Plant growth environment data processing method based on multi-modal sensing and deep learning

Abstract

The invention discloses a plant growth environment data processing method based on multi-modal sensing and deep learning, and relates to the technical field of data processing. The plant growth environment data processing method based on multi-modal sensing and deep learning comprises the following steps: S1, collecting multivariate environment data of a plant growth environment state, normalizing the multivariate environment data, and uploading the normalized multivariate environment data to a multivariate environment database; s2, performing signal enhancement and anomaly identification analysis on the plant physiological gas-electricity data; s3, performing growth environment state analysis by fusing the plant physiological gas-electricity data and the environment data; s4, performing abnormal risk assessment according to the growth environment state analysis result; s5, realizing unified archiving, dynamic visualization and intelligent tracing management of the multi-source environment data; the problems that in the plant growth process, physiological gas-electricity data is prone to being covered, correlation analysis is not accurate, and environment regulation and control intelligence is insufficient are solved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a method for processing plant growth environment data based on multimodal sensing and deep learning. Background Technology

[0002] With the development of precision agriculture, intelligent sensing, and digital planting management, plant growth environment monitoring and intelligent control systems are of great significance for improving the efficiency of modern agricultural production and crop quality. Most existing technologies rely on single monitoring and conventional automatic control of physical environmental data such as temperature, humidity, light, carbon dioxide, and soil moisture.

[0003] For example, the invention patent publication number CN118839248B discloses a reinforcement learning-based method and system for designing carbon-fixing plant combinations, relating to the field of data processing technology. It addresses the problem that when a garden experiences sudden rainfall, some plant combinations absorb water to accelerate carbon fixation, while others may suffer root erosion, causing trees to fall and lose their original carbon-fixing effect. The method includes selecting plant combinations with well-developed root systems or those with high-efficiency carbon-fixing capabilities based on the carbon-fixing needs of different locations in the garden; collecting planting feedback data and corresponding environmental data on sudden events at different locations; comprehensively analyzing the planting feedback data and environmental data to determine whether the plant combination needs adjustment; and further determining whether the plant combination needs to be replaced based on the collected planting feedback data and environmental data.

[0004] For example, the invention patent publication number CN118690143B discloses a multi-dimensional data processing method and system for a cloud-based planting cabin based on big data, involving the field of data processing technology. The method includes: collecting environmental parameters and plant growth parameters; performing time-frequency domain analysis; extracting multi-scale dynamic features; adaptively decomposing environmental parameters; generating eigenfunction models; evaluating data; identifying data anomalies and performing combined modifications to obtain heterogeneous data feature vectors; uploading the data to a cloud server and constructing a high-dimensional data model; performing feature fusion; generating growth state representation vectors and performing nonlinear mapping; constructing a dynamic correlation model and performing serialization modeling; constructing a plant growth prediction model and performing model transfer; solving the model using a multi-objective algorithm to obtain an initial environmental parameter configuration strategy; sending the data to a control module for environmental adjustment; collecting monitoring data and extracting key event features; updating hyperparameters; generating an optimized environmental parameter configuration strategy; and constructing a plant growth knowledge base.

[0005] However, these methods generally suffer from problems such as limited perception dimensions, difficulty in data fusion, and insufficient intelligent analysis. They lack high-sensitivity acquisition and anomaly identification of physiological gaseous signals such as ethylene, carbon monoxide gas, plant microcurrents, and leaf potential, making it difficult to accurately capture key physiological changes and early stresses.

[0006] Therefore, in order to address the above problems, there is an urgent need for a method for processing plant growth environment data based on multimodal sensing and deep learning. Summary of the Invention

[0007] Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a plant growth environment data processing method based on multimodal sensing and deep learning, which solves the problems of easily masked physiological and electrical data during plant growth, inaccurate correlation analysis, and insufficient intelligent environmental regulation.

[0008] Technical solution To achieve the above objectives, this invention provides a plant growth environment data processing method based on multimodal sensing and deep learning, comprising: S1, collecting multivariate environmental data on plant growth environment status, normalizing the multivariate environmental data, and uploading it to a multivariate environmental database; S2, performing signal enhancement and anomaly identification analysis on plant physiological and electrical data, and optimizing sampling strategies and data processing based on the anomaly identification analysis results; S3, performing growth environment status analysis by fusing plant physiological and electrical data and environmental data, and dynamically adjusting adaptive closed-loop control measures for environmental parameters based on the growth environment status analysis results; S4, conducting anomaly risk assessment based on the growth environment status analysis results, and intelligently formulating and executing graded response and precise control strategies based on the risk level; S5, achieving unified archiving, dynamic visualization, and intelligent traceability management of multi-source environmental data. This method solves the problems of easily masked physiological and electrical data during plant growth, inaccurate correlation analysis, and insufficient intelligent environmental control.

[0009] Furthermore, the specific process for collecting multivariate environmental data on plant growth conditions is as follows: Multivariate environmental data is collected, including environmental data and plant physiological and electrophysiological data. Environmental data includes: temperature, humidity, light intensity, carbon dioxide concentration, and soil moisture content. Plant physiological and electrophysiological data includes: ethylene concentration, trace carbon monoxide, plant microcurrent, and leaf potential. Temperature data is acquired using a temperature sensor; humidity data is acquired using a resistive humidity sensor; light intensity data is acquired using a photodiode-type light sensor; carbon dioxide concentration data is acquired using a non-dispersive infrared carbon dioxide sensor; soil moisture content is acquired using a soil capacitive moisture probe; ethylene concentration is acquired using an electrochemical ethylene gas sensor; trace carbon monoxide is obtained using an electrochemically based carbon monoxide gas sensor; plant microcurrent is measured using a high-impedance current amplifier module with non-polar silver chloride electrode patches placed on the plant stem; and leaf potential is continuously collected using a dual-electrode implantation method combined with a high-input-impedance differential amplifier.

