Methods for deploying biosentinels to agricultural fields and remotely monitoring biotic and abiotic stresses in crops

Abstract

The present application relates to a method for deploying biosentinels in agricultural fields and remotely monitoring biotic and abiotic stresses in crops. A variation of the method for interpreting stress in plants includes: acquiring a first image of a first group of sentinel plants in a field; acquiring a second image of a second group of sentinel plants in the field recorded during a first time period; interpreting a first stressor in the first group based on features extracted from the first image captured during the first time period; interpreting a second stressor in the second group based on features extracted from the second image; deriving a model that relates the stressor in the first group to the stressor in the second group based on the first and second stressors; interpreting a third stressor in the first group based on features extracted from a third image captured during the second time period; and predicting a fourth stressor in the second group during the second time period based on the third stressor and the model.

Description

[0001] This application is a divisional application of an application with a filing date of June 22, 2020, application number 202080054171.6, and invention name “Method for deploying biological sentinels to farmland and remotely monitoring biotic and abiotic stresses in crops”.

[0002] CROSS-REFERENCE TO RELATED APPLICATIONS

[0003] This application claims the benefit of U.S. Provisional Application No. 62 / 864,401, filed June 20, 2019, which is incorporated by reference in its entirety. Technical Field

[0004] The present invention relates generally to the field of agriculture, and more particularly to a new and useful method based on biosensors in the field of agriculture for deploying biosensors to agricultural fields and monitoring plant stressors in crops. Summary of the Invention

[0005] This application provides the following:

[0006] 1). A method for interpreting stress in a plant, comprising:

[0007] acquiring a first set of images recorded at a first frequency by a fixed sensor facing a first set of sentinel plants in a field;

[0008] acquiring a second image of a second group of sentinel plants in the field, the second image recorded by the mobile sensor during the first time period;

[0009] interpreting a first pressure of a stressor in the first set of sentinel plants during the first time period based on a first set of features extracted from a first image of the first set of images captured during the first time period;

[0010] interpreting a second pressure of the stressor in the second set of sentinel plants during the first time period based on a second set of features extracted from the second image;

[0011] deriving a stress model based on the first stress and the second stress, the stress model relating the stress of the stressor at the first group of sentinel plants to the stress of the stressor at the second group of sentinel plants;

[0012] interpreting a third pressure of the stressor in the first set of sentinel plants during the second time period based on a third set of features extracted from a third image of the first set of images captured during the second time period;

[0013] predicting a fourth stress of the stressor in the second group of sentinel plants during the second time period based on the third stress and the stress model; and

[0014] In response to the fourth stress in the second group of sentinel plants exceeding a threshold stress, a prompt is generated to address the stressor in plants proximate to the second group of sentinel plants in the agricultural field.

[0015] 2) The method according to 1), further comprising:

[0016] acquiring a fourth image of the farmland recorded by the aerial sensor during the first time period;

[0017] interpreting a first pressure gradient in the agricultural field during the first time period based on a fourth set of features extracted from a region of the fourth image, the region of the fourth image including the first group of sentinel plants, the second group of sentinel plants, and a third group of sentinel plants in the agricultural field; and

[0018] The first pressure gradient in the field during the first time period is corrected based on the first pressure and the second pressure.

[0019] 3) The method according to 1), further comprising:

[0020] acquiring a fourth image of a third group of sentinel plants in the field, the fourth image recorded by the mobile sensor during the first time period;

[0021] interpreting a fourth pressure of the stressor in the third set of sentinel plants during the first time period based on a fourth set of features extracted from the fourth image;

[0022] interpreting a first pressure gradient in the farmland during the first time period based on the first pressure, the second pressure, and the fourth pressure;

[0023] deriving a gradient model based on the first pressure, the second pressure, and the third pressure, the gradient model relating the pressure of the stressor at the first group of sentinel plants, the pressure of the stressor at the second group of sentinel plants, and the pressure of the stressor at the third group of sentinel plants; and

[0024] A second pressure gradient in the field during the second time period is interpreted based on the third pressure in the first set of sentinel plants and the gradient model.

[0025] 4) The method according to 1), wherein generating a prompt for resolving the stressor in plants near the second group of sentinel plants in the farmland further comprises:

[0026] Isolating a first action from a set of actions defined for a sentinel plant, the first action being associated with the stressor; and

[0027] A notification is sent to a computing device of a user associated with the farmland to perform the first action in the farmland to alleviate the stressor.

[0028] 5). According to the method described in 1):

[0029] wherein acquiring the first set of images recorded by the fixed sensor comprises acquiring the first set of images recorded by a camera mounted on a fixed beam, the fixed beam being located at a center of the first set of sentinel plants within the field; and

[0030] Wherein acquiring a second batch of images recorded by the mobile sensor includes acquiring the second batch of images recorded by a camera of a mobile device of a user associated with the farmland.

[0031] 6). The method according to 1):

[0032] wherein acquiring the first set of images of the first set of sentinel plants in the farmland comprises acquiring the first set of images of the first set of sentinel plants including a first cluster of sentinel plants arranged near a center of the farmland; and

[0033] Wherein acquiring the second image of the second group of sentinel plants in the farmland includes acquiring the second image of the second group of sentinel plants including a second cluster of sentinel plants arranged in a row along an edge of the farmland.

[0034] 7). The method according to 1), wherein obtaining the first batch of images of the first group of sentinel plants includes obtaining the first batch of images recorded by the fixed sensor facing the first group of sentinel plants, the first group of sentinel plants includes a group of promoters and a group of reporters forming a group of promoter-reporter pairs, the group of promoter-reporter pairs is configured to emit a signal of the stress of a group of stressors at the sentinel plants, and the group of promoter-reporter pairs includes a first promoter-reporter pair configured to emit a signal of the stress of the stressor at the first group of sentinel plants.

[0035] 8). According to the method described in 1):

[0036] Wherein, acquiring the first batch of images recorded by the fixed sensor facing the first group of sentinel plants in the farmland comprises: acquiring the first batch of images recorded by the fixed sensor facing the first group of sentinel plants in a greenhouse environment; and

[0037] Wherein, obtaining the second image of the second group of sentinel plants in the farmland recorded by the mobile sensor includes: obtaining the second image of the second group of sentinel plants, wherein the second group of sentinel plants are arranged in vertically stacked layers in the greenhouse environment.

[0038] 9) A method for interpreting stress in a plant, comprising:

[0039] acquiring a first set of images recorded at a first frequency by a fixed sensor facing a first set of sentinel plants in a field;

[0040] acquiring a second image of an area of ​​the field including the first set of sentinel plants, the second image recorded by a mobile sensor during a first time period;

[0041] interpreting a first pressure of a stressor in the first set of sentinel plants during the first time period based on a first set of features extracted from a first image of the first set of images captured during the first time period;

[0042] interpreting a first pressure gradient of the stressor in sentinel plants in the region of the agricultural field during the first time period based on a second set of features extracted from the second image;

[0043] deriving a gradient model based on the first pressure of the stressor and the first pressure gradient, the gradient model relating the pressure of the stressor at the first group of sentinel plants to the pressure gradient of the stressor in the region of the agricultural field;

[0044] interpreting a second stress of the stressor in the first set of sentinel plants during the second time period based on a third set of features extracted from a third image of the first set of images captured during the second time period;

[0045] predicting a second pressure gradient of the stressor in the area of ​​the farmland during the second time period based on the second pressure and the gradient model; and

[0046] In response to predicting a third pressure in a sub-region of the agricultural field based on the second pressure gradient and the third pressure exceeding a threshold pressure, generating a prompt to address the stressor in plants occupying the agricultural field adjacent to the sub-region of the agricultural field.

[0047] 10). The method according to 9):

[0048] Wherein, acquiring the second image of the area of ​​the farmland including the first group of sentinel plants comprises: acquiring the second image of the area of ​​the farmland including the first group of sentinel plants and the second group of sentinel plants;

[0049] The method further comprises:

[0050] acquiring a second batch of images recorded at the first frequency by a second fixed sensor facing the second set of sentinel plants in the field; and

[0051] interpreting a third pressure of the stressor in the second set of sentinel plants during the first time period based on a fourth set of features extracted from a fourth image in the second batch of images captured during the first time period; and

[0052] Among them, deriving a gradient model that associates the pressure of the stressor at the first group of sentinel plants and the pressure gradient of the stressor in the area of ​​the farmland based on the first pressure of the stressor and the first pressure gradient includes: deriving a gradient model that associates the pressure of the stressor at the first group of sentinel plants, the pressure of the stressor at the second group of sentinel plants and the first pressure gradient.

[0053] 11). The method according to 9):

[0054] Wherein, acquiring the first batch of images recorded by the fixed sensor facing the first group of sentinel plants in the farmland comprises: acquiring the first batch of images recorded by the fixed sensor facing the first group of sentinel plants in the sentinel plant population;

[0055] wherein acquiring the second image of the area of ​​the farmland including the first group of sentinel plants comprises: acquiring the second image of the sentinel plant population, the second image comprising a group of pixels, each pixel in the group of pixels corresponding to a group of sentinel plants in the sentinel plant population; and

[0056] Wherein, interpreting the first pressure gradient of the stressor in the sentinel plants in the region of the farmland based on the second set of features extracted from the second image comprises:

[0057] interpreting a set of stresses in the sentinel plant population based on groups of features in the second set of features, each group of features extracted from pixels in the set of pixels; and

[0058] Based on each pressure in the set of pressures, a first pressure gradient of the stressor in the sentinel plant population is generated.

[0059] 12). The method according to 9):

[0060] wherein acquiring the first batch of images of the first group of sentinel plants in the farmland comprises acquiring a batch of images of the first group of sentinel plants in the farmland, the sentinel plants in the first group of sentinel plants comprising a first promoter associated with plant dehydration and a first reporter associated with red fluorescence, the first promoter and the first reporter forming a first promoter-reporter pair, the first promoter-reporter pair being configured to signal plant dehydration at the sentinel plants in the first group of sentinel plants;

[0061] wherein interpreting the first stress of the stressor in the first group of sentinel plants based on the first set of features extracted from the first image comprises: interpreting a first stress of plant dehydration in the first group of sentinel plants based on a first set of red fluorescence measurements extracted from the first image;

[0062] wherein interpreting the first pressure gradient of the stressor in the sentinel plants in the region of the agricultural field based on the second set of features extracted from the second image comprises: interpreting a first pressure gradient of plant dehydration in the sentinel plants in the region of the agricultural field based on a second set of red fluorescence measurements extracted from the second image;

[0063] wherein interpreting the second stress of the stressor in the first group of sentinel plants based on the third set of features extracted from the third image comprises: interpreting the second stress of plant dehydration in the first group of sentinel plants based on a third set of red fluorescence measurements extracted from the third image;

[0064] wherein predicting the second pressure gradient of the stressor in the region of the farmland based on the second pressure and the gradient model comprises: predicting a second pressure gradient of plant dehydration in the region of the farmland based on the second pressure and the gradient model; and

[0065] Wherein, in response to the second pressure gradient exceeding a threshold pressure gradient, generating a prompt to resolve the stressor in the plants of the farmland near the area occupying the farmland includes: in response to the second pressure gradient of plant dehydration exceeding the threshold pressure gradient of plant dehydration, generating a prompt to irrigate the plants of the farmland near the area occupying the farmland to a degree corresponding to the second pressure gradient.

[0066] 13). The method according to 9):

[0067] Wherein, acquiring the first batch of images comprises acquiring a first batch of spectral images captured by a first spectrometer;

[0068] The method further includes: obtaining a reporter model that relates solar-induced fluorescence measurements extracted from the spectral image to stress of the stressor of the sentinel plant; and

[0069] Wherein, interpreting the first stress in the first group of sentinel plants includes interpreting the first stress of the stressor based on first solar-induced fluorescence measurements extracted from the first image.

[0070] 14). The method according to 9), wherein interpreting the first stress of the first stressor in the first group of sentinel plants based on the first set of features extracted from the first image comprises:

[0071] extracting a first feature of the first set of features from the first image, the first feature corresponding to a first pixel of the first image;

[0072] extracting a second feature from the set of features from the first image, the second feature corresponding to a second pixel of the first image;

[0073] estimating a representative feature based on a combination of the first feature and the second feature;

[0074] obtaining a reporter sub-model that correlates features extracted from images in the first batch with stress from the first stressor at the first set of sentinel plants; and

[0075] The first stress of the first stressor in the first group of sentinel plants is interpreted based on the representative features and the reporter model.

[0076] 15). The method according to 14):

[0077] Wherein, extracting the first feature from the first image comprises: extracting a first intensity of a first wavelength from a first spectral image of the first group of sentinel plants;

[0078] wherein extracting the second feature from the first image comprises extracting a second intensity of the first wavelength from the first spectral image;

[0079] Wherein, estimating the representative feature comprises: estimating a first average intensity of the first intensity and the second intensity at the first wavelength;

[0080] wherein obtaining a reporter sub-model associating features extracted from the images in the first batch with the pressure of the first stressor at the first group of sentinel plants comprises: obtaining the reporter sub-model associating the average intensity of the first wavelength extracted from the spectral images of the first batch with the pressure of the first stressor at the first group of sentinel plants; and

[0081] Wherein, explaining the first pressure of the first stressor in the first group of sentinel plants based on the representative characteristics and the reporter model includes: explaining the first pressure of the first stressor in the first group of sentinel plants based on the first average intensity and the reporter model.

