Plant physiology monitoring and irrigation control system and method and readable storage medium

Abstract

The invention discloses a plant physiology monitoring and irrigation control system and method and a readable storage medium, and the system employs an ultrasonic sensor to obtain an acoustic emission signal generated by a plant, and carries out the processing of the acoustic emission signal, so as to obtain acoustic characteristics representing the acoustic characteristics of the plant. A multi-modal physiological sensor is adopted to obtain a plant physiological signal and process the plant physiological signal to obtain physiological characteristics representing plant physiological characteristics; and then fusing the acoustic features and the physiological features by using a multi-modal fusion model to obtain a plant feedback index (PFItotal) representing the internal stress state of the plant, predicting a future plant stress trend by using a prediction model, and finally feeding back a stress trend prediction result for controlling an irrigation decision to realize adaptive irrigation adjustment. According to the system, irrigation control can be directly, autonomously and accurately carried out according to physiological needs of plants, early-stage stress detection and control can be achieved several hours before visible symptoms appear, and water consumption can be reduced according to plant types and climate conditions.

Description

A plant physiological monitoring and irrigation control system, method, and readable storage medium Technical Field

[0001] This application relates to the technical fields of plant physiological sensing, artificial intelligence, and smart agriculture. Specifically, it relates to a plant physiological monitoring and irrigation control system, method, and readable storage medium. The system architecture disclosed in this application realizes plant feedback analysis (PFA-1) by integrating acoustic and physiological signals generated by plants to calculate the plant feedback index (PFI_total), so as to achieve early stress state inference and closed-loop control. Background Technology

[0002] In recent years, agricultural irrigation systems and environmental monitoring technologies have undergone significant development. Traditional systems primarily rely on environmental measurements, such as soil moisture, ambient temperature, ambient humidity, or light intensity, to estimate crop water requirements and adjust irrigation timing. These methods form the technological basis for commonly used irrigation controllers, greenhouse management platforms, and climate-adaptive planting systems. While effective to some extent, these systems have a key limitation: they monitor the environment surrounding the plants, relying on indirect environmental measurements rather than signals generated by the plants themselves. Some existing literature attempts to use acoustic or electrical signal measurements for plant monitoring; however, these existing techniques only utilize simple acoustic event detection for irrigation control, such as inferring plant water shortage solely from sound amplitude or the number of events. These existing techniques do not consider high-resolution, plant-specific acoustic characteristics and lack the fusion of multimodal physiological signals.

[0003] Therefore, existing technologies cannot provide accurate, timely, or applicable information that reflects the physiological feedback of plants themselves (such as cavitation events, internal water column tension, vascular stress response, etc.), resulting in low accuracy and reliability of irrigation decisions. Summary of the Invention

[0004] To address the limitations of existing technologies, this application proposes a plant physiological monitoring and irrigation control system, method, and readable storage medium. This system acquires and processes plant acoustic emission signals to obtain acoustic features, and combines this with cross-modal correlation analysis between acoustic and physiological features (electrical, optical, biological, etc.) to identify specific inherent plant stress patterns. Irrigation control is then performed based on the predicted results of these inherent stress patterns, enabling autonomous and precise irrigation control directly based on the plant's physiological needs. Specifically, the system calculates the Plant Feedback Coefficient (PFI_total) as a quantitative plant status indicator obtained from multimodal feature fusion, and utilizes PFI_total time-series prediction for proactive closed-loop irrigation control. Furthermore, the system can achieve early stress detection several hours before visible symptoms appear and can reduce water consumption and save energy based on plant type and climatic conditions.

[0005] This application is achieved through the following technical solution:

[0006] A plant physiological monitoring and irrigation control system, the system comprising:

[0007] The ultrasonic sensing module is configured to: acquire acoustic emission signals from plants and input the acquired acoustic signals into the signal preprocessing module;

[0008] The multimodal physiological sensing module is configured to: acquire plant physiological signals and input the acquired physiological signals into the feature extraction module;

[0009] The signal preprocessing module is configured to: isolate noise from the received acoustic signal, segment the acoustic signal according to a preset time window, and input the segmented acoustic signal into the feature extraction module;

[0010] The feature extraction module is configured to: extract multiple acoustic features for the acoustic signal within each time window, including at least one event-level feature obtained by performing acoustic emission event detection, and form an acoustic feature vector corresponding to each time window; simultaneously extract features from the received physiological signal to obtain physiological features, normalize all the physiological features to form a physiological feature vector, synchronize the physiological feature vector with the acoustic feature vector in time to form a multimodal feature vector, and input the multimodal feature vector to the fusion engine calculation module;

[0011] The fusion engine calculation module is configured to: use a pre-trained multimodal fusion model to perform weighted fusion on the input multimodal features to obtain the plant feedback index PFI_total and form a PFI_total time-series vector, and input the PFI_total time-series vector to the prediction analysis module;

[0012] The predictive analysis module is configured to: use a pre-trained predictive model to generate a predicted value of PFI_total for the prediction window based on the input PFI_total time series vector, and input the predicted value of PFI_total to the closed-loop irrigation controller.

[0013] Furthermore, the closed-loop irrigation controller is configured to: determine the corresponding irrigation decision based on the input PFI_total predicted value, and control the actuator to perform actions according to the irrigation decision to achieve adaptive irrigation adjustment.

[0014] In some embodiments, the ultrasonic sensing module includes:

[0015] At least one broadband ultrasonic microphone or piezoelectric transducer is used to capture plant acoustic emission signals in the frequency range of 20 to 150 kHz.

[0016] An analog front-end amplifier is used to amplify the acoustic emission signals from the plant.

[0017] An anti-aliasing filter is used to filter the amplified acoustic emission signal from the plant.

[0018] And an analog-to-digital converter, used to convert the filtered plant acoustic emission signal into a digitized acoustic signal and input it to the signal preprocessing module.