[0010] Furthermore, the specific process of uploading the normalized multi-modal environmental data to the multi-modal environmental database is as follows: all acquisition nodes establish communication connections with the data acquisition master control unit through industrial bus wired communication technology and low-power wide-area wireless communication technology. Each data point is accompanied by a timestamp and spatial location information. Using the master control clock signal, the asynchronous acquisition modal data are uniformly aligned to the standard time series through interpolation and resampling to eliminate data mismatch caused by time difference. For raw data with abrupt changes, packet loss, and distortion, sliding window absolute median difference, piecewise linear extremum removal, statistical rules, and outlier filtering are used to improve data reliability. For modal data with different physical dimensions, quantile mapping is used to map to the standard interval to achieve standardization and obtain standardized data. All processed multimodal standard data are synchronously archived in the multi-modal environmental database to provide a high-quality data foundation for subsequent intelligent analysis and control.

[0011] Furthermore, the specific process of signal enhancement and anomaly identification analysis of plant physiological electrophysiological data is as follows: The k-th plant physiological electrophysiological data at time t is acquired; the high-frequency component values ​​of the plant physiological electrophysiological data are extracted using a wavelet transform algorithm; first-order difference is performed to obtain the instantaneous changes in the plant physiological electrophysiological data; the mean and standard deviation of the plant physiological electrophysiological data within the sliding window are calculated; the deviation value of the plant physiological electrophysiological data is obtained by subtracting the mean from the real-time plant physiological electrophysiological data, taking the absolute value, and then dividing by the standard deviation; the temporal features of the enhanced plant physiological electrophysiological data and the observed environmental data are extracted as data input. These temporal features include: difference, moving average, and high frequency components. Frequency components are selected through correlation analysis and feature selection to identify key features highly correlated with plant growth status. A plant-sensory environment feature optimization model is constructed by jointly using regression algorithms, neural networks, and global optimization methods for modeling and iterative training. The weight parameters are dynamically adjusted to minimize prediction error, outputting high-frequency feature weight factors, instantaneous fluctuation weight factors, and deviation feature weight factors. The high-frequency feature value is obtained by multiplying the high-frequency component value by the high-frequency component value weight coefficient. The instantaneous fluctuation value is obtained by multiplying the instantaneous change of the environmental data observation value by the instantaneous change weight coefficient. The deviation feature value is obtained by multiplying the deviation value of the plant physiological and electrical data by the deviation weight coefficient. The sum of the three terms is then calculated to obtain the plant physiological and electrical data enhancement value.

[0012] Furthermore, based on the results of anomaly identification and analysis of plant physiological and electrical data, the sampling strategy and data processing procedures were optimized as follows: The enhancement value of plant physiological and electrical data was compared with the anomaly detection threshold for determination. When the enhancement value was less than the anomaly detection threshold, the current sampling and monitoring measures were maintained, the current plant physiological and electrical data was included in the multivariate environmental database, and a fluctuation trend report was generated, reducing redundant communication and device wake-up frequency within the sampling interval. When the enhancement value was greater than or equal to the anomaly detection threshold, the sampling frequency of plant physiological and electrical data was increased, noise diagnosis and zero-point calibration were performed on the sensors collecting the plant physiological and electrical data, and the local environmental status was reviewed in conjunction with adjacent strong modal sensors. Ventilation, exhaust, humidification, and supplemental lighting were adjusted in the areas where abnormal data was observed. All abnormal data and handling operations were simultaneously archived in the multivariate environmental database and pushed to the environmental and electrical data fusion process.

[0013] Furthermore, the specific process of analyzing the growth environment status by fusing plant physiological and electrical data and environmental data is as follows: The enhanced values ​​of plant physiological and electrical data are obtained, and environmental data observations are used to advance the environmental and electrical data fusion process; the enhanced value of the k-th plant physiological and electrical index at time t is calculated and multiplied by the corresponding plant physiological and electrical data fusion weight coefficient to obtain the plant physiological and electrical characteristic value; the enhanced value of the environment data at time t is calculated and multiplied by the corresponding environment data fusion weight coefficient to obtain the environmental characteristic value; these two types of characteristic values ​​are then weighted and combined with the overall fusion weight factor of plant physiological and electrical data factors and conventional environmental data characteristics to obtain the plant growth environment status value.

[0014] Furthermore, the specific process of adaptive closed-loop control measures to dynamically adjust environmental parameters based on the analysis results of the growth environment status is as follows: Real-time comparison of plant growth environment status values ​​with the lower and upper thresholds for anomaly judgment; When the plant growth environment status value is less than the lower threshold for anomaly judgment, extract the soil moisture of the enhanced value of plant physiological gas-electric data, calculate the soil moisture difference, adjust irrigation amount according to the soil moisture difference, extend daily light duration, coordinately regulate gas and temperature and humidity to rise to the growth range, record intervention parameters in real time, and enter the deep evaluation process; When the plant growth environment status value is greater than or equal to the lower threshold for anomaly judgment and less than or equal to the upper threshold for anomaly judgment, maintain the current inspection mode of each control device; When the plant growth environment status value is greater than the upper threshold for anomaly judgment, activate strong light suppression to reduce the power of supplemental lighting equipment, regulate gas balance through ventilation, activate cooling water curtains to implement temperature and humidity reduction control, simultaneously push early warning, and enter the deep evaluation process.

[0015] Furthermore, anomaly risk assessment is conducted based on the results of the plant growth environment status analysis: abnormal plant growth environment status values ​​are obtained, and a multimodal comprehensive risk mapping value is obtained by combining historical sample data of plant growth environment status values ​​with nonlinear regression; the product of the plant growth environment status value and the linear risk weight factor is calculated, and the product of the multimodal comprehensive risk mapping value and the nonlinear risk weight factor is added to obtain the plant growth abnormality risk value.