[0082] 16) A method for interpreting stress in a plant, comprising:

[0083] acquiring a first set of images of a first set of sentinel plants in a field, the first set recorded by a fixed sensor at a first frequency;

[0084] acquiring a second image of the farmland recorded by the aerial sensor during the first time period;

[0085] interpreting a first stress of a first stressor from a set of stressors in the first set of sentinel plants during the first time period based on a first set of features extracted from a first image in the first set of images captured during the first time period;

[0086] interpreting a second stress of the first stressor in the first group of sentinel plants during the first time period based on a second set of features extracted from a region of a second image including the first group of sentinel plants;

[0087] interpreting a first pressure gradient of the first stressor in the agricultural field during the first time period based on a third set of features extracted from a region of the second image;

[0088] deriving a model that relates the pressure of the first stressor at the first group of sentinel plants to the pressure gradient of the first stressor in the field based on the second pressure and the first pressure gradient; and

[0089] The first pressure gradient of the first stressor in the agricultural field during the first time period is corrected based on the first pressure and the model.

[0090] 17) The method according to 16), further comprising:

[0091] interpreting a third stress of the first stressor in the first group of sentinel plants during a second time period after the first time period based on a fourth set of features extracted from a third image of the first group of images captured during the second time period;

[0092] predicting a second pressure gradient of the first stressor in the agricultural field during the second time period based on the third pressure and the model; and

[0093] In response to the second pressure gradient exceeding a threshold pressure gradient, a prompt is generated to address the stressor in plants in the agricultural field.

[0094] 18) The method according to 16)

[0095] wherein acquiring the first batch of images recorded by the fixed sensor comprises acquiring the first batch of images recorded by a camera mounted on a fixed beam located within the farmland; and

[0096] Wherein, obtaining the second image of the farmland recorded by the aerial sensor includes obtaining the second image of the farmland recorded by a satellite.

[0097] 19) The method according to 16) further comprising:

[0098] acquiring a second batch of images of a second group of sentinel plants in the field, the second batch being recorded at a second frequency less than the first frequency;

[0099] interpreting a third stress of a second one of the set of stressors in the second set of sentinel plants during the first time period based on a fourth set of features extracted from a third image in the second batch of images captured during the first time period;

[0100] interpreting a fourth stress of the second stressor in the second group of sentinel plants during the first time period based on a fifth set of features extracted from the region of the second image;

[0101] interpreting a second pressure gradient of the second stressor in the farmland during the first time period based on a sixth set of features extracted from a region of the second image;

[0102] deriving a model that relates the pressure of the second stressor at the second group of sentinel plants to the pressure gradient of the second stressor in the field based on the fourth pressure and the second pressure gradient;

[0103] correcting the second pressure gradient of the second stressor in the field during the first time period based on the third pressure and the model; and

[0104] A pressure map is generated based on a combination of the first pressure gradient of the first stressor and the second pressure gradient of the second stressor.

[0105] 20). The method according to 19):

[0106] Wherein, acquiring the first batch of images of the first group of sentinel plants comprises: acquiring the first batch of images of a first cluster of sentinel plants arranged in the farmland, wherein the sentinel plants in the first cluster of sentinel plants are configured to signal the presence of a fungal stressor via pigment changes;

[0107] wherein acquiring the second batch of images of the second group of sentinel plants comprises: acquiring the second batch of images of a second cluster of sentinel plants arranged in a row along an edge of the farmland, the sentinel plants in the second cluster of sentinel plants being configured to signal the presence of an insect stressor by emitting light;

[0108] wherein interpreting the first stress of the first stressor in the first group of sentinel plants comprises: interpreting the first stress of the fungal stressor in the first cluster of sentinel plants based on pigment changes of the sentinel plants detected in the first image; and

[0109] Wherein, interpreting the third pressure of the second stressor in the second group of sentinel plants includes: interpreting the third pressure of the insect stressor in the second group of sentinel plants based on the luminescence of the sentinel plants in the second cluster of sentinel plants detected in the third image. BRIEF DESCRIPTION OF THE DRAWINGS

[0110] Figure 1 is a flowchart representation of the method;

[0111] Figure 2 is a flowchart representation of the method;

[0112] Figure 3 is a flowchart representation of the method;

[0113] Figure 4 It is a diagram of the method;

[0114] Figure 5 It is a diagram of the method;

[0115] Figure 6 is a diagram of the method; and

[0116] Figure 7 is a schematic representation of a fixed optical sensor.

[0117] Description of Embodiments

[0118] The following description of the embodiments of the present invention is not intended to limit the invention to these embodiments, but rather to enable those skilled in the art to make and use the invention. The variations, configurations, embodiments, example embodiments, and examples described herein are optional and do not exclude the variations, configurations, embodiments, example embodiments, and examples they describe. The invention described herein may include any and all combinations of these variations, configurations, embodiments, example embodiments, and examples.

[0119] 1. Methods

[0120] like Figure 1-3 As shown, method S100 includes: in block S110, accessing a first feed of images recorded at a first frequency by a fixed sensor facing a first group of sentinel plants in a farmland; in block S112, accessing a second image of a second group of sentinel plants in the farmland, the second image being recorded by a mobile sensor during a first time period; in block S120, interpreting a first pressure of a stressor in the first group of sentinel plants during the first time period based on a first set of features extracted from a first image captured during the first time period in the first feed of images; in block S122, interpreting a second pressure of the stressor in the second group of sentinel plants during the first time period based on a second set of features extracted from the second image; in block S130, deriving a pressure based on the first pressure and the second pressure. a force model that correlates a pressure of a stressor at a first group of sentinel plants with a pressure of a stressor at a second group of sentinel plants; in block S140, interpreting a third pressure of the stressor in the first group of sentinel plants during the second time period based on a third set of features extracted from a third image captured during the second time period from the first group of images; in block S150, predicting a fourth pressure of the stressor in the second group of sentinel plants during the second time period based on the third pressure and the pressure model; and, in block S160, generating a prompt to resolve the stressor in plants near the second group of sentinel plants in the field in response to the fourth pressure in the second group of sentinel plants exceeding a threshold pressure.

[0121] like Figure 2 and Figure 3As shown, a variation of method S100 includes: in block S110, obtaining a first set of images recorded at a first frequency by a fixed sensor facing a first group of sentinel plants in a farmland; in block S114, obtaining a second image of an area of ​​the farmland including the first group of sentinel plants, the second image being recorded by a mobile sensor during a first time period; in block S120, interpreting a first pressure of a stressor in the first group of sentinel plants during the first time period based on a first set of features extracted from a first image captured during the first time period in the first set of images; in block S124, interpreting a first pressure gradient of the stressor in the sentinel plants in the area of ​​the farmland during the first time period based on a second set of features extracted from the second image; and in block S132, interpreting a first pressure gradient of the stressor in the sentinel plants in the area of ​​the farmland during the first time period based on the stressor. a first pressure of the stressor and a first pressure gradient, deriving a gradient model relating the pressure of the stressor at the first group of sentinel plants to the pressure gradient of the stressor in the region of the farmland; in block S140, interpreting a second pressure of the stressor in the first group of sentinel plants during the second time period based on a third set of features extracted from a third image captured during the second time period from the first group of images; in block S152, predicting a second pressure gradient of the stressor in the region of the farmland during the second time period based on the second pressure of the stressor and the model; and in block S160, in response to the second pressure gradient, predicting a third pressure in the sub-region of the farmland and the third pressure exceeding a threshold pressure, generating a prompt to resolve the stressor in plants in the farmland near the sub-region of the farmland.

[0122] like Figure 2 and Figure 3 As shown, a variation of method S100 includes: in block S110, obtaining a first set of images of a first group of sentinel plants in a field, the first set of images being recorded by a fixed sensor at a first frequency; in block S114, obtaining a second image of the field recorded by an aerial sensor during a first time period; in block S120, interpreting a first stress of a first stressor in a set of stressors in the first group of sentinel plants during the first time period based on a first set of features extracted from a first image captured during the first time period in the first set of images; in block S122, interpreting a first stressor of a first stressor in a set of stressors in the first group of sentinel plants based on a region of the second image including the first group of sentinel plants In block S124, based on a third set of features extracted from the region of the second image, a second pressure of the first stressor in the farmland during the first time period is interpreted; in block S132, based on the second pressure and the first pressure gradient, a model is derived that relates the pressure of the first stressor at the first group of sentinel plants to the pressure gradient of the first stressor in the farmland; and in block S134, based on the first pressure and the model, the first pressure gradient of the first stressor in the farmland during the first time period is corrected.

[0123] like Figure 3 As shown, a variation of method S100 includes: in block S110, acquiring a first batch of ground-based images recorded at a first frequency by a fixed sensor, the fixed sensor facing a first group of sentinel plants in the farmland; in block S112, acquiring a second batch of ground-based images of a second group of sentinel plants in the farmland recorded at a second frequency by a mobile ground sensor; in block S114, acquiring a third batch of aerial images of the farmland recorded at a third frequency that is less than the first frequency and the second frequency; in block S120, estimating a first stress of a stressor in the first group of sentinel plants at a first time based on a first set of features extracted from the first batch; and in block S122 Based on a second set of features extracted from the second batch, estimating a second pressure of the stressor in the second group of sentinel plants at a second time; interpolating the pressure in plants between the first group and the second group at the second time based on the first pressure and the second pressure; in block S124, calculating a first pressure gradient in the field at the first time based on a third set of features extracted from a region of the third batch, the third set of features depicting the first group of sentinel plants, the second group of sentinel plants, and the third group of sentinel plants in the field; and in block S126, correcting the first pressure gradient in the field at the first time based on the first stressor in the first group and the second stressor in the second group. The method may also include, in block S160, providing a prompt to an operator associated with the field to address the pressure of the stressor in the field based on the first pressure, the second pressure, and the first pressure gradient.

[0124] 2. Application

[0125] Typically, a computer system (e.g., a local computing device, a remote server, a computer network) executes blocks of method S100: identifying a stressor present at a sentinel plant based on a signal (e.g., fluorescence in the electromagnetic spectrum) generated by the sentinel plant, the sentinel plant being genetically modified to signal that environmental conditions are unfavorable for plant health or growth; interpreting the presence and / or magnitude of stressors at other nearby plants based on the signal generated by the sentinel plant; and selectively generating and distributing cues to mitigate stressors at the sentinel plant and / or nearby plants.

[0126] More specifically, sentinel plants can be genetically engineered to contain a set of promoter-reporter pairs configured to trigger the generation of a signal within the sentinel plant in the presence of a specific biotic and / or abiotic stressor to which the sentinel plant is exposed, such as: insect pests; viral diseases; excess or insufficient water; excessive heat or cold; and / or nutrient deficiencies. An optical device can record the optical signals generated by the sentinel plant (e.g., in the form of color or multispectral images); and a computer system can extract features from these images (e.g., intensity at specific wavelengths), interpret the presence and / or magnitude of the specific stressor to which the sentinel plant is exposed based on these features, and interpolate or extrapolate the health and environmental conditions of other nearby plants (e.g., non-sentinel plants; other non-imaged sentinel plants) based on the presence and / or magnitude of the stressor thus indicated by the sentinel plant.

[0127] For example, a computer system can extract the intensities of specific wavelengths corresponding to specific compounds (e.g., proteins) in a sentinel plant and, based on the intensities of these wavelengths (e.g., based on a stored model that associates plant stressors with wavelengths of interest, based on known properties of promoters and reporter genes in the sentinel plant), interpret the stress of exposure to a specific stressor in the sentinel plant before such stressor becomes visually discernible (i.e., with the human eye) in the visible spectrum. The computer system can also interpolate or extrapolate the presence or magnitude of these stressors in other plants near the sentinel plant to predict the overall health of the crop or field.

[0128] 2.1 Application: Sentinel Plant Clusters and Fixed Sensors

[0129] In one example, sentinel plants can be genetically engineered to include a promoter that indicates a fungal stressor found in corn crops. The promoter can be paired with a red fluorescent reporter so that the sentinel plants exhibit red fluorescence when exposed to stress from this fungus exceeds a threshold magnitude and / or exceeds a threshold time period. Sentinel plants exhibiting this characteristic can be planted in clusters throughout a field planted with a commercial non-sentinel corn crop, such as near the center of the field. Optical sensors (e.g., multispectral cameras) mounted on poles within the center cluster sentinel plants can collect images of adjacent sentinel plants, such as every hour or every day, and download these images to a computer system (e.g., via a computer network). The computer system can then extract the amplitude (e.g., intensity) of the wavelengths of the red fluorescent reporter from these images and implement a stored model to interpret the stress (e.g., presence and / or amplitude) of the fungal stressor in the center cluster sentinel plants over time based on the amplitude of these wavelengths.

[0130] Based on the interpreted pressure of the fungal stressor, the computer system can recommend a specific action or set of actions to reduce the pressure of the fungal stressor. More specifically, the computer system can: separate a subset of actions associated with reducing the fungal stressor in the action set; and separate a first action associated with the pressure of the fungal stressor in the action subset. For example, the computer system can recommend a first action for reducing fungal pressure that is above a threshold fungal pressure and a second action for reducing fungal pressure that is below a threshold fungal pressure. In addition, the computer system can recommend mitigation or treatment techniques to be applied to plants near the center cluster sentinel plants (such as within a specific distance from the center of the crop) based on the pressure of the fungal stressor. For example, the computer system can recommend: treating plants (e.g., sentinel plants and non-sentinel plants) within a first radius from the center cluster with a first amount of fungicide at a first frequency; treating plants outside the first radius and within a second radius from the center cluster with a second amount of fungicide that is less than the first amount at a first frequency; and treating plants outside a second radius and within a third radius from the second cluster with a second amount of fungicide at a second frequency that is less than the first frequency. In another example, the computer system can recommend treatments for surrounding plants based on the predicted spread of fungal stress across the crop (e.g., based on previous stress from a fungal stressor). In yet another example, the computer system can recommend collecting samples from soil and plants near the fungal stressor in the cluster of sentinel plants to gather a more precise diagnosis of the type and spread of fungal stress throughout the crop and determine appropriate treatments.