[0019] In some embodiments, the multimodal physiological sensing module includes at least one of the following sensors: electrobiopotential electrode, optical reflection sensor, chlorophyll fluorescence sensor, thermal infrared sensor, humidity microgradient sensor, and soil or matrix conductivity sensor.

[0020] The electrobiopotential electrode is used to detect slow wave and spike activity related to plant signal transduction;

[0021] The optical reflection sensor is used to detect changes in pigments and moisture related to plant stomatal behavior;

[0022] The chlorophyll fluorescence sensor is used to measure optical efficiency parameters.

[0023] The thermal infrared sensor is used to monitor the temperature of plant leaves or canopy to calculate transpiration, heat stress, and stomatal conductance.

[0024] The humidity micro-gradient sensor is used to measure local vapor pressure deficit;

[0025] The conductivity sensor is used to detect moisture and ion availability.

[0026] In some implementations, the signal preprocessing module uses a bandpass filter to remove background noise from the acoustic signal, and segments the denoised acoustic signal according to a preset time window, then inputs the segmented signal to the feature extraction module for subsequent feature extraction.

[0027] In some embodiments, the signal preprocessing module is further configured to:

[0028] Adaptive denoising is performed using a pre-trained machine learning model to remove persistent non-plant acoustic components;

[0029] The machine learning model includes an autoencoder or a denoising autoencoder trained using background noise samples, wherein the background noise samples include at least one of fan noise, HVAC noise, pump noise, and insect noise.

[0030] In some implementations, the feature extraction module extracts multiple acoustic features for the acoustic signal within each time window, including: pulse rate density, peak amplitude envelope, spectral centroid, spectral bandwidth, harmonic decomposition index, event duration, and event interval.

[0031] The extracted acoustic features are used to form an acoustic feature vector corresponding to each time window and stored to represent the acoustic morphology of each time window;

[0032] The physiological features obtained by the feature extraction module from the physiological signals include at least one of the following features: peak frequency, amplitude variability, and slow wave propagation characteristics obtained by analyzing the signals collected by the electrobiopotential electrode; indicators of leaf pigment composition and water status obtained by analyzing the signals collected by the optical reflection sensor; photosynthetic efficiency indicators obtained by analyzing the signals collected by the chlorophyll fluorescence sensor; temperature difference between the leaf and the environment and spatial temperature patterns obtained by analyzing the signals collected by the thermal infrared sensor to indicate transpiration and stomatal conductance; and features related to microclimate vapor pressure and root zone water and nutrient availability obtained by analyzing the signals collected by the humidity microgradient sensor and the conductivity sensor.

[0033] In some implementations, the feature extraction module performs event detection on the acoustic signal within each time window to identify discrete acoustic emission events and calculate at least one event-level feature, including event rate, event amplitude, event duration, or event interval, and stores the event-level feature as part of the acoustic feature vector.

[0034] In some implementations, the fusion principle of the multimodal fusion model is as follows:

[0035] PFI_total = Σ (Wi × Fi_norm);

[0036] Where Wi represents the weight corresponding to the i-th feature, which is a dynamic adaptive weight trained by a supervised machine learning model; Fi_norm represents the i-th feature, which is a normalized acoustic or physiological feature; PFI_total represents the plant feedback index.

[0037] The dynamic adaptive weights can also be dynamically calibrated based on plant species, developmental stage, sensor drift, and environmental change factors.

[0038] And / or, the prediction model can also classify plant stress states based on feature trajectories and mixed feature patterns, that is, the output of the prediction model includes: the future PFI_total predicted value and the type of plant stress state.

[0039] In some implementations, the system calibrates the dynamic adaptive weights using at least one of the following data:

[0040] (i) The parameter set of the species;

[0041] (ii) Parameter set for the growth stage;

[0042] (iii) Sensor drift estimation;

[0043] And, (iv) environmental compensation characteristics.

[0044] In some implementations, the irrigation decisions made by the closed-loop irrigation controller include:

[0045] If the predicted value of PFI_total is greater than the first threshold, increase irrigation;

[0046] If the PFI_total predicted value remains stable within the second threshold error range, maintain the current irrigation decision;

[0047] If the predicted value of PFI_total is less than the third threshold, reduce or suspend irrigation;

[0048] Wherein, the first threshold is greater than the second threshold, the second threshold is greater than the third threshold, and the first threshold, the second threshold and the third threshold are adaptively adjusted according to the plant species, historical plant performance and environmental changes.

[0049] In some embodiments, the closed-loop irrigation controller is further configured to:

[0050] Apply safety constraints to limit the maximum rate of change in irrigation volume for each control cycle and / or enforce minimum irrigation intervals to reduce oscillations in closed-loop control.

[0051] In some embodiments, the system further includes:

[0052] The relevant module is configured to: calculate the temporal cross-correlation between the acoustic feature changes and the physiological feature changes to obtain mixed correlation features, which are used to optimize stress state classification and improve the accuracy of the prediction model;

[0053] And / or, the event clustering model is configured to group acoustic events into cavitation, embolism, dehydration, mechanical impact, or noise types using supervised or unsupervised learning.

[0054] Secondly, this application proposes a method for plant physiological monitoring and irrigation control, implemented based on the system described in any of the above embodiments, the method comprising:

[0055] Real-time acquisition of plant acoustic and physiological signals;

[0056] According to a preset time window, corresponding feature vectors are extracted from the collected plant acoustic signals to form acoustic feature vectors, and physiological features are extracted from the collected plant physiological signals and normalized to form physiological feature vectors. The physiological feature vectors and the acoustic feature vectors are synchronized in time to form multimodal feature vectors.

[0057] The multimodal feature vectors are input into a pre-trained multimodal fusion model, and the PFI_total time-series vector is generated through the multimodal fusion model.

