Method and system for monitoring anomalies in growth phase of a facility facing tomato

Abstract

The application provides a facility tomato growth stage abnormality monitoring method and system, which comprises the following steps: continuously collecting crop environment information and plant image information; determining the current growth stage of the tomato based on the crop environment information and the plant image information, comparing the crop environment information and the plant image information in the recent preset period with standard crop state information, and calculating the crop growth stage deviation; obtaining a symptomatic score based on the current plant image information; obtaining an asymptomatic score based on the crop environment information in the recent preset period; determining the current synergistic abnormality intensity data based on the current crop growth stage deviation, the symptomatic score and the asymptomatic score; determining the current abnormality evolution memory amount based on the synergistic abnormality intensity data in the recent preset period; and determining the current abnormality risk value based on the current synergistic abnormality intensity data and the abnormality evolution memory amount. The application can accurately distinguish normal changes and abnormal changes in different stages.

Description

Technical Field

[0001] This invention relates to the field of smart agriculture technology, and in particular to a method and system for monitoring anomalies during the growth stages of greenhouse tomatoes. Background Technology

[0002] Greenhouse tomatoes are an important crop in modern protected agriculture. Their growth is easily affected by various factors such as temperature, humidity, light, water supply, nutrient status, and disease stress. Abnormalities in the growing environment often directly impact flowering, fruit setting, fruit enlargement, and quality development, and in severe cases, can lead to yield reduction or even plant death. Therefore, timely detection of abnormalities, accurate assessment of their severity, and prompt intervention are crucial for ensuring stable and high yields of greenhouse tomatoes. Greenhouse tomatoes refer to tomatoes cultivated within artificially constructed protective facilities (such as greenhouses and polytunnels).

[0003] Current methods for monitoring tomato growth largely rely on manual inspections, with managers primarily judging whether tomatoes are abnormal by observing leaf color, plant growth, and fruit appearance. While this method is intuitive, it is heavily reliant on experience, leading to issues such as subjectivity, high workload, and delayed detection. This is especially problematic when the planting area is large or the growing environment changes rapidly, making continuous, stable, and accurate monitoring difficult. Even in some planting scenarios where temperature and humidity sensors, soil or root zone moisture sensors have been introduced, these often only collect and display single indicators, lacking the ability to comprehensively analyze multi-source data and thus failing to fully reflect the true growth status of tomatoes.

[0004] Meanwhile, although existing monitoring methods are beginning to incorporate image recognition or sensor data analysis technologies, most still rely on fixed thresholds or single models for judgment, failing to fully consider the differences in the normal state of tomatoes at different growth stages. In fact, the growth characteristics of the seedling stage, flowering and fruit setting stage, fruit enlargement stage, and color-changing and ripening stage are not the same, and the same environmental fluctuations or phenotypic changes have significantly different meanings at different stages. Applying a uniform standard to all stages can easily lead to false alarms or missed alarms.

[0005] Therefore, there is an urgent need for a method that can utilize multi-source information for monitoring, comprehensively reflect the true state of tomatoes, and fully consider the growth stages of tomatoes, thereby achieving accurate identification of abnormal tomato growth. Summary of the Invention

[0006] In view of this, embodiments of the present invention provide a method and system for monitoring anomalies during the growth stages of greenhouse tomatoes, in order to eliminate or improve one or more defects existing in the prior art.

[0007] One aspect of the present invention provides a method for monitoring abnormalities in the growth stages of greenhouse tomatoes. The method includes the following steps: continuously collecting crop environmental information and plant image information; wherein the crop environmental information includes spatial environmental information and root zone environmental information; determining the current growth stage of the tomato based on the crop environmental information and plant image information within the most recent preset time period; matching and comparing the crop environmental information and plant image information within the most recent preset time period with standard crop status information stored in a healthy sample database to calculate the deviation degree of the crop growth stage; obtaining a symptom score based on the current plant image information using a pre-trained image recognition model, wherein the symptom score indicates the degree of obvious symptom abnormalities currently present in the tomato; and utilizing the pre-trained... The time-series autoencoder prediction model obtains latent symptom scores based on crop environmental information within the most recent preset time period. These latent symptom scores represent the degree to which the current spatial environment and / or root zone environment of the tomato deviates from its normal growth trajectory. Based on the current crop growth stage deviation, symptom scores, and latent symptom scores, the model determines the current synergistic anomaly intensity data. Based on the synergistic anomaly intensity data within the most recent preset time period, the model determines the current anomaly evolution memory. Based on the current synergistic anomaly intensity data and the anomaly evolution memory, the model determines the current anomaly risk value. The anomaly evolution memory is used to characterize the upward or downward trend of anomalies within the most recent preset time period, and the anomaly risk value is used to characterize the severity and development trend of the current tomato growth anomaly.

[0008] In some embodiments of the present invention, the step of determining the current growth stage of the tomato based on crop environment information and plant image information within the most recent preset time period includes: extracting spatial environment change features and root zone environment state features based on crop environment information within the most recent preset time period; extracting canopy morphology change features, leaf color features, canopy temperature change features, and growth time sequence features based on plant image information within the most recent preset time period; using some or all of the spatial environment change features, root zone environment state features, canopy morphology change features, leaf color features, canopy temperature change features, and growth time sequence features as multi-source feature inputs; and outputting the current growth stage of the tomato using a pre-trained growth stage recognition model; wherein the growth stage recognition model is pre-trained based on a training set composed of historical multi-source features and corresponding growth stage labels.

[0009] In some embodiments of the present invention, the standard crop status information includes standard spatial environment information, standard root zone environment information, standard plant image information, standard comprehensive growth status value, and standard growth change range corresponding to the current growth stage of the tomato; the step of matching and comparing the crop environment information and plant image information within the most recent preset time period with the standard crop status information stored in the healthy sample bank to calculate the deviation degree of the crop growth stage includes: comparing the current spatial environment information, root zone environment information, and plant image information recording the canopy status with the standard crop status information corresponding to the current growth stage of the tomato, and calculating the spatial environment score, root zone environment score, and canopy score respectively; weighting the spatial environment score, root zone environment score, and canopy score according to preset weights to obtain the current comprehensive growth status value of the tomato, and determining the current growth change range of the tomato based on the comprehensive growth status value within the most recent preset time period; matching and comparing the current comprehensive growth status value and growth change range of the tomato with the standard comprehensive growth status value and standard growth change range respectively, and obtaining the crop growth stage deviation degree through deviation degree calculation.