[0131] 2.2 Applications: Sentinel Plant Clusters and Ground-Based Mobile Sensors

[0132] In the previous example, sentinel plants can be planted in other clusters throughout the field, such as near each corner of the field. A mobile optical sensor mounted on a truck, tractor, or other implement can intermittently (such as multiple times a day per week) capture images of these cluster sentinel plants while traveling along a pathway in the field. The computer system can: acquire these images; implement the above-described methods and techniques to extract the amplitude of the wavelength of the red fluorescent reporter from these images; implement the stored stress model to interpret the stress (e.g., presence and / or amplitude) of the fungal stressor in these cluster sentinel plants based on the amplitude of these wavelengths; pair these stressor diagnostics for these corner clusters with the temporally most recent stressor diagnostics for the center cluster; and compile concurrent stressor diagnostics for the corner and center clusters over time to generate a model that predicts the presence and amplitude of fungi at the corner clusters based on the presence and amplitude of fungi at the center cluster.

[0133] Later, a scanning cycle can: implement the model to predict fungal pressure at the corner clusters based on the fungal pressure derived from the next image of the center cluster; interpolate the fungal pressure across the crop between the center and corner clusters; and generate tips or recommendations for fungal reduction in all or specific areas of the field.

[0134] In one embodiment, a mobile optical sensor can capture images of sentinel plants at different frequencies and at different locations within the crop to achieve greater spatial resolution. The mobile optical sensor can collect these images intermittently and inconsistently (e.g., lower temporal resolution). However, the computer system can utilize data extracted from these images recorded by the mobile optical sensor, combined with consistent data extracted from images recorded by the fixed sensor on a single cluster of sentinel plants, to extend fungal pressure predictions on the crop. Furthermore, the computer system can converge on a more accurate model for predicting time-varying stress on the crop based on the data extracted from these images, such as by incorporating a machine learning algorithm.

[0135] 2.3 Applications: Sentinel Plant Clusters and Aerial Sensors

[0136] In the previous example, an aerial optical sensor may intermittently (such as once every two weeks or once a month) capture an image of the farmland including each cluster of sentinel plants. Based on the aerial image of the farmland, the computer system may interpret the pressure gradient in the farmland and / or the pressure at each cluster of sentinel plants in the farmland. The computer system may distinguish between clusters of sentinel plants and non-sentinel plants in the aerial image, such as by: overlaying a mask on the aerial image, the mask being configured to mask areas in the image corresponding to non-sentinel plants in the farmland; detecting baseline signal characteristics of the sentinel plants, but not associating them with fungal pressure in sub-regions of the image corresponding to sentinel plants in the farmland; and / or matching geolocation tags included in the aerial image with known GPS locations of sentinel plants in the farmland. When matching sub-regions of the aerial image corresponding to multiple clusters of sentinel plants, the computer system may derive the pressure of the fungal stressor for each sub-region and interpolate between the pressures at each sub-region to interpret the pressure gradient across the farmland. In addition, the computer system may interpret the pressure of the fungal stressor at a cluster of sentinel plants based on images recorded by fixed sensors over concurrent time periods. Then, based on the location of the fixed sensors corresponding to a specific sub-area of ​​the agricultural field, the computer system can derive a scalar that relates the pressure of the fungal stressor at that specific sub-area recorded by the fixed sensors to the pressure of the fungal stressor at that specific sub-area recorded by the aerial optical sensor.

[0137] The computer system can then correct (e.g., scale) the pressure at each sub-region or cluster of sentinel plants based on the scalar. Based on these updated pressures, the computer system can generate prompts for reducing fungal pressure in sub-regions of the crop as needed.

[0138] Furthermore, the computer system can derive a gradient model (e.g., a scalar) that relates the pressure of a fungal stressor at a cluster of sentinel plants in the center of the crop, as recorded by a fixed sensor, to the pressure of the stressor at other clusters of sentinel plants (e.g., the gradient of the field). Later, the computer system can: acquire an image of the cluster of sentinel plants recorded by the fixed sensor; interpret the pressure of the fungal stressor at the cluster of sentinel plants based on features extracted from the image; and predict the pressure of the fungal stressor at each sub-region of the field based on the pressure of the fungal stressor at the cluster of sentinel plants and the gradient model.

[0139] 3. Terminology

[0140] As described above, a "sentinel plant" refers herein to a plant that is configured to signal the presence of a particular stressor or a group of stressors in and / or at a plant. Sentinel plants can be genetically modified to include a set of promoter-reporter pairs (e.g., one promoter-reporter pair, three promoter-reporter pairs) that are configured to trigger the sentinel plant to generate a detectable signal or multiple signals in the presence of a particular stressor or a group of stressors. For example, a sentinel plant can be genetically modified to include a first promoter-reporter pair that is configured to trigger the sentinel plant to generate a red fluorescent signal in the presence of a fungus. Thus, the sentinel plant can generate a detectable signal that, when detected, can alert a user associated with the sentinel plant (e.g., a farmer, an agronomist, a botanist) to the presence of a stressor or multiple stressors. In addition, sentinel plants of a first plant type can be configured to signal the presence of a stressor in plants of a first type and / or different types. For example, a sentinel corn plant can be configured to signal the presence of a stressor in a corn plant. In another example, a sentinel tomato plant can be configured to signal the presence of a stressor in a potato plant.

[0141] In one embodiment, sentinel plants can monitor for the presence of stressors (e.g., pests, diseases, dehydration) in other (non-sentinel) plants. Generally, a small number of sentinel plants can be monitored to extract insights into a larger population of plants (e.g., crops). For example, a cluster of sentinel plants can be planted along the outer edge of a crop of plants and monitored for the presence of pests, notifying a user (e.g., farmer, agronomist, botanist) associated with the crop if and / or when a pest population has entered the crop along the outer edge. In another example, sentinel plants of a first plant type (e.g., tomatoes) can be grown in a greenhouse environment (e.g., a glass roof or plant farm) located in a particular area and monitored for the presence of stressors (e.g., dehydration, disease, pests) that are indicative of plant health. Users associated with the greenhouse environment can extract insights from the stressors present at the sentinel plants to inform the planting and / or treatment of other plants (e.g., crops) of the same plant type growing in the particular area.

[0142] As mentioned above, "stressor" herein refers to an abiotic and / or biotic stress that may negatively impact plant health, such as pests, diseases, water, heat, and / or nutrient stress or deficiency. For example, a plant may experience an insect stressor corresponding to the presence of insects or insect populations within the plant, which may hinder plant growth and / or health.

[0143] As described above, "stress" herein refers to the measurable and / or detectable presence of a particular stressor and / or a group of stressors in a plant (e.g., in a cluster of sentinel plants, in a crop of plants). For example, a computer system can detect an insect stressor at a cluster of sentinel plants and estimate insect stress (e.g., measurable presence, distribution, magnitude) at the cluster based on features extracted from images of the cluster of sentinel plants. Thus, stress represents the measurable presence of a particular stressor.

[0144] As described above, "pressure gradient" in this article refers to the distribution of pressure of a stressor (or multiple stressors) on multiple sentinel plants and / or multiple groups (clusters) of sentinel plants in a farmland. For example, a user may initially assign three groups of sentinel plants within a farmland. Later, a computer system may acquire an image of the farmland depicting the three groups of sentinel plants recorded by an aerial sensor (e.g., a satellite). Based on features extracted from an area of ​​the image depicting each group of sentinel plants, the computer system may interpret the pressure gradient of the stressor in the farmland. More specifically, the computer system may: interpret a first pressure of the stressor in the first group of sentinel plants based on features extracted from a first area of ​​the image depicting the first group of sentinel plants; interpret a second pressure of the stressor in the second group of sentinel plants based on features extracted from a second area of ​​the image depicting the second group of sentinel plants; interpret a third pressure of the stressor in the third group of sentinel plants based on features extracted from a third area of ​​the image depicting the third group of sentinel plants; and interpret the pressure gradient of the stressor in the farmland based on the first pressure, the second pressure, and the third pressure. Based on the pressure gradient, the computer system can account for the pressure of stressors at different locations within the field (eg, via interpolation).

[0145] As described above, a "user" in this article refers to a person associated with an agricultural environment that includes sentinel plants, such as a farmland, a plant crop, a greenhouse, a botanical garden, or a laboratory. For example, a user can refer to a farmer associated with a particular farmland. In another example, a user can refer to an agronomist associated with a particular plant crop. In another example, a user can refer to a scientist who studies or develops sentinel plants and / or stressor management in sentinel plants and non-sentinel plants.

[0146] 4. Promoter and reporter pair

[0147] A network of sentinel plants can be deployed to agricultural fields to communicate (e.g., visually, thermally, chemically) biotic and abiotic stressors in nearby crops to a person, such as a farmer, field operator, or agronomist. In particular, when exposed to these plant stressors and stressors, in the presence of certain plant stressors, the sentinel plants can experience, react, and deteriorate in the same or similar measures as comparable non-sentinel plants planted in the crops. Thus, the sentinel plants can serve as accurate sensors and predictors of diseases and / or stressors in these nearby crops. For example, sentinel plants can be deployed to agricultural fields and planted with other non-sentinel plants (such as in clusters of sentinel plants surrounded by non-sentinel plants) in order to detect, measure, and communicate certain stressors in these sentinel plants, which can then be interpolated or extrapolated to stressors in nearby non-sentinel plants.

[0148] In order to generate sentinel plants, plant cells can be genetically modified to combine known reporter genes with certain biological processes. Molecular genetic techniques can be implemented to associate the expression of reporter genes with certain biotic stresses and traits. Therefore, reporter genes can serve as signals for biotic stresses or traits in plant cells. For example, in the presence of a disease or stressor (and in proportion thereto), sentinel plants can be modified to fluoresce (i.e., absorb photons of one frequency and emit photons of a different frequency). In this example, sentinel plants can be modified to fluoresce in the presence of one or more diseases or stressors, such as fungi, bacteria, nematodes, parasites, viruses, insects, high temperature stress, water stress, nutrient stress, phytoplasmal disease, etc. In another example, sentinel plants can be modified to send a signal of the presence of a stressor via the bioluminescence of the sentinel plants. In yet another example, sentinel plants can be modified to send a signal of the presence of a stressor via changes in the pigments of the sentinel plants.

[0149] Plant cells can be genetically engineered to contain promoter and reporter pairs that indicate the presence of certain stressors in plants or plant crops. Promoters contain gene regulatory elements that drive the expression of mRNA at a specific time and location, which is subsequently translated into a functional protein. Promoter activity indicates the natural biological process that occurs when a specific stress is present in a plant. To detect the presence of these stressors, a known reporter gene that expresses a specific signal can be coupled to the selected promoter. Thus, when a plant cell expresses a promoter associated with a particular stressor, the reporter tagged to the promoter is also expressed and, therefore, detectable. Several fluorescent signals naturally exist in plants that have not been genetically modified. These signals can be enhanced through selective breeding and / or other plant selection techniques. Each of these reporter genes is capable of producing a light signal that is distinguishable from the plant itself. Combinations of reporter genes can also be used to indicate the presence of various plant stressors in a plant or crop.

[0150] A promoter and reporter pair can be implemented by tagging one reporter to one promoter. For example, if a red fluorescent protein is tagged to a promoter gene that indicates water stress in a sentinel plant, then when the water level in the plant cell is below a minimum water potential, the promoter gene and, therefore, the red fluorescent protein, can be expressed in the plant cell. A computer system (e.g., a computer network, a remote server) can: acquire an image of a field containing sentinel plants collected by various fixed or mobile, local or remote sensors (e.g., a fixed camera mounted on a pole in the field, a smartphone or tablet, a sensor mounted on a truck or 4x4, a sensor mounted on a drone or crop duster, a sensor mounted on a drone or airplane, a camera integrated into a satellite); extract from the image the intensity of a target wavelength of red fluorescence produced by the reporter protein in the sentinel plant in the presence of a water stressor; and estimate the magnitude of the water stressor in the plant based on the intensity of the target wavelength of red fluorescence in the image. In response to the estimated pressure of the water stressor exceeding a threshold water pressure, the computer system can alert the field operator to address an irrigation problem (e.g., insufficient irrigation) in the area of ​​the field occupied by the sentinel plant. In this example, the computer system can: repeat this process to extract intensities of the target wavelength of red fluorescence from other areas of the same image or other concurrent images depicting other individual or clustered instances of the sentinel plant planted in other areas of the field; estimate the pressure of the water stressor in these other areas of the field based on the intensities of the target wavelength of red fluorescence extracted from other areas of the image and / or from other concurrent images of the field; and interpolate or extrapolate a water pressure gradient across the field based on the locations and pressures of the water stressors indicated in the sentinel plants distributed across the field. Thus, the computer system can notify the field operator to address the irrigation problem for the entire field or a target area of ​​the field based on the water pressure gradient. In addition, the computer system can: repeat the process over time to estimate water pressure or water pressure gradients in a region or an entire field; extrapolate future water pressures in the field based on the region-specific or field-wide water pressures thus derived from continuous images of sentinel plants occupying the field; and then prompt field operators to proactively address predicted future water pressure changes in the field before the changes in water pressure (substantially) affect crop yields in the field.