[0016] Furthermore, the specific process of intelligently formulating and implementing graded response and precise control strategies based on risk levels is as follows: When the risk value of abnormal plant growth is less than the risk threshold: routine inspections and data archiving are performed, and the current environmental data and plant physiological and electrical data are pushed to the next monitoring cycle to ensure the continuous operation of the monitoring closed loop; when the risk value of abnormal plant growth is greater than or equal to the risk threshold: enhanced values ​​of plant physiological and electrical data from multiple environments are extracted, and the differences in soil moisture content, temperature, air humidity, light intensity, carbon dioxide concentration, ethylene concentration, carbon monoxide concentration, plant microcurrent, and leaf potential are calculated respectively, and targeted controls are implemented accordingly: irrigation and drainage are adjusted based on the soil moisture content difference. The system controls temperature based on temperature differences, humidity based on humidity differences, light intensity based on light intensity differences, carbon concentration based on carbon dioxide concentration differences, and ventilation and gas purification based on ethylene and carbon monoxide concentration differences. For plant microcurrent and leaf potential differences, it retrieves data on soil moisture, temperature, air humidity, light intensity, carbon dioxide concentration, ethylene, and carbon monoxide concentrations, compares and analyzes these data to pinpoint the abnormal causes, activates the control module, allocates control intensity according to the proportion of each parameter difference, and pushes an early warning report containing abnormal difference data, cause analysis, and control measures. Each abnormal handling process, intervention parameters, their effects, and verification records are archived in a multivariate environmental database.

[0017] Furthermore, the specific process for achieving unified archiving, dynamic visualization, and intelligent traceability management of multi-source environmental data is as follows: Multi-source environmental data, enhanced values ​​of plant physiological and electrical data, plant growth environment status values, plant growth abnormality risk values, early warning reports, and manual review records are archived into a multi-source environmental database according to timestamps, spatial locations, and unique object codes; observation curves, fluctuation trend reports, in-depth assessment results, intervention execution history, and response effects of each sensor channel are displayed in real time; administrators monitor the entire process from data collection and risk analysis to control execution.

[0018] Beneficial effects The present invention has the following beneficial effects: (1) This invention achieves a comprehensive and accurate perception of crop growth environment and physiological and electrical data by collecting environmental data and plant physiological and electrical data from multiple sources, covering temperature, humidity, light, carbon dioxide, soil moisture content, ethylene, carbon monoxide, plant microcurrent, and leaf potential. This effectively solves the problem of single monitoring parameters and limited information dimensions in the existing technology.

[0019] (2) The present invention uses wavelet transform, sliding window statistics and feature weight optimization algorithm for plant physiological gaseous and electrical data to enhance the ability to strengthen weak physiological signals and identify anomalies, thereby achieving the effect of high-sensitivity mining of weak physiological signals and effectively solving the problem that weak signals are masked by strong signals and abnormal changes are difficult to identify in the prior art.

[0020] (3) This invention integrates plant physiological and electrical data with environmental data, realizes quantitative assessment of growth environment status through multi-feature weight modeling, and implements adaptive closed-loop precise regulation of environmental parameters based on the analysis results, thereby achieving the effect of deep integration and intelligent control of environmental and physiological information, effectively solving the problems of information fragmentation and extensive regulation in the prior art.

[0021] (4) Based on the analysis results of the growth environment status, the present invention conducts multi-level abnormal risk assessment, formulates response strategies at different levels and implements targeted control measures, thereby realizing the effect of intelligent risk warning and graded response to crop growth environment abnormalities, effectively solving the problems of slow risk response and single control in the existing technology.

[0022] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0023] Figure 1 This is a flowchart of the plant growth environment data processing method based on multimodal sensing and deep learning according to the present invention. Figure 2 This is a quantitative analysis diagram of the comprehensive state of plant growth environment based on the plant growth environment data processing method of the present invention, which is based on multimodal sensing and deep learning. Detailed Implementation

[0024] 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.

[0025] Please see Figures 1-2This invention provides a technical solution: a plant growth environment data processing method based on multimodal sensing and deep learning, comprising: S1, collecting multivariate environmental data on the plant growth environment status, normalizing the multivariate environmental data, and uploading it to a multivariate environmental database; S2, performing signal enhancement and anomaly identification analysis on plant physiological and electrical data, and optimizing sampling strategies and data processing based on the anomaly identification analysis results; S3, performing growth environment status analysis by fusing plant physiological and electrical data and environmental data, and dynamically adjusting adaptive closed-loop control measures for environmental parameters based on the growth environment status analysis results; S4, conducting anomaly risk assessment based on the growth environment status analysis results, and intelligently formulating and executing graded response and precise control strategies based on the risk level; S5, achieving unified archiving, dynamic visualization, and intelligent traceability management of multi-source environmental data. This method solves the problems of easily masked physiological and electrical data during plant growth, inaccurate correlation analysis, and insufficient intelligent environmental control.

[0026] Specifically, the process of collecting multivariate environmental data on plant growth conditions is as follows: Taking Cymbidium goeringii as an example, multivariate environmental data is collected, including environmental data and plant physiological and electrophysiological data. Environmental data covers key indicators such as temperature, humidity, light intensity, carbon dioxide concentration, and soil moisture content. Temperature data of the Cymbidium goeringii planting area is acquired in real time using a temperature sensor; air humidity is monitored using a resistive humidity sensor; light intensity received by Cymbidium goeringii leaves is obtained using a photodiode-type light sensor; carbon dioxide concentration is collected using a non-dispersive infrared carbon dioxide sensor; and soil moisture content around the roots is obtained using a soil capacitive moisture probe. Plant physiological and electrophysiological data includes ethylene concentration, trace carbon monoxide, plant microcurrent, and leaf potential. Specifically, the concentration of ethylene released by Cymbidium goeringii during different growth cycles is accurately collected using an electrochemical ethylene gas sensor; trace carbon monoxide content is monitored in real time using an electrochemically based carbon monoxide gas sensor; non-polar silver chloride electrode patches are placed on the Cymbidium goeringii stem and combined with a high-impedance current amplifier module for continuous measurement of plant microcurrent; and a dual-electrode implantation method combined with a high-input-impedance differential amplifier is used for uninterrupted monitoring of leaf potential. All of the above sensors are deployed at key locations in the Cymbidium orchid growing area, collecting data at regular intervals and uploading it to a multi-environmental database via wired or wireless means to ensure the objectivity, real-time nature and continuity of the data, providing a reliable foundation for the accurate monitoring and intelligent control of the Cymbidium orchid's growing environment and physiological state.

