Apparatus, evaluation device, evaluation system and method for analyzing the condition of a plant

Abstract

The aim of the invention is to analyze the condition of a plant. In order to achieve this aim, an apparatus (10) comprises a near-field camera (21) and a spectral analytical measuring device (22). The near-field camera (21) has a first field of view (41, 44) relative to a housing (20) of the apparatus (10). The spectral analytical measuring device (22) has a second field of view (42) relative to the housing (20), wherein the second field of view (42) is contained entirely in the first field of view (41, 44).

Description

[0001] DEVICE, EVALUATION DEVICE, EVALUATION SYSTEM AND METHOD FOR A CONDITION ANALYSIS OF A PLANT

[0002] TECHNICAL FIELD

[0003] The invention relates to devices, evaluation devices or systems, and methods that can be used in connection with a plant condition analysis. The invention particularly relates to such devices, systems, and methods that can support wavelength-resolved data acquisition, which can be used to detect a plant condition, for example, its nutrient and / or water supply, as well as pathological conditions. The invention particularly relates to such devices, systems, and methods that can be used with crop plants.

[0004] BACKGROUND

[0005] Plant condition analysis is of great importance. It serves to ensure adequate nutrient supply and / or to detect possible diseases.

[0006] Conventional methods for analyzing plant health have traditionally relied on human expert assessment. Techniques that use measuring devices to objectively and quantitatively analyze plant health are becoming increasingly important. For example, such techniques can be used to detect over- or under-supply of certain nutrients.

[0007] WO 2019 / 169434 A1, US Pat. No. 7,804,588 B2, US Pat. No. 11,320,307 B2, and DE 10 2018 103 509 disclose exemplary techniques for this. The sensor, designated HarvestLab™ 3000, described in the product brochure "JOHN DEERE TECHNOLOGIES FOR PRECISION AGRICULTURE," YY2314884GER_DE 01 / 23, features a near-infrared sensor.

[0008] Even when using modern measurement techniques (e.g., spectral analysis) for condition analysis, data acquisition can require expert knowledge in the field. Data acquisition at an unfavorable location on the plant can lead to inaccurate condition analysis results. An example of this is the acquisition of spectral analysis data from a location where foreign substances are present on a plant leaf. Data acquisition that relies exclusively on the knowledge and experience of a human expert to select a suitable location can result in problems of a lack of objectivity in the assessment and / or a lack of comparability. Automatic data acquisition at multiple locations on the same plant allows for subsequent selection but has the disadvantage that large amounts of data must be recorded, possibly transmitted, and analyzed.Thus, there is still a need for methods, devices and systems that offer mechanical support, optionally even further automation, in data acquisition, transmission and / or evaluation for plant condition analysis.

[0009] SUMMARY

[0010] The invention is based on the object of providing improved methods, devices, and systems that can be used for analyzing the condition of plants. In particular, the invention is based on the object of providing methods, devices, and systems that provide improvements with regard to the acquisition, transmission, and / or evaluation of spectral analytical information for analyzing the condition of plants. The invention is also based on the object of providing such methods, devices, and systems with which the acquisition, transmission, and / or evaluation of spectral analytical information on crop plants or plants is guided, otherwise supported, and / or simplified.

[0011] According to the invention, a device, an evaluation device or an evaluation system, and a method are provided as defined in the independent claims. The dependent claims define preferred and advantageous embodiments.

[0012] According to one aspect, the invention relates to a device for analyzing the condition of a plant. The device comprises a housing, a near-field camera, and a spectral analysis detection device. The near-field camera is configured to capture at least one image of at least part of the plant. The spectral analysis detection device is configured to capture spectral analysis information of the plant. The near-field camera has a first field of view relative to the housing. The spectral analysis detection device has a second field of view relative to the housing. The second field of view is completely contained within the first field of view. The device has at least one interface configured to output the at least one image and the spectral analysis information or data derived therefrom.

[0013] The device provides various technical effects and advantages. The device enables the capture of at least one image that can be used to support data acquisition by the spectral analysis acquisition device or evaluation of the data acquired by the spectral analysis acquisition device. This can be achieved, for example, by user guidance during data acquisition by the spectral analysis acquisition device, wherein the user guidance depends on the at least one image acquired with the near-field camera. Alternatively or additionally, the image acquired with the near-field camera can be used to automatically control the spectral analysis acquisition device.Since the second field of view is completely contained within the first field of view, it can be ensured that image information that can be captured with the near-field camera is present at all locations for which data can be captured by the spectral analysis acquisition device. This makes the usually more time-consuming capture and / or evaluation of the spectral analysis data more efficient and less error-prone using the at least one image captured with the near-field camera. The at least one image captured with the near-field camera allows for objective assessment and support of the data capture by the spectral analysis acquisition device. Alternatively or additionally, the at least one image captured with the near-field camera allows for objective assessment and support of the data evaluation of the spectral analysis information. This reduces sources of human error.By combining the near-field camera and the spectral analysis device in the device (especially in the same housing), simple and reliable data acquisition for condition analysis can be achieved.

[0014] The device can be configured such that, based on the at least one image, a suitability assessment of the area of ​​the plant shown in the at least one image for the acquisition or evaluation of the spectral analytical information can be determined.

[0015] The device thus makes it possible to assess, based on the at least one image captured with the near-field camera, whether and / or at which locations on the plant (e.g., in which leaf regions) it is appropriate to capture the spectral analytical information before data capture of the spectral analytical information. Data capture of spectral analytical information captured at unsuitable positions on the plant can be avoided or reduced. Alternatively or additionally, the device makes it possible to assess, based on the at least one image captured with the near-field camera, whether it is appropriate to evaluate the spectral analytical information captured at one or more specific plant locations (e.g., leaf regions) before data evaluation of the spectral analytical information.Alternatively or additionally, the device enables the quality of the evaluation result to be assessed based on the at least one image acquired with the near-field camera after a data evaluation of the spectral analytical information. The at least one image acquired with the near-field camera is used for the assessment. This reduces sources of human error. The acquisition and / or evaluation of spectral analytical information acquired, for example, at unsuitable leaf locations can be avoided, which offers particular advantages with regard to processing time. Alternatively or additionally, after evaluation of the spectral analytical information, a quality assessment can be carried out based on the at least one image from the near-field camera in order to sort out potentially falsified evaluation results or to give them a low weighting in relation to other evaluation results of the spectral analytical information.

[0016] To enable the determination of such a suitability assessment, the device can have a controllable light source for illumination during capture of the at least one image with the near-field camera. Alternatively or additionally, the device can be configured to reduce, in particular shield, ambient light influences while the at least one image is captured with the near-field camera.

[0017] This enables the acquisition of at least one image under consistent conditions. This allows for an objective assessment of the suitability, for example, of the plant part depicted in the at least one image (e.g., a leaf) as a whole or of different leaf locations or regions for the acquisition, analysis, or subsequent evaluation of analysis results of the spectral analytical information.

[0018] The spectral analytical detection device can be configured to carry out the detection of the spectral analytical information depending on the at least one image, for example depending on the suitability assessment.

[0019] This allows the acquisition of spectral analytical information based on at least one image captured with the near-field camera. This can be achieved by user guidance during the acquisition of the spectral analytical information or by automatic control of the spectral analytical acquisition device. Data acquisition of spectral analytical information captured at unsuitable positions on the plant can be avoided or reduced. Human error sources are reduced.

[0020] The device can be configured to transmit the spectral analytical information or the data derived therefrom via the interface depending on the at least one image, for example depending on the suitability assessment, or to analyze it depending on the at least one image, for example depending on the suitability assessment.

[0021] This allows the transmission or analysis of spectral analytical information or data derived from it to be carried out based on at least one image acquired with the near-field camera. Analysis or transmission of spectral analytical information acquired at inappropriate positions on the plant can be avoided or reduced. Human error sources are reduced.

[0022] The suitability assessment can indicate the suitability of the plant part (e.g., a leaf) depicted in the at least one image for capturing and / or evaluating the spectral analytical information. The suitability assessment can be a binary value (suitable / unsuitable). The suitability assessment can be a value selected from a range of values ​​that quantifies the suitability of the plant part (e.g., a leaf) depicted in the at least one image for capturing and / or evaluating the spectral analytical information. The device can thus be used to check the suitability of the plant part currently adjacent to the device for capturing the spectral analytical information and / or evaluating it for condition analysis.

[0023] The value can be a value from a continuous range. This can continuously quantify suitability with a value from an interval from a lower limit (e.g., 0%) to an upper limit (e.g., 100%).

[0024] This allows suitability to be quantified.

[0025] The value can be a value from a discrete range, preferably from an ordinal scale. This can indicate various, easily understandable, different ratings.

[0026] This allows suitability to be determined in a way that is easy to understand for a user.

[0027] The suitability assessment can include the suitability of different positions or areas of the plant for determining nutrient concentrations. Alternatively or additionally, the suitability assessment can include the suitability of different positions or areas of the plant for determining a health status. For example, the suitability assessment can include the suitability of different positions or areas of the plant for determining water content and / or pathological conditions. In each case, the determination can be made based on spectral analysis information.

[0028] This allows, depending on which further investigation is to be carried out on the basis of the spectral analytical information, to automatically (in particular processor-assisted) assess on the basis of the at least one image captured with the near-field camera at which positions or areas of the plant the spectral analytical information is to be recorded in order to enable the most unadulterated automatic, spectral analytical determination of nutrient concentrations, state of health, water content and / or pathological conditions.

[0029] The device can be configured such that the spectral analysis detection device acquires the spectral analysis information depending on the suitability of different positions or areas. The device can be configured to control the spectral analysis detection device depending on the at least one image acquired with the near-field camera. The control can comprise an active adjustment of optical components for shifting the second field of view relative to the housing.

[0030] This allows the acquisition of spectral analytical information based on at least one image captured with the near-field camera. This can be achieved by automatically controlling the spectral analytical acquisition device. Data acquisition of spectral analytical information captured at unsuitable positions on the plant can be avoided or reduced. Human error sources are reduced.

[0031] The device may be configured to receive the suitability assessment via the interface.

[0032] This enables the suitability assessment to be determined by at least one computing system (e.g., a computer or server system) remotely from the device. This allows the device to be designed as a compact unit, while the suitability assessment can be determined remotely from the device. To determine the suitability assessment, the device can be configured to transmit the at least one image captured by the near-field camera via the interface and subsequently receive the suitability assessment.

[0033] The device may be configured to determine the suitability assessment based on the at least one image captured by the near-field camera.

[0034] This allows the suitability assessment to be determined locally within the device. This can be particularly advantageous if the device has the necessary computing power.

[0035] The device may comprise a data processing circuit configured to process the at least one image captured with the near-field camera. The data processing circuit may be configured to perform image segmentation of the at least one image to determine the suitability assessment. The data processing circuit may be configured to enable or automatically execute control of the spectral analysis detection device based on the image segmentation.

[0036] This allows at least one image captured by the near-field camera to be processed automatically.

[0037] The suitability assessment can define the suitability of several different positions or areas on a leaf of the plant for the acquisition and / or evaluation of the spectral analytical information.

[0038] This makes it possible to quantitatively determine, based on at least one image captured by the near-field camera, at which positions or areas of the leaf the acquisition and / or evaluation of the spectral analytical information is useful.

[0039] Alternatively or additionally, an evaluation of the spectral analytical information can be carried out on the basis of the at least one image captured by the near-field camera depending on the position or area of ​​the leaf at which the spectral analytical information is or was obtained.

[0040] The suitability assessment can define suitability pixel-by-pixel or region-by-region. This allows the quantitative determination of the positions or regions of the leaf where the acquisition and / or evaluation of spectral analytical information is appropriate, based on at least one image captured with the near-field camera.

[0041] The suitability rating may include a numerical value for multiple pixels of the at least one image captured with the near-field camera, indicating whether reliable condition analysis results can be expected for a location or area defined by the pixel. The numerical value may be a binary value indicating whether or not the location or area is suitable for spectral data acquisition for condition analysis. The numerical value may be selected from a range of values ​​to further quantify suitability.

[0042] This makes it possible to determine which locations or areas are suitable for acquiring spectral analytical information. Alternatively or additionally, it is possible to determine which location or area is best suited for acquiring spectral analytical information.

[0043] The suitability assessment can indicate at which pixels of the at least one image captured by the near-field camera foreign substances (for example liquids, dust or other particles or insects) are superimposed on the imaged plant part and / or at which pixels of the at least one image captured by the near-field camera a degenerate condition is present, for example necrosis, chlorosis, anthocyanosis (i.e. an accumulation of anthocyanins which is noticeable in numerous plant species through discoloration of leaves), disease or mechanical damage.

[0044] This allows the spectral analytical information to be recorded and / or analyzed specifically at a location on the plant where the validity of a condition analysis is neither impaired by foreign substances nor by local degeneration.

[0045] The device may include a device for reducing ambient light influences. The device for reducing ambient light influences may include a screen provided on the housing. The screen may protrude at least partially outward around a viewing window held in the housing.

[0046] This facilitates the acquisition of at least one image under consistent conditions. This allows for an objective, automatic assessment of the suitability of, for example, different leaf locations or areas for the acquisition, analysis, or subsequent evaluation of analysis results of the spectral analytical information.

[0047] The near-field camera can comprise a camera and near-field optics for capturing a plant part in contact with the aperture. This allows the at least one image to be captured under consistent conditions and in a field situation in which the device is already positioned in contact with a potential sample location.

[0048] The device may be configured to capture the at least one image and the spectral analytical information while the device (for example the aperture) is in contact with a leaf of the plant.

[0049] This makes it easier to capture at least one image under consistent conditions.

[0050] The device may comprise a light source for illuminating the at least one image when capturing the near-field camera.

[0051] This facilitates the acquisition of at least one image under consistent conditions. This allows for an objective, automatic assessment of the suitability of, for example, different leaf locations or areas for the acquisition, analysis, or subsequent evaluation of analysis results of the spectral analytical information.

[0052] The device may comprise illumination optics configured to illuminate at least the first field of view during the acquisition of the at least one first image. The illumination optics may be configured such that the light rays from the light source impinge on the sample as parallel beams and are recorded by the camera optics due to the sample reflection.

[0053] This ensures good illumination under consistently reproducible conditions, facilitating the automatic evaluation of at least one image.