[0151] In one variation, multiple promoters can be tagged to a reporter so that sentinel plants output signals for a specific stressor over an extended duration. For example, a group of three promoters associated with water stress can be tagged with a red fluorescent protein reporter. At the initial time, the presence of the first promoter can trigger the expression of red fluorescent protein in response to a specific water pressure. At the second time, as the signal generated by the first promoter decreases, the presence of the second promoter can trigger the continued expression of red fluorescence. And once again, at the third time, the third promoter can trigger the expression of red fluorescence in the plant. Therefore, genetic engineering techniques can be implemented to string together multiple promoter genes and mark this string of promoters with a reporter gene to identify which promoter genes are expressed in the plant, thereby extending the detection window.

[0152] In one embodiment, the sentinel plant can be configured to include a first number of promoters and a second number of reporters that are less than the first number of promoters. For example, expression of red fluorescent protein can signal the presence of a certain water pressure, while expression of yellow fluorescent protein can signal the presence of a certain heat pressure. However, expression of both red fluorescent protein and yellow fluorescent protein can signal the presence of both a certain water pressure and a certain heat pressure, or a third pressure (such as a certain insect pressure). Therefore, the fluorescence of the sentinel plant can be combined with knowledge of disease frequency, common disease location, and common disease time to isolate specific plant stressors present in farmland. In another example, the first, second, and third fluorescent compounds are each coupled to the first, second, and third biological processes, respectively. In addition, the fourth biological process is coupled to the first and second fluorescent compounds; the fifth biological process is coupled to the second and third fluorescent compounds, the sixth biological process is coupled to the first and third fluorescent compounds; and the seventh biological process is coupled to the first, second, and third fluorescent compounds. In this example, detecting all three fluorescent compounds in a plant could signal each of the following: activation of the sixth biological process; activation of the first, second, and third biological processes; activation of the first and fifth biological processes; activation of the fourth and third biological processes; and activation of the sixth and second biological processes. These biological processes can be distinguished, enabling detection of different processes occurring in these plant cells and, therefore, different stressors present in the plant. For example, a computer system could prompt a crop manager to address all possible diseases, or a specific disease that could be catastrophic if not addressed quickly. In another example, a farmer or agronomist could take samples from the plant and test them for each possible disease to initiate the appropriate course of action.

[0153] Similarly, plant cells can be genetically engineered to include a combination of reporters that present different signals in response to different stressors and / or pressures. A computer system can then use a model to interpret these signals, including information that is more than the sum of the parts of the set of reporters, such as: the presence of a fungus in addition to fungal stress; or the ratio of water stress to heat stress.

[0154] 4.1 Sentinel Plants

[0155] like Figure 5 As shown, the sentinel plant includes a first promoter-reporter pair, which includes: a first promoter, which is activated when a first stressor is present in the sentinel plant; and a first reporter, which is coupled to the first promoter and is configured to present a first signal in the electromagnetic spectrum in response to activation of the first promoter by the first stressor.

[0156] In one variation, Figure 5 As shown, the sentinel plant further includes a second promoter-reporter pair, which includes: a second promoter, which is activated when a second stressor is present in the sentinel plant; a second reporter, which is coupled to the second promoter and is configured to present a second signal in the electromagnetic spectrum in response to activation of the second promoter by the second stressor, and the second signal is different from the first signal.

[0157] In one variation, the sentinel plant further includes a third promoter that is activated in the presence of a third stressor in the sentinel plant, and the first reporter and the second reporter are both coupled to the third promoter and are configured to present a third signal in the electromagnetic spectrum in response to activation of the third promoter by the third stressor, the third signal being different from the first signal and the second signal.

[0158] One variation of the sentinel plant includes a first promoter-reporter pair comprising: a first promoter configured to be activated in the presence of a first stressor within a first amplitude range at the sentinel plant; and a first reporter coupled to the first promoter and configured to present a first signal in the electromagnetic spectrum in response to activation of the first promoter by the first stressor. In this variation, the sentinel plant also includes a second promoter-reporter pair comprising: a second promoter configured to be activated in the presence of the first stressor within a second amplitude greater than the first amplitude range at the sentinel plant; and a second reporter coupled to the second promoter and configured to present a second signal in the electromagnetic spectrum in response to activation of the second promoter by the second stressor.

[0159] Another variation of the sentinel plant includes: a first promoter that is activated at a first time and for a first duration in response to the presence of a first stressor in the sentinel plant; a second promoter that is activated at a second time and for a second duration in response to the presence of the first stressor in the sentinel plant, the second time being after the first time and before the end of the first duration; and a reporter coupled to the first promoter and the second promoter, which, in response to activation of the first promoter, expresses a first signal for detecting the first stressor for the first duration; and, in response to activation of the second promoter, expresses a second signal for detecting the first stressor for the second duration.

[0160] 5. Detection

[0161] The computer system can detect and interpret signals generated by sentinel plants by extracting features from images of sensor plants that are associated with the presence of a specific stressor at the sentinel plant.

[0162] In one embodiment, a computer system may acquire digital images (e.g., spectral images) of one or more sentinel plants and / or plant canopies (e.g., sentinel plants and surrounding plants) captured by optical sensors (e.g., multispectral or hyperspectral imaging devices) deployed at one or more sentinel plants and / or plant canopies. For example, Figure 7 As shown, the optical sensor may include: an optomechanical fore optic capable of measuring fluorescent and non-fluorescent targets; and a digital spectrometer or digital camera that records images through the optomechanical fore optic. Thus, a computer system may acquire images recorded by the optical sensor and process these images according to method S100 to detect reporter signals and interpret the stressors present in these plants. More specifically, in this example, the computer system may: acquire images (e.g., spectral images) of sentinel plants recorded by a digital spectrometer; extract wavelengths of compounds of interest from these images; and identify stressors present in the sentinel plants based on these wavelengths.

[0163] The computer system can acquire images of the sentinel plants captured by an optical sensor, such as from a handheld camera, a handheld spectrometer, a mobile phone, a satellite, or from any other device that includes a high-resolution spectrometer, includes filters for specific wavelengths, or is otherwise configured to detect wavelengths of electromagnetic radiation that fluoresce, luminesce, or are transmitted by the sentinel plants in the presence of a particular stressor.

[0164] The computer system can implement different instruments depending on the compound of interest, as different compounds are best observed at different wavelengths under different conditions and may require different detection modes. For example, the computer system can: acquire an image captured by a handheld spectrometer of a sentinel plant configured to emit red fluorescence in the presence of a stressor; and acquire an image captured by a handheld camera of a sentinel plant configured to exhibit a pigmentation change in the presence of a stressor.

[0165] The computer system can acquire images of sentinel plants collected at specific times of day and / or time intervals to maximize the detectability of signals generated by the sentinel plants. For example, for sentinel plants configured to produce a bioluminescent signal in the presence of one or more specific stressors, the computer system can acquire images of sentinel plants collected at night when other signals generated by the sentinel plants and their surroundings are minimized.

[0166] 5.1 Active / Passive Detection

[0167] The computer system can detect and interpret stressors in sentinel plants through active and / or passive detection modes. For example, the computer system can implement passive detection to detect signals generated by sentinel plants in the presence of one or more stressors without excitation of the sentinel plants. Alternatively, the computer system can implement active detection to detect signals generated by sentinel plants in response to excitation of the sentinel plants (e.g., via external illumination) in the presence of one or more stressors. More specifically, the computer system can implement a detection method in which sentinel plants are illuminated in oscillating light for excitation so that the response to the illumination can be isolated.

[0168] In one variation, a computer system detects solar-induced fluorescence signals generated by sentinel plants via narrow wavelength measurements near dark spectral features in incident solar radiation. Narrowband techniques associated with Fraunhofer lines (from absorption in the solar atmosphere) and Telluric lines (from absorption by molecules in the Earth's atmosphere) are able to measure light signals during the day without the need for external lighting. Implementation of this measurement technique allows for both specificity and accuracy in measuring small, fuzzy signals, as well as the ability to collect measurements both on the ground and in the air. Thus, it is possible to collect images of sentinel plants from a wide range of distances. A computer system can detect these solar-induced fluorescence signals and extract insights into the stress of the stressor at the sentinel plant that generated them. For example, Figure 5 and Figure 6As shown, a computer system can: obtain a first batch of spectral images captured by a first spectrometer; interpret a first stress of a stressor in a first group of sentinel plants based on a solar-induced fluorescence measurement extracted from a first image in the first batch of images; obtain a reporter model that associates the solar-induced fluorescence measurement extracted from the spectral images with the stress of the stressor in the sentinel plants; and interpret the first stress in the first group of sentinel plants based on the first solar-induced fluorescence measurement extracted from the first image.

[0169] 5.2 Single Plant Fixed Sensor

[0170] In one embodiment, a computer system can acquire data recorded by fixed sensors from a single sentinel plant in a field or greenhouse. For example, the computer system can acquire images collected by an optical sensor that is configured to be mounted (e.g., clamped) to a leaf or stem of a sentinel plant and capture close-up images of a fluorescent surface on the sentinel plant at a high frequency (e.g., once per minute, once per hour). In these examples, the computer system can upload the images to a remote database via a cellular network, or when the mobile device or vehicle is nearby, the images can be downloaded to the mobile device or vehicle via a local ad hoc wireless network and then uploaded to the remote database from the mobile device or vehicle.

[0171] 5.3 Fixed Cluster Sensor

[0172] In one embodiment, a computer system can acquire images of a group (e.g., a cluster) of sentinel plants collected by a fixed optical sensor facing the group of sentinel plants and mounted (e.g., mounted) in a farmland. For example, a computer system can acquire images of a cluster of sentinel plants in a farmland recorded by an optical sensor mounted on a boom or pole located at the center of the cluster of sentinel plants, the optical sensor capturing close-up images of fluorescent surfaces on the sentinel plants in the cluster of sentinel plants at a high frequency (e.g., once an hour, once a day). The computer system can extract insights from these close-up images of the cluster of sentinel plants to explain the stress of a particular one or more stressors in the cluster of sentinel plants. Furthermore, by interpreting the stress in the cluster of sentinel plants from images recorded by a fixed sensor located at the cluster, the computer system can extract insights into stress in a subregion of the farmland that includes the cluster, as well as in adjacent subregions.

[0173] 5.4 Handheld Sensors

[0174] In another embodiment, the farmer can manually collect data on the sentinel plants on a handheld device. For example, the computer system can obtain images of a first cluster of sentinel plants along the edge of the field collected by a mobile device (e.g., a smartphone) operated by a farmer associated with the field, which captures close-up images of the cluster of sentinel plants at a lower frequency (e.g., once a week, once every two weeks). Additionally or alternatively, the computer system can obtain close-up images of individual sentinel plants in the cluster of sentinel plants. In this implementation, the computer system can upload the images to a remote database via a cellular network, or automatically upload the images via a local or web-based agricultural application executed on the handheld device. The computer system can interpret the stress in the cluster of sentinel plants and / or individual sentinel plants directly from features extracted from these close-up images to generate a high-resolution, short-interval time series representation of the health of the cluster of sentinel plants and / or individual sentinel plants.

[0175] 5.5 Ground-based Mobile Imaging

[0176] Alternatively, the computer system can implement ground-based mobile imaging to extract insights into the health of sentinel plants and clusters of sentinel plants by collecting images from optical sensors mounted on manned or unmanned vehicles. For example, the computer system can acquire images of a cluster of sentinel plants collected by an optical sensor that is configured to be mounted (e.g., loaded) into the bed of a truck operated by a farmer associated with a field that includes the cluster of sentinel plants. In this example, the farmer can drive the truck along the edge of the field to capture images of the cluster of sentinel plants as the truck moves along the edge of the field. The computer system can then upload these images to a remote database, time-stamp and geo-reference them, and retrieve these images at the time of upload or at a later time.

[0177] 5.6 Aerial Imaging

[0178] In one embodiment, a computer system can acquire images of a cluster of sentinel plants, multiple clusters of sentinel plants, and / or a crop of sentinel plants recorded by an aerial sensor configured to capture images of the sentinel plants. For example, the computer system can acquire images of a crop of sentinel plants collected by an optical sensor configured to be mounted (e.g., mounted) on a drone operated by an agronomist associated with the crop. Alternatively, in a non-sentinel crop with multiple clusters of sentinel plants, a farmer can operate a drone or dispatch an autonomous drone to scan the crop area where the sentinel plant clusters are located to collect images of these sentinel plants.

[0179] In another embodiment, the computer system can acquire images of a cluster of sentinel plants, multiple clusters of sentinel plants, and / or crops of sentinel plants recorded by an aerial sensor (e.g., a long-duration, high-altitude UAV or a satellite such as OCO-2 or GOSAT) configured to capture remote images of sentinel plants. For example, the computer system can acquire images collected by a satellite sensor that is configured to collect remote images of sentinel plants at a low frequency (e.g., once a week, once every two weeks, once a month). In another example, the computer system can acquire images collected by a commercial satellite sensor that is configured to collect remote images of sentinel plants at a relatively high frequency (e.g., once a day, multiple times a week).