[0058] The PFI_total time series vector is input into a pre-trained prediction model, and the prediction model generates a PFI_total prediction value of future plant stress trends.

[0059] Based on the PFI_total predicted value of future plant stress trends, corresponding irrigation decisions are generated, and control commands are issued to control the actuators to perform actions in order to carry out adaptive irrigation regulation.

[0060] In some embodiments, the method further includes:

[0061] The multimodal fusion model and / or the prediction model are updated based on newly acquired labeled or pseudo-labeled data to adapt to plant species, growth stage, sensor drift, or seasonal environmental changes.

[0062] Thirdly, this application proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method.

[0063] This application proposes a plant physiological monitoring and irrigation control system. It employs an ultrasonic sensor to acquire and process acoustic emission signals emitted by the plant itself, obtaining acoustic features characterizing the plant's acoustic properties. A multimodal physiological sensor is used to acquire and process physiological signals, obtaining physiological features characterizing the plant's physiological characteristics. Then, a multimodal fusion model is used to fuse the acoustic and physiological features to obtain the Plant Feedback Index (PFI_total), which characterizes the plant's internal stress state. A prediction model is then used to predict the future stress trend of the plant (PFI_total). Finally, irrigation decisions are controlled based on the PFI_total prediction results, achieving adaptive irrigation regulation. This system supports autonomous and precise irrigation control based directly on the plant's physiological needs. It can not only detect early stress hours before visible symptoms appear and adjust accordingly, but also reduce water consumption and save energy based on plant type and climate conditions.

[0064] Furthermore, the system exhibits high robustness, requiring no hardware replacement. It can update and optimize the model simply by using stored sensor characteristic data from different plant species, supporting acoustic and physiological characteristic studies and adaptive irrigation control for various plant species. In addition, the system has a wide range of applications, including irrigation control inside greenhouses, localized outdoor irrigation control, and laboratory research on plant acoustic and physiological characteristics.

[0065] Accordingly, the plant physiological monitoring and irrigation control method and computer-readable storage medium proposed in this application also possess the same technical effects as described above. Attached Figure Description

[0066] The accompanying drawings, which are included to provide a further understanding of the embodiments of this application and form part of this application, do not constitute a limitation on the embodiments of this application. In the drawings:

[0067] Figure 1 is a schematic diagram of the system architecture of the first embodiment of this application;

[0068] Figure 2 is a schematic diagram of the system architecture of the second embodiment of this application;

[0069] Figure 3 is a schematic diagram of the system architecture of the third embodiment of this application;

[0070] Figure 4 is a flowchart of the method proposed in an embodiment of this application;

[0071] Figure 5 is a schematic diagram of the electronic device proposed in an embodiment of this application;

[0072] Figure 6 is a schematic diagram of a computer-readable storage medium proposed in an embodiment of this application.

[0073] Figure reference numerals and corresponding component names:

[0074] 100-System, 101-Ultrasonic sensing module, 102-Multimodal physiological sensing module, 103-Signal preprocessing module, 104-Feature extraction module, 105-Fusion engine computing module, 106-Predictive analysis module, 107-Closed-loop irrigation controller, 108-Related module, 109-Event clustering module, 300-Electronic device, 310-Memory, 320-Processor, 311-Computer program A, 400-Computer-readable storage medium, 411-Computer program B. Detailed Implementation

[0075] In the following, the terms “comprising” or “may include” as used in the various embodiments of this application indicate the presence of a function, operation, or element of the invention and do not limit the addition of one or more functions, operations, or elements. Furthermore, as used in the various embodiments of this application, the terms “comprising,” “having,” and their cognates are intended only to indicate a specific feature, number, step, operation, element, component, or combination of the foregoing and should not be construed as primarily excluding the presence of one or more other features, numbers, steps, operations, elements, components, or combinations of the foregoing, or adding one or more combinations of the foregoing.

[0076] In various embodiments of this application, the expression "or" or "at least one of A and / or B" includes any combination or all combinations of the words listed simultaneously. For example, the expression "A or B" or "at least one of A and / or B" may include A, may include B, or may include both A and B.

[0077] The terms used in the various embodiments of this application (such as "first," "second," etc.) may modify various constituent elements in the various embodiments, but do not limit the corresponding constituent elements. For example, the above terms do not limit the order and / or importance of the elements. The above terms are only used for the purpose of distinguishing one element from other elements. For example, a first user device and a second user device refer to different user devices, although both are user devices. For example, without departing from the scope of the various embodiments of this application, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element.

[0078] It should be noted that if a description is made of "connecting" one component to another, then the first component can be directly connected to the second component, and a third component can be "connected" between the first and second components. Conversely, when a component is "directly connected" to another component, it can be understood that there is no third component between the first and second components.

[0079] The terminology used in the various embodiments of this application is for the purpose of describing particular embodiments only and is not intended to limit the various embodiments of this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which the various embodiments of this application pertain. The terms (such as those defined in a generally used dictionary) are to be interpreted as having the same meaning as in the context of the relevant technical field and are not to be interpreted as having an idealized or overly formal meaning, unless clearly defined in the various embodiments of this application.

[0080] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the embodiments and accompanying drawings. The illustrative embodiments and descriptions of this application are only for explaining this application and are not intended to limit this application.

[0081] Existing technologies mainly achieve irrigation control through indirect environmental sensing or simple acoustic event detection. These methods cannot accurately and promptly assess the internal stress of plants, nor do they comprehensively consider multimodal physiological data, and therefore cannot support adaptive automatic irrigation based on the plant's own signals. To address this, this application proposes a plant physiological monitoring and irrigation control system, as shown in Figure 1. This system 100 includes: an ultrasonic sensing module 101, a multimodal physiological sensing module 102, a signal preprocessing module 103, a feature extraction module 104, a fusion engine calculation module 105, a predictive analysis module 106, and a closed-loop irrigation controller 107.