[0054] The illumination optics may comprise at least one paraboloid reflector and / or at least one facet reflector.

[0055] This enables consistent illumination with parallel light when capturing at least one image.

[0056] The illumination optics can be designed so that the light source acts as a point light source in the sample measuring plane.

[0057] This enables consistent illumination with parallel light when capturing at least one image.

[0058] Alternatively or additionally, the device can be configured so that camera calibration or referencing can be performed using a diffuse white / gray standard during device initialization. This can occur once or repeatedly. During recurring white / gray standard measurements, intensity and / or spectral changes of the light source can be detected and compensated. This facilitates automatic evaluation of the at least one image to determine the suitability assessment.

[0059] The near-field camera may have at least one color channel, at least two color channels, or three or more color channels. The near-field camera may have a camera with three color channels, for example, an RGB camera.

[0060] This facilitates the automatic determination of suitability assessments. For example, a color camera image (e.g., with RGB color channels) can be advantageous, for example, to assess the plant leaf stage or condition or to distinguish between color inhomogeneities and glare over the leaf surface.

[0061] The device can be designed as a movable device.

[0062] This enables its use in the field on living plants.

[0063] The device can be designed as a manually held device.

[0064] This facilitates use in the field on living plants without necessarily having vehicle access to the plant.

[0065] The device can be attached to a vehicle, in particular an agricultural vehicle. Accordingly, according to one embodiment, a vehicle is provided with the device according to the invention mounted thereon.

[0066] This facilitates use in the field on living plants using a vehicle.

[0067] The device can be attached to a robot. Accordingly, according to one embodiment, a robot is provided that has the device according to the invention. The robot can be movably mounted, for example, on a rail system or in another manner.

[0068] This makes it easier to use in greenhouses, for example.

[0069] The device can be attached to a flying object. Accordingly, according to one embodiment, a flying object is provided that has the device according to the invention.

[0070] This facilitates use in the field on living plants using the flying object.

[0071] The vehicle, robot or flying object may comprise a receiver for receiving remote control signals and at least one actuator for moving the vehicle, robot or flying object in response to the remote control signals for positioning the device.

[0072] This enables control from a control position remote from the device.

[0073] The vehicle, robot, or flying object may have a controller configured to move the vehicle, robot, or flying object autonomously or semi-autonomously to position the device. This achieves a higher degree of automation.

[0074] The vehicle, robot or flying object may have a gripper arm for gripping and positioning a plant part relative to the device.

[0075] This allows even parts of the plant that are difficult to access to be correctly positioned in relation to the device.

[0076] According to a further aspect or embodiment, an evaluation device or an evaluation system for analyzing the condition of a plant is provided, comprising: at least one evaluation device interface configured to receive at least one image of at least one part of the plant, and an evaluation circuit. The evaluation circuit is configured to determine, based on the at least one image, a suitability assessment of the region of the plant depicted in the at least one image for the acquisition or evaluation of spectral analytical information.

[0077] The evaluation device or evaluation system provides various technical effects and advantages. The evaluation device or evaluation system uses the at least one image to support data acquisition by the spectral analysis acquisition device or evaluation of the data acquired by the spectral analysis acquisition device. As a result, the at least one image (which may be or may include at least one near-field image) can be used to support the condition analysis based on the spectral analysis information.

[0078] The evaluation circuit may be configured to detect one or more conditions selected from a group consisting of: presence of foreign matter, necrosis, chlorosis, anthocyanosis, damaged plant areas, pest infestation, incorrect positioning of the area of ​​the plant during image capture, crack or break in a part of the plant for suitability assessment.

[0079] This makes it possible to determine which locations or areas are suitable for acquiring or evaluating the spectral analytical information. Alternatively or additionally, it is possible to determine which location or area is best suited for acquiring or evaluating the spectral analytical information.

[0080] The suitability assessment can indicate the suitability of the plant part (e.g., a leaf) depicted in the at least one image for capturing and / or evaluating the spectral analytical information. The suitability assessment can be a binary value (suitable / unsuitable). The suitability assessment can be a value selected from a range of values ​​that quantifies the suitability of the plant part (e.g., a leaf) depicted in the at least one image for capturing and / or evaluating the spectral analytical information. This allows the suitability of the plant part depicted in the at least one image for capturing the spectral analytical information and / or evaluating it for condition analysis to be tested.

[0081] The suitability assessment can indicate pixel-by-pixel or area-by-area which locations or areas are suitable for the acquisition or evaluation of the spectral analytical information.

[0082] This makes it possible to quantitatively determine, based on at least one image, at which positions or areas of a leaf the acquisition and / or evaluation of the spectral analytical information is useful.

[0083] The suitability rating may include a numerical value for multiple pixels of the at least one image captured with the near-field camera, indicating whether reliable condition analysis results can be expected for a location or area defined by the pixel. The numerical value may be a binary value indicating whether or not the location or area is suitable for spectral data acquisition for condition analysis. The numerical value may be selected from a range of values ​​to further quantify suitability.

[0084] This makes it possible to determine which locations or areas are suitable for acquiring spectral analytical information. Alternatively or additionally, it is possible to determine which location or area is best suited for acquiring spectral analytical information.

[0085] The evaluation device or the evaluation system can be configured to take into account a number and / or geometric arrangement of several pixels relative to one another when determining the suitability assessment.For example, the evaluation device or the evaluation system can be configured to evaluate an area as suitable for spectral analysis only if at least two or more conditions are cumulatively met, such as: (a) none of the checked possible interference effects are present at the pixels; (b) the total number of such neighboring pixels satisfies a threshold comparison (e.g., is greater than a threshold); (c) the shape and / or total area of ​​the neighboring pixels satisfies one or more geometric criteria (e.g., a size of more than a predetermined minimum dimension in one or two mutually orthogonal directions, a size that satisfies an inscription of a circle with a predetermined minimum dimension, and / or an aspect ratio of the region defined by the pixels satisfies a further threshold condition).

[0086] This ensures that an area - for example at least the second

[0087] Field of view—is free from potential interference to capture the spectral analytical information. This allows for more precise control, for example, by defining maximum areas where certain interference factors must not be present.

[0088] The suitability assessment can indicate at which pixels of the at least one image captured with the near-field camera foreign substances (for example liquids, dust or other particles or insects) are superimposed on the imaged plant part and / or at which pixels of the at least one image captured with the near-field camera a degenerated condition, for example necrosis, chlorosis, anthocyanosis or mechanical damage, is present.

[0089] This allows the spectral analytical information to be recorded and / or analyzed specifically at a location on the plant where the validity of a condition analysis is neither impaired by foreign substances nor by local degeneration.

[0090] The evaluation device or system may have a human-machine interface. The human-machine interface may be configured to receive a user input specifying one, several, or all of the following: plant stage, leaf stage, plant species, plant variety, at least one nutrient to be analyzed, and at least one disease to be tested.

[0091] This allows the evaluation device or system to evaluate at least one image based on the user input. In particular, this prevents unnecessary, unwanted, or unsuitable evaluations from being performed for the sample in question.

[0092] The evaluation circuit may be configured to perform the suitability assessment depending on the information or information contained in the user input.

[0093] As a result, the evaluation device or the evaluation system can evaluate the at least one image depending on the user input in such a way that those areas or locations (for example of a leaf) are determined at which a condition analysis according to the user input is reliable on the basis of the spectral analytical information.

[0094] The at least one evaluation device interface can be configured to output the suitability assessment for use in the acquisition of spectral analytical information.

[0095] This allows the acquisition of the spectral analytical information by a device that is separate from the evaluation device or system, in accordance with the suitability assessment.

[0096] The at least one evaluation device interface can be configured to transmit the suitability assessment to the device according to an aspect or embodiment of the

[0097] The invention can be transferred to the at least one evaluation device interface. The at least one evaluation device interface can be configured to receive the at least one image and the spectral analysis information from this device.

[0098] As a result, the evaluation device or the evaluation system can interact specifically with the device according to the invention.

[0099] The evaluation device or the evaluation system can be configured to control the at least one evaluation device interface or the human-machine interface in order to enable the user to select leaf areas to be analyzed.

[0100] This enables a user-defined selection of the leaf areas to be analyzed, which can be carried out on the evaluation device or the evaluation system or, under the control of the same, on the device according to the invention.

[0101] The evaluation circuit can be configured to perform the suitability assessment without using spectral analytical information.

[0102] This allows the evaluation device or the evaluation system to determine, on the basis of the at least one image, at which points or areas of the plant the spectral analytical information should be acquired in order to be able to carry out the condition analysis reliably, even before the more time-consuming acquisition of the spectral analytical information.

[0103] The evaluation circuit can be configured to evaluate the at least one image using at least one trained machine learning model for suitability assessment.

[0104] This allows the evaluation device or system to perform the suitability assessment in a data-driven and objective manner. Annotated image data of plants or plant parts is used to train the machine learning model. The annotations indicate which locations or areas are suitable for condition analysis based on spectral analysis information. Expert knowledge is required for creating the annotated training data, but not for applying the trained machine learning model.

[0105] The trained machine learning model may be configured to receive pixel values ​​of the at least one image as input.

[0106] This allows the at least one image to be easily processed by the evaluation device or system. Optional preprocessing of the at least one image, for example, by applying filters, edge detection methods, or other preprocessing techniques, is possible.

[0107] The trained machine learning model can be configured to output the suitability score.

[0108] The evaluation device interface can be configured to receive spectral analytical information from the plant or data derived therefrom. This allows the evaluation device or evaluation system to not only determine the suitability assessment but also perform the condition analysis based on the spectral analytical information.

[0109] The evaluation device or evaluation system can be configured to evaluate the spectral analytical information in order to perform the condition analysis of the plant. The evaluation device or evaluation system can be configured such that the spectral analytical information itself, its evaluation, or an evaluation of the evaluation result is carried out depending on the suitability assessment determined from the at least one image.

[0110] This allows the evaluation device or system not only to determine the suitability assessment but also to use it in the condition analysis based on the spectral analytical information.

[0111] The evaluation circuit can be configured to carry out at least one nutrient analysis based on the spectral analytical information or the data derived therefrom.

[0112] This allows the evaluation device or system to perform a status analysis that includes determining at least one nutrient concentration. The evaluation device or system can be configured to determine, for one or more nutrients, whether the plant is oversupplied or undersupplied with the respective nutrient, depending on the evaluation of the spectral analysis information.

[0113] The evaluation device or evaluation system can be structurally integrated into the housing of the device. The evaluation device or evaluation system can be separate from the device with which the at least one image and the spectral analytical information are acquired.

[0114] According to a further aspect of the invention, a system is provided which comprises the device according to one aspect or embodiment and the evaluation device or the evaluation system according to one aspect or embodiment.

[0115] The effects achieved with the system correspond to the effects described with reference to the device according to the invention and the evaluation device or the evaluation system according to the invention.

[0116] According to a further aspect of the invention, a method for a condition analysis of a plant is provided, the method comprising: receiving at least one image of the plant and determining a suitability assessment of the plant for a detection or evaluation of spectral analytical information of the plant, wherein the suitability assessment is determined based on the at least one image, and wherein the detection and / or evaluation of the spectral analytical information of the plant is carried out depending on the suitability assessment.

[0117] The method provides various technical effects and advantages. The at least one image is used to support data acquisition and / or evaluation of the spectral analytical information. Thus, the at least one image (which may be or may include at least a near-field image) can be used to support the condition analysis based on the spectral analytical information.

[0118] The method can be carried out automatically by the evaluation device, the evaluation system, the apparatus or the system according to one aspect or embodiment.

[0119] The method may further comprise outputting the suitability assessment via a human-machine interface for visualizing the suitability assessment.

[0120] This allows the acquisition of spectral analytical information.

[0121] The method may comprise detecting one or more conditions selected from a group consisting of: presence of foreign matter, necrosis, chlorosis, anthocyanosis, spoiled plant areas, pest infestation, incorrect positioning of the area of ​​the plant during image acquisition, crack or breakage in a part of the plant.

[0122] This makes it possible to determine which locations or areas are suitable for acquiring or evaluating the spectral analytical information. Alternatively or additionally, it is possible to determine which location or area is best suited for acquiring or evaluating the spectral analytical information.

[0123] In the procedure, the suitability assessment can indicate pixel by pixel or area by area which locations or areas are suitable for the acquisition or evaluation of the spectral analytical information.

[0124] This makes it possible to quantitatively determine, based on at least one image, at which positions or areas of a leaf the acquisition and / or evaluation of the spectral analytical information is useful.

[0125] In the method, the suitability assessment for a plurality of pixels of the at least one image captured with a near-field camera may include a numerical value indicating whether reliable condition analysis results can be expected for a location or area defined by the pixel. The numerical value may be a binary value indicating whether or not the location or area is suitable for spectral data acquisition for condition analysis. The numerical value may be selected from a range of values ​​to further quantify the suitability.

[0126] This makes it possible to determine which locations or areas are suitable for acquiring spectral analytical information. Alternatively or additionally, it is possible to determine which location or area is best suited for acquiring spectral analytical information.

[0127] In the method, the suitability assessment can indicate the suitability of the plant part (e.g., a leaf) depicted in the at least one image for capturing and / or evaluating the spectral analytical information. The suitability assessment can be a binary value (suitable / unsuitable). The suitability assessment can be a value selected from a range of values ​​that quantifies the suitability of the plant part (e.g., a leaf) depicted in the at least one image for capturing and / or evaluating the spectral analytical information.

[0128] As a result, the at least one image can be used to check the suitability of the plant part currently adjacent to the device used to capture the at least one image for capturing the spectral analytical information and / or evaluating it for condition analysis.

[0129] The value can be a value from a continuous range. This can continuously quantify suitability with a value from an interval from a lower limit (e.g., 0%) to an upper limit (e.g., 100%).

[0130] This allows suitability to be quantified.

[0131] The value can be a value from a discrete range, preferably from an ordinal scale. This can indicate various, easily understandable, different ratings.

[0132] This allows suitability to be determined in a way that is easy to understand for a user.

[0133] In the method, the suitability assessment can indicate at which pixels of the at least one image captured with the near-field camera foreign substances (for example liquids, dust or other particles or insects) are superimposed on the imaged plant part and / or at which pixels of the at least one image captured with the near-field camera a degenerate condition, for example necrosis, chlorosis, anthocyanosis or mechanical damage, is present.