[0027] In this implementation plan, this step achieves comprehensive, real-time, and complete monitoring of the Cymbidium orchid's growth environment and physiological state through automated multi-parameter, multi-channel data acquisition of the orchid's growing environment and the plant itself. By accurately acquiring environmental data such as temperature, humidity, light, carbon dioxide concentration, and soil moisture content, as well as physiological electrophysiological data such as ethylene concentration, trace carbon monoxide, plant microcurrent, and leaf potential, the dimensionality and sensitivity of the perception of the orchid's growth status are significantly improved. All data are collected into a multivariate environmental database, laying a solid, high-quality data foundation for subsequent signal enhancement, fusion analysis, risk assessment, and precise control, effectively supporting intelligent and scientific cultivation and management of Cymbidium orchids.

[0028] Specifically, the process of uploading normalized multivariate environmental data to the multivariate environmental database is as follows: All collection nodes distributed in the Cymbidium orchid planting area establish stable data communication connections with the data acquisition control unit through industrial bus wired communication technology and low-power wide-area wireless communication technology, ensuring efficient collection of large-area, multi-point data. Each collected data is labeled with a timestamp and spatial location information, realizing spatiotemporal holographic tracking of Cymbidium orchid environmental and physiological signals. The clock signal of the control unit is uniformly used as the reference. Through interpolation and resampling, various asynchronously collected environmental and gas-electric modal data are aligned into a standard time series, completely eliminating time difference and data mismatch problems caused by differences in equipment response speed or acquisition frequency. For abnormal raw data with mutations, packet loss, and distortion, a sliding window absolute median difference and piecewise linear extremum removal algorithm combined with statistical rules are used to remove outliers and filter outliers, effectively improving the reliability and consistency of Cymbidium orchid environmental and physiological data throughout its entire life cycle. Based on this, a quantile mapping method is uniformly adopted for modal data of environmental parameters and gas-electric signals with different physical dimensions to normalize the original observation values ​​to a standard interval, thereby achieving standardized processing of multimodal data. All standardized data processed in the above manner will be synchronously archived into a multivariate environmental database, providing a high-quality, structured data foundation for subsequent signal enhancement, feature fusion, risk assessment, and intelligent control of Cymbidium orchid growth status.

[0029] In this implementation plan, this step achieves efficient and standardized preprocessing and unified management of Cymbidium growth environment and physiological data. Through spatiotemporal annotation, time-series alignment, anomaly removal, and standardized normalization processes, the accuracy and consistency of multi-source heterogeneous data are significantly improved. All processed high-quality data are centrally archived into a multi-dimensional environmental database, providing solid data support for subsequent intelligent analysis, feature fusion, and precise control of Cymbidium growth status, effectively ensuring data integrity and reliability.

[0030] Specifically, the process of signal enhancement and anomaly identification analysis of plant physiological electrophysiological data is as follows: First, the physiological electrophysiological data of Cymbidium goeringii at time t, such as ethylene concentration, trace carbon monoxide, plant microcurrent, and leaf potential, are acquired. For each electrophysiological signal, a wavelet transform algorithm is used to extract high-frequency component values ​​to capture sudden abnormal signals that occur during the growth of Cymbidium goeringii. Then, the instantaneous changes in real-time data are calculated using first-order difference to reflect short-term dynamic fluctuation characteristics. The mean and standard deviation of each electrophysiological channel in the local time interval are dynamically calculated using sliding window statistical technology. The deviation value of the physiological signal of Cymbidium goeringii is obtained by subtracting the mean from the real-time electrophysiological observation value, taking the absolute value, and then dividing by the standard deviation, to measure the degree of outlier. The extracted enhanced values ​​of plant physiological electrophysiological data and the temporal features in the environmental data observation values ​​are used as input. The temporal features include difference, moving average, and high-frequency components. Correlation analysis and feature selection algorithms are used to screen out the core features that are highly correlated with the growth status of Cymbidium goeringii. By combining regression algorithms, neural network models, and global optimization methods, a vegetation-sensory environment feature optimization model was established for feature weights and subjected to multiple rounds of iterative training. The optimization objective was to minimize the prediction error of the model, dynamically adjusting the weight parameters of each feature component. The final output included high-frequency feature weight factors, instantaneous fluctuation weight factors, and deviation feature weight factors. Finally, the high-frequency feature value was obtained by multiplying the high-frequency component value by its weight coefficient; the instantaneous fluctuation value was obtained by multiplying the instantaneous change in environmental data observations by its weight coefficient; and the deviation feature value was obtained by multiplying the deviation value of plant physiological and electrical data by its deviation weight coefficient. The sum of these three values ​​yielded the enhanced value of plant physiological and electrical data, laying a data foundation for subsequent identification of abnormal growth in Cymbidium orchids, environmental status assessment, and precise regulation.

[0031] The specific calculation method for the augmentation value of plant physiological atmospheric and electrical data is as follows: ; In the formula, The enhancement value of plant physiological and electrical data is a quantitative result of the sensitivity of plant physiological and electrical data to weak anomalies after signal enhancement processing. It reflects the key dynamic information of plant physiological stress, early lesions and key information that is difficult to be masked by strong signals. This represents the k-th observation of plant physiological and electrical data at time t, specifically the detected values ​​of ethylene concentration, trace carbon monoxide, plant microcurrent, and leaf potential signal. This represents the absolute value of the high-frequency coefficients after wavelet transform of the observed values ​​of plant physiological gaseous and electrical data. It is used to extract instantaneous changes and local features of the signal, reflecting the activity level of weak anomalous signals. The wavelet transform algorithm is used for extraction. It represents the instantaneous change of plant physiological and electrical data observation values. It is directly calculated by the numerical difference between adjacent time points, reflecting the rate of change of plant physiological and electrical data signals in a short period of time and revealing the dynamic fluctuation of weak signals. This represents the deviation of the k-th plant physiological and electrical monitoring data from the mean within the sliding window, reflecting the degree of deviation of the physiological signal from the recent historical fluctuation range at the current moment. This represents the mean and standard deviation of plant physiological and electrical data within a set sliding window, reflecting the local statistical characteristics of the plant physiological and electrical data. It is obtained by dynamically statistically analyzing each plant physiological and electrical data observation value according to the historical data sliding window. This represents the weighted coefficient of the absolute value of the high-frequency coefficients. This represents the weighting coefficient for instantaneous changes in plant physiological atmospheric and electrical data. This represents the weighting coefficient for deviation values, and its value satisfies... , representing the contribution of each feature to the enhancement value of plant physiological and electrical data, is obtained by global search and fitting of plant sensing environment feature optimization model based on multiple rounds of monitoring data in a multivariate environmental database and a joint data-driven approach of genetic algorithm and Bayesian optimization.

