Strawberry disease and pest early warning and disposal decision method

Abstract

The application discloses a strawberry disease and pest early warning and disposal decision method, and belongs to the field of strawberry disease and pest control. The method comprises the following steps: collecting multi-source time sequence data of target strawberry plants, wherein the multi-source time sequence data comprises natural physiological response data and operation disturbance records; constructing a natural state sequence and a reference rhythm template based on the multi-source time sequence data; and determining a detection period according to the deviation degree of the natural state sequence and the reference rhythm template. The natural state sequence and the reference rhythm template are constructed for the target strawberry plants, so that the abnormal deviation state can be identified before the appearance of a dominant symptom of the strawberry plants, thereby solving the problem that the prior art is difficult to timely find the recessive development process of diseases and pests.

Description

Technical Field

[0001] This invention relates to the field of heating blanket technology, and more specifically, to a method for early warning and decision-making regarding strawberry diseases and pests. Background Technology

[0002] Strawberries are susceptible to diseases and pests such as gray mold, powdery mildew, anthracnose, thrips, and spider mites during greenhouse cultivation. Especially during the flowering and fruit setting period and under high temperature and humidity conditions, the frequency of disease and pest occurrence is high and the speed of spread is fast, which can easily cause fruit damage, reduced yield and deterioration in quality. Existing methods for monitoring and controlling strawberry diseases and pests mostly rely on manual inspection, image recognition, or environmental parameter monitoring. Manual inspection mainly depends on growers' observation of symptoms on the surface of leaves, flowers, and fruits, which is subjective and lacks timeliness. Image recognition methods usually rely on obvious symptoms such as lesions, chlorosis, and curling for classification, making it difficult to identify them in advance when they are latent. Although environmental parameter monitoring methods can reflect changes in external conditions such as temperature and humidity, they are difficult to accurately characterize the plant's dynamic response to abnormal disturbances. Therefore, existing technologies generally have limitations in accurately identifying whether strawberry plants that have not yet developed obvious symptoms have entered the latent development stage of pests and diseases, and it is also difficult to further determine whether the current abnormal state is still in a reversible stage. As a result, it is impossible to determine the effective treatment time in a timely manner, leading to problems such as delayed pest and disease early warning and insufficient targeted treatment decisions. Summary of the Invention

[0003] To address the problems mentioned in the background section, the present invention provides the following technical solution: A method for early warning and decision-making regarding strawberry diseases and pests includes the following steps: Multi-source time-series data of target strawberry plants are collected. The multi-source time-series data includes natural physiological response data and operation disturbance records. A natural state sequence and a reference rhythm template are constructed based on the multi-source time-series data. The time period to be inspected is determined according to the degree of deviation between the natural state sequence and the reference rhythm template. Under preset safety constraints, a small-amplitude trial disturbance is applied to the target strawberry plant during the inspection period, and the response sequence before and after the disturbance is collected. Based on the natural state sequence, the reference rhythm template, and the response sequence, phase offset, response amplitude difference, recovery slope, and recovery duration are extracted to construct controlled response features; The natural state sequence, the controlled response features, and the operational disturbance records are input into a long short-term memory neural network to obtain the damage stress state value, which characterizes the degree of pathogen infection accumulation and the degree of insect ingestion accumulation, as well as the recovery capacity state value, which characterizes the plant's recovery capacity and organ tolerance capacity. Based on the increment of the harmful pressure state value, the decrease of the recovery ability state value, the recovery slope, and the recovery duration within adjacent time windows, a reversible boundary criterion is constructed. Based on the reversible boundary criterion, the target strawberry plant is determined to be in a reversible abnormal state, a transitional dangerous state, or an irreversible progressive state. At the same time, the latest treatment time is calculated. When the target strawberry plant is in the transitional danger state, a trial treatment action is applied, a short-term feedback sequence is collected after the treatment, and the damage stress state value and the recovery capacity state value are corrected according to the short-term feedback sequence. The target treatment path is determined according to the correction result and the latest treatment time. Furthermore, the reference rhythm template is constructed according to the growth stage, organ type and diurnal time segment of the target strawberry plant, and the natural state sequence of the target strawberry plant in the current time window is compared with the corresponding reference rhythm template. When the deviation is greater than the preset deviation threshold for multiple consecutive time windows, the corresponding time interval is determined as the period to be inspected.

[0004] Furthermore, the step of applying a small-amplitude trial disturbance to the target strawberry plants during the inspection period and collecting the response sequences before and after the disturbance includes: Based on the deviation direction of the natural state sequence relative to the reference rhythm template, determine the type of micro-amplitude probing disturbance corresponding to the current abnormal deviation type; The micro-amplitude probing disturbance is controlled within a preset safety window, so that the target strawberry plant remains in a recoverable state during the probing process; The response sequence of the target strawberry plant was continuously collected before, during and after the application of the micro-amplitude probing perturbation to obtain the controlled response change corresponding to the micro-amplitude probing perturbation.

[0005] Furthermore, based on the natural state sequence and the controlled response features, harmful gain features and recovery attenuation features are extracted, and the harmful gain features, the recovery attenuation features and the operational disturbance records are input into the long short-term memory neural network to obtain the harmful pressure state value and the recovery capability state value.

[0006] Furthermore, the reversible boundary criterion is constructed based on the coupling relationship between the increment of the harmful pressure state value and the decrease of the recovery capacity state value within adjacent time windows, and combined with the recovery slope and the recovery duration, to determine the trend of the target strawberry plant's current abnormal state transitioning from a reversible abnormal state to a transitional dangerous state or an irreversible progressive state.

