In an example, the methods include obtaining a plurality of pixel maps from the multispectral camera, with each pixel map of the plurality of pixel maps being obtained at different wavelengths. The vegetation index may include intensities of electromagnetic radiation obtained at different wavelengths from the plurality of pixel maps. For example, pixel maps may be obtained for different spectrums such as 610 nm, 680 nm, 730 nm, 760 nm, 810 nm and 860nm of light. Each spectrum may be recorded at 20nm of full-width half-max detection. This selection of wavelengths includes the Red, Red Edge and Near Infrared bands which may be used in computing corrections for the Vegetation indices such as the normalised difference vegetation index (NDVI) or the normalised difference red edge index (NDRE). An NDVI may be calculated as: NDVI = (NIR - R) / (NIR + R), where R and NIR stand for the spectral reflectance measurements acquired in the red (visible) and near-infrared regions, respectively, and an NDRE may be calculated as: NDRE = (NIR - RE) / (NIR + RE), RE and NIR stand for the spectral reflectance measurements acquired in the red edge (visible) and near-infrared regions, respectively. The red edge may for example be defined as the part of the spectrum centred around 715 nm, or another wavelength in the region between red light and near infrared light. In the present methods, the vegetation index is defined for the region of interest based on the pixel map. However, the vegetation index may also be provided from the map data. The vegetation index may be defined by the pixel map alone or with help of the map data, or the vegetation index may be provided from the map data alone. Should the pixel map of the multispectral camera not be considered to be reliable, for example, due to soil build up or environmental conditions, then the map data may be used instead. When the vegetation index is defined from the pixel map, this vegetation index may be compared with a vegetation index included in the map data to define the plant treatment need of a plant in the region of interest.

[0023] The vegetation index may be defined at different scales and the vegetation index may be expressed in relation to an area, e.g. at a scale ranging from continent-based, country level, regional scale, or a scale covering multiple or individual fields, to a group of plants and down to individual plants. In the present methods, the multispectral camera collects light reflected by surfaces exposed to the light, including plants and non-plant material, and pixel maps collected by the multispectral camera are used to define a vegetation index. Thus, each pixel of the pixel map represents an intensity of electromagnetic radiation, and the pixels in the pixel map are compared to the background light to define the vegetation index. The vegetation index is defined for the region of interest in the present methods. However, the pixel map may also be used to defined vegetation indices for sub-regions within the area of interest, e.g. subregions in the pixel map.

[0024] In the methods, a first region of interest is identified. It is to be understood that the methods may involve defining any number of regions of interests, in particular an agricultural field may contain any number of regions of interest. Thus, when the methods identify a “first region of interest” it is understood that “first” could also be “second”, “third”, “fourth”, etc. Thus, the methods may include defining a plurality of regions of interest for an agricultural field. Each region of interest represents a section, e.g. an area, of the agricultural field. Each region of interest of the plurality of regions of interest may be rectangular and tile the agricultural field. Alternatively, the region of interest may be limited to a row or the like of locations where plants have been planted. Furthermore, the size of the region of interests may be chosen such that they match an effective treating radius of a treating unit and located such that they are centred in relation to the treating unit. Any part of a region of interest may overlap with any part of another region of interest in the plurality of regions of interest. By defining a plurality of regions of interest for an agricultural field, the present methods define plant treatment needs of plants in a region of interest independently of plant treatment needs of plants in other regions of interest in the agricultural field. Thereby, the present method of treating plants in an agricultural field provides optimal treatment of plants in the agricultural field, and the method of producing a treatment prescription map of plants in an agricultural field allows optimal treatment of plants in the agricultural field. This can be compared to providing the same treatment to all plants in the agricultural field. Thus, the present methods can provide optimal treatment of individual plants in an agricultural field.

[0025] The region of interest, in particular each region of interest in a plurality of regions of interest, may have an identifier. The identifier of a region of interest may include information about the agricultural field in which the region of interest is located, e.g. including location data for both the region of interest and the agricultural field, and a number or the like for the region of interest, e.g. with respect to the total number of regions of interest in a plurality of regions of interest.

[0026] The region(s) of interest may also be defined dynamically in the present methods. In an example, the step of defining the vegetation index includes defining a plurality of pixel-vegetation indices for a plurality of pixels in the pixelmap of the first region of interest and scaling the pixel-vegetation indices up. Thus, the pixels in the pixel map may be grouped and a vegetation index may be defined for a group of pixels in a pixel map. A single vegetation index may be defined for the pixel map, or a vegetation index may be defined for a group of pixels in the pixel map. Moreover, if the methods identify a need to define different plant treatment needs within a region, e.g. from groups of pixels for which a vegetation index is defined, and which vegetation index is different from a vegetation index for the whole pixel map or region of interest, initially defined as a single region of interest, the corresponding region of interest may be divided into smaller regions of interest to better define the plant treatment needs for plants in the region of interest. When a region of interest is divided into smaller regions of interest, the smaller regions of interest may be referred to as sub-regions of interest.

[0027] The methods employ a sensor system comprising a multispectral camera, an inertial measurement unit and a GNSS sensor. The multispectral camera is generally configured to collect light in the visible regime and near infrared (NIR) light, so that the multispectral camera sensitive to electromagnetic radiation with a wavelength in the range of 500 nm to 900 nm. The multispectral camera is configured to provide pixel maps with each pixel representing an intensity of electromagnetic radiation. The pixel maps may be provided as a continuous recording, e.g. a video, with 5 to 50 pixel maps per second, or the pixel maps may be provided as still images. The pixel map may generally be rectangular and can be described to have a width and a height and a format being the width divided by the height. The format may be any format commonly used in digital photography, e.g. 3:2, 16:9, 4:3, 1 :1 , etc. However, the format may also be circular or elliptical. The pixel map may have any resolution, but the pixel map commonly has at least 1 million pixels, e.g. at least 10 million pixels, or at least 20 million pixels, although the number of pixels can also be lower than 1 million or higher than 20 million. Regardless of the format, the pixel map may be referred to as a two-dimensional pixel map.