[0010] In some embodiments of the present invention, the method further includes the step of pre-constructing a healthy sample library, specifically including: acquiring standard spatial environment information, standard root zone environment information, and standard plant image information recording canopy status of healthy tomatoes at each growth stage; calculating the standard comprehensive growth status value and standard growth variation range of healthy tomatoes at each growth stage; and storing the standard spatial environment information, standard root zone environment information, standard plant image information, standard comprehensive growth status value, and standard growth variation range of healthy tomatoes at each growth stage into the healthy sample library.

[0011] In some embodiments of the present invention, the plant image information includes a visible light image and a canopy thermal infrared image of the tomato plant; the image recognition model is a two-branch image recognition model, which structurally includes a first branch, a second branch, a fusion layer, and a classification judgment layer. The step of obtaining a symptom score based on the current plant image information using the pre-trained image recognition model includes: the first branch extracts visible light features based on the visible light image, and the second branch extracts infrared light features based on the canopy thermal infrared image; the visible light features and the infrared light features are weighted and fused in the fusion layer, and the weighted fusion result is input into the classification judgment layer, which outputs the symptom score; wherein, the two-branch image recognition model is pre-trained based on a training set consisting of historical visible light features, infrared light features, and corresponding symptom score labels.

[0012] In some embodiments of the present invention, the spatial environment information includes some or all of air temperature, air humidity, carbon dioxide concentration, and light intensity; the root zone environment information includes some or all of the root zone cultivation medium water content, root zone cultivation medium salinity, and root zone temperature; the temporal autoencoder prediction model adopts an encoder-decoder structure; the step of obtaining latent disease scores based on crop environment information within the most recent preset time period using the pre-trained temporal autoencoder prediction model includes: constructing a real monitoring sequence for the current time period using crop environment information within the most recent preset time period; using the real monitoring sequence as input to the temporal autoencoder prediction model; extracting the variation patterns in the real monitoring sequence using the encoder; generating a health sequence reconstruction result for the current time period using the decoder; generating a prediction result for the health status at the next moment; and determining the latent disease score based on the deviation between the real monitoring sequence and the health sequence reconstruction result for the current time period, and the deviation between the prediction result for the health status at the next moment and the actual status at the next moment.

[0013] In some embodiments of the present invention, after determining the current abnormal risk value, the method further includes: mapping the current abnormal risk value to a risk level based on a preset rule, and providing corresponding prompts or alarms according to the risk level; wherein the risk level includes prompt level, attention level, early warning level and alarm level.

[0014] Corresponding to the above methods, the present invention also provides an abnormal monitoring system for greenhouse tomato growth stages, including a processor, a memory, and a computer program / instructions stored in the memory. The processor is used to execute the computer program / instructions, and when the computer program / instructions are executed, the system implements the steps of any of the methods described in the above embodiments.

[0015] In accordance with the above methods, the present invention also provides a computer-readable storage medium having a computer program / instructions stored thereon, which, when executed by a processor, implements the steps of the method as described in any of the above embodiments.

[0016] Corresponding to the above methods, the present invention also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the method as described in any of the above embodiments.

[0017] The proposed method for abnormal monitoring of greenhouse tomatoes at different growth stages can identify the current growth stage of the tomato by combining the environmental and morphological characteristics of different growth stages. Then, it can call the normal growth reference standard of the corresponding stage for analysis, which helps to accurately distinguish between normal and abnormal changes at different stages, reduce misjudgments and omissions caused by stage differences, and improve the accuracy of monitoring abnormal growth of tomatoes.

[0018] Additional advantages, objects, and features of the invention will be set forth in part in the description which follows, and will also become apparent in part to those skilled in the art upon studying the description, or may be learned by practice of the invention. The objects and other advantages of the invention can be realized and obtained by means of the structures specifically pointed out in the specification and drawings.

[0019] Those skilled in the art will understand that the objectives and advantages achievable with this invention are not limited to those specifically described above, and that the above and other objectives achievable with this invention will become clearer from the following detailed description. Attached Figure Description

[0020] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, are not intended to limit the scope of the invention. In the drawings: Figure 1 This is a flowchart of an abnormality monitoring method for greenhouse tomatoes during their growth stages, according to one embodiment of the present invention.

[0021] Figure 2 This is a flowchart illustrating the process of determining the current growth stage of a tomato in one embodiment of the present invention.

[0022] Figure 3 This is a flowchart for calculating the deviation of crop growth stages in one embodiment of the present invention.

[0023] Figure 4 This is a schematic diagram of the computer equipment included in the system. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the embodiments and accompanying drawings. Here, the illustrative embodiments and descriptions of this invention are used to explain the invention, but are not intended to limit the invention.

[0025] It should also be noted that, in order to avoid obscuring the invention with unnecessary details, only the structures and / or processing steps closely related to the solution according to the invention are shown in the accompanying drawings, while other details that are not closely related to the invention are omitted.

[0026] It should be emphasized that the term "including / comprises" as used herein refers to the presence of a feature, element, step, or component, but does not exclude the presence or addition of one or more other features, elements, steps, or components.

[0027] It should also be noted that, unless otherwise specified, the term "connection" in this article can refer not only to a direct connection, but also to an indirect connection involving an intermediary.

[0028] In the following description, embodiments of the invention will be illustrated with reference to the accompanying drawings. In the drawings, the same reference numerals represent the same or similar parts, or the same or similar steps.