[0007] Furthermore, the latest time of treatment is determined based on the growth rate of the harmful stress state value, the decay rate of the recovery ability state value, and the remaining time before the reversible boundary criterion reaches the preset boundary threshold. The latest time of treatment is the latest time point before the target strawberry plant enters the irreversible progression state from the transitional dangerous state.

[0008] Further, the correction of the harmful pressure state value and the recovery capability state value based on the short-time feedback sequence includes: The abnormal fall-off amount and recovery enhancement amount are determined based on the short-time feedback sequence; The harmful pressure state value is corrected based on the abnormal drop amount; The recovery capability status value is corrected based on the recovery enhancement amount.

[0009] Furthermore, the target treatment path is determined based on the corrected harmful pressure state value, the corrected recovery capacity state value, and the latest treatment time; The correction results corresponding to different trial treatment actions or combinations of actions are compared, and the trial treatment action or combination of actions that reduces the harmful pressure state value and increases the recovery ability state value before the latest treatment time is selected as the target treatment path.

[0010] In summary, the present invention has the following beneficial effects: By constructing natural state sequences and reference rhythm templates for target strawberry plants, abnormal deviations can be identified before overt symptoms appear, thus solving the problem that existing technologies cannot detect the latent development process of pests and diseases in a timely manner. By applying a small-amplitude probing disturbance and collecting response sequences during the inspection period, the abnormal state of plants can be transformed from passive observation to active identification, thus solving the problem of delayed early warning caused by existing technologies that rely solely on apparent symptoms or environmental parameters for judgment. By generating damage stress state values ​​and recovery capacity state values ​​based on long short-term memory neural networks, and combining reversible boundary criteria, latest treatment time, and short-term feedback correction to determine the target treatment path, pest and disease early warning and treatment decisions have stronger timeliness and pertinence, thus solving the problem that existing technologies are unable to accurately grasp the timing of treatment and form an effective closed-loop decision. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is an overall flowchart of the strawberry pest and disease control multi-source temporal feature fusion early warning and treatment decision-making method in the embodiments of the present invention; Figure 2 This is a schematic diagram illustrating the principle of identifying latent harmful states and generating harmful pressure state values ​​and recovery capacity state values ​​in an embodiment of the present invention. Figure 3 This is a flowchart illustrating the reversible boundary determination, latest disposal time determination, and disposal decision-making closed loop in an embodiment of the present invention. Detailed Implementation

[0013] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Example

[0014] The following combination Figures 1-3 The present invention will be described in further detail below.

[0015] This invention provides a technical solution: a method for early warning and decision-making regarding strawberry diseases and pests, such as... Figures 1-3 As shown, it includes the following steps: Multi-source time-series data of target strawberry plants are collected. The multi-source time-series data includes natural physiological response data and operation disturbance records. A natural state sequence and a reference rhythm template are constructed based on the multi-source time-series data. The time period to be inspected is determined according to the degree of deviation between the natural state sequence and the reference rhythm template. Under preset safety constraints, a small-amplitude trial disturbance is applied to the target strawberry plant during the inspection period, and the response sequence before and after the disturbance is collected. Based on the natural state sequence, the reference rhythm template, and the response sequence, phase offset, response amplitude difference, recovery slope, and recovery duration are extracted to construct controlled response features; The natural state sequence, the controlled response features, and the operational disturbance records are input into a long short-term memory neural network to obtain the damage stress state value, which characterizes the degree of pathogen infection accumulation and the degree of insect ingestion accumulation, as well as the recovery capacity state value, which characterizes the plant's recovery capacity and organ tolerance capacity. Based on the increment of the harmful pressure state value, the decrease of the recovery ability state value, the recovery slope, and the recovery duration within adjacent time windows, a reversible boundary criterion is constructed. Based on the reversible boundary criterion, the target strawberry plant is determined to be in a reversible abnormal state, a transitional dangerous state, or an irreversible progressive state. At the same time, the latest treatment time is calculated. When the target strawberry plant is in the transitional danger state, a trial treatment action is applied, a short-term feedback sequence is collected after the treatment, and the damage stress state value and the recovery capacity state value are corrected according to the short-term feedback sequence. The target treatment path is determined according to the correction result and the latest treatment time. This embodiment uses a greenhouse for strawberry cultivation as an example. Strawberry plants in the flowering and fruiting stage are selected as monitoring targets. Multiple monitoring areas are divided along the cultivation ridges, with each area containing 1 to 3 adjacent strawberry plants. Environmental data acquisition units, leaf surface condition data acquisition units, volatile matter information acquisition units, and operation recording units are set up within each monitoring area to acquire multi-source time-series data of the target strawberry plants under natural growth conditions. The multi-source time-series data includes at least natural physiological response data and operation disturbance records. Natural physiological response data includes air temperature, air humidity, leaf surface temperature, changes in leaf surface reflectance, changes in local volatile components, and organ morphology changes. Operation disturbance records include information on events such as personnel entry, pruning, harvesting, irrigation, pesticide application, and inspection. During data collection, air temperature and air humidity are collected at 1-minute intervals, leaf surface temperature at 30-second intervals, changes in leaf surface reflectance at 5-minute intervals, changes in volatile components at 2-minute intervals, and changes in organ morphology at 10-minute intervals. Operational events are recorded immediately upon occurrence. After the data collection is completed, all types of data are mapped onto the same time axis to form a multi-source time-series data sequence corresponding to the target strawberry plant.