[0028] Multispectral cameras are readily available, for example from MAPIR, Inc (San Diego, CA, USA). An exemplary multispectral camera from MAPIR, Inc, is a single sensor-based camera with a special Bayer pattern bandpass optical filter with wavelengths capturing lights in Red, Green and NIR bands, i.e. wavelengths in the range of 500 nm to 900 nm. The multispectral camera provides pixel maps that are used to define a vegetation index for a region of interest. The methods may include defining a plurality of regions of interest for an agricultural field so that a pixel map for each region of interest in the plurality of regions of interest of the agricultural field. Regardless whether the methods define a single region of interest or a plurality of regions of interest, the multispectral camera may also be configured to differentiate between different types of plants and materials in a single region of interest.

[0029] The present methods define a plant treatment need of a plant in a region of interest in an agricultural field, and in an example, a plurality of regions of interest is defined in an agricultural field so that the present methods define plant treatment needs of plant in the agricultural field. The plant treatment need may be calculated in a linear combination of the at least the vegetation index, the treatment input, and the map data for that region of interest. Alternatively, a non-linear combination may be used instead to define the plant treatment need. The linear combination may include also other factors such as past weather data, weather prediction, climate data, elevation map data etc. The plant treatment needs, especially when the plant treatment needs are linked to specific regions of interest in the agricultural field, may also be said to define a “prescription map”. Thus, the prescription map includes information about plant treatment needs of plants in an agricultural field, and the prescription map may include different plant treatment needs for different plants in the agricultural field. Correspondingly, the method of treating plants in an agricultural field may also be considered to be a method of producing a prescription map of plants in an agricultural field. In this aspect, the method comprises the same steps as in the first mentioned aspect, although the method need not include the step of actuating a plant treatment unit in accordance with the plant treatment need. Thereby, a final step in the method may be defining a plant treatment need of a plant in the first region of interest based on at least the vegetation index, the treatment input, and the map data, and defining the prescription map. The pixel maps recorded by the multispectral camera represent reflections from the plants at wavelengths in the range of 500 nm to 900 nm. The reflected light is preferable normal daylight, e.g. sunlight, although it is also contemplated that the methods may include exposing the regions of interest to light in the range of 500 nm to 900 nm. For example, the methods may employ a light source configured to provide a light profile, e.g. a spectrum, in the range of 500 nm to 900 nm. When used, the light source is preferably configured to provide a spectrum similar to the spectrum of sunlight at the location where the methods are performed.

[0030] It is preferred that the sensor system further comprises an ambient light sensor configured to provide an ambient light signal indicative of incoming light at wavelengths of light in the range of 500 nm to 900 nm. When an ambient light sensor is used, the methods further comprise basing the definition of the vegetation index also on the ambient light signal. Thereby, a more accurate vegetation index is obtained from the pixel maps combined with the measurements from the ambient light sensor.

[0031] The vegetation index obtained from the multispectral camera is compared to map data, in particular map data for the agricultural field. The map data may be provided from commercial or otherwise available sources, or the map data may be collected and compiled by the user of the methods and the systems, or from other data. For example, map data may be provided from the Copernicus Open Access Hub (presently found at or as provided by of the European Union and the European Space Agency (ESA) for Earth observation. The map data includes information about the area in which the agricultural field is located, and this information is thus generally on a larger scale than the pixel maps provided from the multispectral camera. The map data may also be considered to represent background knowledge of the agricultural field covered by the map data, and the map data comprises at least one of soil type, a water content, soil moisture, a vegetation state, a vegetation index, and a vegetation type. Further map data include soil degradation indicators, such as compaction zones, erosion risks, and organic matter distribution. By comparing the vegetation index of the region of interest, as provided from the pixel map, with map data for the agricultural field, the vegetation index of the region of interest can be compared to map data for the agricultural field, which allows that a more accurate dosage of plant treatment is applied in the region of interest, since the map data represent historical knowledge of the agricultural field.

[0032] The map data may in particular include a time label, e.g. the map data may be average map data for a specific growth season, e.g. spring, summer, autumn or winter, or a specific time relevant for a growth season, e.g. sowing time or harvest time, for the agricultural field or for the landscape in which the agricultural field is located. The map data may represent any period of time up to and including the time the pixel map is obtained. For example, the map data may be obtained one day, one week, one month, etc. before the time the pixel map is obtained or planned to be obtained. By including a time label with the map data, an improved knowledge of the expectations of the farmer for the agricultural field in questions is included in the considerations and thereby better knowledge of the correct dosage of treatment can be employed in the methods.

[0033] The map data for the agricultural field may be provided from any source. For example, the map data may include satellite images and / or aerial image, e.g. as obtained from a vehicle flying above the agricultural field, and aerial images may be obtained from an altitude above the agricultural field of up to 10,000 m, e.g. in the range of 50 m to 5,000 m, or 200 m to 2,000 m. Any information available from a satellite image may also be available from an aerial image, but aerial images provided at an altitude above the agricultural field of up to 10,000 m generally provides information of a higher resolution than is available from a satellite image. Aerial images or satellite images may be used to define a vegetation index for the agricultural field, or aerial images and / or satellite images may provide data about features of the landscape of the agricultural field, e.g. with respect elevations, and thereby also sloping parts of the agricultural field, nearby lakes, rivers or coastlines, or satellite images. Satellite images and aerial images may also provide information about soil type water content, or soil moisture, but information about soil type and / or water content may also be obtained from analysing the soil in the agricultural field. Likewise, information about the vegetation state and / or vegetation type may be provided from aerial images or satellite images, although it is preferred that the vegetation state and / or the vegetation type is / are obtained by observation of the agricultural field at a distance of up to 500 m. The vegetation type may also be based on knowledge of what plants have been sown in the agricultural field.