[0134] This allows the spectral analytical information to be recorded and / or analyzed specifically at a location on the plant where the validity of a condition analysis is neither impaired by foreign substances nor by local degeneration.

[0135] The method may include controlling a human-machine interface to enable receipt of a user input specifying one, several, or all of the following information: plant stage, leaf stage, plant species, plant variety, at least one nutrient to be analyzed, and at least one disease to be tested. Depending on the user input, the at least one image can be evaluated to determine the suitability rating specifically for the user-defined information.

[0136] The procedure may perform the suitability assessment depending on the information or information contained in the user input.

[0137] As a result, depending on the user input, at least one image can be evaluated to determine those areas or locations (e.g. of a leaf) where a condition analysis based on the spectral analytical information is reliable according to the user input.

[0138] The method may comprise outputting the suitability assessment for use in acquiring spectral analytical information.

[0139] This allows the spectral analytical information to be collected in accordance with the suitability assessment.

[0140] In the method, the suitability assessment can be transferred to the device according to one aspect or embodiment of the invention.

[0141] As a result, the evaluation device or the evaluation system can interact specifically with the device according to the invention.

[0142] The method may comprise controlling an evaluation device interface or a human-machine interface to enable the user to select leaf areas to be analyzed.

[0143] This enables a user-defined selection of the leaf areas to be analyzed, which can be carried out on the evaluation device or the evaluation system or, under the control of the same, on the device according to the invention.

[0144] The method allows the suitability assessment to be carried out without using spectral analytical information.

[0145] This makes it possible to determine, on the basis of at least one image, at which points or areas of the plant the spectral analytical information should be recorded in order to be able to carry out the condition analysis reliably, even before the more time-consuming acquisition of the spectral analytical information.

[0146] The method may include evaluating the at least one image using at least one trained machine learning model to perform the suitability assessment.

[0147] This allows for a data-driven and objective suitability assessment. Annotated image data of plants or plant parts is used to train the machine learning model. The annotations indicate which locations or areas are suitable for condition analysis based on spectral analysis. Expert knowledge is required for creating the annotated training data, but not for applying the trained machine learning model.

[0148] The trained machine learning model may be configured to receive pixel values ​​of the at least one image as input.

[0149] This allows the at least one image to be easily processed by the evaluation device or system. Optional preprocessing of the at least one image, for example, by applying filters, edge detection methods, or other preprocessing techniques, is possible.

[0150] The trained machine learning model can be configured to output the suitability score.

[0151] The method may further comprise receiving spectral analytical information of the plant or data derived therefrom.

[0152] This allows not only the suitability assessment to be determined, but also the condition analysis to be carried out based on the spectral analytical information.

[0153] The method may further comprise evaluating the spectral analytical information to perform the plant condition analysis. The spectral analytical information itself, its evaluation, or an evaluation of the evaluation result may be performed depending on the suitability assessment determined from the at least one image.

[0154] This allows not only the suitability assessment to be determined, but also the condition analysis to be carried out based on the spectral analytical information.

[0155] The method may comprise performing at least one nutrient analysis based on the spectral analytical information or the data derived therefrom.

[0156] This allows a status analysis to be performed, which includes the determination of at least one nutrient concentration. For one or more nutrients, depending on the evaluation of the spectral analysis information, it can be determined whether the plant is oversupplied or undersupplied with the respective nutrient.

[0157] The method may comprise using a result of the condition analysis to adjust cultivation conditions.

[0158] This means that the results of the condition analysis can be used to improve the supply of crops or plants based on the condition analysis.

[0159] The method may comprise capturing the at least one image and / or the spectral analytical information by the device according to one aspect or embodiment.

[0160] This allows the effects described with reference to the device according to the invention to be achieved. The method may include receiving the suitability assessment by the device, wherein the spectral analytical information is acquired depending on the suitability assessment.

[0161] This allows the time-consuming acquisition of spectral analytical information to be carried out in a targeted manner using the suitability assessment. The risk of human error and / or acquisition of spectral analytical information at an unsuitable location is reduced.

[0162] According to a further aspect of the invention, a method for cultivating plants is provided, which comprises the method for carrying out a condition analysis, wherein at least one cultivation condition is set depending on a result of the condition analysis.

[0163] This means that the results of the condition analysis can be used to improve the supply of crops or plants based on the condition analysis.

[0164] According to a further aspect of the invention, a method for cultivating plants is provided, which comprises a condition analysis using the device, the evaluation device or the evaluation system according to an aspect or embodiment of the invention.

[0165] As a result, the effects described with reference to the device according to the invention and the evaluation device or the evaluation system of the invention are achieved.

[0166] In the method, the at least one image and the spectral analytical information can be obtained with the device, wherein the evaluation device or the evaluation system uses the at least one image captured with the near-field camera to support the condition analysis on the basis of the spectroscopic information.

[0167] In the method, at least one cultivation condition can be set depending on a result of the condition analysis.

[0168] This means that the results of the condition analysis can be used to improve the supply of crops or plants based on the condition analysis.

[0169] According to a further aspect of the invention, machine-readable instruction code is provided which, when executed by a programmable computing unit, carries out the method according to one aspect or embodiment of the invention.

[0170] According to a further aspect of the invention, a storage medium is provided with machine-readable instruction code stored thereon, which, when executed by a programmable computing unit, carries out the method according to one aspect or embodiment of the invention.

[0171] The devices, methods, systems, and system components according to exemplary embodiments can also be used in other fields of agricultural technology, for example for animal examinations. According to a further aspect, the invention relates to a device for analyzing the condition of animals. The device comprises a housing, a near-field camera, and a spectral analysis detection device. The near-field camera is configured to capture at least one image of at least part of the animals. The spectral analysis detection device is configured to capture spectral analysis information of the animals. The near-field camera has a first field of view relative to the housing. The spectral analysis detection device has a second field of view relative to the housing. The second field of view is completely contained within the first field of view.The device has at least one interface which is configured to output the at least one image and the spectral analytical information or data derived therefrom.

[0172] According to a further aspect or embodiment, an evaluation device or an evaluation system for analyzing the condition of animals is provided, comprising: at least one evaluation device interface configured to receive at least one image of at least a portion of the animals, and an evaluation circuit. The evaluation circuit is configured to determine, based on the at least one image, a suitability assessment of the area of ​​the animals represented in the at least one image for the acquisition or evaluation of spectral analytical information.

[0173] According to a further aspect of the invention, a system is provided which comprises the device for a condition analysis of animals according to one aspect or embodiment and the evaluation device or the evaluation system according to one aspect or embodiment.

[0174] According to a further aspect of the invention, a method for a condition analysis of animals is provided, the method comprising: receiving at least one image of the animals and determining a suitability assessment of the animals for a detection or evaluation of spectral analytical information of the animals, wherein the suitability assessment is determined based on the at least one image, and wherein the detection and / or evaluation of the spectral analytical information of the animals is carried out depending on the suitability assessment.

[0175] Further advantageous features of these devices, methods, evaluation devices and evaluation systems for application to animals correspond to the features and effects explained with reference to plant analysis.

[0176] The devices, methods, systems, and system components according to embodiments of the invention achieve various effects. In particular, the acquisition and / or evaluation of spectral analytical information can be supported using one or more images acquired with a near-field camera. The devices, methods, systems, and system components can be used in various fields. These include, but are not limited to, the condition analysis of crops or cultivated plants in agricultural engineering.

[0177] BRIEF DESCRIPTION OF THE CHARACTERS

[0178] Embodiments of the invention are described with reference to the figures. In the figures, similar or identical reference numerals designate elements with similar or identical design and / or function.

[0179] Figure 1 is a block diagram of a device according to an embodiment.

[0180] Figure 2 shows a schematic representation of the device.

[0181] Figure 3 is a schematic representation of fields of view of a near-field camera and a spectral analytical detection device of the device.

[0182] Figure 4 is a schematic representation of fields of view of a near-field camera and a spectral analytical detection device of the device.

[0183] Figure 5 is a flowchart of a method according to an embodiment.

[0184] Figure 6 is a schematic representation of a system according to an embodiment.

[0185] Figure 7 is a diagram explaining the operation of the system according to an embodiment.

[0186] Figure 8 is a block diagram of an evaluation device or an evaluation system according to an embodiment.

[0187] Figure 9 is a block diagram of a device according to another embodiment.

[0188] Figure 10 is a schematic representation of a plant part to explain the functioning of embodiments.

[0189] Figure 11 is a schematic representation of a human-machine interface to explain the functionality of embodiments.

[0190] Figure 12 is a flowchart of a method according to an embodiment.

[0191] Figure 13 is a diagram to explain the functioning of the evaluation device or the evaluation system according to an embodiment.

[0192] Figure 14 shows an implementation of a suitability assessment using a support vector machine in the evaluation device or the evaluation system according to an embodiment.

[0193] Figure 15 shows an implementation of a suitability assessment using a machine learning model designed as a neural network in the evaluation device or the evaluation system according to one embodiment.

[0194] Figure 16 is a flowchart of a method according to one embodiment. Figure 17 is a schematic representation of spectral analytical information to explain the functionality of the evaluation device or the evaluation system according to one embodiment.

[0195] Figure 18 is a schematic representation of a result of a principal component analysis to explain the functioning of the evaluation device or the evaluation system according to an embodiment.

[0196] Figure 19 is a flowchart of a method according to an embodiment.

[0197] Figure 20 is a block diagram of a device according to another embodiment.

[0198] Figure 21 shows a schematic representation of a near-field camera of the device according to embodiments.

[0199] Figure 22 shows a schematic representation of the device when it is designed as a hand-held device.

[0200] Figure 23 shows a schematic representation of an agricultural vehicle according to an embodiment.

[0201] Figure 24 shows a schematic representation of a robot according to an embodiment.

[0202] Figure 25 shows a schematic representation of a flying object according to an embodiment.

[0203] Figure 26 is a flowchart of a method according to an embodiment.

[0204] DETAILED DESCRIPTION OF EMBODIMENTS

[0205] Embodiments of the invention are described with reference to the figures. In the figures, similar or identical reference numerals designate elements with similar or identical design and / or function.

[0206] While embodiments are described in connection with a nutrient analysis or other condition analysis of a crop or plant, the embodiments are not limited thereto.

[0207] The features of the embodiments can be combined with each other unless this is expressly excluded in the following description.

[0208] Devices, evaluation devices and systems and methods according to embodiments of the invention are configured to use at least one image taken with a near-field camera, which image shows at least part of a plant, to support the acquisition, evaluation or assessment of spectral analytical information.This can be done in different ways, for example by using the at least one image captured with the near-field camera to control a human-machine interface to indicate which positions or areas of, for example, a leaf are suitable for capturing the spectral analytical information, using the at least one image captured with the near-field camera to automatically control a spectral analytical capturing device, using the at least one image captured with the near-field camera to determine whether captured spectral analytical information should be transmitted or evaluated in order to carry out the condition analysis, and / or using the at least one image captured with the near-field camera to subsequently evaluate, for example, weight, a condition analysis carried out on the basis of the spectral analytical information.

[0209] The term "spectral analytical information" as used here encompasses an optical measurement quantity recorded for multiple (two or more) wavelengths at the same spatial measurement area of ​​the plant. The measurement quantity may, for example, include reflectivity, reflected light intensity, scattered light intensity, Raman scattered light intensity, fluorescent light intensity, phosphor light intensity, or other optical measurement quantity recorded for the multiple wavelengths, which can also be determined, for example, in a transmission setup (e.g., by recording transmitted intensity as a function of wavelength) or a transflection setup.

[0210] The term "spectro-analytical detection device" as used here encompasses a device that is capable of and configured to detect an optical measurement variable detected for multiple (two or more than two) wavelengths at the same measurement area of ​​the plant. The spectro-analytical detection device may comprise a spectrometer. However, this is not necessarily required. For example, as an alternative or in addition to using a spectrometer, the spectro-analytical detection device may be configured to detect the optical measurement variable for the multiple wavelengths sequentially over time, for example, by actively irradiating with different wavelengths and determining the intensity of the scattered or reflected light detected in each case.

[0211] The term near-field camera, as used here, includes in particular an image capture device with near-field optics that allows the sample (e.g. a leaf of the plant) to be positioned in proximity to the near-field camera (e.g. in direct contact with an aperture of a device in whose housing the near-field camera is installed).

[0212] The term near-field image as used here refers to an image captured with the near-field camera.

[0213] The term "device for a condition analysis of a plant," as used here, refers to a device that can be used for a condition analysis of a plant. The device is specifically configured to capture at least some, and advantageously all, of the recordings (images and spectral analytical information) required for the condition analysis. It is possible, but not mandatory, for the evaluation for the condition analysis to be carried out by the device itself. Rather, the term "device for a condition analysis of a plant" also includes devices that enable the capture of at least one image with a near-field camera and spectral analytical information on an object (in particular a leaf of a plant or another plant part).

[0214] The term plant, as used here, also includes eukaryotes, especially algae.

[0215] The invention provides techniques that enable the analysis of a plant's condition to be supported using at least one image captured with a near-field camera. For this purpose, a near-field camera, which has near-field optics, and a spectral analysis device (e.g., a spectrometer) can be installed in a housing of a device. The at least one near-field image captured with the near-field camera is automatically analyzed. This can occur either locally in the device or in a separate computing system.

[0216] According to the invention, a device thus comprises both a near-field camera (which may in particular comprise a camera and a near-field lens) and a spectral analysis detection device. The near-field camera has a field of view that may be approximately equal to or larger than the field of view of the spectral analysis detection device. Thus, the at least one image captured with the near-field camera covers at least the area of ​​the object to be measured (in particular a plant part, for example, a leaf) that is used for measuring the spectral analysis information). The device can first capture the at least one image of the object to be measured with the near-field camera.The at least one image can be used to perform an automated evaluation of the object to be measured for properties intended for a condition analysis (such as absence of foreign matter, correct leaf side, correct positioning, absence of necrosis, absence of chlorosis, absence of anthocyanosis). This automated evaluation can be performed by the device itself or remotely from the device by an evaluation device or an evaluation system that is in communication with the device. The result of the evaluation of the at least one image can be used in a condition analysis based on spectral analytical information.