[0082] The ultrasonic sensing module 101 is used to acquire plant acoustic emission signals and transmit the acquired acoustic signals to the signal preprocessing module 103 for noise isolation and segmentation. In one embodiment, the ultrasonic sensing module 101 includes at least one broadband ultrasonic microphone or piezoelectric transducer capable of capturing plant acoustic emission signals (autonomous acoustic emission signals generated by the plant) in the frequency range of 20~150kHz. The at least one broadband ultrasonic microphone or piezoelectric transducer can be placed within a range of 1~20cm from the plant stem, petiole, or leaf surface, depending on the plant's morphology and sound attenuation characteristics. The ultrasonic sensing module 101 integrates an analog front-end amplifier (e.g., a low-noise amplifier), an anti-aliasing filter, and a high-speed analog-to-digital converter (ADC) with a sampling rate between 250kHz and 2MHz. The plant acoustic emission signals are amplified by the analog front-end amplifier, filtered by the anti-aliasing filter, and finally converted into digital acoustic signals by the high-speed ADC and transmitted to the signal preprocessing module 103.

[0083] The multimodal physiological sensing module 102 includes one or more of the following sensors: electrobiopotential electrode, optical reflection sensor, chlorophyll fluorescence sensor, thermal infrared sensor, humidity microgradient sensor, and soil or matrix conductivity sensor.

[0084] Among them, the electrobiopotential electrode can be attached to the surface of plant leaves or stems to detect slow wave and spike activity related to plant signal transduction, and its output is a voltage fluctuation signal.

[0085] Optical reflectance sensors can operate in one or more bands within the 530–700 nm wavelength range to detect pigment and moisture changes associated with stomatal behavior.

[0086] The chlorophyll fluorescence sensor can measure the maximum photochemical efficiency (Fv / Fm) of photosystem II, non-photochemical quenching (NPQ), or other optical efficiency parameters.

[0087] Thermal infrared sensors can monitor the temperature of plant leaves or canopy to estimate transpiration, heat stress, and stomatal conductance.

[0088] A humidity micro-gradient sensor can be placed near the blade surface to measure local vapor pressure deficit.

[0089] Conductivity sensors can be placed in soil or hydroponic solutions to detect the availability of water and ions.

[0090] These sensors generate a continuous signal stream and transmit it to the feature extraction module 104 for feature extraction and time synchronization. The sampling rate is between 1 Hz and 1 kHz depending on the sensing mode.

[0091] The signal preprocessing module 103 uses a bandpass filter (with a typical operating bandwidth of 20~150kHz) to separate the plant acoustic signal from the background noise. In another embodiment, the signal preprocessing module 103 may also use an adaptive noise reduction autoencoder to remove non-plant acoustic components such as continuous greenhouse mechanical noise (e.g., noise generated by fans, HVAC, water pumps, insects, etc.). This adaptive noise reduction autoencoder is trained on background noise samples (e.g., data on fan, pump, HVAC, and insect sounds). The signal preprocessing module 103 divides the denoised acoustic signal into time windows of 1~20ms and transmits them to the feature extraction module 104 for subsequent feature extraction.

[0092] The feature extraction module 104 includes an ultrasound feature extraction unit and a physiological feature extraction unit. The ultrasound feature extraction unit extracts various acoustic features from the acoustic signal within each time window, including: pulse rate density, peak amplitude envelope, spectral centroid, spectral bandwidth, harmonic fragmentation index, event duration, and event interval. It should be noted that acoustic feature extraction can employ existing time-frequency domain analysis methods, such as time-domain feature extraction, short-time Fourier transform, spectral analysis, and burst signal detection algorithms, which will not be elaborated upon here.

[0093] The extracted acoustic features are used to form acoustic feature vectors corresponding to each time window and stored to represent the acoustic morphology of each time window.

[0094] In another embodiment, the feature extraction module 104 further includes an event detection unit that performs event detection on the acoustic signal within each time window to identify discrete acoustic emission events and calculate at least one event-level feature, including event rate, event amplitude, event duration or event interval, and stores the event-level feature as part of the acoustic feature vector.

[0095] The physiological feature extraction unit is used to extract features from the physiological signals collected by the multimodal physiological sensing module 102 to obtain plant physiological features, which include at least one of the following (1)-(5):

[0096] (1) The peak frequency, amplitude variability and slow wave propagation characteristics are extracted by analyzing the electrobioelectric signals collected by the electrobioelectric potential electrode.

[0097] (2) By analyzing the optical reflection signals collected by the optical reflection sensor, indicators indicating the pigment composition and moisture status of the leaves can be obtained, which may include the normalized difference index or band ratio.

[0098] (3) By analyzing the chlorophyll fluorescence signal collected by the chlorophyll fluorescence sensor, the photosynthetic efficiency index is extracted, including but not limited to the maximum photochemical efficiency of photosystem II, photochemical quenching (qP), non-photochemical quenching and other indicators.

[0099] (4) The temperature difference between the blade and the environment and the space temperature pattern are estimated by analyzing the thermal signals collected by the thermal infrared sensor to indicate transpiration and stomatal conductance.

[0100] (5) By analyzing the humidity signal collected by the humidity micro-gradient sensor and the conductivity signal collected by the conductivity sensor, characteristics related to microclimate vapor pressure and root zone water and nutrient availability are obtained. It should be noted that the analysis methods for the signals collected by the above physiological sensors can all be implemented using existing analysis techniques, and will not be elaborated here.

[0101] All physiological features are normalized to form physiological feature vectors, and time-synchronized with acoustic feature vectors (time alignment, which can be achieved by resampling and other techniques) to form multimodal feature vectors that are input into the fusion engine computing module 105.