[0180] The computer system can implement any combination of these data collection methods (e.g., instrumentation, frequency, range) to collect high-quality data that enables rapid, targeted responses to certain plant stressors and, thereby, increases the yield of nearby non-sentinel plants in the same field. For example, the computer system can acquire high-resolution images recorded by a high-resolution optical sensor (e.g., an RGB camera, a multispectral camera or spectrometer, a thermal camera, or an IR camera) mounted on a pole located in the center of a first cluster of sentinel plants in the crop and configured to capture high-resolution images of the sentinel plants at a high frequency (e.g., three times a day) per day and upload these images to a remote database. The computer system can extract features (e.g., intensity at a specific wavelength) from these high-resolution images to explain the stress of the stressor at the first cluster of sentinel plants. In addition, the computer system can acquire low-resolution images recorded by a satellite sensor configured to capture low-resolution images of the entire crop, including multiple clusters of sentinel plants, at a low frequency (e.g., once every two weeks). The computer system can extract features (e.g., intensity at specific wavelengths) from these low-resolution images to explain the stress of the stressor at each sentinel plant in the crop. The computer system can derive a model that relates the stress of the stressor at the first sentinel plant to the stress of the stressor at other clusters in the crop based on the daily behavior of the first cluster and the biweekly behavior of all sentinel plants in the crop; and interpolate the behavior of the crop as a whole in areas with and without sentinel plants.

[0181] 6. Imaging frequency

[0182] The computer system can acquire images of sentinel plants captured at set intervals or specific times of the day to increase the likelihood of detecting signals and detect stressors of these stressors in sensitive plants and crops including sensitive plants at an early stage, before the stressors amplify in magnitude or negatively impact crop yield. For example, the computer system can acquire images of sensitive plants in a crop recorded by an optical sensor to monitor stressors that indicate plant health and, upon detecting these stressors (e.g., above a threshold stress), prompt a user associated with the crop (e.g., a farmer) to alleviate these stressors. Alternatively, a user manually monitoring the crop may not see or detect stressors in the crop until the stressors have significantly damaged the plants in the crop. Thus, the computer system can reduce the risk or likelihood of stressors spreading across all crops and across crops to other fields, and improve overall crop yield. In addition, the sentinel plants can be configured to output signals of relatively large magnitude (e.g., greater intensity) in response to stressors of relatively low magnitude. The sentinel plants can include a promoter that is configured to activate within a few hours of initial infection or defect development in the sensitive plant. The computer system can then detect the signal generated by the activation of the promoter in the sensitive plant. Based on early detection of signals, the computer system can recommend minimal treatment to alleviate stress in sensitive plants.

[0183] A computer system can periodically monitor a group of sentinel plants at a set frequency so that stressors in the sentinel plants are detected early while limiting costs and efforts for users (e.g., farmers, agronomists) associated with the farmland that includes the group of sentinel plants. For example, a computer system can: acquire a batch of images of a group of sentinel plants in a farmland recorded at a set frequency (e.g., twice a day, once a day, once a week); interpret stressors in the group of sentinel plants based on features extracted from a first image in the first batch of images; and, in response to pressure exceeding a threshold pressure, generate a prompt to a user associated with the farmland to address stressors in plants occupying the farmland near the group of sentinel plants. In this example, if the pressure drops below the threshold pressure, the computer system can continue to acquire images at the set frequency in the first batch of images to continue monitoring stressors in the group of sentinel plants. Additionally and / or alternatively, the computer system can generate a prompt to alert the user to stressors. Thus, the computer system enables a user to regularly monitor the health of a sentinel plant and / or multiple sentinel plants in a field associated with the user while minimizing the user's physical travel to the field including the sentinel plants, handling of the sentinel plants, and / or testing of the health of the sentinel plants.

[0184] In one embodiment, a computer system implements both high-frequency measurements and low-frequency measurements to more accurately interpret and predict stressors in sentinel plants and the field containing the sentinel plants. In such an embodiment, the computer system can combine a high-resolution, short-interval time series representation of the sentinel plant's health with features extracted from low-frequency, wider-field images of clusters of plants or the entire field containing the sentinel plant to predict the health of multiple plants or all plants in the field. For example, the computer system can acquire a first batch of images recorded at a first frequency (e.g., twice a day, once a day, once every two weeks) by a fixed sensor facing a first group of sentinel plants in the field. Additionally, the computer system can acquire a second batch of images of the field area including the first group of sentinel plants recorded at a second frequency (e.g., weekly, every two weeks) by a mobile sensor (e.g., deployed by a user associated with the field) at a second frequency less than the first frequency. Based on the images from these sources, the computer system can derive a model that relates features extracted from the images in the first batch of images to stressors in both the first group of sentinel plants and the field area. Thus, the computer system can predict pressure on the field area at a first frequency based on features extracted from images in the first batch.The computer system can periodically confirm and / or calibrate the model based on features extracted from images in the second batch at a second frequency.

[0185] 7. Mark sentinel plants

[0186] The computer system can extract features (e.g., intensity of specific wavelengths) from images of one or more sentinel plants, a cluster of one or more sentinel plants, and / or an agricultural field including sentinel plants to explain stressors in the sentinel plants. To extract the features, the computer system can distinguish between sentinel plants and non-sentinel plants in the images.

[0187] In one embodiment, a computer system can identify locations in a field that include sentinel plants and extract features from an image or image region corresponding to these locations. For example, the computer system can obtain a georeferenced image of multiple clusters of sentinel plants in a field recorded by a ground-based mobile sensor. The computer system can: obtain the location and orientation of the ground-based mobile sensor when the image was captured; obtain a set of GPS coordinates corresponding to the locations of the multiple clusters of sentinel plants in the field; and identify the multiple clusters of sentinel plants in the image based on the location and orientation of the ground-based mobile sensor and the GPS coordinates of the multiple clusters of sentinel plants.

[0188] In another embodiment, a computer system can identify sentinel plants in an image of sentinel plants and non-sentinel plants based on a baseline signal generated only by the sentinel plants. For example, the sentinel plants can be configured to generate a baseline signal within a first wavelength band, where the non-sentinel plants do not generate any signal. Additionally, the sentinel plants can be configured to generate a signal within a second wavelength band in response to stress from a stressor at the sentinel plant, the second wavelength band being different from the first wavelength band. Thus, the computer system can examine a sub-region of an image of a plurality of clusters of sentinel plants or a crop that includes sentinel plants for this baseline signal within the first wavelength band to identify a region of the image that includes the sentinel plants and / or the plurality of clusters of sentinel plants.

[0189] In another embodiment, a computer system can identify sentinel plants in an aerial image of a crop (e.g., sentinel plants and non-sentinel plants) by overlaying the image with a mask configured to hide non-sentinel plants and highlight sentinel plants. For example, the computer system can generate a mask for a field that includes five clusters of sentinel plants distributed throughout the field, the mask defining an opaque layer that includes five transparent areas corresponding to the five clusters. The computer system can then: overlay the mask on the crop image captured by the aerial sensor; apply null pixel values ​​to the crop areas covered by the opaque layer; and extract features (e.g., intensity measurements) from the five transparent areas corresponding to the five clusters of sentinel plants in the crop.

[0190] 7.1 Feature Extraction

[0191] like Figure 4 、 Figure 5 and Figure 6As shown, the computer system can extract features from these images of the sentinel plants to explain the stress in the sentinel plants. For example, the computer system can: obtain a first batch of images of a first group of sentinel plants in a field; and explain the first stress of the stressor in the first group of sentinel plants based on a first set of features extracted from the first image in the first batch of images. More specifically, the computer system can: extract a first feature in the first set of features from the first image, the first feature corresponding to a first pixel of the first image; extract a second feature in the set of features from the first image, the second feature corresponding to a second pixel of the first image; and estimate a representative feature based on a combination of the first feature and the second feature; obtain a reporter model that associates the features extracted from the images in the first batch to the stress of the first stressor at the first group of sentinel plants; and explain the first stress of the first stressor in the first group of sentinel plants based on the representative feature and the reporter model. Thus, based on features extracted from images collected by the optical sensor, the computer system can interpret the stress of a stressor at one or more sentinel plants based on a reporter sub-model that associates characteristics (e.g., intensity of wavelength) to a specific stressor (e.g., insects, high temperature, fungus) and / or stress of a specific stressor.

[0192] 8. Sentinel Plant Distribution

[0193] In one embodiment, each sentinel plant type of a particular crop is configured to generate a signal in response to a single plant stressor, i.e., a sentinel plant type includes a promoter-reporter pair configured to generate a signal in response to a single type of stressor. For example, a first sentinel plant type of a particular crop (e.g., corn) includes a promoter-reporter pair configured to generate a signal in response to fungal stress, and a second sentinel plant type of the particular crop includes a different promoter-reporter pair configured to generate a signal in response to insect stress.

[0194] In another embodiment, promoter-reporter pairs configured to output signals for multiple different stressors are integrated into a single sentinel plant type of a particular crop. For example, a single sentinel plant type of a particular crop contains promoter-reporter pairs configured to produce: a luminescent signal in response to fungal stress; a pigment change in response to insect stress; and a red fluorescent signal in response to phosphorus deficiency. Thus, a single plant or cluster of plants of this sentinel plant type can be sensed to detect multiple discrete stresses.

[0195] In one variation, when a field is planted, sentinel plants can be planted in clusters rather than mixed with seeds of non-sentinel plants. Specifically, rather than mixing seeds of sentinel plants for a particular stressor with non-sentinel seeds of the same or similar plant type prior to planting, these sentinel plant seeds can be planted in clusters in designated sentinel plant seed areas in the field, such as in specific crop rows (e.g., every 50 crop rows) or in target segments of crop rows (e.g., clusters of three rows wide and three meters long, with at least 20 crop rows or 20 meters between sentinel plants in adjacent clusters). Thus, by planting these sentinel plants in clusters adjacent to or surrounded by non-sentinel plants in the same field, the stress-related signals produced by these sentinel plants can exhibit high contrast with adjacent non-sentinel plants and therefore produce a high signal-to-noise ratio for the presence of a particular stressor in the field. For example, by planting multiple instances of sentinel plants in a small area of ​​a field, the red fluorescent reporter output by these sentinel plants can be more easily distinguished from the non-fluorescent background of adjacent non-sentinel plants. Similarly, if multiple sentinel plants are planted in a row in a field, the cluster of sentinel plants can produce a cumulative signal (indicating the presence of insect pressure as it shifts across the crop) characterized by a greater signal-to-noise ratio than a single sentinel plant in the row, and the cluster of sentinel plants can also produce more spatial information about the direction and extent of insect pressure moving across the field than a single sentinel plant in the row.

[0196] Clusters of sentinel plants can be planted in a field with non-sentinel plant crops, wherein the clusters of sentinel plants include at least one sentinel plant for each stressor, or wherein each sentinel plant includes a promoter for each plant stressor. For example, batches of sentinel plant seeds (including at least one seed containing a promoter for at least one stressor) can be planted in clusters in a field with other non-sentinel plants. In another embodiment, clusters of sentinel plant seeds are grouped by promoter. In this embodiment, a first cluster of water stress-sensitive seeds, a second cluster of fungal stress-sensitive seeds, and a third cluster of insect stress-sensitive seeds are planted in a field in discrete groups. In this embodiment, sentinel plant seeds containing the same reporter are planted together in clusters, and when the corresponding stress occurs in the field, the clusters can output stronger, higher amplitude, lower noise signals that are more easily recognized by fixed sensors, local mobile sensors, or remote sensors.

[0197] The locations of the sentinel plant clusters can also be selected so as to be able to detect certain plant stressors with greater accuracy and / or lower noise. In one example, an agronomist or farmer is present in person, such as via sensors mounted on a vehicle or via a handheld device, to collect stressor data from across the field, and multiple clusters of sentinel plants can be planted near the edges of the crop for quick access by the farmer. In this example, because the sentinel plant clusters are located near the edges of the crop, the farmer can collect samples from these sentinel plants and directly test these samples for plant stressors in order to verify the stress indicated by the reporters in these sentinel plant clusters. In another example, sentinel plants are planted in the center of the crop to increase proximity to each plant in the crop, thereby potentially increasing sensing capabilities or the likelihood of detecting diseases that are spreading across crops.

[0198] In yet another example, if one farmer's crops share a border with another farmer's crops, it might be desirable to plant a row of insect stress sentinel plants along the shared border to quickly detect migrating insect populations as they enter the crop. In another example, if there is a lower elevation portion of the crop, a cluster of water stress sentinel plants could be planted in that area to detect when that area is collecting excess water. Clusters could also be planted in the highest elevation portion of the crop, where plant dehydration might be prevalent.

[0199] In the embodiment described above, where sentinel plants are distributed in clusters throughout a field, sentinel plants can be identified and distinguished from non-sentinel plants to improve the efficiency of data collection. For example, if a farmer uses a handheld device to collect images of clusters weekly, markers can be placed in the field to easily locate the clusters. In another example, where satellite imagery is used to collect crop images, the coordinate locations of the clusters can be obtained to collect wavelength measurements of the sentinel plants.

[0200] In another embodiment, sentinel plant seeds are mixed with non-sentinel plant seeds and also planted together in clusters. Clusters of individual sentinel plant seeds can be evenly distributed throughout the crop or in optimized locations. Sentinel plant seeds can be mixed with non-sentinel plant seeds so that approximately 2% of the mixed seeds are sentinel plant seeds. Clusters of sentinel plants can be analyzed more frequently, such as by collecting aerial imagery by drones scanning clusters of sentinel plants daily. Satellites can collect images of the crop as a whole less frequently, collecting data on both clusters of sentinel plants and individual sentinel plants mixed with the rest of the crop. The health of the entire crop or field can be predicted by a computer system based on time-stamped and geo-referenced images of sentinel plants.