[0217] By analyzing at least one image captured by the near-field camera, positions and / or areas of a plant that are unsuitable for a condition analysis based on spectral analysis information can be automatically identified. Reasons for this could include, for example, foreign matter present on the plant, mechanical defects of the plant, diseases, necrosis, chlorosis, or anthocyanosis. Upon detection of such situations that impair the reliability of a condition analysis based on spectral analysis information, at least one automatic action can be taken, for example, one or more of the following actions:

[0218] Control of a human-machine interface to indicate that the reliability of the condition analysis based on the spectral analytical information may be compromised; automatic control of the spectral analytical acquisition device to prevent acquisition of the spectral analytical information at points where the reliability of the condition analysis based on the spectral analytical information may be compromised;

[0219] Preventing further analysis of the spectral analytical information if the reliability of the condition analysis based on the spectral analytical information may be impaired; and / or

[0220] Evaluation of a condition analysis, for example by giving low weight to the analysis result if the condition analysis was carried out on the basis of spectral analytical information, if the reliability of the condition analysis based on the spectral analytical information may be impaired.

[0221] Figure 1 shows a device 10 for analyzing the condition of a plant 2. The device comprises a housing 20. The device 10 comprises a near-field camera 21 arranged on or in the housing 20. The device 10 comprises a spectral analysis detection device 22 arranged on or in the housing 20.

[0222] The near-field camera 21 is configured to capture at least one image of at least a part (e.g., a leaf 3) of the plant 2. The spectral analysis detection device 22 is configured to capture spectral analysis information of the plant. The near-field camera 21 has a first field of view 41 relative to the housing 20. The spectral analysis detection device 22 has a second field of view 42 relative to the housing. The second field of view 42 is completely contained within the first field of view 41. For example, the second field of view 42 and the first field of view 41 can be substantially identical. However, the second field of view 42 can also be smaller than the first field of view 41, with the second field of view 42 being contained within the first field of view.

[0223] The device has at least one interface 23, 24 configured to output the at least one image and the spectral analytical information. The at least one interface 23, 24 can also be configured to output data derived from the at least one image and / or the spectral analytical information. The at least one interface can have a data interface 23. The data interface 23 can have a wireless communication interface. The device 10 can be configured to output the at least one image captured with the near-field camera 21 (also referred to here as the near-field image) or data derived therefrom via the data interface 23.

[0224] The device 10 can be configured to receive an evaluation result determined from the at least one near-field image via the data interface. The evaluation result can include a suitability assessment indicating whether the plant section (e.g., leaf section) circumferentially surrounded by a diaphragm 29 of the device is suitable for data acquisition of the spectral analytical information and / or whether a condition analysis of the spectral analytical information acquired in this plant section can be expected to yield reliable results.

[0225] The evaluation result of the at least one near-field image can be used by the device 10 in various ways to influence the acquisition, evaluation, or subsequent assessment of the spectral analytical information. In this way, the device 10 can, for example, support the acquisition and / or evaluation of the spectral analytical information to enable a reliable condition analysis.

[0226] The device 10 can be configured to control the spectral analytical detection device 22 depending on the evaluation result of the at least one near-field image. For example, detection of the spectral analytical information can be prevented or made dependent on a dedicated user input if the evaluation result of the at least one near-field image indicates that the plant section adjacent to the device 10 is unsuitable for detecting and / or evaluating the spectral analytical information. The spectral analytical detection device 22 can alternatively or additionally be controlled such that the detection of the spectral analytical information occurs at a measuring point in the first field of view 41 at which, based on the evaluation result of the at least one near-field image, such detection is expedient with regard to the condition analysis.The control of the spectral analysis detection device 22 can be carried out by a detection device controller 27 of a processing circuit 26 of the device 10. The control of a human-machine interface 24 to enable dedicated user confirmation depending on the evaluation result of the at least one near-field image can be carried out by an interface controller 28 of the processing circuit 26 of the device 10.

[0227] The device 10 can be configured to control the human-machine interface 24 of the device depending on the evaluation result of the at least one near-field image. For example, the human-machine interface 24 can be controlled to indicate visually, acoustically, and / or tactilely whether it is advisable to acquire the spectral analytical information from the plant section currently adjacent to TI of the device 10. The human-machine interface 24 can have a display device that can be controlled by the interface controller 28 such that, with respect to the plant section currently adjacent to the device 10, the measurement points suitable for acquiring the spectral analytical information are displayed (for example, in the form of overlay(s) on the near-field image).

[0228] The device 10 can alternatively or additionally be configured to transmit the spectral analytical information via the data interface 23 for performing the condition analysis. The device 10 can use the evaluation result of the at least one near-field image to selectively transmit the spectral analytical information only if it is determined from the evaluation result of the at least one near-field image that the spectral analytical information is determined at a measurement point in the first field of view 41 that is suitable for the condition analysis.

[0229] The device 10 thus provides a device with which it can be checked whether a detection and / or evaluation of spectral analytical information in the plant section (for example leaf section) adjacent to the device 10 can be expected to yield reliable results of a condition analysis based on the spectral analytical information.

[0230] To perform the various control and processing functions, the device 10 has a processing circuit 26. The processing circuit 26 may comprise one or more integrated circuits to control the data interface 23, the human-machine interface, the spectral analysis detection device 22, and / or the near-field camera 21. The one or more integrated circuits may, for example, comprise any one or any combination of the following circuits or circuit components: an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a processor (e.g., a GPU ("graphic processor unit") or TPU ("tensor processor unit")), a controller, one or more quantum gates, a circuit for quantum information processing, and other integrated circuits.

[0231] The device 10 can have a storage system 25. The storage system 25 can store machine-readable instruction code which, when executed by the processing circuit 26, causes the functions and steps disclosed herein to be carried out. If the evaluation of the at least one near-field image is performed by the processing circuit 26, i.e., locally in the device 10, the storage system 25 can also store parameters that enable an evaluation of the at least one near-field image to determine whether the imaged plant part is suitable for detecting and / or evaluating the spectral analytical information. The parameters stored in the storage system 25 for evaluating the at least one near-field image can, for example, include one or more of the following parameters: filter parameters; definition of decision boundaries of a support vector model; parameters of a trained machine learning model.Further examples are described in more detail with reference to the evaluation device or the evaluation system.

[0232] The condition analysis based on the spectral analytical information can be performed locally in the device 10 or advantageously in a computer system remote from the device 10. The condition analysis, which may, for example, include the determination of nutrient concentrations based on the spectral analytical information, can be performed using techniques familiar to the person skilled in the art, such as those described, for example, in WO 2019 / 169434 A1, US Pat. No. 7,804,588 B2, and US Pat. No. 1,132,307 B2. The condition analysis can be performed using principal component techniques, partial least squares regression (PLS) techniques, or other techniques that allow a conclusion to be drawn as to whether a nutrient concentration for at least one, and advantageously several, nutrients lies within or outside a target range.

[0233] Figure 2 shows a schematic perspective view of the device 10. Various geometric shapes can be used for the housing of the device 10 and the aperture 29. Advantageously, the aperture 29 defines a continuous contact surface 34 for direct contact with the plant section (e.g., leaf). The aperture 29 reduces the influence of extraneous light.

[0234] A near-field optics system 31 of the near-field camera 21 is arranged such that the near-field optics system 31 images reflected and / or scattered light within the field of view surrounded by the contact surface 34 onto a detector surface (e.g., a light-sensitive semiconductor chip, such as a CMOS or CCD chip) of a camera. A light source system is arranged to actively illuminate the field of view of the near-field camera surrounded by the contact surface 34. An illumination optics system 33 can provide illumination, for example, with parallel light.

[0235] The spectral analytical detection device 22 may have optical components 32 in order to direct a sample beam onto the plant section adjacent to the contact surface 34 within the field of view surrounded by the contact surface 34 and to detect the reflected or scattered light of the sample beam.

[0236] Figures 3 and 4 show the first field of view 41 of the near-field camera 21 and the second field of view 42 of the spectral analysis detection device 22. The second field of view 42 may be substantially identical to the first field of view 41 or may be smaller than the first field of view 41. In both cases, the first field of view 41 of the near-field camera 21 contains the second field of view 42 of the spectral analysis detection device 22.

[0237] The device 10 can be designed such that a position and / or size of the second field of view 42 in an object plane 35 (which can be defined by the contact surface 34) can be changed. For example, a measurement position of the spectral analysis detection device can be moved by a displacement 43. Advantageously, the position and / or size of the measurement position at which the spectral analysis information is acquired can then be determined depending on the evaluation of the at least one near-field image. The processing device 26 can control the spectral analysis detection device 22 depending on the evaluation of the at least one near-field image, optionally also depending on a user input dependent on the near-field image evaluation.This allows spectral analytical information to be acquired in an area within the plant section adjacent to the contact surface 34, which area is determined, based on the evaluation of the near-field image, to be suitable for the acquisition of spectral analytical information for the purpose of condition analysis.

[0238] The position of the second field of view 42 relative to the first field of view 41 can be determined and known during field use. For example, calibration can be performed at the factory, and the corresponding position information can be stored non-volatilely in the device 10. Alternatively or additionally, the calibration can also be performed repeatedly, for example, after commissioning of the device 10, in order to determine the position information regarding the position of the second field of view 42 relative to the first field of view 41. This position information can be used by the device 10, for example, to acquire the spectral analytical information at a location that was determined to be suitable based on the suitability assessment referenced to the near-field image.

[0239] Alternatively or additionally, the position information can also be used to visualize the position of the second field of view 42 relative to the first field of view 41. The position can, for example, be displayed as superimposed information on the near-field image to visualize a measurement spot of the spectral analysis detection device. For this purpose, the human-machine interface 24 can be controlled depending on the position information.

[0240] The position of the second field of view 42 relative to the first field of view 41 can be determined in various ways. Corresponding techniques are known to those skilled in the art. For example, the position determination can be based on one of the following techniques:

[0241] Coupling light into a light path of the spectroscopic detection device to illuminate a sample (which may, for example, be a dedicated calibration sample) and detecting the position of the spot (corresponding to the measurement spot of the spectroscopic detection device) in the near-field image.

[0242] Positioning an optical fiber in front of the viewing window of device 10 and moving the light exit end of the optical fiber in two dimensions to scan the surface of the viewing window. Based on the intensities detected by the spectral analysis detection device, the position of the measurement spot of the spectral analysis detection device relative to the first field of view 41 can be determined.

[0243] Both techniques provide information for each pixel in the near-field image (first field of view 41) indicating how much information from this pixel will reach the spectral analysis acquisition device. This calibration information can be used in the acquisition and / or evaluation of the spectral analysis information, for example, for visualizing and / or evaluating the evaluation results. While an embodiment of the device 10 with an aperture 29 was described with reference to Figures 2, 3, and 4, other devices for reducing ambient light influences can alternatively or additionally be provided. For example, modulation techniques can be used in which the active illumination is temporally modulated and multiple near-field images are acquired during the temporal modulation.The multiple near-field images can be combined with information about the temporal modulation of the illumination to computationally suppress ambient light influences.

[0244] Figure 5 shows a flowchart of a method 50. The method 50 can be executed automatically by the device 10.

[0245] In step 51, a near-field image is acquired. At least one near-field image can be acquired of a plant section directly adjacent to the device 10 (for example, to the contact surface 34 of the aperture 29).

[0246] In step 52, a spectral analysis of the plant section is guided and / or controlled and / or otherwise supported based on an evaluation of the at least one near-field image. This can be done in various ways, as already explained. In steps 51 and 52, the device 10 can remain in contact with the plant section. A relative position of the device 10 to the adjacent plant section can remain substantially unchanged. Step 52 can comprise the acquisition of spectral analysis information based on the evaluation of the at least one near-field image. The acquisition of the spectral analysis information can be selective depending on the result of the evaluation of the at least one near-field image.The acquisition of the spectral analytical information can also be performed independently of the result of the evaluation of the at least one near-field image, whereby the result of the evaluation of the at least one near-field image is used to determine whether the spectral analytical information is usable for the condition analysis or how the spectral analytical information should be evaluated (e.g., weighted) in the condition analysis. The evaluation or weighting can also include a complete rejection or exclusion of spectral analytical information.

[0247] In step 53, a result of the spectral analysis examination (for example, a detected spectrum or wavelengths of local maxima (peaks) of the detected spectrum) is provided for use in the condition analysis. Step 53 may involve performing the condition analysis based on the detected spectral analysis information acquired during the spectral analysis examination. Alternatively or additionally, step 53 may involve transmitting the spectral analysis information to a computer system remote from the device 10. The computer system remote from the device 10 may be configured as an evaluation device or evaluation system according to embodiments of the invention in order to perform the condition analysis based on the spectral analysis information.

[0248] Figure 6 is a schematic block diagram of a system 1 according to an embodiment. The system 1 has a device 10 according to an embodiment. The system 1 has a computer system 4 that is separate from the device 10. The computer system 4 can have a mobile communication terminal (for example, a mobile phone with a display device) and / or one or more computers or servers, which can also be located remotely from the device 10. A display device 5 can be coupled to the computer system 4. In any case, the computer system 4 can be selectively communicatively connected to the device 10. For this purpose, data can be transmitted between the device 10 and the computer system 4 via a communication system that has a wide area network 6, a wireless or wired point-to-point connection, a local area network, and / or a wireless communication system.The data transmission between the device 10 and the computer system 4 can be configured such that data can be transmitted from the device 10 to the computer system 4. Advantageously, bidirectional data transmission is provided, in which, for example, a result of an evaluation of a near-field image and / or a result of the condition analysis based on the spectral analysis information is transmitted from the computer system 4 to the device 10. During field use, the device 10 is positioned near the plant 2 to be examined.

[0249] In one embodiment, system 1 comprises a communication terminal that is different from device 10 and is at least selectively communicatively connected to it. The communication terminal can be configured to perform the functions of the evaluation device disclosed here. The communication terminal can be configured to transmit acquired data (e.g., near-field images and / or spectral analysis information) and / or information derived therefrom to a storage system remote from device 10.

[0250] In a further embodiment, the communication terminal can be configured to transmit the data acquired by device 10 (e.g., near-field images and / or spectral analysis information) to a computer or server that executes the functions of the evaluation device or evaluation system disclosed here. The communication terminal can receive the evaluation results (e.g., the suitability assessment) from the evaluation device or evaluation system. The evaluation results can be used to control device 10.