[0016] In the initial stage of system operation, a reference rhythm template is established using healthy strawberry plants. Specifically, 20 to 30 healthy strawberry plants that have not shown lesions, abnormal insect populations, significant chlorosis, or stunted growth for 7 consecutive days are selected from the same greenhouse. Their multi-source time-series data are categorized and organized according to growth stage, organ type, and diurnal time interval. Growth stage is divided into at least the vegetative growth period, flowering period, and fruit setting period; organ type is divided into at least leaves, flowers, and fruits; and diurnal time interval is divided into at least four segments: 06:00 to 12:00, 12:00 to 18:00, 18:00 to 24:00, and 00:00 to 06:00. For the data of healthy plants within the same category, the normal fluctuation range, normal response amplitude, and normal recovery time within the corresponding time interval are extracted, and a corresponding reference rhythm template is formed based on this. Subsequently, the natural state sequence of the target strawberry plant within the current time window is compared with the corresponding reference rhythm template. When changes in leaf surface temperature, humidity response, leaf surface reflectance, or volatile components deviate from the reference rhythm template for multiple consecutive time windows, the corresponding time interval is determined as the period to be inspected. In this embodiment, when significant deviations occur in three consecutive time windows, a subsequent trial detection process is triggered.

[0017] After determining the inspection period, a slight experimental disturbance is applied to the target strawberry plant. This slight disturbance is set within a safe range that does not disrupt the normal growth of the strawberry, and different disturbance methods are selected based on the type of abnormal deviation. When the target strawberry plant exhibits humidity-related abnormalities, a local ventilation pulse or a local humidity pulse is applied preferentially; when the target strawberry plant exhibits sluggish thermal response, a slight temperature difference pulse is applied preferentially; when the target strawberry plant exhibits abnormal spectral reflectance, a short-duration spectral pulse is applied preferentially. In this embodiment, the wind speed of the local ventilation pulse is controlled between 0.5 m / s and 1.2 m / s, and the duration is controlled between 20 s and 60 s; the local humidity pulse is controlled to increase or decrease the relative humidity by 2% to 5% from the baseline value, and the duration is controlled between 60 s and 180 s; the duration of the short-duration spectral pulse is controlled between 10 s and 40 s; and the slight temperature difference pulse is controlled to cause a local temperature change of 0.3℃ to 1.0℃, and the duration is controlled between 30 s and 120 s. Within 5 minutes before, during, and 20 minutes after the application of the micro-amplitude probing disturbance, the changes in leaf temperature, leaf reflectance, volatile components, and local morphological changes of the target strawberry plant were continuously collected to form a response sequence before and after the disturbance.

[0018] After obtaining the response sequence, the changes in the target strawberry plants before and after the trial disturbance are analyzed to extract controlled response features. These controlled response features include at least whether there is a significant lag in the response, whether the response amplitude is significantly larger or smaller than normal, whether the recovery speed slows down after the disturbance ends, and whether the time required to recover to the baseline range is prolonged. The multi-source temporal changes of the target strawberry plants under natural conditions, the controlled response features after the trial disturbance, and the operational disturbance records within the corresponding time periods are input into a long short-term memory (LSTM) neural network. The LSTM neural network outputs two state results: one state result characterizes the current level of damage stress in the target strawberry plant, reflecting the degree of abnormal accumulation caused by pathogen infection or insect feeding; the other state result characterizes the current level of recovery ability in the target strawberry plant, reflecting the strength of the plant's ability to recover to a normal state after being disturbed.

[0019] After obtaining the damage stress state value and recovery capacity state value, the following factors are further considered in conjunction with the changing trends of damage stress and recovery capacity within adjacent time windows, the recovery speed after the disturbance ends, and the recovery time required to determine whether the current anomaly of the target strawberry plant is still in a reversible stage. When the damage stress continues to rise and the recovery capacity continues to decline, the target strawberry plant is determined to have transitioned from a reversible abnormal state to a transitional dangerous state; when the damage stress further increases, the recovery capacity further decreases, and the recovery time continues to prolong, the target strawberry plant is determined to have entered an irreversible progressive state. Based on this, the latest intervention time is estimated according to the current anomaly evolution rate to determine the final time boundary where effective intervention with low-cost measures can still be carried out.

[0020] When a target strawberry plant is determined to be in a transitional danger state, a trial treatment is applied. This trial treatment can be one or more of the following: localized dehumidification, targeted purging, localized isolation, and removal of abnormal organs. In this embodiment, localized dehumidification is controlled to reduce local air humidity by 3% to 6% within 10 minutes; targeted purging wind speed is controlled between 0.8 m / s and 1.0 m / s, with a duration of approximately 60 seconds; localized isolation is used for short-term isolation of a single plant or fruit; and removal of abnormal organs is used to remove flowers or fruits showing obvious abnormal signs. After the trial treatment, a short-term feedback sequence is collected for 15 to 30 minutes, and the anomaly reduction and recovery enhancement are determined based on this sequence. If the abnormal fluctuations significantly decrease and the recovery speed significantly accelerates after treatment, the damage stress state value is lowered and the recovery capacity state value is raised accordingly; if the abnormal changes are not significant or the recovery speed does not improve sufficiently after treatment, the original state judgment is maintained or the subsequent treatment level is increased.