[0032] In this implementation plan, this step achieves highly sensitive extraction and quantitative analysis of subtle physiological signal changes by enhancing and identifying multiple features of the plant physiological and electrophysiological data of Cymbidium goeringii. Through wavelet transform, first-order difference, and sliding window statistical methods, sudden anomalies and dynamic fluctuations during the growth process of Cymbidium goeringii can be effectively captured, significantly improving the ability to identify potential stress, early lesions, and key physiological changes. The enhanced electrophysiological signals obtained thus provide crucial foundational data for subsequent environmental condition modeling, risk assessment, and intelligent regulation, achieving efficient integration of Cymbidium goeringii physiological information from raw data collection to intelligent analysis.

[0033] Specifically, the optimization process for sampling strategies and data processing based on the analysis results of abnormal plant physiological and electrical data anomalies is as follows: First, the current enhancement value of plant physiological and electrical data is compared with the anomaly detection threshold for judgment. When the detection result shows that the enhancement value of electrical data is less than the anomaly detection threshold, it indicates that the physiological state of Cymbidium goeringii has no obvious abnormal fluctuations, and the current routine sampling and monitoring measures are maintained. The real-time collected plant physiological and electrical data is incorporated into a multivariate environmental database, and a fluctuation trend report of electrical signals is generated. At the same time, the sampling interval is appropriately extended to reduce redundant communication and equipment wake-up frequency, thereby reducing energy consumption and data processing burden. If the enhancement value of electrical data is greater than or equal to the anomaly detection threshold, it indicates that Cymbidium goeringii may be experiencing physiological stress and abnormal reactions. In this case, the sampling frequency of the signal channel is increased to achieve high-resolution tracking of abnormal signals. At the same time, noise diagnosis and zero-point calibration are performed on the relevant electrical sensors to ensure signal accuracy and reliability. In addition, the temperature, humidity, and carbon dioxide environmental sensors adjacent to Cymbidium goeringii are linked for local status review to detect whether there are synchronous fluctuations. Intelligent linkage of ventilation, exhaust, humidification, and supplemental lighting in abnormal observation areas is implemented to quickly eliminate local stress. All abnormal data and their handling actions are archived into a multi-dimensional environmental database and pushed as key inputs to the subsequent environmental gas and electricity data fusion and intelligent decision-making process, realizing an intelligent closed loop from signal anomaly detection to adaptive control.

[0034] In this implementation plan, intelligent dynamic optimization of sampling strategies and data processing workflows is achieved through real-time comparison of enhanced values ​​of physiological gaseous and electrical data of Cymbidium orchids with anomaly detection thresholds. Based on actual fluctuations in the gaseous and electrical signals, the sampling frequency can be adjusted, sensor status optimized, and environmental control equipment coordinated to enable rapid tracking and precise intervention of physiological anomalies. All key data and processing records are archived in a multi-dimensional environmental database, ensuring an efficient closed-loop process and data traceability for Cymbidium orchid growth management, laying the foundation for subsequent fusion analysis and intelligent decision-making.

[0035] Specifically, the process of analyzing the growth environment status by integrating plant physiological and electrophysiological data and environmental data is as follows: First, the enhanced values ​​of real-time collected Cymbidium goeringii plant physiological and electrophysiological data and various environmental data observation values ​​are acquired and input into the environmental and electrophysiological data fusion analysis process. For each moment, the enhanced values ​​of Cymbidium goeringii physiological and electrophysiological indicators are multiplied by their respective fusion weight coefficients obtained by automatic fitting through historical data, and the results are accumulated to obtain the plant physiological and electrophysiological characteristic values ​​representing the current physiological state. Similarly, the enhanced values ​​of environmental data such as temperature, humidity, light, carbon dioxide, and soil moisture are multiplied by their respective environmental data fusion weight coefficients and accumulated to obtain the current environmental characteristic values ​​of Cymbidium goeringii. The above two types of characteristic values ​​are then weighted and combined with the automatically optimized electrophysiological data fusion weight factors and environmental data fusion weight factors to finally obtain a comprehensive environmental status value that fully reflects the current growth environment and physiological state of Cymbidium goeringii, providing a quantitative analysis basis for subsequent growth status assessment, risk identification, and intelligent regulation.

[0036] The specific calculation method for plant growth environment status values ​​is as follows: ; In the formula, It represents the plant's environmental status value, which comprehensively reflects the overall health level and growth conditions of the current environment and plant status under multiple modes; : The enhancement value of the kth plant physiological gas-electric data; The weighting coefficient for the k-th plant physiological and electrical data fusion represents the importance of each weak signal in multimodal fusion. It is obtained by combining historical plant physiological and electrical data enhancement values ​​with a regression evaluation algorithm. : The i-th regular environmental data observation at time t; The fusion weight coefficient of the i-th regular environment data reflects the contribution of each regular environment data to the overall fusion feature, and is obtained by training and optimization in combination with the feature selection algorithm; The overall fusion weighting factor of plant physiological atmospheric and electrical data factors and conventional environmental data characteristics, with a value range of [value range missing]. And satisfy The results were obtained by further optimizing the data based on feature contribution analysis, objective function setting, cross-validation, and Bayesian optimization processes on a large number of historical monitoring samples. This data reflects the actual contribution of environmental and physiological characteristics to the assessment of plant growth environment status.