[0251] Figure 7 is a signal diagram to explain the functioning of the device 10 and the evaluation device or the evaluation system 4 according to an embodiment.

[0252] The device 10 performs a near-field image acquisition 61 to capture at least one near-field image of a plant section adjacent to the device 10. The near-field image 62 or data derived therefrom can be transmitted to the evaluation device or the evaluation system 4.

[0253] The evaluation device or the evaluation system 4 performs a near-field image evaluation 63 of the at least one near-field image. The near-field image evaluation 63 can comprise image analysis techniques known in the art. For example, one or more of the following techniques can be used: filtering (for example, using edge-enhancing filters); segmentation (for example, depending on color information and / or intensity information); processing using a trained machine learning model (as explained in more detail below). The result of the near-field image evaluation 63 can indicate whether the plant section adjacent to the device 10 is sufficiently intact and free of foreign matter to perform a condition analysis based on spectral analytical information acquired on this plant section.The evaluation result can thus include a suitability assessment, represented by a binary value (suitable / unsuitable), for the entire plant section depicted in at least one near-field image. The evaluation result can also indicate which areas and / or positions on the plant section are sufficiently intact and free of foreign matter to perform a condition analysis based on spectral analysis information acquired on this plant section.

[0254] The device 10 can receive data 64 dependent on the evaluation result from the evaluation device or the evaluation system 4. The data 64 can have a binary value (suitable / unsuitable) that indicates whether the plant section depicted in the at least one near-field image is suitable for the acquisition and / or analysis of spectral analytical information for the respective condition analysis.

[0255] The device 10 can acquire 65 the spectral analytical information depending on the data 64. During the time interval in which the near-field image acquisition 61 and the acquisition 65 of the spectral analytical information take place, the device 10 remains positioned in a substantially consistent relative position to the plant 2. The acquisition 65 can depend on the data 64 in different ways, as already explained.

[0256] The device 10 can transmit the spectral analysis information 66 or data dependent thereon (for example, a position of local maxima of a detected spectrum) to the evaluation device or the evaluation system 4. The evaluation device or the evaluation system 4 can then perform the state analysis based on the spectral analysis information 66 or data dependent thereon. A result of the state analysis can optionally be transmitted to the device 10 and output at the human-machine interface of the device 10.

[0257] Various modifications are possible. For example, the near-field image analysis 63 can also be performed locally in the device 10. The result of the near-field image analysis can then optionally be transmitted to the evaluation device or the evaluation system 4 and used by the evaluation device or the evaluation system 4 in the condition analysis, which is performed on the basis of the spectral analysis information 66. In a further embodiment, the condition analysis 67 can also be performed locally in the device 10. Further embodiments are possible in which the processing of the at least one near-field image and / or the spectral analysis information can be distributed in different ways between the device and one or more optional additional computers.

[0258] For example, it is possible for a near-field image recorded by device 10 to be sent to a first computing device (for example, a communications terminal such as a mobile phone) for processing. A local interface such as USB or Bluetooth can be used for this purpose. The first computing device can send the near-field image via a WLAN or cellular interface to a second computing device (for example, one or more computers or servers, which can also be arranged in a cloud architecture). The second computing device can evaluate the near-field image to determine its suitability for spectral analytical data acquisition. The suitability assessment can be transmitted to the first computing device and used by the first computing device or device 10 to acquire the spectral analytical information, for example in the form of generating a recommended course of action.If the at least one plant part is suitable, the spectral analytical information can be acquired by device 10. The spectral analytical information can preferably be evaluated by the first computing device, optionally also by the second computing device. If the first computing device performs the evaluation of the spectral analytical information, the evaluation result of the spectral analytical information can be sent to the second computing device for archiving.

[0259] Figure 8 is a schematic block diagram representation of an evaluation device 4. While the evaluation device 4 is shown as being implemented in a device, the various functions and components can also be realized in a distributed evaluation system, for example in a server or computer system.

[0260] The evaluation device 4 has an evaluation device interface 41 which is configured to receive the at least one near-field image and the spectral analytical information from the device 10. The evaluation device interface 41 can also be configured to output an evaluation result of the at least one near-field image (in particular a suitability assessment for the acquisition of the spectral analytical information) and / or a result of the state analysis based on the spectral analytical information.

[0261] The evaluation device 4 has at least one display interface 44. The evaluation device 4 can control the display device 5 via the display interface 44 to output an evaluation result of the at least one near-field image and / or the result of the condition analysis. Alternatively or additionally, the display device 5 can be controlled via the display interface 44 to enable various user-defined settings, such as entering the plant variety or species, entering the condition analysis to be performed, etc.

[0262] The evaluation device 4 has an evaluation circuit 45. The evaluation circuit 45 can be configured to perform the near-field image evaluation 46 to determine whether the plant section positioned on the device 10 is suitable for the acquisition and / or evaluation of the spectral analysis information. The evaluation circuit 45 can be configured to perform a condition analysis 46 based on the spectral analysis information.

[0263] The evaluation device 4 can be configured to determine an evaluation result of the near-field image and transmit it to the device 10 via the evaluation device interface 41. The evaluation result can include a suitability assessment indicating whether the plant section (e.g., leaf section) positioned on the device 10 is suitable for data acquisition of the spectral analytical information and / or whether a condition analysis of the spectral analytical information acquired in this plant section can be expected to yield reliable results. As explained above, the suitability assessment can be a single (e.g., binary) value that refers to the plant section as a whole.Other discrete value ranges are possible, for example, a suitability rating for the plant section that expresses suitability in one of three possible values ​​(e.g., in a traffic light system) or four possible values ​​(e.g., "fault-free," "still suitable," "less suitable," and "unsuitable"), or more than four possible values. The suitability rating can also be selected from continuous value ranges (e.g., from 0% to 100%), as already explained.

[0264] The suitability assessment can indicate, in each case with respect to the first field of view 41, which areas and / or positions of the plant section adjacent to the device 10 are suitable for the acquisition of the spectral analytical information.

[0265] The evaluation device 4 can thus support a check with which it can be determined whether a detection and / or evaluation of spectral analytical information in the plant section (for example leaf section) adjacent to the device 10 can be expected to produce reliable results of a condition analysis based on the spectral analytical information.

[0266] The status analysis 47 can, for example, include the determination of nutrient concentrations based on the spectral analytical information. The status analysis 47 can be carried out using techniques familiar to the person skilled in the art, such as those described, for example, in WO 2019 / 169434 A1, US 7804588 B2, and US 11320307 B2. The status analysis 47 can be carried out using principal component techniques, PLS techniques, or other techniques that allow a conclusion to be drawn as to whether, for at least one and advantageously several nutrients, a nutrient concentration lies within a target range or outside a target range. The status analysis 47 can also quantitatively determine the nutrient concentration for one or more nutrients. This value can then be used, for example, to check whether the nutrient concentration is within the target range (normal supply) or outside the target range (undersupply or oversupply).

[0267] To perform the various processing functions, the evaluation device 4 has the evaluation circuit 45. The evaluation circuit 45 can comprise one or more integrated circuits to perform the near-field image evaluation 46 and / or the state analysis 47. The one or more integrated circuits can, for example, comprise any one or any combination of the following circuits or circuit components: an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a processor, a GPU, a TPU, a controller, one or more quantum gates, a circuit for quantum information processing, and other integrated circuits.

[0268] The evaluation device 4 can have a memory system 43. The memory system 43 can store machine-readable instruction code which, when executed by the evaluation circuit 45, causes the functions and steps disclosed here to be carried out. If the evaluation of the at least one near-field image is performed by the evaluation circuit 45, the memory system 43 can store parameters that enable an evaluation of the at least one near-field image to determine whether the imaged plant part is suitable for detecting and / or evaluating the spectral analytical information. The parameters stored in the memory system 43 for evaluating the at least one near-field image can, for example, include one or more of the following parameters: filter parameters; definition of decision boundaries of a support vector model; parameters of a trained machine learning model.Further examples are described in more detail with reference to the evaluation device or the evaluation system.

[0269] Figure 9 shows a modification of the device 10. The device 10 is configured such that the near-field image analysis can be performed locally within the device 10. The processing circuit 26 of the device 10 is configured to perform the near-field image analysis 36. The near-field image analysis 36 can be implemented as already explained above. In particular, the near-field image analysis 36 can determine, using automated image processing techniques (for example, the use of one or more trained machine learning models), whether the plant section adjacent to the device 10 and accessible to the spectral analysis detection device is suitable for detecting the spectral analysis information.

[0270] The functioning of the device 10 and / or the evaluation device or the evaluation system 4 in the near-field image evaluation is described in more detail with reference to Figure 10, Figure 11, Figure 12, Figure 13, Figure 14 and Figure 15.

[0271] Figure 10 shows an example of a leaf 3. A leaf section of the leaf 3 is in contact with the contact surface 34 during use of the device 10. The leaf 3 can have regions 61, 62 in which, due to foreign substances such as dirt, pollen or liquids or living organisms such as insects, the acquisition of the spectral analytical information for the condition analysis entails an increased risk of falsifying the analysis results. The leaf 3 can have regions 63 in which, due to mechanical damage, necrosis, chlorosis, anthocyanosis or spoiled leaf areas, the acquisition of the spectral analytical information for the condition analysis entails an increased risk of falsifying the analysis results. The near-field image evaluation detects whether such regions and / or positions are present at the measuring point of the spectral analytical detection device.

[0272] Figure 11 shows, by way of example, an output of the evaluation results via a display device, such as the human-machine interface 24 of the device 10 and / or the display device 5. Areas 60 suitable for acquiring the spectral analytical information can be annotated superimposed on the near-field image. Alternatively or additionally, the unsuitable areas can also be annotated. In this way, the acquisition of the spectral analytical information is guided in such a way that, based on objective criteria, acquisition at an unsuitable location is avoided.

[0273] By using the near-field image, further condition analysis (which can be carried out conventionally using spectral analytical information) can be supported automatically and in a data-based manner.

[0274] Figure 12 is a flowchart of a method 70. The method 70 can be executed automatically by the device 10 and / or the evaluation device or the evaluation system 4.

[0275] At 71, at least one near-field image captured by the near-field camera 21 of the device 10 is received.

[0276] At 72, a suitability rating is determined by processing the at least one near-field image. The suitability rating may be represented by a binary value (suitable / unsuitable) for the entire plant section depicted in the at least one near-field image. The suitability rating may be a value (e.g., an integer value) selected from three or more discrete values; this value may, for example, indicate gradations of suitability, for example, based on a value selected from an ordinal value scale. The suitability rating may be a value that varies within a continuous numerical range, for example, between a lower limit and an upper limit (such as 0% to 100%).The suitability assessment may also indicate which areas and / or positions on the plant section are sufficiently intact and free of foreign matter to carry out a condition analysis based on spectral analytical information collected on that plant section.

[0277] In 73, the suitability assessment is used for further condition analysis. This can be done in different ways, as already explained.

[0278] Figure 13 is a schematic diagram 80 for further illustrating the processing of the at least one near-field image. The at least one near-field image can be processed to evaluate the suitability of a plant section adjacent to the device 10 for the acquisition and / or analysis of the spectral analytical information.

[0279] The suitability rating 82 can be determined based on the absence of various possible error sources. The plant section can be rated as suitable if all verified error sources are absent. More complex techniques can be used. For example, the plant section can be rated as suitable if the area of ​​the plant section surrounded by the contact surface 34 that is free of all verified error sources is greater than the threshold value.

[0280] The suitability assessment 82 influences the condition analysis 81. This can be done in various ways, as already explained in detail. For example, the suitability assessment 82 can be used to ensure that the collection and / or evaluation of spectral analytical information is only carried out on plant sections assessed as suitable. Alternatively or additionally, the suitability assessment 82 can be used to mark the result of the condition analysis (e.g., as reliable / limited reliability) or to evaluate it (e.g., to weight it).

[0281] To determine the suitability assessment, one, several or all of the following checks can be carried out based on at least one near-field image:

[0282] Foreign matter detection 83: Detection of liquid and / or solid foreign matter (such as dust or pollen) in the plant section positioned in the first field of view or the second field of view; Necrosis detection 84: Detection of necrotic sites in the plant section positioned in the first field of view or the second field of view;

[0283] Chlorosis detection and / or anthocyanosis detection: Detection of areas with chlorosis or anthocyanin abnormalities in the plant section positioned in the first field of view or in the second field of view;

[0284] Detection of spoiled areas 85 in the plant section positioned in the first field of view or in the second field of view;

[0285] Pest detection 86: Detection of pests or insects in the plant section positioned in the first field of view or in the second field of view;

[0286] Position verification 87: Detection of incorrect positioning of the device on the plant. For this purpose, it can be determined, for example, whether the device is positioned on a plant section of the correct type (e.g., a leaf at a desired leaf stage) based on automatic detection of the plant section in the first field of view or in the second field of view.

[0287] Crack / Break Detection 88: Detection of mechanical damage in the plant section positioned in the first field of view or the second field of view.

[0288] Restricting the check to the second field of view can be achieved using position information determined during a calibration. This position information can indicate the position of the second field of view relative to the second field of view. This position information can indicate the strength of the contribution of several pixels of the near-field image to the acquisition of the spectral analytical information. The determination of position information during a calibration has already been described.

[0289] By limiting the review to the second field of view, the suitability assessment can be provided and used more quickly.

[0290] A check via the first field of view can be particularly advantageous if the position of the second field of view in the first field of view can be changed in a controllable manner.

[0291] Optionally, additionally or alternatively, further potential sources of error can be automatically checked based on the at least one near-field image. For example, the at least one near-field image can be checked for consistency with user specifications. If the plant species or variety for which the condition analysis is to be performed is specified by an input enabled via the human-machine interface 24 or by the evaluation device or the evaluation system 4, the near-field image can be used to ensure that the device 10 is positioned on a plant of the desired plant species or variety. If this is not the case, the device is deemed unsuitable.Alternatively or additionally, if the type of condition analysis to be carried out is determined by an input made possible via the human-machine interface 24 or by the evaluation device or the evaluation system 4 (for example by selecting nutrients whose concentration is to be determined based on the spectral analytical information, or determining the water supply), it can be ensured on the basis of the near-field image that the device 10 is located on a plant section suitable for this type of condition analysis.