[0034] When the map data include a vegetation index, the vegetation index may also be referred to as a “global vegetation index” while the vegetation index obtained from the pixel map may be referred to as a “local vegetation index” in the present context. Thus, the global vegetation index includes a vegetation index of at least the agricultural field in which the region of interest is located, and the local vegetation index includes a vegetation index of up to a specific region of interest, e.g. a sub-region in the region of interest.

[0035] It is especially preferred to include two or more of soil type, water content, soil moisture, vegetation state, vegetation index, and vegetation type in the map data. Exemplary combinations of map data include soil type and water content, soil moisture, soil type and vegetation index, water content and vegetation index.

[0036] The sensor system comprises an inertial measurement unit. The inertial measurement unit has at least one sensor to track the movement of the sensor system, e.g. as mounted on a tractor or the like. For example, the methods may include defining a starting point, e.g. relative to the agricultural field, and the methods may include tracking the sensor system relative to the starting point. It is especially preferred that the sensor system comprises both an accelerometer and a gyroscope. This allows tracking of the movement of the sensor system both with respect to the starting point, and also provides information of the orientation of the sensor system, e.g. with respect to a horizontal plane. The sensor system and the act of tracking of the sensor system relative to the starting point in the agricultural field work together with the GNSS sensor to more accurately keep track of the position of the sensor system in the agricultural field. The inertial measurement unit generally has a higher resolution than the GNSS sensor. For example, the inertial measurement unit can track the position of the sensor system with an accuracy of 5 m or less, e.g. an accuracy of 0.5 m to 5 m, or 1 m to 2 m. Thereby, the inertial measurement unit, and especially the inertial measurement unit in combination with the GNSS sensor, together with the definition of the region of interest as it is obtained from the pixel map allow a highly specific treatment of plants in the region of interest. Thus, the treatment of plants in the region of interest can be matched exactly to the needs of the plants based on the vegetation index in combination with the map data.

[0037] The GNSS sensor may operate based on one or more of the satellite navigation systems known as the global positioning system (GPS), GLONASS, the BeiDou Navigation Satellite System (BDS), the Galileo system, or the QuasiZenith Satellite System (QZSS).

[0038] Sensors units combining an inertial measurement unit with GNSS sensors are commercially available, e.g. from CubePilot PTY LTD (breakwater VIC, Australia). CubePilot PTY LTD provides units under the name Here, e.g. HerePro, Here 4, and Here 3. For example, the Here 4 unit includes a GNSS sensor (operating under all major GNSS systems) with an inertial measurement unit having an accelerometer, a gyroscope and a magnetometer.

[0039] The methods comprise providing a treatment input for the agricultural field selected from at least one of a crop type, a crop stage, an operation type, and an operation strategy. The treatment input generally reflects the farmer’s aim with the agricultural field, and therefore the methods may include providing the crop type and an operation type, optionally together with an operation strategy. The vegetation index is used together with the treatment input and the map data to define a plant treatment need of a plant in the region of interest, and the plant treatment need is applied to the region of interest, in particular using a plant treatment unit. The plant treatment need may be any need of a plant commonly considered for a plant being grown in a field. Thus, the plant treatment need may include sowing, seeding, fertilisation, irrigation, treatment with plant treatment agent, e.g. a pesticide, such as a herbicide, a fungicide, or an insecticide. The plant treatment need may also include the information that no active treatment is needed for a specific plant.

[0040] In general, the vegetation index, e.g. the local vegetation index, is compared to the map data. The map data may include one or more threshold values for any of the soil type, the water content, the vegetation state, the vegetation index, e.g. the global vegetation index, and the vegetation type, and the plant treatment need may reflect whether the local vegetation index is above or below a threshold value for the map data.

[0041] In an example, the step of defining a plant treatment need of a plant in a region of interest includes comparing the vegetation index, e.g. the local vegetation index, to a predefined vegetation threshold value, e.g. the global vegetation index, and if the predefined vegetation threshold value exceeds the vegetation index setting the plant treatment need for the first region of interest to a value that does not cause the plant treatment unit to be actuated. This facilitates a treatment of an agricultural field with a higher resolution or in greater detail therefore conserving resources such as water, fertiliser or other chemical agents.

[0042] The methods, and thereby also the systems, may further employ an RGB camera. An RGB camera may be configured to provide digital images, and the methods may further comprise obtaining a digital image for the region of interest and basing the definition of the plant treatment need on the digital image, the pixel map and the map data. Thus, in this example, the information available to the sensor system includes a digital image. A digital image, i.e. as obtainable from an RGB camera, provides additional information than is obtainable from the multispectral camera. For example, the digital image can be used to identify specific plants, especially specific individual plants, which are different from plants grown in the agricultural field. For example, a digital image may be compared with a database of plants, e.g. a database of plants as defined by the species and appearance of the species, to identify a plant that is unwanted in the agricultural field, and the plant treatment need for the region of interest having the unwanted plant may include actions directed at the unwanted plant, e.g. directed at only the unwanted plant. An action directed at an unwanted plant may include the removal, e.g. physical removal, of the unwanted plant, or treating the unwanted plant with an appropriate herbicide, e.g. a herbicide specific for the unwanted plant. An example of an unwanted plant is a thistle, although many more unwanted plants are known to the farmer. A digital image obtainable from an RGB camera may also be used to identify non-plant items that should not be present in an agricultural field. Unwanted non-plant items include insects, birds and other animals. For example, the plant treatment need for the region of interest having an unwanted insects may include applying an appropriate insecticide to the region of interest, in particular to a sub-region of interest, having the unwanted insects.