[0021] After completing the state correction, the system compares multiple candidate treatment actions or combinations of actions by combining the corrected harmful pressure state value, the corrected recovery capacity state value, and the latest treatment time. It selects the treatment method that can reduce the harmful pressure state value and increase the recovery capacity state value before the latest treatment time as the target treatment path. Taking a certain monitoring area as an example, this area showed delayed leaf temperature recovery and abnormal local humidity response for three consecutive time windows. Based on this, the system determined the period to be monitored and applied a local ventilation pulse. After the disturbance ended, the leaf temperature recovery rate in this area was significantly slower than that of a healthy template, and the changes in volatile components did not stabilize within the normal time. The neural network output showed that the harmful pressure level in this area was high and the recovery capacity level was low. Based on this, the system determined that the area was in a transitional danger state and estimated that the latest treatment time was approximately 4 hours from the current time. Subsequently, local dehumidification and targeted purging were implemented in sequence in this area, and short-term feedback sequences continued to be collected. Feedback results indicated that the leaf surface temperature recovered more quickly and abnormal fluctuations significantly decreased. Therefore, the system identified "localized dehumidification combined with targeted purging, and localized isolation of abnormal flowers and fruits" as the target treatment approach. Subsequent follow-up results showed that the area did not continue to progress towards an irreversible state.

[0022] The sensor type, sampling period, time window length, micro-amplitude perturbation parameters, and exploratory treatment action parameters used in this embodiment can all be adjusted according to the greenhouse size, strawberry variety, and cultivation method. Any technical solution that employs the following approach—first constructing a natural state sequence and reference rhythm template for the target strawberry plant, then applying micro-amplitude perturbations and collecting response sequences during the inspection period, followed by using a long short-term memory neural network to obtain the damage stress state value and recovery capacity state value, and determining the latest treatment time and target treatment path based on the trends of these two values—can be used to implement this solution. Example

[0023] like Figures 1-3 As shown, the reference rhythm template is constructed according to the growth stage, organ type and diurnal time segment of the target strawberry plant, and the natural state sequence of the target strawberry plant in the current time window is compared with the corresponding reference rhythm template. When the deviation is greater than the preset deviation threshold for multiple consecutive time windows, the corresponding time interval is determined as the period to be inspected. The process of applying a small-amplitude trial perturbation to the target strawberry plants during the specified inspection period and collecting the response sequences before and after the perturbation includes: Based on the deviation direction of the natural state sequence relative to the reference rhythm template, determine the type of micro-amplitude probing disturbance corresponding to the current abnormal deviation type; The micro-amplitude probing disturbance is controlled within a preset safety window, so that the target strawberry plant remains in a recoverable state during the probing process; The response sequence of the target strawberry plant was continuously collected before, during and after the application of the micro-amplitude probing perturbation to obtain the controlled response change corresponding to the micro-amplitude probing perturbation. Based on the natural state sequence and the controlled response features, harmful gain features and recovery attenuation features are extracted, and the harmful gain features, the recovery attenuation features and the operational disturbance records are input into the long short-term memory neural network to obtain the harmful pressure state value and the recovery capability state value.

[0024] This embodiment uses a strawberry cultivation scenario in a solar greenhouse as an example. Strawberry plants in the flowering and fruiting stage are selected as monitoring targets, and several monitoring units are divided according to the arrangement of the cultivation ridges. Each monitoring unit contains two adjacent strawberry plants. Environmental data acquisition devices, leaf surface condition data acquisition devices, volatile matter information acquisition devices, and operation recording devices are deployed in each monitoring unit to collect information on air temperature, air humidity, leaf surface temperature, changes in leaf surface reflectance, changes in local volatile components, and operational events such as pruning, harvesting, irrigation, pesticide application, and inspection. Air temperature and humidity are collected at 1-minute intervals, leaf surface temperature at 30-second intervals, changes in leaf surface reflectance at 5-minute intervals, and changes in volatile components at 2-minute intervals. Operational events are recorded immediately upon occurrence. After collection, all data are mapped onto the same timeline, and a natural state sequence of the target strawberry plant is generated with a 15-minute time window.

[0025] To construct a reference rhythm template, healthy strawberry plants that showed normal growth, no lesions, abnormal insect populations, significant chlorosis, or organ wilting within a continuous 7-day period were selected from the same greenhouse as healthy samples. These healthy samples were categorized according to growth stage, organ type, and diurnal time period. Growth stage was divided into vegetative growth, flowering, and fruit setting; organ type into leaves, flowers, and fruits; and diurnal time period into four segments: 06:00-12:00, 12:00-18:00, 18:00-24:00, and 00:00-06:00. For healthy samples within the same category, the range of leaf temperature fluctuations, air humidity changes, leaf surface reflectance changes, volatile component changes, and the time required for them to return to a stable state were statistically analyzed, and corresponding reference rhythm templates were formed based on these findings. During system operation, the system first calls up a matching reference rhythm template based on the target strawberry plant's growth stage, the currently monitored organ, and the current time period. Then, it compares the plant's natural state sequence within the current time window with the called reference rhythm template. If changes in leaf temperature, humidity response, leaf reflectance, or volatile components exceed the normal fluctuation range for three consecutive time windows, the corresponding time period is determined as the period to be inspected. In this embodiment, the determination of the period to be inspected is triggered when the overall deviation exceeds 0.32 for three consecutive time windows.