[0029] To overcome the problems of existing technologies: 1) lack of comprehensive analysis capabilities for multi-source data, making it difficult to fully reflect the true growth status of tomatoes; 2) insufficient consideration of the differences in normal state of tomatoes at different growth stages; 3) crude alarm methods, failing to meet actual planting management needs. Specifically, the crudeness of existing alarm methods refers to the fact that current technologies typically only provide simple normal or abnormal judgments, lacking hierarchical analysis of the degree of abnormality development, and unable to distinguish between slight fluctuations, persistent abnormalities, and severe abnormalities. This approach is not conducive to managers rationally arranging the sequence of actions and is prone to false alarms due to instantaneous fluctuations, affecting the actual application effect of the system.

[0030] To overcome the problems existing in the prior art, this invention aims to propose a method and system for anomaly monitoring during the growth stages of greenhouse tomatoes. Addressing the needs for anomaly monitoring and grading assessment during the growth stages of greenhouse tomatoes, it constructs a closed-loop technical route of "multi-source sensing—growth stage identification—anomaly detection—risk grading—automatic alarm." This system enables continuous anomaly monitoring and grading assessment of tomato growth status through unified collection and collaborative analysis of crop environmental information (including spatial environmental information and root zone environmental information) and plant image information.

[0031] Figure 1 This is a flowchart of an anomaly monitoring method for greenhouse tomatoes during their growth stages, according to an embodiment of the present invention. The method includes the following steps: Step S100: Continuously collect crop environmental information and plant image information; wherein, crop environmental information includes spatial environmental information and root zone environmental information.

[0032] Step S200: Determine the current growth stage of the tomato based on the crop environment information and plant image information within the most recent preset time period, match and compare the crop environment information and plant image information within the most recent preset time period with the standard crop status information stored in the healthy sample database, and calculate the deviation of the crop growth stage.

[0033] Step S300: Using a pre-trained image recognition model, obtain a symptom score based on the current plant image information. The symptom score is used to indicate the degree of obvious symptom abnormalities currently present in the tomato. Using a pre-trained temporal autoencoder prediction model, obtain a latent symptom score based on crop environmental information within the most recent preset time period. The latent symptom score is used to indicate the degree to which the current spatial environment and / or root zone environment of the tomato deviates from the normal growth trajectory.

[0034] Step S400: Determine the current synergistic anomaly intensity data based on the current crop growth stage deviation, symptom score, and latent symptom score; determine the current anomaly evolution memory based on the synergistic anomaly intensity data within the most recent preset time period; and determine the current anomaly risk value based on the current synergistic anomaly intensity data and the anomaly evolution memory. The anomaly evolution memory is used to characterize the upward or downward trend of anomalies within the most recent preset time period, and the anomaly risk value is used to characterize the severity and development trend of current tomato growth anomalies.

[0035] Using this embodiment of the invention, the proposed method for abnormal monitoring of greenhouse tomato growth stages can identify the current growth stage of the tomato by combining the environmental and morphological characteristics of different growth stages. Then, it can call the normal growth reference standard of the corresponding stage for analysis, which helps to accurately distinguish between normal and abnormal changes in different stages, reduce misjudgments and omissions caused by stage differences, and help improve the accuracy of monitoring abnormal tomato growth.

[0036] Figure 2 This is a flowchart illustrating the process of determining the current growth stage of a tomato in one embodiment of the present invention. In some embodiments of the present invention, step S200, which determines the current growth stage of the tomato based on crop environmental information and plant image information within a recent preset time period, is as follows: Figure 2 As shown, it can specifically include: Step S210: Extract spatial environmental change features and root zone environmental status features based on crop environmental information within the most recent preset time period, and extract canopy morphology change features, leaf color features, canopy temperature change features, and growth time sequence features based on plant image information within the most recent preset time period.

[0037] Step S220: Input some or all of the spatial environment change features, root zone environment state features, canopy morphology change features, leaf color features, canopy temperature change features, and growth time sequence features as multi-source features, and output the current growth stage of the tomato using a pre-trained growth stage identification model; wherein, the growth stage identification model is pre-trained based on a training set composed of historical multi-source features and corresponding growth stage labels.

[0038] Taking the extraction process of spatial environment change characteristics as an example, the steps are as follows: 1) Obtain the time-series data stream of spatial environment information within the most recent preset time period, and perform noise reduction and missing value interpolation preprocessing. 2) Calculate the temporal statistical characteristics of each environmental factor, including but not limited to the mean, range, variance, and first-order difference (used to characterize the severity of environmental fluctuations). 3) Calculate the cumulative effect indicators of light and temperature / humidity (such as daily accumulated temperature and daily accumulated solar radiation) using the sliding window method. 4) Concatenate the above statistical characteristics and cumulative indicators to construct a spatial environment change feature vector. The feature extraction principles of the above spatial environment change characteristics, root zone environmental state characteristics, canopy morphology change characteristics, leaf color characteristics, canopy temperature change characteristics, and growth temporal characteristics are similar and belong to common knowledge in this field. All multi-source feature extraction based on the concept of this invention falls within the protection scope of this solution.

[0039] By employing this embodiment of the invention, a growth stage identification model can be trained using multiple dimensions, including crop spatial environment, root zone environment, and crop morphology, combined with historical data, which helps to achieve accurate identification of tomato growth stages.

[0040] In some embodiments of the present invention, the standard crop status information includes standard spatial environment information, standard root zone environment information, standard plant image information, standard comprehensive growth status value, and standard growth change range corresponding to the current growth stage of the tomato.

[0041] Correspondingly, Figure 3 This is a flowchart illustrating the calculation of crop growth stage deviation in one embodiment of the present invention. Step S200 involves matching and comparing crop environmental information and plant image information from the most recent preset time period with standard crop status information stored in a healthy sample database to calculate the crop growth stage deviation. Figure 3 As shown, it includes: Step S230: Compare the current spatial environment information, root zone environment information, and plant image information recording the canopy status with the standard crop status information corresponding to the current growth stage of the tomato, and calculate the spatial environment score, root zone environment score, and canopy score respectively.