[0102] The fusion engine calculation module 105 uses a pre-trained multimodal fusion model to weightedly fuse multimodal features to obtain the plant feedback index (PFI_total) and forms the PFI_total time-series vector, which is then input into the prediction and analysis module 106. The fusion principle is expressed as follows:

[0103] PFI_total = Σ (Wi × Fi_norm);

[0104] Where Wi represents the weight corresponding to the i-th feature, which is a dynamic adaptive weight trained by a supervised machine learning model; Fi_norm represents the i-th feature, which is a normalized acoustic or physiological feature.

[0105] The Plant Feedback Index (PFI_total) can reflect the intrinsic stress state of plants, including dehydration, salt stress, heat imbalance, nutrient deficiency, or mechanical disturbance.

[0106] In one embodiment, the multimodal fusion model includes multiple input feature channels for receiving multimodal feature structured data (normalized representation), at least one fusion layer for fusing the features of the multiple input feature channels into a unified representation, namely the plant feedback index (PFI_total), the fusion layer assigning adjustable weights to different feature channels so that the model can dynamically reflect the relative influence of each feature on the plant's physiological state, and an output layer for outputting PFI_total to generate a PFI_total time-series vector.

[0107] Alternatively, the fusion layer can be implemented using a neural network-based fusion layer, a weighted fusion based on an attention mechanism, or other machine learning structures capable of feature fusion.

[0108] In another embodiment, the fusion engine computing module 105 may further consider factors such as plant species, developmental stage, sensor drift, and environmental changes to dynamically calibrate the weights. Specifically, at least one of the following data can be used to calibrate the dynamically adaptive weights: (i) the parameter set of the species; (ii) the parameter set of the growth stage; (iii) sensor drift estimation; and (iv) environmental compensation features.

[0109] The predictive analysis module 106 utilizes a pre-trained predictive model, such as a recurrent neural network, an autoregressive integral moving average model (ARIMA model), or a gradient boosting sequence model, to generate a predicted PFI_total value for a prediction window (a future time interval) based on the input PFI_total time series vector. This predicted value is then input into the closed-loop irrigation controller 107, thus enabling the prediction of the plant's own stress trends. Optionally, the prediction window can be configured according to the model, ranging from 30 minutes to 24 hours.

[0110] In another embodiment, the prediction analysis module 106 can also use a pre-trained prediction model to distinguish various stress types, such as dehydration, salt stress, heat stress, nutrient imbalance and mechanical disturbance, based on feature trajectories and mixed feature patterns, thereby realizing the classification of plant stress states.

[0111] In one embodiment, the prediction model includes an input layer for receiving the PFI_total time-series vector output by a multimodal fusion model, one or more intermediate layers for modeling nonlinear relationships between features, and an output layer for generating prediction results. Optionally, the output generated by the prediction model includes, but is not limited to: future plant stress trend prediction results (i.e., PFI_total predicted values) and the type of plant stress state.

[0112] The training process for the prediction model and the multimodal fusion model includes:

[0113] Training datasets are obtained from plants under known or controlled conditions. Each training sample includes: multimodal sensor feature data and the corresponding plant stress state or evaluation index.

[0114] The model is trained using the training dataset. During training: model parameters are initialized; training samples are input into the model to generate predicted outputs; the loss value is calculated by comparing the predicted output with the expected output; and the model parameters, including feature weights, are iteratively updated until conditions such as loss convergence or the number of training iterations are reached are met. After training, the optimal model parameters are obtained.

[0115] In another embodiment, the system 100 supports periodic or incremental updates of model parameters using newly collected sensor data, thereby enabling the system 100 to adapt to various plant species, growth stages, sensor drift, or seasonal environmental changes.

[0116] The closed-loop irrigation controller 107 receives the PFI_total predicted value and the selectable stress type, and determines the corresponding irrigation decision accordingly. Based on the irrigation decision, it controls the actuator (such as a solenoid valve, pump, or atomizing nozzle) to perform actions to adjust the irrigation volume, irrigation duration, and frequency.

[0117] In one embodiment, the irrigation decision of the closed-loop irrigation controller 107 includes:

[0118] If the predicted value of PFI_total is greater than the first threshold, increase the irrigation amount or frequency;

[0119] If the PFI_total predicted value remains stable within the second threshold error range, maintain the current irrigation decision;

[0120] If the predicted value of PFI_total is less than the third threshold, reduce or suspend irrigation.

[0121] Among them, the first threshold is greater than the second threshold, the second threshold is greater than the third threshold, and the specific values ​​of the first threshold, the second threshold and the third threshold can be adaptively adjusted according to the plant species, historical plant performance and environmental changes, so as to achieve continuous optimization of water resource use.

[0122] In another embodiment, the closed-loop irrigation controller further imposes safety constraints to limit the maximum rate of change of irrigation volume in each control cycle and / or enforce a minimum irrigation interval to reduce oscillations in the closed-loop control.

[0123] As shown in Figure 2, in another embodiment, the system 100 further includes:

[0124] The correlation module 108 is used to calculate the temporal cross-correlation between changes in acoustic features and changes in physiological features such as electrical, optical, thermal or humidity features, in order to obtain mixed correlation features. These mixed correlation features are used to optimize stress state classification and improve the accuracy of prediction models.

[0125] As shown in Figure 3, in another embodiment, the system 100 further includes:

[0126] The event clustering module 109 can group acoustic events into cavitation, embolism, dehydration, and mechanical shock types using supervised or unsupervised learning methods, providing interpretable support for agricultural decision-making and research analysis. For example, unsupervised clustering algorithms (such as DBSCAN, K-means, etc.) can be used to cluster acoustic events into categories such as cavitation, embolism, dehydration, mechanical shock, or noise, and cluster labels are used to enhance interpretability and the accuracy of stress classification.