[0026] After determining the time period to be inspected, the system selects the type of micro-amplitude probing disturbance based on the direction of deviation. If the current anomaly is mainly manifested as high air humidity and slow leaf temperature recovery, local ventilation pulses are preferred; if the current anomaly is mainly manifested as delayed leaf temperature changes and abnormal ambient temperature fluctuations, micro-amplitude temperature difference pulses are preferred; if the current anomaly is mainly manifested as abnormal changes in leaf reflectance, short-time spectral pulses are preferred; if the current anomaly is mainly manifested as local humidity anomalies with no significant temperature changes, local humidity pulses are used. To ensure that the trial disturbances do not cause additional damage to the strawberry plants, all micro-level trial disturbances are controlled within a preset safety window. Specifically, the wind speed of the local ventilation pulse is controlled between 0.6 m / s and 1.0 m / s, and the duration is controlled between 30 s and 50 s; the local humidity pulse is controlled to change the relative humidity by 2% to 4% from the baseline value, and the duration is controlled between 60 s and 120 s; the micro-temperature difference pulse is controlled to change the local temperature by 0.5℃ to 0.8℃, and the duration is controlled between 40 s and 90 s; and the duration of the short-time spectral pulse is controlled between 15 s and 30 s. The principle for setting the safety window is that after the trial disturbance ends, the target strawberry plant should be able to return to the baseline state within the normal recovery time range, without irreversible changes such as leaf curling, flower dehydration, or fruit surface damage.

[0027] Five minutes before, during, and 20 minutes after the application of a micro-perturbation, changes in leaf temperature, leaf reflectance, volatile components, and organ appearance of the target strawberry plant were continuously collected to form a response sequence before and after the perturbation. The system extracted controlled response features from this sequence, including the order of response occurrence, the magnitude of response changes, the speed of recovery after the perturbation, and the recovery time required. Specifically, if the target strawberry plant failed to recover to the normal fluctuation range corresponding to the reference rhythm template for a prolonged period after the perturbation, its recovery speed was considered slowed; if its response amplitude was significantly greater than that of a healthy sample under the perturbation, an abnormal amplification response was identified; if its response time significantly lagged behind the response time of the corresponding healthy sample in the reference rhythm template, a phase shift was identified. These controlled response features, together with the natural state sequence, were used to characterize the dynamic change trend of the target strawberry plant under latent damage conditions.

[0028] After extracting the controlled response features, the system further extracts harmful gain features and recovery decay features from the natural state sequence and controlled response features. The harmful gain features characterize whether the abnormal changes continue to intensify during the inspection period, focusing on phenomena such as increased leaf temperature lag, deepening humidity anomalies, increased amplitude of abnormal reflections, and strengthened abnormal fluctuations in volatile components. The recovery decay features characterize whether the plant's recovery ability decreases after the initial disturbance, focusing on phenomena such as slower recovery speed, prolonged recovery time, and decreased stability after recovery. Simultaneously, the operational disturbance records during the inspection period are time-aligned with the above two types of features to reflect the potential impact of operational behaviors such as pruning, harvesting, irrigation, pesticide application, and inspection on the current abnormal state.

[0029] Subsequently, the damage gain features, recovery decay features, and operational disturbance records are input into a long short-term memory (LSTM) neural network. The LSM neural network employs a two-layer temporal structure, with the input time step corresponding to the aforementioned 15-minute time window. The network outputs two state results. The first state result is the damage stress state value, characterizing the degree of abnormal accumulation in the target strawberry plant caused by pathogen infection or insect feeding. The second state result is the recovery capacity state value, characterizing the strength of the target strawberry plant's ability to recover to a normal state after being subjected to a trial disturbance.

[0030] In actual operation, when the causative gain characteristic continues to increase and the recovery attenuation characteristic increases simultaneously, the causative pressure state value output by the system increases, while the recovery capability state value decreases; conversely, when the abnormal changes gradually weaken and the recovery process tends to normalize, the causative pressure state value output by the system decreases, while the recovery capability state value increases.

[0031] Taking a specific monitoring unit as an example, strawberry plants within this unit exhibited delayed leaf temperature recovery and abnormal local humidity fluctuations for three consecutive time windows. Based on this, the system determined this period to be the time to be inspected. Since the anomalies were primarily characterized by humidity deviation and delayed thermal response, the system applied a local ventilation pulse with a wind speed of 0.8 m / s for 40 seconds. During continuous data collection after the disturbance ended, it was found that the leaf temperature recovery rate of this plant was significantly slower than that of a healthy template at the same stage, and the changes in volatile components failed to stabilize again within the normal recovery time. This led to the extraction of high causative gain features and strong recovery attenuation features. After inputting these features, along with the corresponding disturbance records for the time period, into a long short-term memory neural network, the system obtained a high causative stress state value and a low recovery capacity state value for the strawberry plants in this monitoring unit. This indicates that although the plants have not yet shown obvious symptoms, they already possess a strong tendency for latent causative accumulation, providing a basis for subsequent reversible boundary determination and treatment timing.

[0032] In this embodiment, the number of samples, time window length, intensity of micro-amplitude perturbation, duration of response sequence acquisition, and number of layers and input dimensions of the long short-term memory neural network of the reference rhythm template can all be adjusted according to the strawberry variety, greenhouse size, and cultivation mode. As long as the technical solution of constructing the reference rhythm template according to the growth stage, organ type, and diurnal time segment, applying controlled micro-amplitude perturbation during the period to be tested, and obtaining the damage stress state value and recovery capacity state value according to the natural state sequence, controlled response characteristics, and operational perturbation records, this embodiment can be realized. Example