[0042] Step S240: The spatial environment score, root zone environment score and canopy score are weighted according to preset weights to obtain the current comprehensive growth status value of the tomato. The current growth change range of the tomato is determined based on the comprehensive growth status value in the most recent preset time period.

[0043] Step S250: Match and compare the current comprehensive growth status value and growth change range of the tomato with the standard comprehensive growth status value and standard growth change range, and calculate the deviation of the crop growth stage through deviation calculation.

[0044] Using this invention, it is possible to statistically analyze the crop environment information and plant image information of healthy tomato plants based on historical data, and identify the deviation between the current state of the tomato and the standard state by comparison, thereby determining the probability of abnormal growth of the tomato plant from the perspective of deviation from the growth stage.

[0045] In some embodiments of the present invention, the method further includes the step of pre-constructing a healthy sample library, specifically including: acquiring standard spatial environment information, standard root zone environment information, and standard plant image information recording canopy status of healthy tomatoes at each growth stage; calculating the standard comprehensive growth status value and standard growth variation range of healthy tomatoes at each growth stage; and storing the standard spatial environment information, standard root zone environment information, standard plant image information, standard comprehensive growth status value, and standard growth variation range of healthy tomatoes at each growth stage into the healthy sample library.

[0046] Using this embodiment of the invention, standards can be pre-established based on the status information of healthy tomato plants at each growth stage. In subsequent steps, the probability of abnormal growth of tomato plants can be assessed by calculating the deviation between the current growth stage status information of the tomato and the standard status information.

[0047] In some embodiments of the present invention, the plant image information includes visible light images of tomato plants and thermal infrared images of the canopy.

[0048] Accordingly, the above image recognition model is a two-branch image recognition model. Structurally, the two-branch image recognition model includes a first branch, a second branch, a fusion layer, and a classification and judgment layer. The step of obtaining the symptom score based on the current plant image information using the pre-trained image recognition model includes: the first branch extracts visible light features based on the visible light image, and the second branch extracts infrared light features based on the canopy thermal infrared image; the visible light features and the infrared light features are weighted and fused in the fusion layer, and the weighted fusion result is input into the classification and judgment layer, which outputs the symptom score; wherein, the two-branch image recognition model is pre-trained based on a training set consisting of historical visible light features, infrared light features, and corresponding symptom score labels.

[0049] Using this embodiment of the invention, it is possible to identify abnormal phenomena that have already been manifested, such as yellowing, curling, and wilting of leaves.

[0050] In some embodiments of the present invention, the spatial environment information includes some or all of the air temperature, air humidity, carbon dioxide concentration and light intensity, and the root zone environment information includes some or all of the root zone cultivation medium water content information, root zone cultivation medium salinity information and root zone temperature information.

[0051] Accordingly, the time-series autoencoder prediction model adopts an encoder-decoder structure. The step of obtaining the latent disease score based on crop environmental information within the most recent preset time period using the pre-trained time-series autoencoder prediction model includes: constructing a real monitoring sequence for the current time period using crop environmental information within the most recent preset time period; using the real monitoring sequence as the input of the time-series autoencoder prediction model; extracting the change patterns in the real monitoring sequence using the encoder; generating a health sequence reconstruction result for the current time period using the decoder; generating a prediction result for the health status at the next moment; and determining the latent disease score based on the deviation between the real monitoring sequence and the health sequence reconstruction result for the current time period, and the deviation between the prediction result for the health status at the next moment and the actual status at the next moment.

[0052] This invention, by combining changes in environmental conditions, root zone moisture, salinity, and canopy temperature, can detect early signs of abnormality that have not yet become apparent. Clearly, compared to existing methods that primarily rely on manual observation or single image recognition, this invention can identify potential risks in tomato growth much earlier, thus providing conditions for early intervention.

[0053] In some embodiments of the present invention, after determining the current abnormal risk value, the method further includes: mapping the current abnormal risk value to a risk level based on a preset rule, and providing corresponding prompts or alarms according to the risk level; wherein the risk level includes prompt level, attention level, early warning level and alarm level.

[0054] This invention enables a comprehensive assessment and tiered alarm system for anomalies, integrating deviations from the current stage, explicit and implicit symptoms, and the ongoing evolution of abnormalities. It avoids simply outputting a single result of "normal" or "abnormal." This approach reduces false alarms caused by momentary fluctuations and provides more targeted alarm information based on the severity of the anomaly, better meeting the practical needs of greenhouse tomato cultivation and management.

[0055] This invention focuses on greenhouse tomatoes. Through continuous acquisition and collaborative analysis of crop environmental information, substrate information, and plant image information, it establishes corresponding normal growth reference standards based on the tomato's current growth stage. It also identifies both existing abnormalities and early signs of abnormality, enabling continuous monitoring, tiered assessment, and alarm handling of tomato growth abnormalities. The basic approach of this invention is as follows: First, continuously acquire crop environmental information and plant image information during the tomato's growth process; then, identify the current growth stage of the tomato; next, retrieve standard crop state information formed under normal growth conditions corresponding to that stage and compare it with the current tomato growth state information; finally, combine existing abnormalities and early signs of abnormality to calculate the abnormal risk value and output the corresponding alarm level. In the field of data acquisition, continuous acquisition refers to the process by which an acquisition system continuously reads and records signal data at a fixed sampling frequency within a set time period.

[0056] In some embodiments of the present invention, a method for monitoring anomalies during the growth stage of greenhouse tomatoes includes the following process: Process 1: Multi-source data acquisition and preprocessing for greenhouse tomatoes First, various crop environment sensing units and image acquisition units are deployed within the greenhouse for tomato cultivation. The crop environment sensing units include air temperature, air humidity, carbon dioxide concentration, and light intensity sensors for collecting spatial environmental information, and root zone cultivation medium moisture content and conductivity sensors for collecting root zone environmental information. The image acquisition units include an RGB camera and a thermal infrared camera mounted above the planting area. Using these sensors, air temperature, air humidity, carbon dioxide concentration, light intensity, root zone cultivation medium moisture content, and root zone cultivation medium salinity information are collected at fixed time intervals. Simultaneously, visible light images (RGB images) of the tomato plants and canopy thermal infrared images are collected at fixed time intervals.