[0037] In the first round of multivariate environmental data acquisition, the fusion weighting factor α was 0.45, the fusion weighting factor β was 0.55, the plant physiological and electrical characteristics were 12.5, and the environmental characteristic was 25.3, resulting in a calculated plant growth environment state value of 19.5. In the second round, the fusion weighting factor α was 0.47, the fusion weighting factor β was 0.53, the plant physiological and electrical characteristics were 14.2, and the environmental characteristic was 23.9, resulting in a calculated plant growth environment state value of 19.3. In the third round, the fusion weighting factor α was 0.50, and the fusion weighting factor β was 0.5. 0. Plant physiological and electrical characteristics: 11.7; Environmental characteristics: 26.1; Calculated plant growth environment status: 18.9. In the fourth round, the fusion weighting factor α is 0.52, the fusion weighting factor β is 0.48, the plant physiological and electrical characteristics are 15.6, the environmental characteristics are 27.4; Calculated plant growth environment status: 21.3. In the fifth round, the fusion weighting factor α is 0.42, the fusion weighting factor β is 0.58, the plant physiological and electrical characteristics are 13.3, the environmental characteristics are 24.7; Calculated plant growth environment status: 19.9; In the 6th round, the fusion weight factor α was 0.44, the fusion weight factor β was 0.56, the plant physiological and electrical characteristics were 16.8, the environmental characteristics were 28.2, and the calculated plant growth environment status was 23.2; In the 7th round, the fusion weight factor α was 0.46, the fusion weight factor β was 0.54, the plant physiological and electrical characteristics were 12.1, the environmental characteristics were 22.8, and the calculated plant growth environment status was 17.9; In the 8th round, the fusion weight factor α was 0.49, the fusion weight factor β was 0.51, and the plant physiological and electrical characteristics were 16.8, the environmental characteristics were 28.2, and the calculated plant growth environment status was 23.2; The physiological and electrical characteristic value is 14.9, the environmental characteristic value is 26.5, and the calculated plant growth environment state value is 20.8. In the 9th round, the fusion weight factor α is 0.51, the fusion weight factor β is 0.49, the plant physiological and electrical characteristic value is 17.2, the environmental characteristic value is 29.1, and the calculated plant growth environment state value is 23.0. In the 10th round, the fusion weight factor α is 0.46, the fusion weight factor β is 0.54, the plant physiological and electrical characteristic value is 15.0, the environmental characteristic value is 25.6, and the calculated plant growth environment state value is 20.7.

[0038] Table 1. Data on plant growth environment status: like Figure 2 As shown in Table 1, this is a quantitative analysis diagram of the comprehensive state of plant growth environment provided by a plant growth environment data processing method based on multimodal sensing and deep learning according to an embodiment of this application. Figure 2It can be seen that by dynamically optimizing and integrating weighting factors α and β based on real-time monitoring data, the plant growth environment status values ​​detected in each round better reflect the current comprehensive growth status: the higher the plant growth environment status value corresponding to each detection round, the closer the environment of the Cymbidium goeringii is to the ideal growth range, and the more robust the plant's vitality and metabolic function; values ​​falling into the medium range indicate that the Cymbidium goeringii is experiencing mild stress interference, but has not yet affected its overall health; if the status value shows a continuous decrease, it indicates that the Cymbidium goeringii has potential physiological abnormalities, which should trigger a system warning and timely intervention. By intuitively mapping different value ranges to health status labels of normal, mild stress, and potential disease, the quantitative analysis chart of the comprehensive plant growth environment status can help managers interpret the plant status in real time, scientifically guide environmental regulation, and achieve early detection, early warning, and precise management.

[0039] In this implementation plan, a comprehensive evaluation index capable of quantifying the growth environment and physiological state of Cymbidium goeringii was established by integrating plant physiological and electrophysiological data and environmental data. Various enhanced signals and observational data were subjected to feature extraction and weighted combination based on automatically optimized weights, achieving efficient integration of multi-dimensional physiological and environmental information and real-time state modeling. The resulting growth environment state values ​​provide a scientific data foundation for intelligent monitoring, dynamic evaluation, and precise control of Cymbidium goeringii.

[0040] Specifically, the adaptive closed-loop control measures for dynamically adjusting environmental parameters based on the analysis results of the growth environment status are as follows: The calculated growth environment status value of the Cymbidium goeringii is compared in real time with the set lower and upper thresholds for anomaly detection. When the environmental status value is lower than the lower threshold, it indicates that the Cymbidium goeringii's growth environment is in an unfavorable range. The soil moisture signal is extracted from the enhanced values ​​of plant physiological gaseous and electrical data, and the difference between the current soil moisture and the optimal growth range is dynamically calculated. Based on this, the irrigation system in the Cymbidium goeringii's root zone is precisely adjusted to achieve targeted water replenishment. Simultaneously, the daily light duration is appropriately extended, and gaseous parameters and temperature and humidity are linked for control to ensure the Cymbidium goeringii's environment quickly returns to a suitable range. All intervention parameters are recorded and archived in real time and enter a deep evaluation process to continuously track the recovery of the Cymbidium goeringii. If the growth environment status value is between the lower and upper thresholds for anomaly detection, it indicates that the Cymbidium goeringii is in a normal growth range, and all control equipment maintains its current inspection and routine control mode. If the environmental condition value exceeds the upper limit threshold for anomaly detection, the system will activate strong light suppression measures, such as reducing the power of supplemental lighting equipment, adjusting the gas balance through the intelligent ventilation system, and simultaneously activating the cooling water curtain system to control the temperature and humidity. The system will also push out early warnings for abnormal conditions and promptly enter the next round of in-depth evaluation process to ensure that the growth environment of Cymbidium orchids is always in the optimal state.

[0041] This implementation plan achieves automated closed-loop control based on the analysis results of the Cymbidium orchid's growth environment. It can precisely and dynamically adjust multiple environmental parameters such as irrigation, supplemental lighting, ventilation, and cooling according to real-time changes in comprehensive status values, ensuring that the Cymbidium orchids are always within their optimal growth range. Simultaneously, it records the intervention process and sends out early warnings, realizing intelligent regulation, anomaly response, and full-process tracking of the Cymbidium orchid's growth environment, providing efficient and reliable intelligent management support for continuous healthy growth.