[0127] The system 100 proposed in this application acquires and processes plant acoustic emission signals to obtain acoustic features. It then identifies specific inherent plant stress patterns through cross-modal correlation analysis between these acoustic and physiological features. Based on the prediction results of these inherent stress patterns, it adjusts irrigation, enabling autonomous and precise irrigation control directly based on the plant's physiological needs. Specifically, the system calculates PFI_total as a quantitative plant state index obtained from multimodal feature fusion and uses PFI_total time-series prediction for active closed-loop irrigation control. This system can not only achieve early stress detection and timely adjustment several hours before visible symptoms appear, but also reduce water consumption by 10-45% based on plant type and climate conditions, saving energy.

[0128] In another embodiment, the model is optimized using historical sensor data of various plant species (e.g., tomatoes, tobacco, beans, succulents, crops, etc.) stored in the memory to support multiple plant species.

[0129] In another embodiment, the system 100 can be deployed in a greenhouse, connecting multiple ultrasonic and physiological sensor nodes to a central processing unit to achieve irrigation control within the greenhouse.

[0130] In another embodiment, the system 100 can be integrated into a mobile robot platform that travels along crop rows, collects plant feedback data, and controls localized irrigation areas.

[0131] In another embodiment, the system 100 can be deployed in a laboratory for controlled experimental studies of model plants, such as Arabidopsis thaliana, thereby enabling detailed studies of the acoustic and physiological characteristics of the plants.

[0132] The system 100 proposed in this application has high robustness. It can update and optimize the model simply by using sensor feature data of different plant species stored in the memory. This enables the study of acoustic and physiological characteristics of various plant species and adaptive irrigation control. The system 100 has a wide range of applications. It can be used for irrigation control in greenhouses, for irrigation control in local outdoor areas, and for laboratory research on the acoustic and physiological characteristics of plants.

[0133] In another embodiment, this application also proposes a method for plant physiological monitoring and irrigation control, as shown in Figure 4. The method includes the following steps:

[0134] Step 201: Real-time acquisition of plant acoustic signals and plant physiological signals;

[0135] Step 202: Extract corresponding acoustic features from the collected plant acoustic signals according to the preset time window to form an acoustic feature vector, extract physiological features from the collected plant physiological signals and normalize them to form a physiological feature vector, and synchronize the physiological feature vector with the acoustic feature vector in time to form a multimodal feature vector.

[0136] Step 203: Input the multimodal feature vector into the pre-trained multimodal fusion model to generate the PFI_total time series vector;

[0137] Step 204: Input the PFI_total time series vector into the pre-trained prediction model to generate the PFI_total predicted value of future plant stress trends.

[0138] Step 205: Based on the PFI_total predicted value of future plant stress trends, generate corresponding irrigation decisions and issue control commands to control the actuators to perform actions for adaptive irrigation regulation.

[0139] Optionally, step 201 can utilize the ultrasonic sensing module 101 and the multimodal physiological sensing module 102 of the system 100 to achieve real-time acquisition of plant acoustic signals and plant physiological signals.

[0140] Optionally, before performing acoustic feature extraction, step 202 requires preprocessing of the acquired acoustic signal, such as noise reduction and segmentation, which can be performed using the preprocessing method described in the signal preprocessing module 103 of the system 100.

[0141] Optionally, the extraction of acoustic and physiological features in step 202 is implemented in the same way as described in the feature extraction module 104 of the system 100 above, and will not be repeated here.

[0142] Optionally, the models used in steps 203 and 204 are the multimodal fusion model and prediction model described in system 100 above, and their structure and training methods are the same as those described in system 100 above, and will not be repeated here.

[0143] Optionally, the irrigation decision used in step 205 is the same as the irrigation decision described in the closed-loop irrigation controller 107 of the system 100 above, and will not be repeated here.

[0144] In another embodiment, the method proposed in this application further includes:

[0145] A pre-trained prediction model is used to classify plant stress states based on feature trajectories and mixed feature patterns. The classification results include, but are not limited to, dehydration, salt stress, heat stress, nutrient imbalance, and mechanical disturbance.

[0146] In another embodiment, the method proposed in this application further includes:

[0147] The parameters of the model (multimodal fusion model and / or predictive model) are updated periodically or incrementally using newly collected sensor data to adapt to various plant species, growth stages, sensor drift or seasonal environmental changes.

[0148] In another embodiment, the method proposed in this application further includes:

[0149] The temporal cross-correlation between changes in acoustic characteristics and changes in physiological characteristics such as electrical, optical, thermal, or humidity characteristics is calculated to obtain mixed correlation features. These mixed correlation features are used to optimize the classification of plant stress states and improve the accuracy of prediction models.

[0150] In another embodiment, the method proposed in this application further includes:

[0151] Acoustic events can be grouped into cavitation, embolism, dehydration, and mechanical shock types using supervised or unsupervised learning methods, providing interpretable support for agricultural decision-making and research analysis. For example, unsupervised clustering algorithms (such as DBSCAN and K-means) can be used to cluster acoustic events into categories such as cavitation, embolism, dehydration, mechanical shock, or noise, with cluster labels used to enhance interpretability and the accuracy of stress classification.

[0152] In another embodiment, this application also proposes an electronic device, as shown in FIG5. The electronic device 300 includes: a memory 310, a processor 320, and a computer program A311 stored in the memory 310 and executable on the processor 320. When the processor 320 executes the computer program A311, it performs the following steps:

[0153] Real-time acquisition of plant acoustic and physiological signals;

[0154] According to the preset time window, the corresponding feature vectors are extracted from the collected plant acoustic signals to form acoustic feature vectors, and physiological features are extracted from the plant physiological signals and normalized to form physiological feature vectors. The physiological feature vectors and acoustic feature vectors are synchronized in time to form multimodal feature vectors.