[0201] In one embodiment, sentinel plants can be transplanted as seedlings into a crop. For example, a sentinel strawberry plant can initially be transplanted as a seedling into a field of strawberry plants. In another implementation, a sentinel plant can be sown as seed into a crop. For example, a sentinel soybean plant can initially be sown as seed into a soybean crop. In another embodiment, a sentinel plant can be grafted onto an existing perennial crop. For example, a sentinel grape scion sensor can be grafted into a grape producing vine.

[0202] 8.1 Variant: Sterile Sentinel Plants

[0203] Sentinel plants can be genetically modified to be sterile or non-flowering. Sterile sentinel plants can be grown in either GMO or non-GMO crops because they do not reproduce. A small portion of a field can be planted with sterile sentinel plant seeds, while the remaining crop can be planted with standard non-sentinel plant seeds. For example, a farmer growing a corn crop might plant 2%-5% of the crop as genetically modified sterility-sensitive corn plants and the remaining 95%-98% as standard non-sterile corn plants. Before planting, the different sentinel plant types can be mixed together in appropriate ratios so that sterile seeds comprise approximately 2%-5% of the total seeds planted. As the crop grows, the sterile plants will be randomly distributed throughout the crop, producing a roughly even distribution of sterile plants throughout the crop. In this example, by having a single sentinel plant type containing all selected promoters, or by separating the promoters into different plant seeds, each plant stressor can be tested in every area of ​​the crop. In this embodiment, it may be advantageous to integrate all selected promoters into a single plant, allowing multiple plants with the same reporter to be in close proximity, thereby increasing the signal strength produced by the reporter.

[0204] The percentage of sterile sentinel plant seeds in a seed mixture can be manipulated to optimize crop yield. Sterile sentinel plant seeds result in a yield loss for farmers because sterile plants do not bear fruit. However, farms can use the data collected from the sentinel plants to improve the yield of the next crop. For example, a farmer may plant a corn crop with 100% non-sterile corn seeds ("normal" corn seeds) and expect an average crop yield of 88% over a ten-year period, assuming that an average of 12% of the crop may be lost or failed due to disease and other stresses over the long term. To increase yield over that time period, the farmer may plant a mixture of 5% sterile sentinel corn seeds and 95% non-sterile non-sentinel corn seeds in the field. Although there may be an initial 5% yield loss due to the application of sterile sentinel corn seeds, these sterile sentinel plants can enable early detection and response to various stresses (which previously caused the crop to lose an average of 10% yield over several years), allowing the farmer to reduce losses due to disease and other stresses to less than 1%, thereby increasing the overall average yield over many years to approximately 94%.

[0205] In one embodiment, sterile sentinel plant seeds replace some and / or all of the refuge seeds present in a seed mixture. For example, a seed mixture can be mixed to include a first percentage (e.g., 2% to 10%) of sterile sentinel refuge seeds and a second percentage of GMO seeds, the refuge seeds being configured to prevent pathogens and weeds from developing resistance to the GMO seeds. In this example, sterile sentinel plant seeds can be incorporated into the seed mixture as refuge seeds, thereby limiting any crop yield losses resulting from the implementation of sterile sentinel plants.

[0206] Similarly, the stressor signals emitted by these sterile sentinel plants can enable farmers to respond quickly, initially reducing average crop losses from 10% to 5%, allowing farmers to initially achieve the same average yield, but enabling the computer system to collect a relatively large amount of data from these deployed sterile sentinel plants. Over time, as the computer system collects additional stressor information from the fields based on the signals generated by the deployed sterile sentinel plants over multiple seasons, the computer system can recommend a smaller ratio of sterile to non-sterile plants while continuing to output preemptive prompts to address early stressors across the fields, thereby enabling farmers to reduce yield losses due to combining both sterile plants and stressors across the fields and, therefore, achieve a higher average yield for the crop over time. Thus, the computer system can indicate a target minimum ratio of sterile sentinel plant seeds to non-sterile non-sentinel seeds to be planted across the fields in order to achieve minimum stress sensing capability for long-term yield protection while minimizing immediate yield losses.

[0207] 8.2 Non-sterile sentinel plants

[0208] In one embodiment, the seeds of these sentinel plants are non-sterile. In this variation, the non-sterile sentinel plant seeds can also be planted in clusters according to the methods and techniques described above for sterile sentinel plant seeds, along with non-sentinel plants of the same fruit or similar crop type, to maintain high signal-to-noise ratio and sensing capabilities for such crops while limiting overall sowing costs (e.g., because sensing seeds cost more than non-sentinel seeds of the same fruit).

[0209] Alternatively, in this variation, the sensing trait can be integrated into the non-sterile GMO plant genome as part of a GMO stack already present in a GMO seed, which can then be planted to produce an entire crop of sentinel plants. However, in this variation, these non-sterile sentinel plant seeds can be configured to generate several different signals representing a set of stresses and can be planted in clusters across a field, as described above, where all plants in a cluster contain the same one or more promoter-reporter pairs configured to generate a signal for a specific biotic or abiotic stressor (or a specific set of biotic and / or abiotic stressors). For example, non-sterile sentinel plant seeds containing the same promoter-reporter pair are planted along the entire length of a crop row across a field, and non-sterile sentinel plant seeds in two adjacent crop rows contain different promoter-reporter pairs configured to generate signals for different biotic or abiotic stressors; in this example, this row pattern containing seeds with different promoter-reporter pairs is repeated along the entire length of the field. In another example, non-sterile sentinel plant seeds containing the same promoter-reporter pair are planted in linear clusters, such as in adjacent five-meter segments of five consecutive crop rows, with non-sterile sentinel plant seeds in adjacent clusters containing different promoter-reporter pairs configured to produce signals in response to different biotic or abiotic stressors; in this example, this grid surrounding multiple clusters of non-sterile sentinel plant seeds containing the same promoter-reporter pair is repeated along the entire length and width of the field.

[0210] Thus, by clustering non-sterile sentinel plants into one- or two-dimensional groups of plants configured to generate signals for the same stressor, the crop as a whole can generate high-amplitude signals characterized by a high signal-to-noise ratio for a variety of different biotic and / or abiotic stressors in discrete rows or discrete areas of the field. As described above, the stressors indicated by these rows or clusters of plants configured to generate signals for the same stressor can then be interpolated or extrapolated across the field to predict stress for the entire crop.

[0211] Thus, in this variation, because each plant in the field exhibits sensing capabilities and can directly monitor the entire crop, the computer system can generate a stress map of biotic and / or abiotic stressors for the crop as a whole based on the signals generated by these plants over a period of time (e.g., a day) and detected by fixed sensors or mobile local sensors or remote sensors. By repeating this process over time to develop new stress maps for the field, the computer system can monitor stressors across the field over time and provide data and / or recommendations to proactively mitigate these stressors. The computer system can also implement this process to update the stress map of the field after a stressor treatment is applied to the field, thereby enabling the field operator to directly evaluate the efficacy of the stressor treatment and make more informed treatment decisions for the field in the future. Furthermore, once a specific treatment is applied to the field based on these interpreted stresses, the computer system can continue to measure and detect the signals generated by the sentinel plants and thereby evaluate the efficacy of the specific treatment based on the new stresses interpreted from these signals.

[0212] 8.3 Plant Grafting

[0213] In one embodiment, rather than planting sentinel plants as seeds (such as row crops), sentinel plants can be grafted onto existing plants. Grafting may be useful for perennial crops and other high-value crops, such as almond trees or grapevines. The scion or leafy portion of the sentinel plant can be grafted onto a portion of the desired plant, such as the middle portion of the trunk. For example, a scion of a sentinel grapevine can be grafted into the trunk of a mature grapevine so that the scion portion of the mature grapevine can implement the sensing technology, providing an indication of the health of the mature grapevine. Because grafting sentinel plants into existing plants is initially a more time-consuming process, the grafting method may be useful for perennial crops that do not need to be replanted every year. These plants are pruned at the end of each season, but the sensing ability is still there when the leaves are in bloom the next season. Therefore, the graft only requires a single application to be maintained for the life of the plant.

[0214] The location of sensors in these perennial or high-value crops can also be optimized, similar to row crops. Multiple grafts can be applied to one plant to include each selected promoter and reporter in each grafted plant. Alternatively, specific reporter grafts can be selected for plants in different crop regions based on the likelihood that certain plant stressors will occur in different crop regions. When grafts are applied to growing plants, it may be beneficial to position the sensor near the edge of the crop for ease of application.

[0215] 8.4 Controlled Environment Agriculture Applications

[0216] In one variation, sentinel plants can be grown in a controlled environment such as a greenhouse (e.g., a glass roof or plant farm) or another enclosed growing structure. Sentinel plants grown in a controlled environment can be regularly monitored to detect stressors at the sentinel plants. In one embodiment, sentinel plants can be grown in an enclosed growing structure via vertical farming.

[0217] Sentinel plants grown in these controlled environments can be transplanted to other locations (e.g., commercial farmland) to serve as sentinel plants. Alternatively, sentinel plants grown in controlled environments can be monitored to detect stress from one or more stressors under specific controlled environmental conditions (e.g., climate, region, presence of other plants) within the controlled environment. A computer system can interpret stress in these sentinel plants in a greenhouse environment and, based on the stress in the sentinel plants, extract insights into plants under similar environmental conditions (e.g., in a farmland).

[0218] Because the greenhouse environment is smaller, the computer system can monitor sentinel plants in the controlled environment more frequently than sentinel plants located in a farmland. Consequently, the computer system can extract further insights into these sentinel plants grown in the controlled environment. For example, by accounting for the daily stress of a particular stressor in sentinel plants in the greenhouse, the computer system can more accurately converge on a model that relates features extracted from captured images of the sentinel plants to stress from the particular stressor. The computer system can then better model stress from the particular stressor in a farmland that includes sentinel plants of the same type and / or that includes these sentinel plants after they have been transplanted by a user associated with the farmland.

[0219] 9. Output

[0220] The computer system can: acquire images (e.g., spectra) of sentinel plants; extract features in these sentinel plants that are indicative of stressors and stress corresponding to these stressors; interpolate or extrapolate the stress of specific stressors in these sentinel plants to other plants (e.g., sensitive plants and non-sentinel plants) in the same field (and nearby fields); and then generate real-time prompts or treatment decisions for these crops to improve the efficiency of crop treatment and maintenance over time and maintain or increase the yield of the field.

[0221] In one embodiment, a computer system: extracts wavelength measurements of a particular compound in a region of an image depicting a cluster of sentinel plants; and converts these wavelength measurements into pressure (e.g., presence, magnitude) of one or more particular stressors in the cluster of sentinel plants. For example, if the computer system detects a particular wavelength of a compound associated with a fungal disease in the region of the image, the computer system can obtain a model that relates the wavelength of the compound of interest to fungal stressors, and then pass the intensity of the wavelength in the region of the image into the model to estimate fungal pressure (e.g., pressure as a percentage) in the cluster of sentinel plants. Based on the fungal pressure of a particular sentinel plant, the computer system can generate a prediction of fungal pressure in non-sentinel plants surrounding or near the cluster of sentinel plants.

[0222] In the aforementioned example, to generate a model that relates wavelength intensity to stress from a stressor, a farmer can collect samples from the leaves or soil of sentinel plants to detect plant stressors. The samples can be tested to determine the specific type of stressor and stress present on the leaves, while the wavelengths of compounds associated with the disease in the plant can be measured from the collected images. A model describing the relationship between the detection wavelength and stress amplitude for the compound of interest can then be generated based on this empirical data (e.g., by a computer system). The computer system can then automatically (and autonomously) predict stress for the entire crop based on features extracted from images of a cluster of sentinel plants, rather than based on physical samples collected by the farmer. Alternatively, the model can be generated based on laboratory data before the sentinel plants are deployed in the field, and can be associated with the deployed sentinel plants during the subsequent growing season.

[0223] In crops with multiple clusters of sentinel plants or with sentinel plants distributed throughout the crop, a computer system can acquire images collected both on the ground and in the air to output a stress map of the crop. The stress map can show the location of specific diseases and stressors and can be updated or combined to show the spread or elimination of specific stresses over time. The map can show interpolated stress data for areas of the crop without sentinel plants. In one embodiment, images can be collected multiple times daily from a camera on a pole located at the center of a cluster of sentinel plants. In addition, satellite images of the entire crop, including other clusters of sentinel plants, can be collected every two weeks. Data collected daily from a single cluster can be used to model the behavior of other clusters based on biweekly wavelength measurements of disease compounds in the remaining clusters. Crop areas between clusters, or "non-sentinel" areas, can also be modeled by interpolation (e.g., via a machine learning algorithm). To confirm the presence of a stressor and interpret the pressure of that stressor, farmers can collect samples of the sentinel plants themselves or the soil around the sentinel plants.

[0224] For example, a computer system can obtain a batch of images from a remote database, the first batch of images being time-stamped and geo-referenced and uploaded to the remote database via a wireless network at a frequency of one image per hour from a device on a pole located at the center of a first cluster of sentinel plants in a farmland; obtain satellite images of a farmland including a set of multiple clusters of sentinel plants, the satellite images being collected every two weeks; interpret the stress of the stressor in the first cluster based on a model that associates features extracted from the batch of images to stressors and stressor stress; interpolate the stress of a set of multiple clusters and all plants (e.g., sterile and non-sterile plants) in the farmland based on the model and the batch of images and satellite images from the remote database; generate a stress map including the following items: the location of the pressure in the farmland, the magnitude of the pressure, the location of the sentinel plant cluster, a first timestamp indicating the time when the map was generated, and a second timestamp indicating the time represented by the map; generate prompts or treatment recommendations for the farmland based on the pressure map; and, transmit the pressure map and the corresponding prompts or treatment recommendations to an operator of the farmland.