[0057] Then, the various data collected at the same time are bound together according to a unified timestamp, and can be organized by plant number, planting row number, and greenhouse number to form a time-series observation dataset for individual tomato plants and individual monitoring areas. Each collection does not yield isolated data, but rather a set of comprehensive observation results that collectively reflect the current state of the tomato, thus providing complete input for subsequent analysis. Based on stage one, a multi-source input foundation required for tomato growth status analysis can be established.

[0058] Furthermore, in a specific embodiment of the present invention, a preprocessing process is also included for the raw data collected by the crop environment sensing unit and the image acquisition unit. For example, for air temperature, air humidity, carbon dioxide concentration, light intensity, root zone cultivation medium water content information, and root zone cultivation medium salinity information, outlier removal, missing data repair, and time series smoothing are required to help reduce the impact of occasional fluctuations and sampling errors. For visible light images (RGB images) and canopy thermal infrared images, brightness correction, temperature calibration, background removal, and plant area extraction are required to avoid interference from supports, ropes, ground, and changes in external light intensity in subsequent judgments.

[0059] Process Two: Identifying Tomato Growth Stages In some specific embodiments of the present invention, spatial environmental change features and root zone environmental state features can be extracted based on crop environmental information over a continuous period of time, and canopy morphological change features, leaf color features, canopy temperature change features, and growth time sequence features can be extracted based on plant image information over a corresponding continuous period of time. The root zone environmental state features may include the root zone water and fertilizer supply status.

[0060] Furthermore, the current growth stage of the tomato can be determined based on the changing patterns of the aforementioned characteristics within the most recent preset time period. For tomatoes, the growth stages can be divided into seedling stage, flowering and fruit setting stage, fruit enlargement stage, and color-changing and ripening stage. Since the normal growth of tomatoes differs at different stages—for example, during the seedling stage, environmental stability is more important, while during the fruit enlargement stage, root zone water supply and canopy heat status are more crucial—if the current stage is not determined first, and the same set of judgment criteria is used for all stages, normal changes are easily misjudged as abnormal. Therefore, this embodiment first identifies the stage before proceeding with subsequent analysis.

[0061] Optionally, a growth stage identification model with multi-source features as input can be pre-constructed. This growth stage identification model can adopt an existing time-series graph neural network architecture, using the aforementioned spatial environment change features, root zone environmental state features, canopy morphology change features, leaf color features, canopy temperature change features, and growth temporal features as multi-source feature inputs. The training set used in its pre-training phase includes historical multi-source features and corresponding growth stage labels. It should be noted that the improvement of the growth stage identification model mainly lies in the construction of the training set; the model training method and model architecture can refer to existing classification or prediction models.

[0062] Once the current growth stage of the tomato is determined, it can provide a basis for subsequent anomaly detection using staged standards.

[0063] Step 3: Multi-source collaborative characterization and sample library construction for tomato growth stages In one specific embodiment of the present invention, a healthy sample library is pre-constructed to store standard crop status information of the same type of tomato under normal growth conditions. The standard crop status information stored in the healthy sample library is used as a control reference for the current monitoring object (tomato plant).

[0064] After identifying the current growth stage of the tomato plant, its spatial environment score, root zone environment score, and canopy score are first calculated. Then, a healthy sample database for the corresponding growth stage is accessed, and a multi-source collaborative representation under stage constraints is established. Standard crop state information stored in the healthy sample database is used as a reference for calculation of the spatial environment score, root zone environment score, and canopy score. The spatial environment score reflects the degree to which the environmental (air) conditions of the plant's location are close to the normal state at that stage; the root zone environment score reflects the degree to which the environmental (moisture, salinity) conditions of the root zone are close to the normal state at that stage; and the canopy score reflects the degree to which the plant morphology, tomato leaf color, and canopy thermal state (canopy temperature changes) are close to the normal state at that stage.

[0065] Optionally, the current comprehensive growth status value can be calculated based on the spatial environment score, root zone environment score, and canopy score. The comprehensive growth status value can be used to comprehensively represent the growth status of tomatoes.

[0066] The overall growth status value of tomatoes is calculated using the following formula: in, Indicates the first The overall growth state value at time t. Indicates the spatial environment score. Indicates the root region environment score. Indicates the canopy score. These represent the weight coefficients corresponding to the current growth stage. Different weight configurations are used for different growth stages, rather than always using a fixed ratio for judgment.

[0067] Furthermore, after obtaining the current comprehensive growth status value, it can be compared with the standard crop status information (e.g., healthy reference status value) of healthy tomatoes at the same growth stage to determine the degree of deviation of the current plant from the standard crop status information. Simultaneously, a normal growth trend of tomatoes can be constructed using continuous historical observation sequences, and the difference between the current state and the normal growth trajectory can be compared to determine whether the current growth deviates from the normal evolutionary pattern. That is, both the difference between the current state and the reference value and the difference between the current growth rate and the normal growth rate are compared to determine the degree of deviation from the crop growth stage.

[0068] In a specific embodiment of the present invention, the formula for matching and comparing the comprehensive growth status value of tomatoes with the standard comprehensive growth status value of healthy tomatoes at the same growth stage is expressed as follows: in, Indicates the first Deviation from the crop growth stage at any given moment This represents the standard composite growth state value corresponding to the current growth stage. This indicates the magnitude of change in the current growth state. This represents the reference range of change during normal growth at this stage. and It represents the adjustment coefficient for the deviation between the two parts at the current stage. It is necessary to pay attention to both "whether the current growth is deviating" and "whether the recent growth changes are deviating", so as to avoid the problem of one-sided judgment caused by looking only at the state at a certain moment.