[0033] like Figures 1-3 As shown, the reversible boundary criterion is constructed based on the coupling relationship between the increment of the damage stress state value and the decrease of the recovery ability state value within adjacent time windows, and combined with the recovery slope and the recovery duration, to determine the trend of the target strawberry plant's current abnormal state from a reversible abnormal state to a transitional dangerous state or an irreversible progressive state. The latest time of treatment is determined based on the growth rate of the harmful pressure state value, the decay rate of the recovery ability state value, and the remaining time for the reversible boundary criterion to reach the preset boundary threshold. The latest time of treatment is the latest time point before the target strawberry plant enters the irreversible progression state from the transitional dangerous state. The step of correcting the harmful pressure state value and the recovery capacity state value based on the short-time feedback sequence includes: The abnormal fall-off amount and recovery enhancement amount are determined based on the short-time feedback sequence; The harmful pressure state value is corrected based on the abnormal drop amount; The recovery capability status value is corrected based on the recovery enhancement amount; The target treatment path is determined based on the corrected harmful pressure state value, the corrected recovery capacity state value, and the latest treatment time; Compare the correction results corresponding to different trial treatment actions or combinations of actions, and select the trial treatment action or combination of actions that reduces the harmful pressure state value and increases the recovery ability state value before the latest treatment time, and determine it as the target treatment path; This embodiment is based on Embodiment 2. For the target strawberry plant that has completed reference rhythm template comparison, micro-amplitude trial perturbation application, and long short-term memory neural network state output, the system updates the damage stress state value and recovery capacity state value every 15 minutes, and simultaneously records the average recovery speed and recovery time required within that time window. To determine whether the current abnormal state of the target strawberry plant is still in a reversible stage, the system does not directly rely on the state value at a single moment, but instead combines the damage stress change trend, recovery capacity change trend, and recovery performance after trial perturbation within multiple adjacent time windows to establish a reversible boundary criterion.

[0034] In practice, a continuous judgment unit is formed by the current time window and its three preceding time windows. First, the changes in the harmful stress state value within this continuous judgment unit are compared. If it continuously increases, and the increase in each time window is more than 1.2 times the fluctuation range of the corresponding healthy sample, the target strawberry plant is considered to have an abnormal accumulation trend. Then, the changes in the recovery capacity state value within this continuous judgment unit are compared. If it continuously decreases, and the decrease in each time window is more than 1.1 times the fluctuation range of the corresponding healthy sample, the target strawberry plant is considered to have a recovery decline trend. Simultaneously, the recovery speed and recovery time after each trial disturbance within this continuous judgment unit are combined to determine whether the target strawberry plant can still return to the baseline state within the normal timeframe after being subjected to a minor disturbance. If the recovery speed continues to decrease and the recovery time continuously increases, the judgment level of the current abnormal state shifting towards a high-risk direction is further increased.

[0035] In this embodiment, the abnormal state of the target strawberry plant is divided into three levels: reversible abnormal state, transitional dangerous state, and irreversible progressive state. The specific judgment rules are as follows: when the damage stress state value increases but the upward trend is not obvious, the recovery capacity state value fluctuates slightly, and after the trial disturbance ends, it can return to the baseline fluctuation range within 1.2 times the normal recovery time of a similar healthy sample, it is judged as a reversible abnormal state; when the damage stress state value increases continuously, the recovery capacity state value decreases continuously, and after the trial disturbance ends, the recovery time extends to between 1.2 and 1.8 times the normal recovery time of a similar healthy sample, it is judged as a transitional dangerous state; when the damage stress state value increases rapidly and the recovery capacity state value decreases rapidly, and after the trial disturbance ends, it still cannot return to the baseline fluctuation range even at a time greater than 1.8 times the normal recovery time of a similar healthy sample, it is judged as an irreversible progressive state. To facilitate automatic system execution, this embodiment further sets a comprehensive boundary score as an auxiliary criterion. The comprehensive boundary score is jointly formed by the upward trend of harmful pressure, the downward trend of recovery ability, the degree of abnormality in recovery speed, and the degree of prolongation of recovery time. Among them, the upward trend of harmful pressure and the downward trend of recovery ability have the main weight, while the degree of abnormality in recovery speed and the degree of prolongation of recovery time have the secondary weight. When the comprehensive boundary score is not higher than 0.35, it is judged as a reversible abnormal state; when the comprehensive boundary score is greater than 0.35 but not higher than 0.65, it is judged as a transitional dangerous state; when the comprehensive boundary score is greater than 0.65, it is judged as an irreversible progressive state.

[0036] After determining the state level, the system further determines the latest time for intervention. This latest time is not a pre-set fixed time, but is dynamically estimated based on the abnormal evolution rate of the target strawberry plant. Specifically, the system reads the trends of the damage stress state value and the recovery capacity state value over the last four time windows, and combines this with the remaining space between the current comprehensive boundary score and the irreversible advancement threshold to calculate the remaining time required for the target strawberry plant to reach the irreversible advancement state while maintaining its current evolution trend. To avoid misjudging the intervention time due to occasional noise, this embodiment uses the average rate of change over the last four time windows as the prediction basis. If the estimation result shows that the remaining time is greater than 6 hours, it is determined that there is still a relatively wide intervention window; if the estimation result shows that the remaining time is between 2 and 6 hours, the target strawberry plant is listed as a priority intervention target; if the estimation result shows that the remaining time is less than 2 hours, the target strawberry plant is listed as an emergency intervention target, and the fastest intervention action is prioritized. In this embodiment, when the target strawberry plant is determined to be in a transitional dangerous state and the remaining reversible intervention time is less than 4 hours, the system immediately initiates a trial intervention process.