[0042] Specifically, the process of anomaly risk assessment based on the analysis results of the growth environment status is as follows: First, the current abnormal plant growth environment status value is acquired in real time, and sample data of various growth environment status values ​​of Cymbidium goeringii within the historical monitoring period are retrieved. A nonlinear regression method is used to train the data to obtain a multimodal comprehensive risk mapping value that comprehensively reflects the abnormal characteristics of multidimensional environmental and physiological information. The current growth environment status value is multiplied by a data-driven linear risk weighting factor to obtain a linear risk component, reflecting the direct impact of environmental changes on the risk of Cymbidium goeringii. Simultaneously, the multimodal comprehensive risk mapping value is multiplied by a nonlinear risk weighting factor to obtain a nonlinear risk component, comprehensively considering the coupled risk effects of complex physiological environments. The two results are then added together to obtain the current abnormal plant growth risk value of Cymbidium goeringii. The risk assessment results can be used for intelligent graded response and precise regulation, providing scientific risk warnings and decision-making basis for the healthy growth of Cymbidium goeringii.

[0043] The specific calculation method for the risk value of abnormal plant growth is as follows: ; In the formula, This represents the risk value for abnormal plant growth, reflecting the abnormal risks associated with the plant's growth environment and physiological state. It represents the plant's environmental status value, which comprehensively reflects the overall health level and growth conditions of the current environment and plant status under multiple modes; The multimodal comprehensive risk mapping value is based on the nonlinear transformation result of the plant growth environment state value, obtained by combining historical sample data with nonlinear regression, and represents the comprehensive risk probability. This represents a non-linear risk weighting factor. The nonlinear risk weighting factor comprehensively reflects the contribution of linear and nonlinear information to the final risk assessment. The parameters are obtained by adapting on the training and validation sets through an objective function that minimizes the risk prediction error.

[0044] In this implementation plan, this step achieves a quantitative assessment of the growth risk of Cymbidium goeringii by comprehensively modeling the historical and current abnormal information of the growth environment data. By combining linear and nonlinear factors, the abnormal risk value of the current growth of Cymbidium goeringii is accurately calculated, providing a scientific basis for subsequent graded response, early warning push and intelligent control, which greatly improves the risk prevention and control capabilities and safety of Cymbidium goeringii cultivation and management.

[0045] Specifically, the process of intelligently formulating and implementing graded response and precise control strategies based on risk levels is as follows: When the calculated abnormal plant growth risk value is lower than the risk threshold, it indicates that the current growth environment of the Cymbidium orchid is safe and in good condition. Routine equipment inspections and environmental data archiving are then performed, pushing the environmental data and plant physiological and electrical data of the current cycle to the next monitoring cycle, achieving continuous, closed-loop health monitoring. When the detected abnormal risk value of the Cymbidium orchid is greater than or equal to the risk threshold, the latest enhanced values ​​of plant physiological and electrical data from the multivariate environmental database are extracted, and the differences are calculated for various key parameters, including soil moisture content, temperature, air humidity, light intensity, carbon dioxide concentration, ethylene concentration, carbon monoxide concentration, plant microcurrent, and leaf potential. For each difference, precise control measures are implemented: such as adjusting irrigation or drainage systems based on soil moisture differences, intelligently linking heating and / or cooling equipment based on temperature differences, activating humidification or dehumidification measures based on air humidity differences, adjusting supplemental lighting or shading equipment based on light intensity differences, adjusting CO2 supply or ventilation dilution based on carbon dioxide concentration differences, and linking exhaust and gas purification equipment based on ethylene and carbon monoxide concentration differences. For physiological abnormalities in plant microcurrents and leaf potentials, all environmental and physiological signal data are comprehensively retrieved, and the abnormal causes are precisely located through comparative analysis. The control intensity of each parameter is intelligently allocated, and an early warning report containing the difference, cause analysis, and control measures is generated and pushed to the administrator. All abnormal handling processes, intervention parameters and effects, and review records are fully archived in a multi-dimensional environmental database to ensure that the entire process of Cymbidium orchid environmental management is traceable, scientific, and transparent.

[0046] In this implementation plan, this step achieves intelligent graded response and precise control based on the abnormal risk level of Cymbidium orchids. It can analyze the real-time differences of various environmental and physiological parameters and selectively coordinate multiple control devices such as irrigation, ventilation, heating, supplemental lighting, and gas purification to quickly correct abnormalities in the growth environment and physiology of Cymbidium orchids. Through full-process recording and archiving, the scientific nature, traceability, and intelligent closed-loop management of intervention measures are ensured, significantly improving environmental safety and growth protection capabilities during the cultivation of Cymbidium orchids.

[0047] Specifically, the process of achieving unified archiving, dynamic visualization, and intelligent traceability management of multi-source environmental data is as follows: Taking the intelligent cultivation of Cymbidium goeringii as an example, the multi-source environmental data collected from various monitoring points, the enhanced values ​​of plant physiological and electrical data, the plant growth environment status values, the plant growth abnormality risk values, various generated early warning reports, and all manually reviewed records are all archived into a multi-source environmental database according to unified data standards, based on timestamps, spatial locations, and unique object codes, achieving full-process data traceability. The supporting data visualization platform can dynamically display the observation curves of various sensor channels of Cymbidium goeringii in real time, generating visualized reports including enhanced electrical and electrical signal trajectories, environmental parameter fluctuation trends, in-depth assessment results, historical intervention execution, and control response effects. Administrators or decision-makers can intuitively monitor the real-time status of the Cymbidium goeringii's growth environment, historical risk changes, and the entire process of intervention through a large visualization screen. It supports multi-dimensional data filtering, trend analysis, and anomaly tracing, providing strong intelligent data support for scientific decision-making and refined operation and maintenance of Cymbidium goeringii's environmental management.