[0069] The above method involves continuously collecting environmental data, root zone status data, and plant image data of greenhouse tomatoes, identifying the current growth stage of the tomatoes, and calling upon the corresponding normal growth reference standards for that stage to conduct targeted analysis of the current status. This makes the anomaly judgment stage-adaptive, thereby improving the accuracy of monitoring results.

[0070] Step 4: Detection of Obvious and Latent Symptoms After obtaining the deviation results of the current tomato status from the healthy baseline (i.e. the deviation of the crop growth stage), a joint judgment is made from two directions: "abnormalities that have been shown" and "early signs that have not yet been shown".

[0071] In summary, the visible abnormality detection uses RGB and thermal infrared images of tomato plants to identify outward-facing abnormalities such as leaf yellowing, leaf curling, localized wilting, fruit abnormalities, and canopy thinning, and outputs confidence results for visible abnormalities. The latent symptom detection uses sequences of environmental, substrate, and canopy temperature changes within a continuous time window to identify signs of heat stress, water shortage stress, and nutrient imbalance that have formed before visible symptoms appear, and outputs latent abnormality scores. Using this dual-channel monitoring method, both existing growth abnormalities can be identified, and potential risks that are yet to manifest can be detected in advance, thus achieving early warning of tomato growth abnormalities.

[0072] The following is the detailed process for detecting obvious abnormalities and latent symptoms: Firstly, the detection of symptomatic abnormalities employs a dual-branch image recognition model. One branch processes the visible light image of the tomato plant, primarily extracting information such as leaf color, texture, edge morphology, and canopy density variations to identify visible light features like yellowing, curling, localized wilting, and abnormal fruit color. The other branch processes the thermal infrared image, mainly extracting canopy temperature distribution and localized thermal anomalies to identify infrared features such as localized temperature increases, concentrated heat spots, and stressed canopy areas. The features extracted from the two branches (visible and infrared features) are weighted and fused in a fusion layer before being input into a classification and judgment layer, outputting the probability of a symptomatic abnormality in the current tomato plant. For ease of subsequent risk assessment, this embodiment defines this probability as a symptomatic score. A higher symptom score indicates a more pronounced abnormality in the tomato's appearance. The visible light image branch uses a convolutional neural network to extract color and morphological features, while the thermal infrared image branch uses a lightweight convolutional network to extract temperature distribution features. Subsequently, an attention fusion module integrates the two feature sets, highlighting key information related to the abnormality and reducing the influence of background and irrelevant areas. The symptom score can be expressed as: in, This represents the probability of color anomalies obtained from a visible light image. This represents the probability of morphological anomalies obtained from visible light images. This indicates the probability of thermal anomalies obtained from thermal infrared images. In this way, the symptom score is no longer given by a single image feature, but is determined by color changes, morphological changes and thermal anomalies. Therefore, it is more stable than relying solely on leaf color or solely on temperature thresholds.

[0073] Secondly, the latent symptom detection primarily targets the early identification of anomalies in tomatoes before obvious external changes appear. The latent symptom detection employs a time-series autoencoder prediction model. This model takes crop environmental information over a continuous period (which may include some or all of the following: air temperature, air humidity, light intensity, carbon dioxide concentration, root zone water content, root zone salinity, and average canopy temperature) as input to learn the normal variation patterns of healthy tomatoes at their current growth stage. When the input sequence matches a healthy growth pattern, the model can reconstruct and predict the sequence well; however, when the input sequence deviates from the healthy pattern, such as persistent water shortage, increased root zone salinity, continuously high canopy temperature, or abnormal environmental fluctuations, the model's reconstruction and prediction errors will significantly increase. The system uses this error to characterize the degree of latent anomaly and converts it into a latent symptom score. This time-series autoencoder prediction model employs an encoder-decoder structure. The encoder extracts the changing patterns from continuously monitored crop environmental information, while the decoder reconstructs the expected temporal trend under healthy conditions and, in conjunction with the prediction unit, predicts the normal state at the next moment. The system integrates the differences between the current actual sequence and the model reconstruction and prediction results to obtain a latent disease score, which can be expressed as: in, This represents the actual monitoring sequence for the current time period. This represents the model's reconstruction results of the healthy sequence for the current time period. Indicates the actual state at the next moment. This indicates a prediction of the plant's health status at the next moment. A higher latent symptom score means that although the tomato may not yet show obvious signs of yellowing, curling, or wilting leaves, its root water supply, salt balance, heat status, or environmental adaptation status has already begun to deviate from the normal growth trajectory.

[0074] By combining the aforementioned symptom detection model and latent symptom detection model, both "visible problems" and "problems that have not yet manifested" can be covered simultaneously. When the symptom score is high, it indicates that the tomato (a type of tomato) has already shown obvious abnormalities; when the latent symptom score is high but the symptom score is still low, it indicates that the tomato is more likely in an early risk stage; when both are high, it indicates that the abnormality has developed from early signs into a more obvious problem.

[0075] Step 5: Abnormal Risk Assessment and Classification After obtaining the crop growth stage deviation, symptom score, and latent symptom score, a stage-sensitive mapping is performed on these three types of abnormal information based on the current growth stage of the tomato. This is because abnormal signals of the same magnitude have different effects on tomato growth at different stages. For example, during the fruit enlargement stage, latent water shortage and heat stress tend to affect fruit development more quickly, while during the color-changing and ripening stage, symptom-related leaf and fruit surface abnormalities more directly reflect the problem. Therefore, this embodiment first maps the crop growth stage deviation, symptom score, and latent symptom score to the risk response space of the current stage, obtaining the stage-corrected crop growth stage deviation, symptom score, and latent symptom score, expressed as follows: in, , , These represent the crop growth stage deviation, manifest symptom score, and latent symptom score, respectively, after correction based on the current growth stage. This indicates the sensitivity coefficient corresponding to that stage. Through this type of processing, minor anomalies will not be over-amplified, while more sensitive anomaly signals in high-risk stages will be highlighted more clearly.