[0292] Each of the aforementioned detection tasks 83-88 can be performed using known techniques of computer-aided image analysis. For example, segmentation techniques can be used to detect necrotic, spoiled, or mechanically damaged areas, or foreign matter and pests. The segmentation techniques can use edge filters and / or color-dependent techniques. A check to determine whether the device 10 is applied to a plant of the desired variety or species and / or to a plant section suitable for the respective condition analysis (e.g., a leaf of a specific leaf stage) can be determined by determining feature vectors and comparing them with a database or by other techniques that provide a comparison metric between images. Such techniques are familiar to those skilled in the art.

[0293] To detect the identified error sources, techniques such as support vector methods (Figure 14) can also be used, which preferably convert derived features of the at least one image into a classification decision regarding the at least one error source. Other trained machine learning models, such as neural networks (Figure 15) or VisionTransformer (ViT), can also be used to detect the identified error sources. The parameterization of the corresponding detection logic for detecting the various error sources can be data-driven using training images. The training images can each be annotated to indicate which of the error sources to be detected is or are present in the corresponding training image.Training can be performed as supervised learning, semi-supervised learning, or unsupervised learning. The parameterization of the detection logic(s) that achieves good or even optimal performance on a test dataset can be determined using well-known optimization methods such as gradient descent.

[0294] Figure 14 shows an example of recognition logic based on a support vector model. Images are converted into feature vectors of features Fi, F2, with application-relevant information being preferably retained and application-irrelevant information being discarded. Suitable techniques for this purpose are familiar to those skilled in the art, for example, histograms of oriented gradients (HOG) or descriptors of scale-invariant feature transforms (SIFT). Activation features from layers of existing image classification models such as AlexNet or VGG are also possible. In this space of features, decision boundaries 94, 95, 96 can be defined using the support vector model. Depending on the respective error sources present, the feature vectors lie in different regions 91, 92, 93 of the feature space.During training, the decision boundaries and, optionally, the suitable features are automatically determined before the support vector model is used. The corresponding parameters can be stored in the device 10 and / or the evaluation device or the evaluation system 4 or stored in a way that is accessible to them. During use, the device 10 or the evaluation device or the evaluation system 4 can evaluate the at least one near-field image. For this purpose, the feature vector for the at least one near-field image can be determined, and, with reference to the decision boundaries, the area of ​​the feature space in which the feature vector for the at least one near-field image lies and the type of error source (if any) present can be determined. In a further embodiment, the feature vector can only be used to check whether the feature vector for the at least one near-field image lies in the area of ​​the feature space assigned to the "suitable" assessment.

[0295] Figure 15 schematically shows a machine learning model 100 that can be used to evaluate the near-field image. The machine learning model 100 has an input (e.g., an input layer) 101, one or more hidden layers 103, and an output (e.g., an output layer) 102. Instead of or in addition to the hidden layers 103, stacks of other processing units can also be present. The machine learning model 100 can be configured to receive pixel values ​​of the at least one near-field image and / or data or features derived therefrom (e.g., detected edges) as input. The machine learning model 100 can be trained to provide as output a value indicating the presence or absence of a specific error source (e.g., one of the error sources explained with reference to Figure 13).The machine learning model 100 may be trained to provide as output a value indicating the suitability rating (e.g., as a binary value suitable / unsuitable).

[0296] To check for various error sources, multiple machine learning models can optionally be trained. This allows machine learning models to be trained, each trained to detect one or more possible error sources. This can further increase reliability.

[0297] The machine learning models can be trained separately for different plant species or plant varieties. This is particularly useful for error sources whose visual appearance may depend on the plant variety or plant species. Examples include necrosis, chlorosis, anthocyanosis, or damaged leaf spots. For error sources whose visual appearance is not or only slightly dependent on the plant variety or plant species, a common machine learning model can be trained that is applicable to multiple plant varieties or plant species. One example of this is the detection of foreign substances.

[0298] Machine learning models suitable for automatic image processing are known to those skilled in the art. Suitable machine learning models include, for example, so-called "convolutional neural networks" (CNNs) or other so-called "deep learning" (DL) techniques. The at least one near-field image can be processed using at least one DL technique, for example, CNNs or "VisionTransformer" (ViT), both of which are known to those skilled in the art.

[0299] In a specific example, a ResNeXt50 machine learning model can be used, with its parameters preferably already pre-trained by training on another, preferably larger, dataset and used as initialization. Adaptation can be performed separately for each recognition task, depending on the recognition task described below using binary annotated images: o Plant variety / plant species recognition

[0300] ■ Goal: Avoid accidental measurement of plants that do not match the stored nutrient analysis algorithm

[0301] ■ Input data: Near-field image(s)

[0302] ■ Input data: Plant variety or species identifier o Leaf age detection

[0303] ■ Objective: To avoid accidentally measuring leaves that do not match the required leaf age for the respective plant species

[0304] ■ Input data: Near-field image(s)

[0305] ■ Initial data: Leaf age as one of several possible options (e.g.

[0306] Classification into classes "young", "young ripened" and "ripe") o Water drop / water film detection

[0307] ■ Objective: To avoid accidental measurement of leaves which are due to

[0308] Water deposits would produce a distorted measurement result

[0309] ■ Input data: Near-field image(s)

[0310] ■ Initial data: Water drops or water film present yes / no o Detect necrotic leaf spots

[0311] ■ Aim: To avoid accidental measurement of leaves that contain necrotic

[0312] have shares

[0313] ■ Input data: Near-field image(s)

[0314] ■ Initial data: Leaf contains necrotic areas yes / no o Detect spoiled leaf areas ■ Goal: Avoid accidentally measuring leaves that have otherwise spoiled parts

[0315] ■ Input data: Near-field image(s)

[0316] ■ Initial data: Leaf contains spoiled areas yes / no lice detection

[0317] ■ Objective: To avoid accidentally measuring leaves on which aphids are sitting, which in turn would influence the nutrient concentration of the measurement

[0318] ■ Input data: Near-field image(s)

[0319] ■ Initial data: Aphid present yes / no n / Dust / foreign body detect

[0320] ■ Objective: Avoid accidental measurement of leaves, which may cause additional

[0321] have (non-living) foreign bodies that would distort the measurement

[0322] ■ Input data: Near-field image(s)

[0323] ■ Initial data: Leaf contains pollen / dust / foreign matter yes / no Detect right / bent leaf

[0324] ■ Objective: To avoid the acquisition of spectral analytical information partially or completely on an abaxial side (if acquisition on the adaxial side is desired) or on the adaxial side (if acquisition on the abaxial side is desired), which could lead to erroneous results (e.g. different texture of the back side)

[0325] ■ Input data: Near-field image(s)

[0326] ■ Output data: Sheet twisted / bent yes / no Detect the sheet side

[0327] ■ Objective: To avoid the acquisition of spectral analytical information over the entire area of ​​the

[0328] back of the sheet, which could lead to incorrect results (e.g. different texture of the back)

[0329] ■ Input data: Near-field image(s)

[0330] ■ Output data: Wrong page yes / no detect torn / torn page

[0331] ■ Objective: To avoid scanning the background where the sheet is torn

[0332] ■ Input data: Near-field image(s)

[0333] ■ Output data: Sheet broken / torn yes / no the positioning

[0334] ■ Aim: Check correct positioning of the sample under a measuring window • Example: Too close to the leaf edge, not filling the entire measuring window

[0335] ■ Input data: Near-field image(s)

[0336] ■ Initial data: Incorrectly positioned yes / no o Correct pressing of the sheet on a sample holder of the device 10

[0337] ■ Objective: Check the correct pressure of the sample on the measuring window

[0338] • Example: For example, protruding curved leaves on the glass

[0339] ■ Input data: Near-field image(s)

[0340] ■ Initial data: not properly pressed or too much object - air distance or too curved sheet

[0341] For each of these described recognition logics, a DL ("deep learning") model can be developed, for example, which predicts an output value for each possible recognition state with respect to the corresponding error source. For example, the prediction of two values ​​for binary cases yes / no, or several for a classification task into more than two possible classes. The parameters of the DL model can, for example, be determined iteratively using known optimization methods with respect to known target metrics such as cross-entropy with annotated training data.

[0342] The detection logic can also be implemented as a detection model (preferably a DL detection model), so that, for example, the positions of the at least one near-field image are determined using parametric descriptions such as bounding boxes, which indicate the presence of error sources. One example is the detection of all aphids on the leaf in the current (first or second) field of view to detect aphid infestation of the sample. In a specific example, a RetinaNet machine learning model can be used.

[0343] The detection logic can also be implemented as a segmentation model (preferably a DL segmentation model), so that, for example, the positions that indicate the presence of error sources are determined with pixel precision for the at least one near-field image. One example is a semantic segmentation of all pixels in the current (first or second) field of view where pollen is present. In a specific example, a Mask2Former machine learning model can be used.

[0344] When training the machine learning model(s), augmentation techniques can be applied to artificially generate more images available for training. Augmentation can involve one, several, or all of the following operations, which can be applied to annotated images: random cropping, random rotations, random flipping (left-right and / or up-down), random changes in brightness, and random changes in contrast. For training and testing, the annotated images can be divided (preferably before augmentation) into a training dataset (e.g., 70% of the images), a validation set (e.g., 15% of the images), and a test set (e.g., 15% of the images). Especially when training classifiers, it is advantageous to divide the dataset into training, validation, and test sets separately for each annotated class.This is especially beneficial for classes with very unequal amounts of images.

[0345] Augmentation can be used to reduce or eliminate imbalances in the number of annotated images belonging to different classes. Alternatively, a numerically underrepresented class can be selected more frequently during training to ensure images from all classes are used at least approximately equally during training.

[0346] For a machine learning model designed to perform detection (e.g., an aphid detector), it is advantageous for the annotated images to contain at least 100 images of the object to be detected from one plant species. The images can show different leaf stages. For an aphid detector applicable to multiple plant species, at least 10 additional images of the object to be detected can be used in the annotated training data for each additional plant species. This allows for the training of a general detector (applicable to multiple plant species) (e.g., a general aphid-on-leaf detector).

[0347] For a machine learning model that is to perform segmentation (e.g., a pollen segmenter), it is advantageous for the annotated images to contain at least 25 images with pixel annotations. It is advantageous if some images contain no pollen but visually similar symptoms (such as burnt leaf spots). Optionally, some additional images (e.g., at least 5) can contain only leaves but no pollen and no visually pollen-like symptoms. Optionally, additional images can contain other known disturbances (such as a kink, water on the leaf, etc.). This training data can be used for the segmenter for the first plant species.

[0348] If application to additional plant species is desired, at least five pixel-annotated images with pollen from each additional plant species can be advantageously used in the training dataset. This allows a general pollen-on-leaf segmentation model to be trained.

[0349] For a machine learning model intended to perform a classification (e.g., a water film classification), the training data can advantageously comprise at least 1,000 images, of which at least 250 depict a leaf with a water film. The images can be distributed across multiple growth stages (plant and / or leaf stages). Further improvements can be achieved with correspondingly larger training data sets. For example, it may be advantageous to use at least five times, at least ten times, or at least approximately twenty times the number of images mentioned in the previous paragraphs during training. As explained, the data sets are divided into training, validation, and test sets.

[0350] The results of the different detection techniques can be logically linked. One or more filters can be used to implement the detection logic. For example, a classification can be made as unsuitable if at least one aphid was detected during aphid detection, or if a segmentation to determine pollen-covered areas reveals more than a certain area (e.g., 2 mm 2 ) is covered with pollen.

[0351] An assessment as suitable can be made if a certain possible disturbance factor (e.g. aphid infestation) is present within the first field of view 41, but not in the possibly smaller second field of view 42.

[0352] By capturing and evaluating at least one near-field image, the status analysis can be avoided in the presence of unsupported error cases. This avoids the risk of incorrect operation of the device 10 based on objective criteria. Expert knowledge in the field use of the device 10 is no longer necessarily required. Error case detection is partially or fully automated.

[0353] Figure 16 is a flowchart of a procedure 120 that can be used to utilize the evaluation result of the at least one near-field image. The procedure 120 can be used to implement step 73 of the method 70 of Figure 12. The procedure 120 can be executed automatically by the device 10 and / or the evaluation device or the evaluation system 4.

[0354] At 121, the spectral analysis detection device is activated. This can be done depending on the evaluation result of the at least one near-field image, as already explained.

[0355] At 122, acquired spectral analysis information is evaluated. This allows a nutrient concentration of at least one, and advantageously several, nutrients to be determined and / or the health status of the plant to be assessed.

[0356] At 123, the evaluation result is provided. This can be done locally at the device 10 or via the evaluation unit or system 4. The evaluation result can be used to adjust the cultivation conditions of plants so that a good, as optimal as possible, supply is achieved.

[0357] Figure 17 shows exemplary spectral analysis information for illustrative purposes, each of which is configured as a spectrum 131, 132 over a wavelength range. The wavelength range can extend over at least part of the visible and / or near-infrared spectrum. The wavelength range can, for example, be within a range of 300 nm and 2000 nm or of 500 nm and 1650 nm, optionally also covering these ranges.

[0358] The spectral analytical information may comprise an optical measurement quantity (e.g., an intensity of reflected, scattered, or Raman-scattered electromagnetic radiation or fluorescence radiation) as a function of wavelength for several wavelengths.

[0359] Depending on the concentration of various nutrients, the spectral analytical information changes. For example, an undersupply of certain nutrients leads to a reduction in the local maxima ("peaks") characteristic of these nutrients. This can be used to quantitatively determine the concentration of various nutrients, an under- or oversupply of water, or other characteristics related to the plant's health based on the spectral analytical information.

[0360] The condition analysis based on the spectral analytical information is influenced by the at least one near-field image. In particular, the at least one near-field image can be used to prevent errors due to the spectral analytical information being acquired at an unsuitable location on the plant.

[0361] The processing of spectral analytical information to determine health status can be carried out in various ways. For example, the techniques described in WO 2019 / 169434 A1, US Pat. No. 7804588 B2, and US Pat. No. 1,132,0307 B2 can be used. In general, in a phase in which the processing logic for processing the spectral analytical information is parameterized, laboratory test results obtained from the respective leaves can be used in addition to the spectral analytical information to correlate the spectral analytical information with nutrient concentrations or other health data.