[0043] The methods employ a multispectral camera, and the methods may also employ an RGB camera. Int the present context, the multispectral camera and the RGB camera may be referred to collectively as “cameras”. Thus, when a camera is mentioned without being specified as a multispectral camera or an RGB camera, this camera may be either of the multispectral camera or an RGB camera.

[0044] Both the multispectral camera and an RGB camera, when used, may be soiled as the sensor system is moved over the agricultural field. In the present context, soiling may include that a lens of a camera is subjected to rain, mud, dirt, earth particles, or the like from the surroundings. The camera or cameras may include scrapers, wipers, or the like to remove soiling. In an example, the steps of defining a vegetation index for the first region of interest and / or the step of obtaining a digital image for the region of interest includes estimating a soiling value for the multispectral camera and / or the RGB camera. For example, the methods may include estimating a soiling value for the multispectral camera and only using the map data if the soiling value does not exceed a pre-defined soiling threshold value. The soiling value may be defined for individual pixels or groups of pixels in a pixel map or a digital image as obtained from the camera so that the soiling value can be said to be defined for the camera. The soiling value may include a pre-defined threshold value that indicates if a pixel or group of pixels is / are useful in defining the vegetation index or the digital image. Based on the soiling value, the pixel map and / or the digital image may be adjusted to take the soiling into account. For example, pixels or groups of pixels in a pixel map may be ignored in defining the vegetation index for the pixel map. In general, as long as less than 50% of the pixels in a pixel map exceed the soiling value, the soiling does not have effect on the vegetation index defined for a region of interest.

[0045] These methods advantageously increase the reliability as soiling of the camera is recognized and the vegetation index is calculated based on the available map data.

[0046] The methods may use an ambient light sensor. An ambient light sensor allows to more reliably evaluate the signals received from other sensors and compare them to older signals recorded at different lighting conditions. This is especially useful for the multispectral camera and also for the RGB camera, when this is used. The determination of the vegetation index may thereby have an increased reliability, and likewise detection of unwanted plants or non-plant item using the RGB camera may also have increased reliability.

[0047] In another aspect, the invention relates to a system for the treatment of plants. The system is configured to perform any example of the methods, especially the method of treating plants in an agricultural field, of the invention, and the system may comprise any one or all of the units described for examples of the methods. The present method of treating plants in an agricultural field includes the step of actuating a plant treatment unit in accordance with the plant treatment need defined in the method of treating plants in an agricultural field. In the present context, “actuating” may involve any act commonly done to a plant in an agricultural feed in order to improve the conditions of the plant. In the present context, the word “action” may be used for a treatment of a plant. Thus, the step of actuating a plant treatment unit generally means performing an action, e.g. performing an action in accordance with the plant treatment need. Furthermore, an action may involve be to unwanted plants in order to improve the conditions of a plant of interest in the agricultural field.

[0048] The action for a treatment of a plant may involve sowing, irrigation, fertilisation or applying, e.g. spraying, a plant with a plant treatment agent. The action may also be the removal of a plant, e.g. an unwanted plant, or an insect.

[0049] The method of treating plants in an agricultural field uses a plant treatment unit, and a plant treatment unit is included with the system. The plant treatment unit may include a container for liquids and / or powders, a distributor for liquids and / or powders, connectors between the container and the distributor, and actuators to apply a liquid and / or a powder via the distributor. In the present context, seeds for sowing are regarded as a powder.

[0050] In the method of treating plants in an agricultural field, the plant treatment unit may move in the agricultural field allowing substantially all plants in the agricultural field, in particular plants in all of a plurality of regions of interest to be treated via the plant treatment unit. For example, the methods, including the method of producing a treatment prescription map of plants in an agricultural field, may include providing a tractor where the multispectral camera, the inertial measurement unit, the GNSS sensor and the plant treatment unit are mechanically fastened to the tractor. For example, the tractor may comprise an arm extending from the tractor, e.g. a pivotable arm extending from the tractor at a pivotable link, where the arm has the treatment unit. In an example, the plant treatment unit includes a plurality of subunits which are individually controlled. For example, the plant treatment unit may comprise a plurality of distributor for liquids and / or powders and actuators allowing the distributors of the plurality of distributor to be controlled individually. The individually controllable subunits may for example be mounted along an arm configured to extend from the tractor. The tractor may have a forward movement direction, and the arm may extend in a substantially horizontal plane in a direction at an angle, especially a right angle, to the forward movement direction of the tractor.

[0051] By having a plurality of subunits which are individually controlled, especially when the plurality of subunits is mounted along an arm extending from a tractor at an angle to the forward movement direction of the tractor, each subunit can provide a different treatment compared to another subunit, and thereby the agricultural field can be treated at a high resolution of regions of interest or subregions of regions of interest. This method advantageously allows to treat the plant in the field with a high resolution meaning that each subunit can be adjusted to provide the right amount of treatment for the area that is covered by every subunit.