[0155] The multimodal feature vectors are input into a pre-trained multimodal fusion model to generate the PFI_total time series vector;

[0156] The PFI_total time series vector is input into a pre-trained prediction model to generate PFI_total predictions of future plant stress trends.

[0157] Based on the PFI_total predicted value of future plant stress trends, corresponding irrigation decisions are generated, and control commands are issued to control the actuators to perform actions for adaptive irrigation regulation.

[0158] Optionally, when the processor 320 executes the computer program A311, it can implement any of the embodiments in the corresponding examples of the above-described plant physiological monitoring and irrigation control method.

[0159] It should be noted that the electronic device proposed in this application embodiment is a device used to implement the above-mentioned plant physiological monitoring and irrigation control method. Therefore, based on the above-mentioned plant physiological monitoring and irrigation control method proposed in this application embodiment, those skilled in the art can understand the specific implementation method and various variations of the electronic device in this application embodiment. Therefore, the specific implementation method of the above-mentioned plant physiological monitoring and irrigation control method will not be described in detail here. Any electronic device used by those skilled in the art to implement the above-mentioned plant physiological monitoring and irrigation control method falls within the scope of protection of this application.

[0160] In another embodiment, this application also proposes a computer-readable storage medium, as shown in FIG6, on which a computer program B411 is stored, which, when executed by a processor, performs the following steps:

[0161] Real-time acquisition of plant acoustic and physiological signals;

[0162] According to the preset time window, the corresponding feature vectors are extracted from the collected plant acoustic signals to form acoustic feature vectors, and physiological features are extracted from the plant physiological signals and normalized to form physiological feature vectors. The physiological feature vectors and acoustic feature vectors are synchronized in time to form multimodal feature vectors.

[0163] The multimodal feature vectors are input into a pre-trained multimodal fusion model to generate the PFI_total time series vector;

[0164] The PFI_total time series vector is input into a pre-trained prediction model to generate PFI_total predictions of future plant stress trends.

[0165] Based on the PFI_total predicted value of future plant stress trends, corresponding irrigation decisions are generated, and control commands are issued to control the actuators to perform actions for adaptive irrigation regulation.

[0166] Optionally, when the computer program B411 is executed by the processor, it can implement any of the embodiments corresponding to the above-described plant physiological monitoring and irrigation control method.

[0167] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0168] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0169] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more blocks of the flowchart illustrations and / or one or more blocks of the block diagrams.

[0170] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and / or one or more block diagrams.

[0171] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide steps for implementing the functions specified in one or more flowcharts and / or one or more block diagrams.

[0172] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of this application. It should be understood that the above description is only a specific embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A plant physiological monitoring and irrigation control system, characterized in that, The system (100) includes: an ultrasonic sensing module (101) configured to: acquire plant acoustic emission signals and input the acquired acoustic signals to a signal preprocessing module (103); a multimodal physiological sensing module (102) configured to: acquire plant physiological signals and input the acquired physiological signals to a feature extraction module (104); a signal preprocessing module (103) configured to: isolate noise from the received acoustic signals, segment the acoustic signals according to a preset time window, and input the segmented acoustic signals to the feature extraction module (104); and a feature extraction module (104) configured to: extract multiple acoustic features for the acoustic signals within each time window, including at least one event-level feature obtained by performing acoustic emission event detection, to form an acoustic feature vector corresponding to each time window, and simultaneously extract features from the received physiological signals to obtain physiological features, and normalize all the physiological features to form a physiological feature vector, and input the physiological feature vector... The acoustic feature vector is synchronized with the time to form a multimodal feature vector, and the multimodal feature vector is input to the fusion engine calculation module (105); the fusion engine calculation module (105) is configured to: use a pre-trained multimodal fusion model to perform weighted fusion on the input multimodal features to obtain the plant feedback index PFI_total and form a PFI_total time-series vector, and input the PFI_total time-series vector to the prediction analysis module (106); the prediction analysis module (106) is configured to: use a pre-trained prediction model to generate the PFI_total prediction value of the prediction window according to the input PFI_total time-series vector, and input the PFI_total prediction value to the closed-loop irrigation controller (107); and the closed-loop irrigation controller (107) is configured to: determine the corresponding irrigation decision according to the input PFI_total prediction value, and control the actuator to perform actions according to the irrigation decision to achieve adaptive irrigation adjustment.

2. The plant physiological monitoring and irrigation control system according to claim 1, characterized in that, The ultrasonic sensing module (101) includes: at least one broadband ultrasonic microphone or piezoelectric transducer for capturing plant acoustic emission signals in the frequency range of 20~150kHz; an analog front-end amplifier for amplifying the plant acoustic emission signals; an anti-aliasing filter for filtering the amplified plant acoustic emission signals; and an analog-to-digital converter for converting the filtered plant acoustic emission signals into digitized acoustic signals and inputting them to the signal preprocessing module (103).

3. The plant physiological monitoring and irrigation control system according to claim 1, characterized in that, The multimodal physiological sensing module (102) includes at least one of the following sensors: an electrobioelectric electrode, an optical reflection sensor, a chlorophyll fluorescence sensor, a thermal infrared sensor, a humidity microgradient sensor, and a soil or substrate conductivity sensor; the electrobioelectric electrode is used to detect slow wave and peak activity related to plant signal transduction; the optical reflection sensor is used to detect pigment and water changes related to plant stomatal behavior; the chlorophyll fluorescence sensor is used to measure optical efficiency parameters; the thermal infrared sensor is used to monitor plant leaf or canopy temperature to calculate transpiration, heat stress, and stomatal conductance; the humidity microgradient sensor is used to measure local vapor pressure deficit; and the conductivity sensor is used to detect water and ion availability.