[0362] Alternatively or additionally, techniques such as cluster or principal component analysis can be used. Figure 18 illustrates this as an example for a principal component analysis 140. Principal components can be determined from the spectrum acquired in each case, and their assignment to different regions 143, 144 can be determined within a space spanned by the principal components. Depending on the nutrient concentration, the principal components lie in a region in which spectra for plants with similar nutrient concentrations are grouped together in clusters 141, 142. The nutrient concentrations for each cluster 141, 142 can be quantitatively determined in a training phase prior to field use of the device 10. In addition to the spectral analysis information, laboratory test results obtained on the respective leaves can be used to correlate the spectral analysis information with nutrient concentrations or other condition data.Alternatively or additionally, machine learning techniques can be used to derive quantitative information about nutrient concentration(s) and / or other factors influencing plant condition (such as water supply) from the spectral analysis information. To determine nutrient concentrations, methods such as least squares regression, PLS regression, support vector regression, Gaussian process regression, k-nearest neighbor regression, random forest regression, multilayer perceptrons, and / or CNNs (convolutional neural networks) can be used. At least one machine learning model can be trained to receive the spectral analysis information as input and output one or more nutrient concentrations as output.When training at least one machine learning model, spectral analysis information annotated with the laboratory-determined nutrient concentration for one or more nutrients is used. ML classification techniques can also be used, the output of which predicts an assignment of the nutrient concentration to two or more different value ranges. These can, for example, represent normal supply, undersupply, and oversupply of the respective nutrient. The machine learning models can then be trained in such a way that they do not regress on a quantitative nutrient concentration value, but rather, as a classifier, indicate the assignment to possible states such as normal supply, undersupply, and oversupply. For this purpose, support vector machines, a k-nearest neighbor model, a random decision forest model, MLPs, CNNs, or other models can be trained as classifiers.

[0363] For a time-efficient execution of the condition analysis, various parameters can be specified by the user. For example, the device 4 or the evaluation device or the evaluation system 4 can be configured to allow the user to specify the plant species or variety and nutrients of interest via a human-machine interface. The evaluation of the near-field image and / or the condition analysis based on the spectral analysis information can then be carried out depending on the user-defined plant species or variety and the nutrients to be determined.

[0364] Figure 19 is a flowchart of a method 150. The method 150 can be executed automatically by the device 10 and / or the evaluation device or the evaluation system 4.

[0365] At 151, a human-machine interface is activated to enable user input. The user input can, for example, specify the plant species or plant variety and nutrients of interest. Alternatively or additionally, the user input can specify which checks should be performed on the at least one near-field image before the spectral analytical information can be acquired.

[0366] At 152, spectral analytical data is acquired depending on the user input and the evaluation of at least one near-field image. For example, the spectral analytical acquisition device can be controlled such that the spectral analytical information is acquired selectively only if the evaluation of the near-field image reveals that the plant section positioned on the device is suitable for determining the concentration of the specified nutrients from the spectral analytical information.

[0367] At 153, the evaluation result is provided. This can be done via the human-machine interface 24 of the device 10 or a human-machine interface of the evaluation device or the evaluation system 4. The provision of the evaluation result can result in an automatic or semi-automatic adjustment of cultivation conditions.

[0368] Figure 20 shows additional features of the device 10 that are usable in the devices, methods and / or systems disclosed herein.

[0369] The spectral analysis detection device 22 of the device 10 can comprise a sample illumination source 160, a spectrometer 162, and optical components 161, 163. The optical components 162 can comprise at least one controllable component 161. The spectral analysis detection device 22 can be configured such that the two fields of view 42 of the spectral analysis detection device 22 can be adjusted by controlling the controllable component 161. The adjustment can be performed depending on the at least one near-field image.

[0370] The device 10 can have a positioning system for determining the position of the device 10 in a reference system. The positioning system can have a global navigation satellite system (GNSS) 164. The device 10 can be configured to determine the position of the device 10 using the positioning system upon acquisition of the at least one close-up image and / or the spectral analysis information. The respectively determined position can be used to determine a correlation between plant position and the respective plant condition during the condition analysis of several plants in an area.

[0371] Figure 21 is a schematic representation of a configuration of the near-field camera 21. The near-field camera 21 has a camera chip 171 (for example, a CMOS or CCD chip) and a near-field optics 172, which is only shown schematically. The device 10 can be configured to provide active illumination upon acquisition of the at least one near-field image. An illumination device 173 of the device 10 can have a light source 174 that can be selectively activated upon acquisition of the at least one near-field image. An illumination optics, which can have a paraboloidal or facet reflector 175 and further optical components 176, can provide illumination, for example, with parallel light.

[0372] The near-field optics 172 can be adaptive. Alternatively or additionally, the device 10 can have multiple near-field optics, from which at least one can be selected for capturing the at least one near-field image.

[0373] With reference to the figures, devices, evaluation devices, evaluation systems, systems, and methods have already been described in detail that determine information about a sample status and use it in the condition analysis of plants. A near-field camera with a short-focal-length lens for near-field image acquisition, including an illumination optics unit for sample illumination, image processing techniques for automated analysis of the acquired near-field images, a computing unit for executing the image processing techniques, and a storage system for storing the learned image processing techniques can be used.

[0374] Below, various further modifications and features are described that can optionally be used to support the technical effects already described.

[0375] Field of view: The first field of view 41 can essentially correspond to the second field of view 42. A larger field of view or field of view 41 of the near-field camera compared to the measurement range of the spectral analysis device has the advantage that the illuminated area of ​​the spectrometer can be inadvertently shifted during the measurement, while the (more global) camera analysis is still valid. Furthermore, leaf changes at the edge of the measurement range of the spectral analysis device can also be detected.

[0376] Color channels: A color camera image (e.g., RGB color channels) from the near-field camera can be advantageous, for example, to assess the plant leaf stage or condition or to distinguish between color inhomogeneities and glare above the leaf surface. The camera image can contain several color channels in the visible spectral range and one or more color channels in the near-infrared spectral range.

[0377] Design of the illumination optics: The illumination optics 175, 176 for the sample illumination should preferably be designed such that the light rays from the light source / lamp burner impinge on the sample as parallel beams and are recorded by the camera optics due to the sample reflection. This can be achieved using paraboloidal and / or facet reflectors, or the light source can act as a point light source in the sample measurement plane.

[0378] Camera calibration: Camera calibration or referencing can be performed using a diffuse white / gray standard during device initialization (either once or recurringly). With recurring white / gray standard measurements, intensity and / or spectral changes of the light source can be detected and compensated. The inventive device according to each of the embodiments disclosed here can have an integrated white standard (e.g., made of Spectralon) for adjusting the channel gain factors (white balance) and adjusting the exposure time.

[0379] Enabling user-defined analysis specifications: Each analysis takes time. Some users do not want to perform all possible sources of error and risk accidentally measuring a non-supporting sample in order to achieve faster measurement times. Therefore, it can be advantageous for users to be able to specify which tests are performed based on the at least one near-field image and how precisely they should be performed. For example, a human-machine interface can be controlled to allow user input specifying the recognition tasks to be performed (e.g., only checking for plant variety accuracy before spectral data acquisition).

[0380] Visualization of affected areas in the near-field image: If a repeat measurement is necessary, it can be helpful to visually show the user which areas of the image were classified as faulty. This can be derived, for example, from the analysis of the machine learning models. The user can be supported by visualizing the areas relevant to the analysis result – for example, using techniques familiar to experts such as GradCam / XRAI, etc.

[0381] Error case detection determined pixel-by-pixel or region-by-region in the near-field image: A "spatial" analysis is also possible to determine which areas in the image can be attributed to the respective problem classes, which can be achieved, for example, as pixel-precise segmentation. This offers several advantages. Better visualization for the user is possible. More precise control is possible, for example, by defining maximum areas that may have certain undesired characteristics. These advantages can offset the additional annotation effort in the training phase and the longer computing time. The classification therefore does not have to refer to the plant part in the first or second field of view in its entirety, but can also refer to pixels or regions contained therein.

[0382] Aggregation across multiple positions of the device 10: The disclosed techniques have so far been described primarily with regard to determining a suitability assessment for a condition analysis of precisely one plant. It is evident that the techniques can be used repeatedly or—when using multiple devices 10—in parallel on different plants, for example, in a cultivation area. The results from multiple recording positions can be aggregated and further processed with respect to a reference system (e.g., a field, a forest area, or another cultivation area). For example, a frequency of positive aphid analyses across an entire field and / or a frequency of leaf necrosis and / or chlorosis and / or anthocyanosis, which may indicate a locally poor nutrient medium, can be considered.The positioning system 164, for example a GPS system or another GNSS, can be used for spatial assignment.

[0383] Output (e.g., visualization) of uncertainties: In addition to predicting the suitability for capturing the spectral analytical information, the (un)certainty of the analysis of the at least one near-field image can also be determined and used. The uncertainties can be determined in various ways and output as qualitative or quantitative uncertainty measures. For example, the uncertainties can be determined based on initial values ​​of classifiers, based on uncertainties determined during training, based on a scatter of image analysis results with multiple trained machine learning models, based on a scatter of nutrient concentrations determined on different leaves or different plants based on images, or in other ways.The uncertainties can be output via a human-machine interface and / or used to automatically control the spectral analysis acquisition device.

[0384] Some or all of the image analysis steps may involve determining a confidence or uncertainty for the analysis result, i.e., the uncertainty or confidence in the suitability for capturing the spectral analytical information determined based on at least one near-field image. The confidence or uncertainty may be determined, for example, using one, several, or all of the following techniques:

[0385] Uncertainty can be derived directly from the output values ​​of a machine learning model. For example, in a classification procedure, the entropy over single-class probabilities can be used as an uncertainty measure. Alternatively, the difference between the most certain and second-most certain classes can be used as an uncertainty measure.

[0386] An uncertainty measure can be determined by estimating the similarity of at least one image to the training data. For this purpose, so-called "activation shaping" for CNNs ("convolutional neural networks") can be used, for example. Separate models can also be used to determine anomalies or deviations.

[0387] Uncertainties can be determined by mapping them to the uncertainty of a test dataset (during training). For example, the mean error on a test dataset can be determined and used as the expected uncertainty for each output value determined with the machine learning model. The above techniques can be used particularly for evaluations that solve classification tasks.

[0388] For detection tasks, for example, confidence or uncertainty can be determined by having the machine learning model have an output that also outputs a confidence value. The machine learning model can be trained and used in such a way that, in addition to a detection result, it also outputs the associated confidence or uncertainty.

[0389] For segmentation tasks, confidence or uncertainty can be determined using one or more of the following techniques:

[0390] Uncertainty can be derived directly from the output values ​​of a machine learning model. For example, the entropy can be averaged over individual class probabilities for all pixels in a region of the same class.

[0391] Uncertainty can be derived from a variation of the output values ​​upon repeated, random variation of the input image (augmentation during application or so-called "test time augmentation" techniques) or from intermediate calculations (for example, by so-called "test time drop-out" techniques).

[0392] Alternatively or additionally, in detection and / or segmentation tasks, an uncertainty measure can be determined by estimating the similarity of at least one image area (e.g., a region or box) to image areas in the training data. Separate models can also be used to determine anomalies or deviations.

[0393] The results of the image analysis steps can be filtered or weighted based on the associated confidence or uncertainty. For example, only those results whose confidence is above a threshold can be used further (e.g., for visualization). Also, only the k most certain results can be output. The threshold or the number k can be user-defined. For example, in the visualizations shown in Figure 11, only those areas whose confidence is above a threshold or that are among the k most certain results can be visualized.

[0394] Design and mounting of the device 10: The device 10 can be designed as a portable, in particular manually held device 10.

[0395] Figure 22 schematically shows an embodiment of the device 10 as a manually held device having a structure 39 for holding the device 10.

[0396] Figure 23 shows a further embodiment in which a vehicle 180 according to one embodiment has one or more devices 10 mounted thereon. Accordingly, the device 10 or the devices 10 can be configured with a support structure for attachment to an agricultural vehicle.

[0397] Figure 24 shows a further embodiment in which a robot 185 comprises the device 10. A base 186 of the robot 185 can be stationary or movable, for example, along a rail system. A controller 187 can control one or more actuators for positioning the device 10.

[0398] Figure 25 shows a further embodiment in which a flying object 189 comprises the device 10. The flying object 189 can be remotely controllable and / or configured for autonomous flight operation.

[0399] In yet another embodiment, the device 10 can be configured as a stationary system. The device 10 can be mounted with a stationary support near a conveyor on which plants are transported past the device 10.

[0400] Evaluation of the at least one near-field image: The evaluation of the at least one near-field image can be performed in various ways, as already explained. In particular, classifiers can be used to distinguish defective from defect-free plant sections. A large number of recognition techniques (e.g., CNNs, ViTs, support vector machines with features defined based on expert knowledge, for example, but not limited to these) can be used. Various techniques can also be used for training, for example, gradient-based optimization methods such as stochastic gradient descent (SGD) or an optimization technique with an adaptive learning rate (such as ADAM), fully supervised, semi-supervised, and other training methods.

[0401] Spectral analysis detection device: The spectral analysis detection device 22 can comprise a spectrometer. However, other configurations are possible that enable the detection of an optical measurement variable at two or more (for example, at least ten or more) wavelengths. For this purpose, filter-based systems (configured to block light components), single-wavelength diodes, or other configurations can be used as an alternative or in addition to the use of a spectrometer.

[0402] Type of use of the evaluation of the at least one near-field image for capturing and / or evaluating the spectral analytical information: The at least one near-field image and its evaluation (in particular the suitability assessment) can be used in various ways for capturing, evaluating, or assessing a condition analysis based on spectral analytical information. These include, for example, the following techniques:

[0403] Using at least one near-field image to decide whether or when to acquire spectral analytical information: Initially, only near-field images can be acquired and evaluated continuously or in response to a trigger signal to assess the presence of defects. Only when the evaluation shows that the sample condition is within the intended range is the spectral measurement performed. The result can be output, for example, via an optical, acoustic, and / or tactile signal. Acquiring a near-field image prior to the spectroscopic measurement can save time if a detected anomaly is present, as the unnecessary calibration and spectroscopic measurement step can be omitted.