[0052] The multispectral camera, and the RBG camera and the ambient light sensor, when these are used, may be mounted on the tractor. The multispectral camera may be mounted on the tractor for the multispectral camera to depict an area in front of the tractor, e.g. with respect to the forward movement direction of the tractor. An RGB camera may also be mounted on the tractor for the RGB camera to depict an area in front of the tractor. In particular, the multispectral camera and the RGB camera may depict the same area in front of the tractor, although the multispectral camera and the RGB camera need not depict areas of the same size.

[0053] The systems of the disclosure include a processing unit configured to receive an image signal from the multispectral camera, e.g. a pixel map, receive a location signal from at least one of the GNSS sensor and the inertial measurement unit, receive a map data signal from the memory unit, and receive a treatment input from a user. In the present context, reception of treatment input from a user may involve that the user manually provides the treatment input to the treatment input unit, or the treatment input unit may contain predefined treatment input, e.g. a plurality of predefined treatment input contained in a list or matrix of predefined treatment input, so that the user can select a specific predefined treatment input, e.g. from the plurality of predefined treatment input. The treatment input unit may also be configured to select a predefined treatment input based on input previously entered by the user into the treatment input unit, or the treatment input unit may be configured to select a predefined treatment input based on the growth season, e.g. spring, summer, autumn or winter, or a specific time relevant for a growth season, e.g. sowing time or harvest time, and the predefined treatment input may also consider data regarding a specific agricultural field or for the landscape in which an agricultural field is located. The processing unit is programmed to define a prescription map according to the present method, e.g. to define a plant treatment need of a plant in a region of interest, e.g. from data in the processing unit, and the processing unit is configured to send an instruction signal to the plant treatment unit based on the received signals.

[0054] The system has a memory unit configured to store map data, and the memory unit is in data communication with the processing unit. The memory unit may be a physical part of the system, e.g. the memory unit may be in data communication with the processing unit via a cabled connection, or the memory unit may communicate with the processing unit via a wireless communication protocol. For example, the memory unit may be integrated with a server in wireless communication with the processing unit.

[0055] When the memory unit is in wireless data communication with the processing unit, the memory unit may contain its own data processor unit.

[0056] In a specific example, a single memory unit, especially a memory unit having a data processor unit, is in wireless data communication with a plurality of systems of the present disclosure, e.g. the processing units of a plurality of systems of the present disclosure. For example, the map data may include vegetation indices, e.g. local vegetation indices, collected from a plurality of agricultural fields, e.g. a plurality of agricultural fields located in the same region. Thereby, a regional map data for agricultural fields in the region is provided.

[0057] Any embodiment of any aspect of the invention may be used in any other aspect of the invention, and any advantage for a specific embodiment applies equally when an embodiment is used in a specific aspect.

[0058] Brief description of the drawings

[0059] In the following the invention will be explained in greater detail with the aid of an example and with reference to the schematic drawings, in which Figure 1 shows a system of the invention;

[0060] Figure 2 shows a chart of an example of the method of the invention.

[0061] The invention is not limited to the embodiment / s illustrated in the drawings. Accordingly, it should be understood that where features mentioned in the appended claims are followed by reference signs, such signs are included solely for the purpose of enhancing the intelligibility of the claims and are in no way limiting on the scope of the claims.

[0062] The term “comprising” as used in this specification and claims means “consisting at least in part of”. When interpreting statements in this specification and claims which include the term “comprising”, other features besides the features prefaced by this term in each statement can also be present. Related terms such as “comprise” and “comprised” are to be interpreted in a similar manner.

[0063] Detailed Description

[0064] The present invention relates to a method of treating plants 1 in an agricultural field 11 and to a system 100 for the treatment of plants 1 , in particular a system 100 for the treatment of plants 1 in the method of the invention. The system 100 is shown in Figure 1 .

[0065] Figure 1 shows a tractor 30 carrying a sensor system 20 with the multispectral camera 21 , the inertial measurement unit 22, and the global navigation satellite system (GNSS) sensor 23. The multispectral camera 21 , the inertial measurement unit 22, and the GNSS sensor 23 are integrated into a single unit thus being the sensor system 20. The sensor system 20 is mounted on top of the tractor 30 so that the multispectral camera 21 faces an area in front of the tractor 30, and so that the GNSS sensor 23 faces the sky. The sensor system

[0066] 20 also has an ambient light sensor 25, which faces the sky.

[0067] The inertial measurement unit 22 together with the GNSS sensor 23 are contained in a single unit, specifically a Here 4 unit from CubePilot PTY LTD (breakwater VIC, Australia).

[0068] The multispectral camera 21 is a single sensor based camera with a special Bayer pattern bandpass optical filter from MAPIR, Inc (San Diego, CA, USA)

[0069] Two regions of interest 12 are shown in Figure 1 as areas in the agricultural field 11. In the method, pixel maps are obtained from the multispectral camera

[0070] 21 and the regions of interest 12 are defined in the pixel maps, and thereby the regions of interest 12 are related to the agricultural field 11 as areas in the agricultural field 11 .

[0071] The sensor system 20 further has an RGB camera 26, which faces the same general direction as the multispectral camera 21. The RGB camera 26 is a global shutter RGB camera 26 based on sensor AR0234 which captures images in visible light conditions. The RGB camera 26 provides a digital image of the agricultural field 11 , and the digital image may represent one or more regions of interest 12 or part of a region of interest 12. The processing unit is configured to analyse the digital image for unwanted plants 13, so that an unwanted plant 13 can be treated individually with a liquid or powder from the container.

[0072] The tractor 30 further has an arm 31 that carries three distributors 32. The distributors 32 represent the plant treatment unit 24, although the arm 31 may also be considered to be part of the plant treatment unit 24.