4. The plant physiological monitoring and irrigation control system according to claim 1, characterized in that, The signal preprocessing module (103) uses a bandpass filter to remove background noise from the acoustic signal and segments the denoised acoustic signal according to a preset time window. The segmented signal is then input to the feature extraction module (104) for subsequent feature extraction.

5. A plant physiological monitoring and irrigation control system according to claim 4, characterized in that, The signal preprocessing module (103) is further configured to perform adaptive denoising using a pre-trained machine learning model to remove persistent non-plant acoustic components; wherein the machine learning model includes an autoencoder or a denoising autoencoder trained using background noise samples, the background noise samples including at least one of fan noise, HVAC noise, pump noise and insect noise.

6. A plant physiological monitoring and irrigation control system according to claim 3, characterized in that, The feature extraction module (104) extracts multiple acoustic features for the acoustic signal within each time window, including: pulse rate density, peak amplitude envelope, spectral centroid, spectral bandwidth, harmonic decomposition index, event duration, and event interval. The extracted acoustic features form an acoustic feature vector corresponding to each time window and are stored to represent the acoustic morphology of each time window. The physiological features obtained by the feature extraction module (104) from the physiological signal include at least one of the following features: peak frequency, amplitude variability, and slow wave propagation characteristics obtained by analyzing the signals collected by the electrobiopotential electrode; indicators indicating leaf pigment composition and water status obtained by analyzing the signals collected by the optical reflection sensor; photosynthetic efficiency indicators obtained by analyzing the signals collected by the chlorophyll fluorescence sensor; temperature difference between the leaf and the environment and spatial temperature patterns obtained by analyzing the signals collected by the thermal infrared sensor to indicate transpiration and stomatal conductance; and features related to microclimate vapor pressure and root zone water and nutrient availability obtained by analyzing the signals collected by the humidity microgradient sensor and the conductivity sensor.

7. A plant physiological monitoring and irrigation control system according to claim 1, characterized in that, The feature extraction module (104) performs event detection on the acoustic signal within each time window to identify discrete acoustic emission events and calculate at least one event-level feature, including event rate, event amplitude, event duration or event interval, and stores the event-level feature as part of the acoustic feature vector.

8. A plant physiological monitoring and irrigation control system according to any one of claims 1-7, characterized in that, The fusion principle of the multimodal fusion model is: PFI_total = Σ (Wi × Fi_norm); where Wi represents the weight corresponding to the i-th feature, which is a dynamically adaptive weight trained by a supervised machine learning model; Fi_norm represents the i-th feature, which is a normalized acoustic or physiological feature; PFI_total represents the plant feedback index; the dynamically adaptive weight can also be dynamically calibrated according to plant species, developmental stage, sensor drift, and environmental change factors; and / or, the prediction model can also classify plant stress states according to feature trajectories and mixed feature patterns, that is, the output of the prediction model includes: the future PFI_total prediction value and the type of plant stress state.

9. A plant physiological monitoring and irrigation control system according to claim 8, characterized in that, The dynamic adaptive weights are calibrated using at least one of the following data: (i) a parameter set of the species; (ii) a parameter set of the growth stage; and (iii) sensor drift estimation. And, (iv) environmental compensation characteristics.

10. A plant physiological monitoring and irrigation control system according to claim 8, characterized in that, The irrigation decisions adopted by the closed-loop irrigation controller (107) include: increasing irrigation if the predicted PFI_total value is greater than a first threshold; maintaining the current irrigation decision if the predicted PFI_total value remains stable within the error range of a second threshold; and reducing or suspending irrigation if the predicted PFI_total value is less than a third threshold. Wherein, the first threshold is greater than the second threshold, the second threshold is greater than the third threshold, and the first threshold, the second threshold, and the third threshold are adaptively adjusted according to plant species, historical plant performance, and environmental changes.

11. A plant physiological monitoring and irrigation control system according to claim 10, characterized in that, The closed-loop irrigation controller (107) is also configured to: apply safety constraints to limit the maximum rate of change of irrigation volume in each control cycle and / or enforce a minimum irrigation interval to reduce oscillations in the closed-loop control.

12. A plant physiological monitoring and irrigation control system according to claim 8, characterized in that, The system (100) further includes: a correlation module (108) configured to: calculate the temporal cross-correlation between the acoustic feature changes and the physiological feature changes to obtain mixed correlation features, the mixed correlation features being used to optimize stress state classification and improve the accuracy of the prediction model; and / or an event clustering model (109) configured to: group acoustic events into cavitation type, embolism type, dehydration type, mechanical impact type or noise type using supervised or unsupervised learning methods.

13. A method for monitoring plant physiology and controlling irrigation, characterized in that, The method is implemented based on the system (100) according to any one of claims 1-12, comprising: real-time acquisition of plant acoustic signals and plant physiological signals; extraction of corresponding feature vectors from the acquired plant acoustic signals according to a preset time window to form an acoustic feature vector, and extraction of physiological features from the acquired plant physiological signals and normalization to form a physiological feature vector; time synchronization of the physiological feature vector and the acoustic feature vector to form a multimodal feature vector; input of the multimodal feature vector into a pre-trained multimodal fusion model to generate a PFI_total time-series vector; input of the PFI_total time-series vector into a pre-trained prediction model, and generation of a PFI_total predicted value of future plant stress trends through the prediction model; generation of corresponding irrigation decisions based on the PFI_total predicted value of future plant stress trends, and issuance of control commands to control the actuator to perform actions according to the irrigation decisions, so as to perform adaptive irrigation adjustment.

14. The plant physiological monitoring and irrigation control method according to claim 13, characterized in that, The method further includes updating at least one parameter of the multimodal fusion model and / or the prediction model based on newly acquired labeled or pseudo-labeled data to adapt to plant species, growth stage, sensor drift, or seasonal environmental changes.

15. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of claim 13 or 14.

Images