[0404] Using the at least one near-field image to trigger user confirmation: If the evaluation of the near-field image reveals that a problem exists, a human-machine interface can be triggered to enable user confirmation. The spectral analytical information can be acquired selectively only if the user confirmation via the human-machine interface confirms that the spectral analytical information should be acquired despite the problem.

[0405] Evaluation of the at least one near-field image after acquisition of the spectral analytical information: In this embodiment, the device 10 is configured such that the spectral analytical information is acquired independently of the evaluation of the at least one near-field image. An analysis of the spectral analytical information can also optionally be carried out independently of the evaluation of the at least one near-field image. The spectral analytical information and the near-field image can be evaluated independently of one another. Both results can be provided to the user, for example, as a result of the condition analysis in combination with an indication of the expected reliability of the condition analysis determined from the at least one near-field image. For this purpose, a traffic light-like system can be used, for example, to output the reliability via the human-machine interface 24 or display device 5.Alternatively or additionally, a visualization of the problem found in the sample (e.g., "aphid in view") can be performed.

[0406] Using the evaluation result of the at least one near-field image at a later time: The evaluation result of the at least one near-field image can also be used or even determined after the measurement. This allows, for example, the subsequent filtering out or marking of invalid measurements from statistics for historical analyses or for analyses that aggregate results from individual measurements.

[0407] Areas of application: The disclosed devices, evaluation devices, evaluation systems, systems, and methods can be used for various applications. In particular, the devices, evaluation devices, evaluation systems, systems, and methods can be used on crops or cultivated plants. The plants on which the techniques can be used according to the invention can, in particular, also include forestry plants (e.g., trees), hybrids, and / or genetically modified plants.

[0408] Examples of such plants can be:

[0409] • Crops

[0410] • Fodder plants

[0411] • Fiber plants

[0412] • Oil plants

[0413] • Ornamental plants

[0414] • Industrial crops, such as o Oilseeds o Tobacco o Hemp o Hops o Aromatic, culinary and medicinal plants o Seeds for herbaceous oil plants o Seeds for linseed (and consequently fibre flax) o Energy crops o Plants used for the production of feedstocks for renewable energy production

[0415] The condition analysis may include, but is not limited to, the qualitative and / or quantitative determination of nutrient concentrations. For example, the condition analysis may include analyses regarding one, several, or all of the following plant conditions:

[0416] Nutrient supply problems, for example o Nutrient deficiency o Toxicity biotic stress, for example due to o Nematodes o Insects o Arachnids o Fungi o Bacteria o Viruses

[0417] Abiotic stress, for example due to o drought o excessive water supply and / or waterlogging o salinity o temperature-related stress (cold, frost, heat) o UV light o metal toxicity o mechanical stress (e.g. pressure, squeezing, pressing).

[0418] The results of the evaluation of at least one near-field image and the condition analysis performed based on the spectral analysis information can be used to improve plant conditions. This can be achieved through automatic or semi-automatic adjustment of artificially added substances (such as approved fertilizers or approved insecticides) and artificially added water. Numerous other applications are possible, for example, for monitoring plant health with regard to environmental and regulatory issues.

[0419] Further areas of application: Even if embodiments have been described in the context of a condition analysis of plants, the devices, evaluation devices and evaluation systems and methods can also be used elsewhere, for example for image-based determination of the suitability of an agricultural product or of living organisms for the acquisition of spectral analytical information for further analysis.

[0420] Figure 26 is a flowchart of a method 190 that may be performed using the apparatus 10.

[0421] At 191, data is acquired using the device according to the invention. The data can be evaluated by the device 10 and / or the evaluation device or the evaluation system 4.

[0422] At 192, the results of the evaluation are used to determine measures to improve conditions for the plants. The measures can then be implemented to improve conditions.

[0423] The disclosed devices, evaluation devices, and evaluation systems, systems, and methods provide various technical effects. In particular, they offer improvements with regard to the acquisition, transmission, and / or evaluation of spectral analytical information for plant condition analysis. The methods, devices, and systems can be used to support the acquisition, transmission, and / or evaluation of spectral analytical information on crop plants or cultivated plants.

[0424] Embodiments of the invention can be used in particular in connection with plant cultivation in agriculture, which is a central economic and industrial sector of humanity. Global trends such as a growing population and a simultaneous decrease in the availability of agricultural land lead to a constantly growing demand for crop yields per area. At the same time, the unsustainable exploitation of soils leads to an increased need for targeted fertilization in order to maintain or even increase previous yields. Furthermore, a growing awareness of sustainable management and increasing regulatory pressure are creating new requirements for precisely tailored fertilization, in particular to avoid over-fertilization.

[0425] Of particular importance for the growth of any crop is the appropriate supply of nitrogen (N), as it is the most important element of chlorophyll and thus essential for optimal metabolism. In addition to N, other macronutrients are also relevant, such as phosphorus (P) and potassium (K), which can often be added to the soil together with N as so-called NPK fertilizers, as well as calcium (Ca), magnesium (Mg), and sulfur (S). Mg is the central element in the chlorophyll ring.

[0426] Additionally, depending on the plant type, other nutrients may also be relevant for optimal growth and yield, although these are often required in significantly smaller amounts and are therefore also referred to as micronutrients. These include, for example, boron (B), molybdenum (Mo), copper (Cu), manganese (Mn), zinc (Zn), iron (Fe), and chlorine (Cl).

[0427] Embodiments of the invention can contribute to ensuring optimal conditions with regard to nutrient supply throughout the entire growth phase of a plant. Only in this way can nutrient deficiencies or nutrient oversupply be appropriately responded to through adapted fertilizer applications (through a specific fertigation program in the substrate or soil growth, or as additional soil and / or foliar fertilization). Embodiments of the invention allow for regular monitoring of plants. Compared to laboratory-based leaf analysis, the condition analysis can be performed quickly. This is particularly desirable for fast-growing plants and plant species with rapid fruit ripening (e.g., strawberries and lettuce).Embodiments of the invention thus also address the objective that there should not be a delay of several days between measurement and availability of the evaluation result, as otherwise the measurement result would already be invalid and would not allow for any meaningful adjustment of the nutrient application. Rapid evaluation is also desirable when environmental factors change rapidly (e.g., temperature, precipitation, etc.), as otherwise the condition would change significantly between the measurement and the time of analysis.

[0428] Embodiments of the invention enable data acquisition for nutrient analysis at the intended application site without the need to transmit physical samples. Embodiments of the invention address the fact that more disruptive factors can occur in field use than in a laboratory scenario. Examples of these include: the presence of animals on the leaves (e.g. aphids, tripods, mites), the presence of water on the leaves (rain, water film, dew, fog), measuring leaves of the wrong plant species, measuring leaves of the wrong leaf age, measuring curled leaves, measuring the wrong side of the leaf (usually the underside of the leaf), measuring in a locally inhomogeneous and non-representative area of ​​the leaf blade (e.g. due to fertilizer or pesticide burn or plant spot formation), measuring contaminated leaf blades (e.g.covered by pollen or dirt), measuring over an area that is mostly covered by midribs or lateral ribs, and / or measuring small leaves whose coverage of the measuring window is difficult to assess with the naked eye, measuring diseased leaves.

[0429] Embodiments of the invention enable a nutrient analysis designed to minimize the user's influence on an invalid measurement result. Embodiments of the invention offer precautions for error-free, robust measurements with high repeatability and user-friendliness. Embodiments of the invention, in particular, offer measures that ensure that the susceptibility to errors when checking the leaf to be measured is reduced, so that ultimately only suitable leaves are used for nutrient analysis.

[0430] Embodiments of the invention can thus serve to provide a non-destructive, automated and at the same time fast (almost in real time) solution for detecting deficiencies and / or oversupplies of several nutrients in plants, wherein the solution is particularly user-friendly and has a high repeatability.

[0431] While exemplary embodiments have been described with reference to the figures, modifications can be implemented in further exemplary embodiments. For example, the modifications already explained can be used cumulatively or alternatively. The evaluation device 4 does not necessarily have to be provided separately from the device 10, but can also be structurally connected to it or integrated into the device 10.

[0432] While embodiments have been described that can be used in crops or cultivated plants, the disclosed techniques can also be used in other fields of application. The present disclosure also encompasses embodiments with any combination of features mentioned or shown for various embodiments. It also encompasses individual features in the figures, even if they are shown there in connection with other features and / or are not mentioned above or below. Furthermore, the alternatives to embodiments described in the figures and the description and individual alternatives of their features may be excluded from the subject matter of the invention or from the disclosed subject matter.

[0433] The terms "comprising" and "having" and derivatives thereof indicate a non-exhaustive relationship and do not exclude the presence of other elements or steps. The indefinite article "a" or "an" and derivatives thereof do not exclude the presence of a plurality of the corresponding elements. The functions of several features listed in the claims may be fulfilled by one unit or step.

[0434] A machine-readable instruction code that can be executed by a programmable circuit to perform methods according to embodiments may be stored and / or distributed on a suitable medium, such as an optical storage medium or a solid-state medium provided together with or as part of other hardware. The instruction code may also be distributed in another form, such as a modulated data signal sequence.

[0435] Embodiments of the invention provide improved techniques for analyzing the condition of plants.

Claims

CLAIMS 1. A device (10) for a condition analysis of a plant (2), comprising: a housing (20), a near-field camera (21) configured to capture at least one image of at least part of the plant (2), a spectral-analytical detection device (22) configured to capture spectral-analytical information (131, 132) of the plant, and at least one interface (23, 24) configured to output the at least one image and the spectral-analytical information (131, 132) or data derived therefrom, wherein the near-field camera (21) has a first field of view (41, 44) relative to the housing (20), wherein the spectral-analytical detection device (22) has a second field of view (42) relative to the housing (20), and wherein the second field of view (42) is completely contained in the first field of view (41, 44).

2. Device (10) according to claim 1, wherein the device (10) is configured such that, based on the at least one image, a suitability assessment of the region of the plant represented in the at least one image for the acquisition or evaluation of the spectral analytical information (131, 132) can be determined.

3. Device (10) according to claim 2, wherein the spectral analytical detection device (22) is configured to carry out the detection of the spectral analytical information (131, 132) selectively depending on the suitability assessment, and / or the device (10) is configured to transmit or analyze the spectral analytical information (131, 132) or the data derived therefrom via the interface selectively depending on the suitability assessment.

4. Device (10) according to claim 2 or claim 3, wherein the suitability assessment comprises a suitability of different positions or areas (60) of the plant for determining nutrient concentrations and / or a state of health, in particular a Water content and / or pathological conditions, based on the spectral analytical information (131, 132), wherein the device (10) is configured such that the spectral analytical detection device (22) carries out the detection of the spectral analytical information (131, 132) depending on the suitability of different positions or areas (60).

5. Device (10) according to one of claims 2 to 4, wherein the device (10) is configured to receive the suitability assessment via the at least one interface (23).

6. Device (10) according to one of the preceding claims, further comprising a device (29) for reducing ambient light influences, wherein optionally the device (29) for reducing ambient light influences comprises a diaphragm (29) provided on the housing (20) which projects at least partially outwards around a viewing window held in the housing (20).

7. Device (10) according to one of the preceding claims, wherein the device (10) is designed as a movable device (10), wherein optionally the device (10) is designed as a manually held device (10).

8. Evaluation device or system (4) for a condition analysis of a plant (2), comprising: at least one evaluation device interface (41, 44) which is configured to receive at least one image of at least a part of the plant (2), and an evaluation circuit (45) which is configured to determine, based on the at least one image, a suitability assessment of the region of the plant shown in the at least one image for a detection or evaluation of spectral analytical information (131, 132).

9. Evaluation device or system according to claim 8, wherein the evaluation circuit (45) is arranged to detect one or more conditions for suitability assessment which are selected from a group consisting of: presence of foreign substances, necrosis, chlorosis, anthocyanosis, spoiled plant areas, pest infestation, incorrect positioning of the area of ​​the plant during image acquisition, crack or breakage in a part of the plant.

10. Evaluation device or system according to claim 8 or claim 9, further comprising a human-machine interface (5) arranged to receive a user input specifying one, several or all of the following information: Plant stage, leaf stage, plant species, Plant variety, at least one nutrient to be analyzed, at least one disease to be tested, wherein the evaluation circuit (45) is configured to carry out the suitability assessment depending on the information or information contained in the user input.

11. Evaluation device or system according to one of claims 8 to 10, wherein the at least one evaluation device interface (41, 44) is set up to output the suitability assessment for use in the acquisition of spectral analytical information (131, 132), wherein the at least one evaluation device interface (41, 44) is optionally set up to transmit the suitability assessment to the device (10) according to one of claims 1 to 7, and / or wherein optionally the evaluation device or the evaluation system is set up to control the at least one evaluation device interface (41, 44) in order to enable a selection of the leaf regions to be analyzed by the user.

12. Evaluation device or system according to one of claims 8 to 11, wherein the evaluation circuit (45) is configured to carry out the suitability assessment without using spectral analytical information (131, 132), and / or wherein the evaluation circuit (45) is configured to evaluate the at least one image for the suitability assessment using at least one trained machine learning model (100) which is configured to receive pixel values ​​of the at least one image or values ​​derived therefrom as input.

13. Evaluation device or system according to one of claims 8 to 12, wherein the evaluation device interface (41) is configured to receive spectral analytical information (131, 132) of the plant or data derived therefrom, and wherein the evaluation circuit (45) is configured to carry out at least one nutrient analysis based on the spectral analytical information (131, 132) or the data derived therefrom.

14. System comprising: the device (10) according to any one of claims 1 to 7 and the evaluation device or the evaluation system (4) according to one of claims 8 to 13.

15. Method for a condition analysis of a plant (2), comprising: Receiving at least one image of the plant (2) and Determining a suitability assessment of the plant for a recording or evaluation of spectral analytical information (131, 132) of the plant (2), wherein the suitability assessment is determined based on the at least one image, and wherein the recording and / or evaluation of the spectral analytical information (131, 132) of the plant is carried out depending on the suitability assessment.

16. The method of claim 15, further comprising outputting the suitability assessment via a human-machine interface (24) for visualizing the suitability assessment.

17. The method according to claim 15 or claim 16, wherein the suitability assessment comprises a spatially resolved assessment of the suitability of different positions or regions (60) of the plant (2) for determining nutrient concentrations and / or a state of health, in particular a water content and / or pathological conditions, based on the spectral analytical information.

Images