[0225] After generating a stress map based on the measured wavelengths of specific compounds in plants, a computer system can prompt a farmer in the field to take certain actions to combat plant stressors. In one embodiment, a farmer can plant a row of insect sentinel plant seeds at the edge of a soybean field to monitor the boundary between the farmer's crop and adjacent crops. Each day, an optical device mounted on a pole in the row of sentinel plants can capture images of the sentinel plants. Based on these images, the computer system can measure the wavelengths of compounds associated with insect-related diseases and display a specific insect stress amplitude at the edge of the map corresponding to the row of sentinel plants. Based on the insect pressure amplitude and the time the images were collected, the computer system can display a predicted current insect pressure amplitude for the area surrounding the crop and prompt the farmer to make certain decisions, such as whether to treat the crop with insecticide for insect infestation based on the pressure amplitude readings; which areas of the crop need to be treated for insect infestation; and the degree of treatment to be applied to different areas of the crop. After the initial processing, as more images are collected and more data becomes available, the computer system can update the stress map and prompt the farmer to implement an updated treatment plan using this new information and make improved treatment decisions for future insect-related diseases. The output stress map provides a way for farmers to be alerted to crop diseases or stresses as they occur and to obtain predictions of what might happen in response to certain treatments or in response to no treatment being applied. Over time, as more data is collected and various treatments are applied to crops based on stressors indicated by signals output by sentinel plants across the field, the computer system can develop models to predict how plants and plant stressors will respond to certain treatments, such as how the magnitude of the signal output by sentinel plants for a known stressor will change in response to a treatment of a particular magnitude applied to the field.

[0226] The computer system can generate real-time prompts or treatment decisions for these crops in order to improve the efficiency of crop treatment and maintenance over time and maintain or increase the yield of the farmland. For example, in response to interpreting the stress of a particular stressor in a group of sentinel plants as being above a threshold stress, the computer system can generate a prompt to address the particular stressor in plants near the group of sentinel plants. More specifically, the computer system can: isolate a first action associated with the particular stressor in a set of actions defined for the sentinel plants; and send a notification to a computing device of a user associated with the farmland to perform the first action in the farmland to alleviate the particular stressor. Thus, the computer system can update the user (e.g., an agronomist, farmer, field owner) with information about plant health and / or recommend treatments to alleviate the stress of the stressor in the plants.

[0227] 9.1 Pressure Model

[0228] In one variation, Figure 1 As shown, a computer system can derive a stress model that relates stress from a particular stressor at a first group of sentinel plants (e.g., a single sentinel plant, a cluster of sentinel plants) to stress from the particular stressor at a second group of sentinel plants. By developing the stress model, the computer system can minimize data collection from all sentinel plants in a particular area (e.g., a field) by relating stress in sentinel plants within a single group of sentinel plants to other groups of sentinel plants in the field.

[0229] For example, a computer system may: acquire a first set of images recorded at a first frequency by a fixed sensor (e.g., a camera mounted on a beam in the center of a field) facing a first group of sentinel plants in the field; acquire a second set of images of a second group of sentinel plants in the field, the second images recorded by a mobile sensor (e.g., a camera of a user's mobile device associated with the field) during a first time period; interpret a first stress of a stressor in the first group of sentinel plants during the first time period based on a first set of features extracted from a first image in the first set of images captured during the first time period; and interpret a second stress of a stressor in the second group of sentinel plants during the first time period based on a second set of features extracted from the second image. Based on the first stress interpreted at the first group of sentinel plants and the second stress interpreted at the second group of sentinel plants, the computer system may derive a stress model that relates the stress of the stressor at the first group of sentinel plants to the stress of the stressor at the second group of sentinel plants.

[0230] Once the computer system derives the stress model, the computer system can continue to acquire images from the first group to interpret the stress at the first group of sentinel plants and the second group of sentinel plants based on the model. For example, during the second time period, the computer system can: interpret a third stress of the stressor in the first group of sentinel plants based on a third set of features extracted from a third image captured during the second time period from the first group of images; and predict a fourth stress of the stressor in the second group of sentinel plants during the second time period based on the third stress and the model. Thus, the computer system can predict the stress at the second group of sentinel plants based on the images of the first group of sentinel plants from the first group without acquiring additional images of the second group of sentinel plants. Optionally, the computer system can continue to collect images of the second group of sentinel plants at a second frequency that is less than the first frequency to ensure the accuracy of the stress model and update the stress model over time. In addition, the computer system can collect images of other groups of sentinel plants and develop additional stress models that associate stress in sentinel plants in these other groups of sentinel plants on a particular area with the first group of sentinel plants in the field, thereby being able to predict stress from a particular stressor in the group of sentinel plants on the field based on information extracted from the images of the first group of sentinel plants.

[0231] Based on the predicted fourth stress at the second group of sentinel plants, the computer system can generate a prompt or send a notification to a user associated with the agricultural field. For example, in response to the fourth stress in the second group of sentinel plants exceeding a threshold stress, the computer system can generate a prompt to address a stressor in plants near the second group of sentinel plants in the agricultural field.

[0232] 9.2 Gradient Model

[0233] In one variation, Figure 2 and Figure 3 As shown, a computer system can derive a gradient model that relates the pressure of a particular stressor at a first group of sentinel plants (e.g., a single sentinel plant, a cluster of sentinel plants) to the pressure at a subregion of the field that includes the first group of sentinel plants (e.g., a pressure gradient in the field). By developing the gradient model, the computer system can minimize data collection for all sentinel plants in a particular region (e.g., a field) by relating the pressure gradient in the particular region (e.g., the pressure in sentinel plants across the particular region) to a single group of sentinel plants in the field. Furthermore, the computer system can correct for bias in the interpreted pressure at the first group of sentinel plants based on the gradient model.

[0234] For example, a computer system may: acquire a first set of images recorded at a first frequency by a fixed sensor (e.g., a camera mounted on a pole in a farmland) that faces a first group of sentinel plants in the farmland; acquire a second set of images of an area of ​​the farmland including the first group of sentinel plants, the second images being recorded by a mobile sensor (e.g., an aerial sensor, a drone, a satellite) during a first time period; interpret a first pressure of a stressor in the first group of sentinel plants during the first time period based on a first set of features extracted from first images captured during the first time period in the first set of images; interpret a first pressure gradient of the stressor in the sentinel plants in the area of ​​the farmland during the first time period based on a second set of features extracted from the second images; and derive a gradient model based on the first pressure of the stressor and the first pressure gradient, the gradient model correlating the pressure of the stressor at the first group of sentinel plants with the pressure gradient of the stressor in the area of ​​the farmland.

[0235] After deriving the gradient model, the computer system may correct the first pressure gradient based on the first pressure of the stressor at the first group of sentinel plants and the gradient model. Furthermore, the computer system may predict the pressure gradient of a particular stressor based on features extracted from the images in the first group. For example, the computer system may interpret a second pressure of the stressor in the first group of sentinel plants during the second time period based on a third set of features extracted from a third image captured during the second time period from the first group of images; and predict a second pressure gradient of the stressor in a region of the field during the second time period based on the second pressure and the model.

[0236] Based on this pressure gradient, the computer system can monitor the pressure at different sub-regions of the farmland. If the computer system predicts high pressure for a particular stressor at a particular sub-region of the farmland, the computer system can mark the sub-region and generate a prompt to a user associated with the farmland to address the particular stressor in the sub-region. For example, in response to a second pressure gradient predicting a third pressure in a sub-region of the farmland and the third pressure exceeding a threshold pressure, the computer system can generate a prompt to address the stressor in plants occupying the farmland near the sub-region of the farmland. In addition, based on the pressure gradient, the computer system can generate a pressure map. The computer system can include this pressure map in the prompt to the user.

[0237] In addition, the computer system can improve the gradient model by accounting for pressure from an additional group of sentinel plants in the field. In one embodiment, the entire field is a sentinel plant (e.g., there are no non-sentinel plants). In this embodiment, the computer system interprets the first pressure gradient based on features extracted from a second image recorded by the mobile sensor. The computer system can combine the low-resolution pressure gradient data for the entire field of sentinel plants with the high-resolution pressure data for the first group of sentinel plants to develop a more accurate gradient model for predicting the pressure gradient of the entire field.

[0238] In another embodiment, in which multiple clusters of sentinel plants are planted in a field of non-sentinel plants, the computer system may interpret a first pressure gradient based on features extracted from an area of ​​a second image recorded by a mobile sensor, an area including the first group of sentinel plants and (at least) the second group of sentinel plants. In this embodiment, the computer system may interpret the pressure of the specific stressor at the first group of sentinel plants based on the first image, and interpret the second pressure of the specific stressor at the first group of sentinel plants based on the second image. The computer system may then: derive a gradient model that relates the pressure of the specific stressor at the first group of sentinel plants to the pressure gradient of the first stressor in the field based on both the second pressure extracted from the second image and the first pressure gradient; and correct the first pressure gradient of the specific stressor in the field based on the first pressure and the model.

[0239] 9.3 Annual Model

[0240] The computer system can use data corresponding to a specific farmland or crop to develop an annual model for modeling the pressure of stressors in a specific farmland. For example, during the first season, for a specific crop, the computer system can extract insights into: the movement of water on the specific crop; the amount of sunlight across the crop (e.g., daily, weekly, monthly, seasonal); and the timing of pressure from other stressors (such as insects, fungi, and nutrient deficiencies). The computer system can input each of these insights into an annual model to predict the crop conditions at the beginning of the next season and throughout the next season. Then, at the beginning of the next season, the computer system can predict the initial conditions of the crop based on the model. In addition, the computer system can recommend farming practices to users associated with the crop based on these predicted initial conditions, such as the type of seed mixture to be planted and / or different mixtures of soil to be laid. As the season continues, the system can update the annual model accordingly.

[0241] Furthermore, based on the annual model, the computer system can predict and / or recommend agricultural products and / or treatments that are best suited for the farmland. For example, the computer system can predict a first stressor pressure on plants in the farmland at a specific time based on the annual model. Based on the predicted first stressor, the user can apply a new treatment to the plants at the beginning of the season to alleviate the predicted first stressor. Later, based on data recorded by sensors in the farmland, the computer system can interpret a second stressor pressure on the plants in the farmland at the specific time. If the second stressor pressure is less than the predicted first stressor pressure, the computer system can update the annual model accordingly and / or recommend new treatments in the future to address the stressor pressure.

[0242] 10. Single Sentinel Plant

[0243] In one variation, a computer system can extract insights from a single sentinel plant (e.g., in a non-sentinel plant crop in a greenhouse) to: monitor stressors in plants in a field; develop models for predicting plant behavior over time; develop models for predicting plant responses to various stressors present at sentinel plants; develop models for interpreting stressors at sentinel plants based on measurements; test the effects of treatments for various stressors present at a single sentinel plant; and / or develop models of plant responses to those treatments.

[0244] For example, a single sentinel plant or a single cluster of sentinel plants can be grown in a non-sentinel plant crop. The single sentinel plant (or single cluster of sentinel plants) can be monitored for the presence of a stressor at the sentinel plant. For example, a computer system can acquire data (e.g., images) recorded by a sensor (e.g., a smartphone) and interpret a first stressor for a specific stressor at the sentinel plant based on features extracted from the data. Based on the first stressor interpreted at the single sentinel plant, the computer system can extract insights into plants near the single sentinel plant and / or plants within the non-sentinel plant crop. In addition, the computer system can recommend a specific treatment to the plants in the crop based on the interpreted first stressor. When the user applies the specific treatment, the computer system can interpret a second stressor to confirm the efficacy of the specific treatment.

[0245] In another example, sentinel plants can be grown in a greenhouse. A computer system can acquire data recorded by optical sensors in the greenhouse (e.g., hyperspectral images) to extract a first set of measurements (e.g., intensity of wavelengths) that indicate plant health. A user (e.g., a user associated with the greenhouse) can collect samples from the sentinel plants to confirm the health of the sentinel plants and / or the presence of any stressors at the sentinel plants. In this example, if the user interprets the sentinel plants as healthy and, based on the collected samples, interprets that there is no stress from a particular stressor at the sentinel plants, the computer system can associate the first set of measurements to healthy plants that do not exhibit stress from the particular stressor and store that information in a model. Subsequently, the user can subject the sentinel plants to stress from a particular stressor (e.g., drought). The computer system can again acquire data recorded by the optical sensors in the greenhouse to extract a second set of measurements (e.g., intensity of wavelengths) corresponding to the sentinel plants. The computer system can then associate the second set of measurements of the sentinel plants with the stress from the particular stressor introduced by the user at the sentinel plants and store that information in a model. Thus, over time, the computer system can develop models that relate measurements extracted from data recorded by optical sensors in the greenhouse to the stress of specific stressors at the sentinel plants.

[0246] In yet another example, a computer system can extract insights into the efficacy of plant treatments over time. For example, sentinel plants can be grown in a greenhouse where plants are arranged in vertical stacks (e.g., via vertical farming). The computer system can extract measurements from data (e.g., images) recorded by sensors in the greenhouse to extract insights into the health of the plants. The computer system can interpret a first pressure of a particular stressor at the sentinel plant based on a first set of measurements extracted from data recorded by the sensor at a first time. The computer system can then notify a user associated with the greenhouse of the first pressure. The user can then apply a particular treatment to plants in the greenhouse near the sentinel plant to alleviate the first pressure. Subsequently, the computer system can interpret a second pressure of the particular stressor at the sentinel plant based on a second set of measurements extracted from data recorded by the sensor at a second time (e.g., 24 hours after applying the particular treatment). Based on the first pressure and the second pressure, the computer system can derive a model representing the pressure of the particular stressor over time in response to the application of the particular treatment. Thus, the computer system can derive a model for predicting the response of plants to various treatments and / or agricultural techniques.