[0037] During the exploratory treatment phase, the system pre-sets multiple candidate treatment actions and combinations thereof to select the most suitable target treatment path for the current abnormal state based on short-term feedback results. The candidate treatment actions include one or more of the following: localized dehumidification, targeted purging, localized isolation, removal of abnormal organs, and biocontrol spraying. Localized dehumidification is used to rapidly reduce the localized high humidity environment, controlling the relative humidity of the target strawberry area to decrease by 3% to 6% within 10 minutes; targeted purging is used to improve airflow near the leaves, controlling the formation of a localized airflow of 0.8 m / s to 1.0 m / s around the target strawberry plant for a duration between 60 and 120 seconds; localized isolation is used to block direct contact or localized diffusion paths between abnormal organs and surrounding organs; removal of abnormal organs is used to remove obviously abnormal flowers or fruits without affecting the normal growth of the entire plant; and biocontrol spraying is used to apply low-dose biological control agents to a localized area to observe the short-term response of the abnormal state to biocontrol intervention. In this embodiment, in order to avoid the disposal process itself causing new disturbances to the dual-state value judgment, multiple candidate disposal actions are implemented in sequence, with an interval of 20 to 30 minutes between each disposal action. Before implementing a new candidate disposal action, a short-term feedback collection of the previous candidate disposal action should be completed first.

[0038] After each candidate treatment action or combination of actions is applied, the system continues to collect short-term feedback sequences for 15 to 30 minutes. The short-term feedback sequences focus on four types of changes: first, whether the leaf surface temperature recovery rate accelerates; second, whether changes in leaf surface reflectance return to the fluctuation range corresponding to the healthy template; third, whether changes in local volatile components weaken; and fourth, whether the apparent state of the organ changes from abnormal to stable. The system determines the amount of anomaly reduction and recovery enhancement based on these changes. If the damage-related anomalies significantly weaken after treatment, such as a decrease in leaf surface temperature lag, a reduction in abnormal humidity fluctuations, and a reduction in volatile component anomalies, the anomaly reduction is considered significant. If the target strawberry plant returns to the recovery rate and stable state corresponding to the healthy template after treatment, the recovery enhancement is considered significant. Conversely, if the above changes are not significant after treatment, the candidate treatment action is considered to have a weak corrective effect on the current abnormal state. After obtaining the abnormal decline and recovery enhancement, the system corrects the original harmful pressure state value and recovery capacity state value. Specifically, the larger the abnormal decline, the greater the corresponding decrease in the harmful pressure state value; the greater the recovery enhancement, the greater the corresponding increase in the recovery capacity state value. In this embodiment, if within 15 minutes after the implementation of a candidate treatment action, the leaf surface temperature recovery rate increases by more than 20% of the average recovery rate of a healthy template, and the abnormal fluctuation of local volatile components decreases by more than 30% of the original abnormal amplitude, it is considered a significantly effective feedback, and the harmful pressure state value is decreased by 0.08 to 0.15, and the recovery capacity state value is increased by 0.06 to 0.12; if the feedback change after treatment only reaches half of the above standards, it is considered a generally effective feedback, and the harmful pressure state value is decreased by 0.03 to 0.07, and the recovery capacity state value is increased by 0.02 to 0.05; if the feedback change after treatment does not reach half of the above standards, no significant correction is made to the dual state values, or only a minor correction is made. In this way, the system can make closed-loop corrections on the latent damage and recovery ability of the target strawberry plants based on short-term feedback after actual treatment, rather than solely on the prediction results before treatment.

[0039] After completing the dual-state value correction, the system further determines the target treatment path. Specifically, the system compares the correction results corresponding to different candidate treatment actions or combinations of actions, while also considering the time constraint of the latest treatment time. If a candidate treatment action can reduce the harmful pressure state value, but the execution time is long and may exceed the latest treatment time, it will not be selected first. If a candidate treatment action can significantly reduce the harmful pressure state value and significantly increase the recovery ability state value before the latest treatment time, it will be listed as the priority option. If no single candidate treatment action can meet the requirements, the effects of different action combinations in the short-time feedback phase are compared, and the action combination that can be completed within the treatment window and can simultaneously improve the dual-state values ​​is selected as the target treatment path. In this embodiment, the system uses "degree of improvement of dual-state values," "time required to complete the action," and "implementation resource consumption" as the basis for path selection. Among them, the degree of improvement of dual-state values ​​has the highest weight, the time required to complete the action has the second highest weight, and the implementation resource consumption has the lowest weight.

[0040] Taking a target strawberry plant in a monitoring unit as an example, after generating the dual-state value as described in Example 2, the plant exhibited a continuous increase in the damage stress state value, a continuous decrease in the recovery capacity state value, and a continuous extension of the recovery time within four consecutive time windows. Based on this, the system determined that the plant was in a transitional danger state and estimated that the latest treatment time was approximately 3.5 hours from the current time. Subsequently, the system sequentially implemented local dehumidification, fixed-point purging, and a combination of "local dehumidification plus fixed-point purging," and collected short-time feedback sequences for each. The results showed that when only local dehumidification was applied, the leaf surface temperature recovery rate was improved, but the reduction in local volatile abnormal changes was not significant; when only fixed-point purging was applied, humidity fluctuations were reduced, but the recovery rate of leaf surface reflection abnormalities was still slow; while after implementing "local dehumidification plus fixed-point purging," the leaf surface temperature recovery rate was significantly accelerated, humidity abnormal fluctuations were reduced, volatile abnormal changes were weakened, and the apparent state of the organs tended to stabilize. Based on this, the system corrected the dual-state values. After correction, the damage stress state value decreased significantly, while the recovery capacity state value increased significantly. Furthermore, this action combination could be executed before the latest possible intervention time. Therefore, "local dehumidification plus targeted purging, and local isolation of abnormal flowers and fruits" was determined as the target treatment path. Subsequent tracking results showed that the target strawberry plant did not continue to progress towards an irreversible state within the following 6 hours. This indicates that the reversible boundary determination, latest possible intervention time estimation, short-term feedback correction, and target treatment path determination methods used in this embodiment can effectively intervene in the latent development process of strawberry pests and diseases.