[0076] After completing the phase correction, the system further constructs a synergistic risk prediction system for manifest and latent symptoms. Since manifest and latent symptom scores are not independent, a high latent symptom score indicates that the internal growth of the tomato has begun to deviate from its normal trajectory; conversely, a rising manifest symptom score indicates that the abnormality has progressed from early signs to the external appearance. At this point, the risk should not simply be additive but should exhibit synergistic enhancement. Therefore, the synergistic anomaly intensity data for the current moment is generated using the following method: in, Indicates the first Data on the intensity of collaborative anomalies at any given time. This represents the amplification coefficient of latent symptoms on overall abnormality at the current growth stage. The first term in the formula indicates that if the stage deviation is already significant, and the latent symptom score is also high, it suggests that the current growth deviation is likely not a normal fluctuation, but rather an accumulating abnormal process, thus amplifying the deviation at the crop growth stage. The second term indicates that when both visible and latent symptoms increase simultaneously, the system adds an additional synergistic term, causing a non-linear enhancement of the risk value.

[0077] In actual monitoring, a short-term shading, local temperature fluctuation, or instantaneous change in illumination may temporarily increase the intensity of coordinated anomalies. To avoid the system being overly sensitive to occasional fluctuations, this invention further introduces a persistent evolutionary memory factor to describe whether the anomaly has persisted and shown an upward trend in the recent period. Specifically, the system constructs anomaly evolutionary memory based on the changes in coordinated anomaly intensity data between the current and previous moments, combined with historical risk memory. : in, Indicates the first The amount of memory generated by abnormal evolution at any given moment. This represents the amount of abnormal evolutionary memory from the previous moment. This represents the memory retention coefficient at the current growth stage. If the current anomaly intensity is higher than the previous moment, the system considers the anomaly to be continuing to develop and writes the increased portion into memory; if the current anomaly intensity does not continue to increase, the memory size will not be further amplified. This is how it is obtained. It no longer simply counts "how many times the anomaly occurred", but places greater emphasis on "whether the anomaly has continued to worsen".

[0078] Ultimately, the early warning risk value is related not only to the current intensity of the anomaly but also to its continued development trend, and the coupling relationship between the two is also considered. The anomaly risk value can be expressed as: in, Indicates the first The abnormal risk value at any given time. Let represent the normalized mapping function. The first two terms in the formula represent the direct contributions of the current anomaly intensity and the continuous evolution trend to the risk, respectively, while the third term represents the coupling effect between the two. If the current anomaly is already very strong and continues to worsen, the system will significantly increase the final risk value, which is more in line with the judgment logic in actual planting that "when the anomaly is obvious and deteriorating, it should be quickly upgraded to a high-level warning."

[0079] The abnormal risk value reflects the severity and trend of current growth abnormalities in tomatoes. Subsequently, based on the abnormal risk value... Risk levels can be categorized by size. For example, the current status of a tomato plant can be divided into four levels: alert, attention, warning, and alarm.

[0080] Step 6: Hierarchical Automatic Alarm and Result Feedback Update Optionally, after completing the risk classification, corresponding alarm information can be automatically generated based on the current risk level. For example, when the risk is at a low level, a prompt message can be issued to remind managers to continue monitoring; when the risk increases further, it can enter a key monitoring state; when the warning level is reached, a warning message is sent to managers; when the alarm level is reached, the greenhouse management platform reminder, mobile push notification, and / or on-site alarm prompt are triggered simultaneously. In this way, the monitoring results are not just limited to the analysis level, but can directly serve the tomato planting management process. The alarm information clearly provides the greenhouse number, planting area number, current growth stage of the tomato, anomaly type, anomaly level, alarm time, and handling suggestions.

[0081] Furthermore, after management personnel have completed their review and processing, they can record and provide feedback on the results of manual review and handling. This not only records the circumstances of each anomaly but also the effectiveness of subsequent actions. As usage increases, this feedback information can be used to update the growth stage identification model, revise health reference standards, optimize anomaly judgment rules, and adjust alarm strategies. This helps to gradually improve accuracy and stability in subsequent use, forming a complete closed loop from monitoring, identification, and classification to feedback optimization.

[0082] Corresponding to the above method, the present invention also provides an abnormal monitoring system for greenhouse tomato growth stages. The system includes a computer device, which includes a processor and a memory. The memory stores computer instructions, and the processor is used to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the system implements the steps of the method described above.

[0083] Figure 4 See the schematic diagram of the computer equipment included in the system. Figure 4 The computer device 00 includes: a processor 01, a memory 02, and a computer program stored on the memory 02 and executable on the processor 01. When the processor 01 executes the computer program, it implements the method steps proposed in any of the above embodiments.

[0084] The processor 01 is connected to the memory 02, such as via a bus 03. The processor 01 can be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. The processor 01 can also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc. The bus 03 may include a pathway for transmitting information between the aforementioned components. The bus 03 can be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The bus 03 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 The text uses only a single thick line to represent a bus, but this does not imply that there is only one bus or one type of bus. Memory 02 stores a computer program corresponding to the human factors data server access control method described in the above embodiments of this application. This computer program is executed under the control of processor 01. Processor 01 executes the computer program stored in memory 02 to implement the content shown in the aforementioned method embodiments.

[0085] Corresponding to the methods described above, the present invention also provides a computer-readable storage medium having a computer program / instructions stored thereon, which, when executed by a processor, implements the steps of the method as described in any of the above embodiments. The computer-readable storage medium may be a tangible storage medium, such as random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, register, floppy disk, hard disk, removable storage disk, CD-ROM, or any other form of storage medium known in the art.

[0086] Corresponding to the above methods, the present invention also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the method as described in any of the above embodiments.

[0087] Those skilled in the art will understand that the exemplary components, systems, and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software, or a combination of both. Whether implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention. When implemented in hardware, it can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this invention are programs or code segments used to perform the desired tasks. The programs or code segments can be stored in a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried in a carrier wave.

[0088] It should be clarified that the present invention is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of the present invention.

[0089] In this invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with or in place of features of other embodiments.