[0247] The computer systems and methods described herein may be at least partially embodied and / or implemented as a machine configured to receive a computer-readable medium storing computer-readable instructions. Instructions may be performed by a computer-executable component integrated with an application, applet, host, server, network, website, communication service, communication interface, hardware / firmware / software elements of a user's computer or mobile device, a wristband, a smart phone, or any suitable combination thereof. Other computer systems and methods of embodiments may be at least partially embodied and / or implemented as a machine configured to receive a computer-readable medium storing computer-readable instructions. Instructions may be performed by a computer-executable component integrated with a computer-executable component integrated with the above-mentioned type of device and network. Computer-readable media may be stored on any suitable computer-readable medium, such as RAM, ROM, flash memory, EEPROM, optical device (CD or DVD), hard drive, floppy disk drive, or any suitable device. The computer-executable component may be a processor, but any suitable dedicated hardware device may (alternatively or additionally) execute instructions.

[0248] As those skilled in the art will recognize from the previous detailed description and from the accompanying drawings and claims, modifications and changes may be made to the embodiments of the invention without departing from the scope of the invention as defined in the appended claims.

Claims

1. A method for interpreting stress in a plant, comprising: Acquire a batch of images of the first set of sentinel plants in agricultural environments; interpreting a first pressure of a stressor in the first set of sentinel plants during a first time period based on a first set of features extracted from a first image in the batch of images captured during the first time period; acquiring a second image of a second set of sentinel plants in the agricultural environment, the second image being recorded during the first time period; interpreting a second pressure of the stressor in the second set of sentinel plants during the first time period based on a second set of features extracted from the second image; deriving a stress model based on the first stress and the second stress, the stress model relating the stress of the stressor at the first group of sentinel plants to the stress of the stressor at the second group of sentinel plants; interpreting a third pressure of the stressor in the first set of sentinel plants during a second time period based on a third set of features extracted from a third image in the batch of images captured during a second time period; and predicting a fourth stress of the stressor in the second group of sentinel plants during the second time period based on the third stress and the stress model; Therein, the stress is the measurable and / or detectable presence of a specific stressor and / or a group of stressors in the plant.

2. The method according to claim 1, further comprising: In response to the fourth stress in the second group of sentinel plants exceeding a threshold stress, a prompt is generated to address the stressor in plants proximate to the second group of sentinel plants in the agricultural environment.

3. The method according to claim 2, in, Generating prompts for addressing the stressor in plants adjacent to the second set of sentinel plants includes: Isolating a first action from a set of actions defined for a sentinel plant, the first action being associated with the stressor; and generating a prompt to perform the first action to address the stressor in plants adjacent to the second set of sentinel plants; and The method also includes sending the prompt to a computing device accessed by a user associated with the agricultural environment.

4. The method according to claim 1, in, Acquiring the batch of images of the first group of sentinel plants in the agricultural environment includes: acquiring the batch of images of the first group of sentinel plants in a farmland; and Wherein, acquiring the second image of the second group of sentinel plants in the agricultural environment includes: acquiring the second image of the second group of sentinel plants in the farmland.

5. The method according to claim 1, in, Acquiring the batch of images of the first set of sentinel plants includes: acquiring the batch of images recorded by a ground-based sensor facing the first set of sentinel plants; and The method further comprises: acquiring a fourth image of the agricultural environment recorded by the aerial sensor during the first time period; interpreting a first pressure gradient in the agricultural environment during the first time period based on a fourth set of features extracted from a region of the fourth image, the region of the fourth image depicting the first group of sentinel plants, the second group of sentinel plants, and a third group of sentinel plants in the agricultural environment; and correcting the first pressure gradient in the agricultural environment during the first time period based on the first pressure and the second pressure; The first pressure gradient is a distribution of pressure of the stressor on the first group of sentinel plants, the second group of sentinel plants and the third group of sentinel plants in the agricultural environment.

6. The method according to claim 1, in, Acquiring the batch of images of the first group of sentinel plants comprises: acquiring the batch of images recorded at a target frequency by a fixed sensor facing the first group of sentinel plants; and Wherein, acquiring the second image of the second group of sentinel plants recorded during the first time period includes: acquiring the second image of the second group of sentinel plants recorded by a mobile sensor during the first time period.

7. The method according to claim 6, further comprising: acquiring a fourth image of a third group of sentinel plants in the agricultural environment, the fourth image recorded by the mobile sensor during the first time period; interpreting the initial pressure of the stressor in the third set of sentinel plants during the first time period based on a fourth set of features extracted from the fourth image; deriving a gradient model based on the first pressure, the second pressure, and the initial pressure, the gradient model relating the pressure of the stressor at the first group of sentinel plants, the pressure of the stressor at the second group of sentinel plants, and the pressure of the stressor at the third group of sentinel plants; and interpreting a second pressure gradient in the agricultural environment during the second time period based on the third pressure in the first set of sentinel plants and the gradient model; The second pressure gradient is the distribution of the pressure of the stressor on the first group of sentinel plants, the second group of sentinel plants and the third group of sentinel plants in the agricultural environment.

8. The method according to claim 6: in, Acquiring the batch of images recorded by the fixed sensor includes: acquiring the batch of images recorded by a camera mounted on a fixed beam, the fixed beam being centered within the agricultural environment at the first set of sentinel plants; and Wherein, acquiring the second image recorded by the mobile sensor includes acquiring the second image recorded by a user associated with the agricultural environment via a camera integrated into a mobile device.

9. The method according to claim 1: in, Acquiring the batch of images of the first set of sentinel plants comprises: acquiring the batch of images of the first set of sentinel plants including a first cluster of sentinel plants arranged near a center of the agricultural environment; and Wherein, acquiring the second image of the second group of sentinel plants includes: acquiring the second image of the second group of sentinel plants including a second cluster of sentinel plants arranged along an edge of the agricultural environment.

10. The method according to claim 1, wherein Acquiring the batch of images of the first group of sentinel plants includes acquiring the batch of images of the first group of sentinel plants comprising a first group of promoter-reporter pairs configured to signal stress of a first group of stressors at the sentinel plants, the first group of promoter-reporter pairs comprising a first promoter-reporter pair configured to signal stress of the stressor in the first group of stressors at the first group of sentinel plants.

11. A method for interpreting stress in a plant, comprising: acquiring a batch of images recorded by a fixed sensor facing a first set of sentinel plants in an agricultural environment; interpreting a first pressure of a stressor in the first set of sentinel plants during a first time period based on a first set of features extracted from a first image in the batch of images captured during the first time period; acquiring a second image of an area of ​​the agricultural environment including the first set of sentinel plants, the second image recorded by a mobile sensor during the first time period; interpreting a first pressure gradient of the stressor in the agricultural environment during the first time period based on a second set of features extracted from the second image; deriving a model based on the first pressure of the stressor and the first pressure gradient, the model relating the pressure of the stressor at the first group of sentinel plants to the pressure gradient of the stressor in the region of the agricultural environment; interpreting a second stress of the stressor in the first set of sentinel plants during the second time period based on a third set of features extracted from a third image in the batch of images captured during the second time period; and predicting a second pressure gradient of the stressor in the area of ​​the agricultural environment during the second time period based on the second pressure and the model; wherein said stress is the measurable and / or detectable presence of a specific stressor and / or a group of stressors in the plant; and The stress gradient is the distribution of stress of one or more stressors on a plurality of sentinel plants and / or a plurality of groups of sentinel plants in the agricultural environment.

12. The method according to claim 11, further comprising: In response to the second pressure gradient predicting that a third pressure in the sub-area of ​​the agricultural environment exceeds a threshold pressure, a prompt is generated to address the stressor in plants proximate the sub-area of ​​the agricultural environment.

13. The method according to claim 11: in, Acquiring the second image of the area of ​​the agricultural environment including the first group of sentinel plants comprises: acquiring the second image of the area of ​​the agricultural environment including the first group of sentinel plants and the second group of sentinel plants; and Wherein, interpreting the first pressure gradient based on the second set of features extracted from the area of ​​the second image including the first group of sentinel plants includes: interpreting the first pressure gradient based on the second set of features extracted from the area of ​​the second image including the first group of sentinel plants and the second group of sentinel plants.

14. The method according to claim 11: in, Acquiring the second image of the agricultural environment comprises: acquiring the second image of the sentinel plant population, the second image comprising a set of pixels, each pixel in the set of pixels comprising a set of sentinel plants in the sentinel plant population; and Wherein, interpreting the first pressure gradient of the stressor in the sentinel plant in the agricultural environment based on the second set of features extracted from the second image comprises: For each pixel in the set of pixels: extracting a subset of features from the second set of features; and interpreting stress from a set of stressors in a corresponding set of sentinel plants in the sentinel plant population based on a subset of the features; and A first pressure gradient of the stressor in the sentinel plant population is generated based on the set of pressures.

15. The method according to claim 11: in, Acquiring the batch of images recorded by the fixed sensor includes acquiring the batch of images of the first group of sentinel plants in the agricultural environment, the sentinel plants in the first group of sentinel plants: comprising a first promoter associated with plant dehydration; and comprising a first reporter associated with red fluorescence, and wherein the first reporter is configured to signal dehydration of the plant; wherein interpreting the first stress of the stressor in the first group of sentinel plants based on the first set of features extracted from the first image comprises: interpreting a first stress of plant dehydration in the first group of sentinel plants based on a first set of red fluorescence measurements extracted from the first image; wherein interpreting the first pressure gradient of the stressor in the sentinel plants in the region of the agricultural environment based on the second set of features extracted from the second image comprises: interpreting a first pressure gradient of plant dehydration in the sentinel plants in the region of the agricultural environment based on a second set of red fluorescence measurements extracted from the second image; wherein interpreting the second stress of the stressor in the first group of sentinel plants based on the third set of features extracted from the third image comprises: interpreting the second stress of plant dehydration in the first group of sentinel plants based on a third set of red fluorescence measurements extracted from the third image; and Wherein, predicting the second pressure gradient of the stressor in the area of ​​the agricultural environment based on the second pressure and the model includes predicting a second pressure gradient of plant dehydration in the area of ​​the agricultural environment based on the second pressure and the model.

16. The method according to claim 11: in, Acquiring the batch of images includes acquiring a batch of spectral images captured by a first spectrometer; The method further includes: obtaining a reporter model that relates solar-induced fluorescence measurements extracted from the spectral image to stress of the stressor of the sentinel plant; and Wherein, interpreting the first stress in the first group of sentinel plants includes interpreting the first stress of the stressor based on first solar-induced fluorescence measurements extracted from the first image.

17. A method for interpreting stress in a plant, comprising: acquiring a first set of images of a first set of sentinel plants in a field, the first set of images being recorded at a first frequency; interpreting a first stress of a first stressor from a set of stressors in the first set of sentinel plants during a first time period based on a first set of features extracted from a first image in the first set of images captured during a first time period; acquiring a second image of the farmland recorded during the first time period; interpreting a second stress of the first stressor in the first group of sentinel plants during the first time period based on a second set of features extracted from a region of a second image including the first group of sentinel plants; interpreting a first pressure gradient of the first stressor in the agricultural field during the first time period based on a third set of features extracted from the second image; deriving a model that relates the pressure of the first stressor at the first group of sentinel plants to the pressure gradient of the first stressor in the agricultural field based on the second pressure and the first pressure gradient; and correcting the first pressure gradient of the first stressor in the agricultural field during the first time period based on the first pressure and the model; wherein said stress is the measurable and / or detectable presence of a specific stressor and / or a group of stressors in the plant; and The pressure gradient is the distribution of pressure of one or more stressors on multiple sentinel plants and / or multiple groups of sentinel plants in the farmland.

18. The method according to claim 17, further comprising: interpreting a third stress of the first stressor in the first group of sentinel plants during a second time period after the first time period based on a fourth set of features extracted from a third image of the first group of images captured during the second time period; predicting a second pressure gradient of the first stressor in the farmland during the second time period based on the third pressure and the model; and Responsive to the second pressure gradient exceeding a threshold pressure, generating a prompt to address the first stressor in plants in a sub-region of the agricultural field, the second pressure gradient predicting a fourth pressure of the first stressor in the sub-region of the agricultural field.

19. The method according to claim 17: in, Acquiring the first set of images includes acquiring the first set of images recorded by a fixed sensor located within the agricultural field; and Wherein, obtaining the second image of the farmland includes obtaining the second image of the farmland recorded by an aerial sensor.

20. The method of claim 17, further comprising: acquiring a second batch of images of a second group of sentinel plants in the field, the second batch recorded at a second frequency; interpreting a third stress of a second one of the set of stressors in the second set of sentinel plants during the first time period based on a fourth set of features extracted from a third image in the second batch of images captured during the first time period; interpreting a fourth stress of the second stressor in the second group of sentinel plants during the first time period based on a fifth set of features extracted from a second region of the second image including the second group of sentinel plants; interpreting a second pressure gradient of the second stressor in the agricultural field during the first time period based on a sixth set of features extracted from the second image; deriving a second model that relates the pressure of the second stressor at the second group of sentinel plants to the pressure gradient of the second stressor in the field based on the fourth pressure and the second pressure gradient; correcting the second pressure gradient of the second stressor in the agricultural field during the first time period based on the third pressure and the second model; and A pressure map is generated based on a combination of the first pressure gradient and the second pressure gradient.

Images