[0073] The system 100 further has a memory unit not shown configured to store map data and a processing unit not shown configured to process data from the sensor system 20 and the memory unit. Specifically, the system 100 has a computing unit with both of the memory unit and the processing unit in the form of an NVIDIA Jetson Orin Nx (NVIDIA, Santa Clara, CA, USA) with a FORECR ORNX carrier board. The carrier board supports the mounting of submodules such as Wi-Fi module, LTE module and CAN-BUS Module and also the use of USB ports for connection to the multispectral camera 21 and an RGB camera 26. The processing unit sends an instruction signal to the plant treatment unit 24 based on the received signals. The instruction signal actuates the plant treatment unit 24 to thereby perform the corresponding step of the method.

[0074] The multispectral camera 21 and RGB camera 26 are connected to the computing unit via the USB and stream data at 25fps and 1920p resolution each. The data stream from the multispectral camera 21 is split into individual Red, Green and NIR bands and is further used in computing a normalised difference vegetation index (NDVI). The pixel maps are extracted from the video stream at variable frequency depending on the speed of the tractor 30 and the pixel maps are further processed using developed algorithms to extract further information and location.

[0075] The ambient light sensor 25 is connected to the computing unit over i2C and sends the irradiance values for different spectrums such as 610 nm, 680 nm, 730 nm, 760 nm, 810 nm and 860nm of light, each with 20nm of full-width halfmax detection. The Here 4 unit with the inertial measurement unit 22 and the GNSS sensor 23 is connected to the computing unit via the CAN bus, and the Here 4 sends raw accelerometer, gyroscope and magnetometer data to the computing unit at 50 Hz, and GPS data at 5 Hz where it is further filtered using EKF filtering and fused together to estimate the local and global state of the sensor.

[0076] Once the pixel map, and the digital image, if used, is extracted from the multispectral camera 21 video stream, the pixel map is time synchronised with the data from the Here 4 unit. This step ensures that the frame of the pixel map is properly aligned with the correct pose information, including its 3D orientation and 3D translation.

[0077] In normal use, the multispectral camera 21 is pre-calibrated using a checkboard pattern method. This calibration process helps determine the intrinsic and extrinsic parameters of the multispectral camera 21 , which are important for accurate pose estimation. After computing the camera pose, a geometric reprojection is performed to estimate the position of the target object in the camera frame. This provides per-pixel information on the object's position and pixel value across the three colour bands.

[0078] In addition to the pose estimation, the fact that the multispectral camera 21 may capture a significant amount of unnecessary information that is not related to the vegetation may also be taken into consideration. Thus, to filter out this unwanted data and improve the efficiency of further calculations, the region of Interest 12 is defined. Typically, the width of the region of Interest 12 is determined by the working width of the plant treatment unit 24. The NDVI is then calculated on the pixels within the region of Interest 12, especially only on the pixels within the region of Interest 12. To ensure compatibility with a grid system that covers the entire agricultural field, the NDVI values may be upscaled to a 0.5-meter resolution before being stored.

[0079] The storage system, e.g. the computing unit with the memory unit, of the system 100 of Figure 1 operates with a local grid database with a predefined resolution of x*x meters, an it is used to store the NDVI values per grid cell, along with their respective locations in the field.

[0080] The computing unit may have a user interface (not shown), or the computing unit may communicate, e.g. wirelessly, with an app on a portable electronic device, e.g. a smart phone, where the farmer using the system can enter the treatment input for the agricultural field 11 .

[0081] The user provides input such as the field number, crop type, crop stage, operation type, operation strategy, average rate, and deviation. This input triggers an Agronomic Algorithm, which utilises satellite-based maps downloaded from the Copernicus hub as map data. The Agronomic Algorithm processes the maps locally to generate a custom prescription map based on the user's inputs according to the present method.

[0082] The computing unit may have a display, or information may be shown to the farmer via the app on the portable electronic device. The computing unit or the app, via the portable electronic device, may provide the plant treatment need to the user who can control the tractor and the plant treatment unit 24, or the computing unit may control the plant treatment unit 24.

[0083] The system of Figure 1 further comprises a cloud server module (not shown) that stores the prescription map and moreover, the cloud server may be used to host an app, e.g. the app on an portable electronic device. The computing unit can send and receive data to the cloud server via a cellular or local Wi-Fi connection.

[0084] The tractor 30 has a container not shown for holding a liquid, which may be a plant treatment agent, e.g. a pesticide, a fertiliser or water, or the container may hold a powder, which may be a fertiliser, a pesticide, or seeds. The arm 31 contains pipes not shown providing fluid communication between the container and the distributors 32, and the tractor 30 has pumps not shown and actuators not shown for applying the liquid or powder from the container to the distributors 32 to distribute the liquid or powder on regions of interest 12 of the agricultural field 11 according to the plant treatment needs of the plants 10 in the respective regions of interest 12.

[0085] Thus, the instruction signal from the processing unit actuates the plant treatment unit 24 in accordance with the plant treatment needs of plants 10 in the regions of interest 12.

[0086] The system may employ the following “modules”, which in the present context are mathematical functions / static transformations which enable the extraction of the position of sections of each individual distributor 32 relative to the sensor, e.g. the multispectral camera 21 , the Inertial measurement unit 22, the RGB camera 25, and the GNSS sensor 23, as appropriate, and the target object position relative to the sensor respectively. In the description of the modules, the “sensor” may one or more of the multispectral camera 21 , the Inertial measurement unit 22, the RGB camera 25 and the GNSS sensor 23.

[0087] Position of target object

[0088] This module captures the position of the target object in the region of interest 12 with respect to the sensor.