[0090] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various modifications and variations of the embodiments of the present invention are possible. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for monitoring anomalies during the growth stages of greenhouse tomatoes, characterized in that, The method includes: Continuously collect crop environmental information and plant image information; among which, crop environmental information includes spatial environmental information and root zone environmental information; Based on the crop environment information and plant image information within the most recent preset time period, the current growth stage of the tomato is determined. The crop environment information and plant image information within the most recent preset time period are matched and compared with the standard crop status information stored in the healthy sample database to calculate the deviation of the crop growth stage. A pre-trained image recognition model is used to obtain a symptom score based on the current plant image information. The symptom score is used to indicate the degree of obvious symptom abnormalities in tomatoes. A pre-trained temporal autoencoder prediction model is used to obtain a latent symptom score based on crop environmental information within the most recent preset time period. The latent symptom score is used to indicate the degree to which the current spatial environment and / or root zone environment of tomatoes deviates from the normal growth trajectory. The current synergistic anomaly intensity data is determined based on the current crop growth stage deviation, symptom score, and latent symptom score. The current anomaly evolution memory is determined based on the synergistic anomaly intensity data within the most recent preset time period. The current anomaly risk value is determined based on the current synergistic anomaly intensity data and the anomaly evolution memory. The anomaly evolution memory is used to characterize the upward or downward trend of anomalies within the most recent preset time period. The anomaly risk value is used to characterize the severity and development trend of current tomato growth anomalies.

2. The method according to claim 1, characterized in that, The step of determining the current growth stage of the tomato based on crop environmental information and plant image information within the most recent preset time period includes: Based on crop environmental information within the most recent preset time period, spatial environmental change features and root zone environmental status features are extracted. Based on plant image information within the most recent preset time period, canopy morphology change features, leaf color features, canopy temperature change features, and growth time sequence features are extracted. The tomato's current growth stage is output using a pre-trained growth stage identification model, which takes some or all of the following characteristics as multi-source feature inputs: spatial environment change characteristics, root zone environment state characteristics, canopy morphology change characteristics, leaf color characteristics, canopy temperature change characteristics, and growth time sequence characteristics.

3. The method according to claim 1, characterized in that, The standard crop status information includes standard spatial environment information, standard root zone environment information, standard plant image information, standard comprehensive growth status value, and standard growth change range corresponding to the current growth stage of the tomato. The step of matching and comparing crop environmental information and plant image information within the most recent preset time period with standard crop status information stored in the healthy sample database, and calculating the deviation of crop growth stage, includes: The current spatial environment information, root zone environment information, and plant image information recording the canopy status are compared with the standard crop status information corresponding to the current growth stage of tomatoes, and the spatial environment score, root zone environment score, and canopy score are calculated respectively. The spatial environment score, root zone environment score and canopy score are weighted according to preset weights to obtain the current comprehensive growth status value of the tomato. The current growth change of the tomato is determined based on the comprehensive growth status value in the most recent preset time period. The current comprehensive growth status value and growth change range of tomatoes are matched and compared with the standard comprehensive growth status value and standard growth change range, and the deviation of crop growth stage is obtained by deviation calculation.

4. The method according to claim 2, characterized in that, The method also includes a step of pre-constructing a healthy sample library, specifically including: acquiring standard spatial environment information, standard root zone environment information, and standard plant image information recording canopy status of healthy tomatoes at each growth stage; calculating the standard comprehensive growth status value and standard growth variation range of healthy tomatoes at each growth stage; and storing the standard spatial environment information, standard root zone environment information, standard plant image information, standard comprehensive growth status value, and standard growth variation range of healthy tomatoes at each growth stage into the healthy sample library.

5. The method according to claim 1, characterized in that, The plant image information includes visible light images of tomato plants and thermal infrared images of the canopy. The image recognition model is a two-branch image recognition model. Structurally, the two-branch image recognition model includes a first branch, a second branch, a fusion layer, and a classification and judgment layer. The step of obtaining the symptom score based on the current plant image information using the pre-trained image recognition model includes: The first branch extracts visible light features based on visible light images, and the second branch extracts infrared light features based on canopy thermal infrared images. The visible light features and the infrared light features are weighted and fused in the fusion layer, and the weighted fusion result is input into the classification and judgment layer, which outputs the symptom score. The dual-branch image recognition model is pre-trained based on a training set consisting of historical visible light features, infrared light features and corresponding symptom score labels.

6. The method according to claim 1, characterized in that, The spatial environment information includes some or all of the air temperature, air humidity, carbon dioxide concentration, and light intensity; the root zone environment information includes some or all of the root zone cultivation medium water content, root zone cultivation medium salinity, and root zone temperature. The temporal autoencoder prediction model adopts an encoder-decoder structure. The step of obtaining latent disease scores based on crop environmental information within the most recent preset time period using the pre-trained temporal autoencoder prediction model includes: A real monitoring sequence for the current time period is constructed using crop environmental information within the most recent preset time period. This real monitoring sequence is used as the input to a time-series autoencoder prediction model. An encoder is used to extract the variation patterns in the real monitoring sequence, and a decoder is used to generate a health sequence reconstruction result for the current time period and a prediction result for the health status at the next time moment. Based on the deviation between the real monitoring sequence and the health sequence reconstruction result for the current time period, and the deviation between the prediction result for the health status at the next time moment and the actual status at the next time moment, the latent disease score is determined.

7. The method according to claim 1, characterized in that, After determining the current abnormal risk value, the method further includes: mapping the current abnormal risk value to a risk level based on preset rules, and providing corresponding prompts or alarms according to the risk level; wherein, the risk level includes prompt level, attention level, early warning level and alarm level.

8. An anomaly monitoring system for greenhouse tomato growth stages, comprising a processor, a memory, and a computer program / instructions stored in the memory, characterized in that, The processor is configured to execute the computer program / instructions, and when the computer program / instructions are executed, the system implements the steps of the method as described in any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method as described in any one of claims 1 to 7.

10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 7.

Images