[0041] In this embodiment, the reversible boundary scoring threshold, the prediction window length for the latest treatment time, the types of candidate treatment actions, the acquisition duration of the short-term feedback sequence, and the correction range of the dual-state values ​​can all be adjusted according to the strawberry variety, greenhouse structure, climate conditions, and management methods. Any technical solution that establishes a reversible boundary criterion based on the changing trends of the damage stress state value and the recovery capacity state value, and determines the target treatment path based on the latest treatment time, the short-term feedback sequence, and the corrected dual-state values, can be used to implement this embodiment.

[0042] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0043] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the present invention. Various changes and modifications can be made to the present invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed.

Claims

1. A method for early warning and decision-making regarding strawberry diseases and pests, characterized in that, Includes the following steps: Multi-source time-series data of target strawberry plants are collected. The multi-source time-series data includes natural physiological response data and operation disturbance records. A natural state sequence and a reference rhythm template are constructed based on the multi-source time-series data. The time period to be inspected is determined according to the degree of deviation between the natural state sequence and the reference rhythm template. Under preset safety constraints, a small-amplitude trial disturbance is applied to the target strawberry plant during the inspection period, and the response sequence before and after the disturbance is collected. Based on the natural state sequence, the reference rhythm template, and the response sequence, phase offset, response amplitude difference, recovery slope, and recovery duration are extracted to construct controlled response features; The natural state sequence, the controlled response features, and the operational disturbance records are input into a long short-term memory neural network to obtain the damage stress state value, which characterizes the degree of pathogen infection accumulation and the degree of insect ingestion accumulation, as well as the recovery capacity state value, which characterizes the plant's recovery capacity and organ tolerance capacity. Based on the increment of the harmful pressure state value, the decrease of the recovery ability state value, the recovery slope, and the recovery duration within adjacent time windows, a reversible boundary criterion is constructed. Based on the reversible boundary criterion, the target strawberry plant is determined to be in a reversible abnormal state, a transitional dangerous state, or an irreversible progressive state. At the same time, the latest treatment time is calculated. When the target strawberry plant is in the transitional danger state, a trial treatment action is applied, a short-term feedback sequence is collected after the treatment, and the damage stress state value and the recovery capacity state value are corrected based on the short-term feedback sequence. The target treatment path is determined based on the correction result and the latest treatment time.

2. The strawberry pest and disease early warning and treatment decision-making method according to claim 1, characterized in that, The reference rhythm template is constructed according to the growth stage, organ type and diurnal time segment of the target strawberry plant. The natural state sequence of the target strawberry plant in the current time window is compared with the corresponding reference rhythm template. When the deviation is greater than the preset deviation threshold for multiple consecutive time windows, the corresponding time interval is determined as the period to be inspected.

3. The strawberry pest and disease early warning and treatment decision-making method according to claim 2, characterized in that, The process of applying a small-amplitude trial perturbation to the target strawberry plants during the specified inspection period and collecting the response sequences before and after the perturbation includes: Based on the deviation direction of the natural state sequence relative to the reference rhythm template, determine the type of micro-amplitude probing disturbance corresponding to the current abnormal deviation type; The micro-amplitude probing disturbance is controlled within a preset safety window, so that the target strawberry plant remains in a recoverable state during the probing process; The response sequence of the target strawberry plant was continuously collected before, during and after the application of the micro-amplitude probing perturbation to obtain the controlled response change corresponding to the micro-amplitude probing perturbation.

4. The strawberry pest and disease early warning and treatment decision-making method according to claim 3, characterized in that, Based on the natural state sequence and the controlled response features, harmful gain features and recovery attenuation features are extracted, and the harmful gain features, the recovery attenuation features and the operational disturbance records are input into the long short-term memory neural network to obtain the harmful pressure state value and the recovery capability state value.

5. The strawberry pest and disease early warning and treatment decision-making method according to claim 4, characterized in that, The reversible boundary criterion is constructed based on the coupling relationship between the increment of the harmful pressure state value and the decrease of the recovery capacity state value within adjacent time windows. Combined with the recovery slope and the recovery duration, it determines the trend of the target strawberry plant's current abnormal state transitioning from a reversible abnormal state to a transitional dangerous state or an irreversible progressive state.

6. The strawberry pest and disease early warning and treatment decision-making method according to claim 5, characterized in that, The latest time for treatment is determined based on the growth rate of the harmful stress state value, the decay rate of the recovery capacity state value, and the remaining time for the reversible boundary criterion to reach the preset boundary threshold. The latest time for treatment is the latest time point before the target strawberry plant enters the irreversible progression state from the transitional dangerous state.

7. The strawberry pest and disease early warning and treatment decision-making method according to claim 6, characterized in that, The step of correcting the harmful pressure state value and the recovery capacity state value based on the short-time feedback sequence includes: The abnormal fall-off amount and recovery enhancement amount are determined based on the short-time feedback sequence; The harmful pressure state value is corrected based on the abnormal drop amount; The recovery capability status value is corrected based on the recovery enhancement amount.

8. The strawberry pest and disease early warning and treatment decision-making method according to claim 7, characterized in that, The target treatment path is determined based on the corrected harmful pressure state value, the corrected recovery capacity state value, and the latest treatment time; The correction results corresponding to different trial treatment actions or combinations of actions are compared, and the trial treatment action or combination of actions that reduces the harmful pressure state value and increases the recovery ability state value before the latest treatment time is selected as the target treatment path.

Images