[0089] Sensor Location (local + global)

[0090] This module takes the raw sensor data from the sensor module and combines it with other data sources from inertial measurement unit 22 and the GNSS sensor 23 to determine the absolute position of the target object in the world coordinates.

[0091] Position of sprayer attached to tractor

[0092] This module determines the position of the distributor 32 attached to the tractor 30. This information is used to accurately apply the spray to the target object. Position of sections of sprayer (X1 , X2, Xn)

[0093] This module determines the position of each individual section of the distributor 32. This information is used to control the spray valves on the distributor 32 and apply the correct amount of spray to each section of the region of interest 12 of the agricultural field 11 . Thus, the system of Figure 1 further computes the current position of each section and requests NDVI and Prescription data of the respective grid matching the section position. The module further computes the state of each section based on following factors.

[0094] Section ON / OFF (NDVI threshold)

[0095] If the NDVI at the grid cell / section position is below the threshold NDVI value the section is turned OFF, and vice a versa. This ensures that the field with spots where there is no vegetation (NDVI<0.2) the section should be turned OFF.

[0096] Section Prescription rate (Prescription Map)

[0097] This module determines the appropriate spray rate for the current section of the distributor 32 based on the prescription data that was received from the Store.

[0098] ISOBUS component

[0099] The ISOBUS is a standardised communication protocol that allows agricultural implements to communicate with each other and with the tractor 30. Hence the system computed section states are being forwarded to the ISOBUS module which further converts the information in the right protocol to control the implement or distributor 32.

[0100] The flow chart in Figure 2 provides a detailed explanation of an example of the methodology implemented by the system 100 to generate the prescription map. The prescription map may be generated irrespective of the availability of internet connection, presence of clouds, or the condition of the sensor system 20. The purpose of this methodology is to ensure that the sensor system 20 functions seamlessly under various circumstances that could potentially disrupt the operation of the farmer. For instance, it takes into account scenarios where the satellite data is obscured by clouds, the internet connection is unavailable, or the sensor cameras 21 , 26 are compromised due to dust or mist on the lens. By considering these factors, the system 100 guarantees the redundancy in operation of the sensor system 20 regardless of any hindrances that may arise. Reference signs list

[0101] 10 Plant

[0102] 11 Agricultural field

[0103] 12 Region of interest 13 Unwanted plant

[0104] 100 System

[0105] 20 Sensor system

[0106] 21 Multispectral camera

[0107] 22 Inertial measurement unit 23 Global navigation satellite system (GNSS) sensor

[0108] 24 Plant treatment unit

[0109] 25 Ambient light sensor

[0110] 26 RGB camera

[0111] 30 Tractor 31 Arm

[0112] 32 Distributor

Claims

P A T E N T C L A I M S1 . A method of treating plants (10) in an agricultural field (11 ), the method comprising the steps of: providing a sensor system (20) comprising a multispectral camera (21 ) sensitive to electromagnetic radiation with a wavelength between 500 nm to 900 nm configured to provide pixel maps with each pixel representing an intensity of electromagnetic radiation, an inertial measurement unit (22) having at least one of an accelerometer, a gyroscope and a magnetometer, a global navigation satellite system sensor (23), defining a plurality of regions of interest of the agricultural field (11 ), providing map data for the agricultural field (11 ), the map data comprising at least one of a soil type, a water content, soil moisture, a vegetation state, a vegetation index, and a vegetation type, obtaining a pixel map from the multispectral camera (21 ), identifying a first region of interest in the pixel map based on positioning data from the inertial measurement unit (22) and the global navigation satellite system sensor (23), defining a vegetation index for the first region of interest based on the pixel map of the first region of interest or the map data, providing a treatment input for the agricultural field (11 ) selected from at least one of a crop type, a crop stage, an operation type, and an operation strategy, defining a plant treatment need of a plant (10) in the first region of interest based on at least the vegetation index, the treatment input, and the map data, and actuating a plant treatment unit (24) in accordance with the plant treatment need.

2. The method of treating plants (10) in an agricultural field (11 ) according claim 1 , wherein defining a plant treatment need for the first region of interest includes comparing the vegetation index to a predefined vegetation threshold value and if the predefined vegetation threshold value exceeds the vegetationindex setting the plant treatment need for the first region of interest to a value that does not cause the plant treatment unit (24) to be actuated.

3. The method of treating plants (10) in an agricultural field (11 ) according claim 1 or 2, wherein the sensor system (20) further comprises an RGB camera configured to provide digital images, and the method further comprises obtaining a digital image for the region of interest and basing the definition of the plant treatment need on the digital image, the pixel map and the map data.

4. The method of treating plants (10) in an agricultural field (11 ) according any one of claims 1 to 3, wherein defining a vegetation index for the first region of interest includes estimating a soiling value for the multispectral camera (21 ) and only using the map data if the soiling value does not exceed a pre-defined soiling threshold value.

5. The method of treating plants (10) in an agricultural field (11 ) according to any one of claims 1 to 4, wherein the sensor system (20) further comprises an ambient light sensor configured to provide an ambient light signal indicative of incoming light at wavelengths of light in the range of 500 nm to 900 nm, and the method further comprises basing the definition of the vegetation index also on the ambient light signal.

6. The method of treating plants (10) in an agricultural field (11 ) according to any one of claims 1 to 5, wherein the method further includes providing a tractor and wherein the multispectral camera (21 ), the inertial measurement unit (22), the global navigation satellite system sensor (23) and the plant treatment unit (24) are mechanically fastened to the tractor.

7. The method of treating plants (10) in an agricultural field (11 ) according to any one of claims 1 to 6, wherein the plant treatment unit (24) comprises a plurality of subunits which are individually controlled.

8. The method of treating plants (10) in an agricultural field (11 ) according to any one of claims 1 to 7, wherein defining the vegetation index includes defining a plurality of pixel-vegetation indices for a plurality of pixels in the pixelmap of the first region of interest and scaling the pixel-vegetation indices up.

9. A system for the treatment of plants (10), the system comprising: a memory unit configured to store map data, a multispectral camera (21 ) sensitive to electromagnetic radiation with a wavelength between 500 nm and 900 nm, an inertial measurement unit (22) having at least one of an accelerometer, a gyroscope and a magnetometer, a global navigation satellite system sensor (23), a plant treatment unit (24), a treatment input unit configured to receive a treatment input from a user, a processing unit configured to, receive an image signal from the multispectral camera (21 ), receive a location signal from at least one of the global navigation satellite system sensor (23) and the inertial measurement unit (22), receive a map data signal from the memory unit, receive a treatment input from a user, and send an instruction signal to the plant treatment unit (24) based on the received signals.

10. The system for the treatment of plants (10) according to claim 9, wherein the processing unit is configured to define a plant treatment need of a plant (10) as defined in any one of claims 1 to 8.

11. A method of producing a treatment prescription map of plants (11 ) in an agricultural field (10), the method comprising the steps of: providing a sensor system comprising a multispectral camera configured to provide two-dimensional pixel maps with each pixel representing an intensity of light of a wavelength in the range of 500 nm to 900 nm,an inertial measurement unit having at least one of an accelerometer, a gyroscope and a magnetometer, a global navigation satellite system sensor, defining a plurality of regions of interest of the agricultural field and providing each region of interest with an identifier, providing map data for the agricultural field, the map data comprising at least one of a soil type, a water content, soil moisture, a vegetation state and a vegetation type, obtaining a two-dimensional pixel map from the multispectral camera, identifying a region of interest in the two-dimensional pixel map based on positioning data from the inertial measurement unit and the global navigation satellite system sensor and linking the two-dimensional pixel map with the identifier of the region of interest, defining a vegetation index for the region of interest based on the two- dimensional pixel map of the region of interest and the map data for the agricultural field, providing a treatment input for the agricultural field selected from at least one of a crop type, a crop stage, an operation type, and an operation strategy, defining a plant treatment need of a plant in the region of interest based on at least the vegetation index and the treatment input, and linking the plant treatment need with the identifier for the region of interest to produce the treatment prescription map.

12. The method of producing a treatment prescription map of plants (11 ) in an agricultural field (10) according claim 11 , wherein defining a plant treatment need for the first region of interest includes comparing the vegetation index to a predefined vegetation threshold value and if the predefined vegetation threshold value exceeds the vegetation index setting the plant treatment need for the first region of interest to a value that does not cause the plant treatment unit (24) to be actuated.

13. The method of producing a treatment prescription map of plants (11 ) in an agricultural field (10) according claim 11 or 12, wherein the sensor system(20) further comprises an RGB camera configured to provide digital images, and the method further comprises obtaining a digital image for the region of interest and basing the definition of the plant treatment need on the digital image, the pixel map and the map data.

14. The method of producing a treatment prescription map of plants (11 ) in an agricultural field (10) according any one of claims 11 to 13, wherein defining a vegetation index for the first region of interest includes estimating a soiling value for the multispectral camera (21 ) and only using the map data if the soiling value does not exceed a pre-defined soiling threshold value.

15. The method of producing a treatment prescription map of plants (11 ) in an agricultural field (10) according to any one of claims 11 to 14, wherein the sensor system (20) further comprises an ambient light sensor configured to provide an ambient light signal indicative of incoming light at wavelengths of light in the range of 500 nm to 900 nm, and the method further comprises basing the definition of the vegetation index also on the ambient light signal.

16. The method of producing a treatment prescription map of plants (11 ) in an agricultural field (10) according to any one of claims 11 to 15, wherein the method further includes providing a tractor and wherein the multispectral camera (21 ), the inertial measurement unit (22), the global navigation satellite system sensor (23) and the plant treatment unit (24) are mechanically fastened to the tractor.

17. The method of producing a treatment prescription map of plants (11 ) in an agricultural field (10) according to any one of claims 11 to 16, wherein the plant treatment unit (24) comprises a plurality of subunits which are individually controlled.

18. The method of producing a treatment prescription map of plants (11 ) in an agricultural field (10) according to any one of claims 11 to 17, wherein defining the vegetation index includes defining a plurality of pixel-vegetationindices for a plurality of pixels in the pixel-map of the first region of interest and scaling the pixel-vegetation indices up.

19. A system for producing a treatment prescription map of plants (11 ) in an agricultural field (10), the system comprising: a memory unit configured to store map data, a multispectral camera (21 ) sensitive to electromagnetic radiation with a wavelength between 500 nm and 900 nm, an inertial measurement unit (22) having at least one of an accelerometer, a gyroscope and a magnetometer, a global navigation satellite system sensor (23), a treatment input unit configured to receive a treatment input from a user, a processing unit configured to, receive an image signal from the multispectral camera (21 ), receive a location signal from at least one of the global navigation satellite system sensor (23) and the inertial measurement unit (22), receive a map data signal from the memory unit, and receive a treatment input from a user.

20. The system for producing a treatment prescription map of plants (11 ) in an agricultural field (10) according to claim 19, wherein the processing unit is configured to define a plant treatment need of a plant (10) as defined in any one of claims 11 to 18.

Images