[0048] In this implementation plan, this step achieves unified archiving, dynamic visualization, and intelligent traceability management of Cymbidium goeringii's full-cycle, multi-source environmental and physiological data. It can not only collect all key data and processing records, but also display monitoring curves, intervention history, and evaluation reports in real time, greatly improving the data transparency, anomaly response efficiency, and scientific level of operation and maintenance decision-making in Cymbidium goeringii's growth environment management.

[0049] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0050] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for processing plant growth environment data based on multimodal sensing and deep learning, characterized in that, Includes the following steps: S1. Collect multivariate environmental data on the plant growth environment status, and upload the multivariate environmental data to the multivariate environmental database after normalization. S2 involves signal enhancement and anomaly identification analysis of plant physiological and electrical data, followed by optimization of sampling strategies and data processing based on the anomaly identification analysis results. S3 analyzes the growth environment status by integrating plant physiological and electrical data with environmental data, and dynamically adjusts adaptive closed-loop control measures for environmental parameters based on the analysis results. S4 assesses abnormal risks based on the results of growth environment status analysis, and intelligently formulates and executes graded response and precise control strategies according to the risk level. S5 enables unified archiving, dynamic visualization, and intelligent traceability management of multi-source environmental data.

2. The method for processing plant growth environment data based on multimodal sensing and deep learning according to claim 1, characterized in that: The specific process for collecting multivariate environmental data on the plant growth environment is as follows: Collect multivariate environmental data, which includes environmental data and plant physiological and atmospheric data. Environmental data include: temperature, humidity, light intensity, carbon dioxide concentration, and soil moisture content; plant physiological and electrical data include: ethylene concentration, trace amounts of carbon monoxide, plant microcurrents, and leaf potential. Temperature data is acquired using a temperature sensor; humidity data is acquired using a resistive humidity sensor; light intensity data is acquired using a photodiode-type light sensor; carbon dioxide concentration data is acquired using a non-dispersive infrared carbon dioxide sensor; and soil moisture content is acquired using a soil capacitive moisture probe. Ethylene concentration was collected using an electrochemical ethylene gas sensor, trace amounts of carbon monoxide were obtained using an electrochemical carbon monoxide gas sensor, plant microcurrents were measured using a high-impedance current amplifier module with non-polar silver chloride electrode patches placed on the plant stem, and leaf potentials were continuously collected using a dual-electrode implantation method combined with a high-input-impedance differential amplifier.

3. The method for processing plant growth environment data based on multimodal sensing and deep learning according to claim 1, characterized in that: The specific process of normalizing the multivariate environmental data and uploading it to the multivariate environmental database is as follows: All data acquisition nodes establish communication connections with the data acquisition main control unit through industrial bus wired communication technology and low-power wide-area wireless communication technology. Each data item is accompanied by a timestamp and spatial location information. Using the master clock signal, the asynchronous acquisition modal data are uniformly aligned to the standard time series through interpolation and resampling, eliminating data mismatch caused by time difference; For raw data with mutations, packet loss, and distortion, we use sliding window absolute median difference, piecewise linear removal of extrema, statistical rules, and filtering of outliers to improve data reliability. For modal data with different physical dimensions, a quantile mapping method is used to map them to standard intervals to obtain standardized data. All processed multimodal standard data are synchronously archived in a multivariate environmental database to provide a high-quality data foundation for subsequent intelligent analysis and control.

4. The method for processing plant growth environment data based on multimodal sensing and deep learning according to claim 1, characterized in that: The specific process of signal enhancement and anomaly identification analysis of plant physiological and electrophysiological data is as follows: The k-th plant physiological gaseous and electrical data at time t is obtained. The high-frequency component values ​​of the plant physiological gaseous and electrical data are extracted using the wavelet transform algorithm. The instantaneous change of the plant physiological gaseous and electrical data is obtained by performing first-order difference. Calculate the mean and standard deviation of plant physiological and electrical data within a sliding window. Subtract the mean from the real-time plant physiological and electrical data, take the absolute value, and then divide by the standard deviation to obtain the deviation value of the plant physiological and electrical data. Temporal features were extracted from the augmented values ​​of plant physiological atmospheric and electrical data and environmental observations as data inputs. Temporal features included: difference, moving mean and high-frequency components. Key features highly correlated with plant growth status were screened out through correlation analysis and feature selection. A plant-sensory environment feature optimization model was constructed by jointly using regression algorithm, neural network and global optimization method for modeling and iterative training. The weight parameters were dynamically adjusted to output high-frequency feature weight factor, instantaneous fluctuation weight factor and deviation feature weight factor with the goal of minimizing prediction error. The high-frequency component value is calculated by multiplying it by the high-frequency component value weighting coefficient to obtain the high-frequency characteristic value. The instantaneous change of the environmental data observation value is multiplied by the instantaneous change weighting coefficient to obtain the instantaneous fluctuation value. The deviation value of the plant physiological gaseous and electrical data is multiplied by the deviation weighting coefficient to obtain the deviation characteristic value. The sum of the three items is then calculated to obtain the enhancement value of the plant physiological gaseous and electrical data.

5. The method for processing plant growth environment data based on multimodal sensing and deep learning according to claim 1, characterized in that: The specific process for optimizing the sampling strategy and data processing based on the anomaly identification and analysis results of plant physiological and electrical data is as follows: The enhancement values ​​of plant physiological gaseous and electrical data were compared with the anomaly detection threshold for determination. When the enhancement value of plant physiological and electrical data is less than the abnormal detection threshold, the current sampling and monitoring measures are maintained, the current plant physiological and electrical data are included in the multivariate environmental database and a fluctuation trend report is generated, and redundant communication and device wake-up frequency within the sampling interval are reduced. When the enhanced value of plant physiological gaseous and electrical data is greater than or equal to the anomaly detection threshold, the sampling frequency of plant physiological gaseous and electrical data is increased, noise diagnosis and zero-point calibration are performed on the sensors that collect plant physiological gaseous and electrical data, and the local environmental conditions are reviewed in conjunction with adjacent strong modal sensors. Ventilation, exhaust, humidification and supplemental lighting are adjusted in the area where abnormal data is observed. All abnormal data and handling operations are synchronously archived in the multivariate environmental database and pushed to the environmental gaseous and electrical data fusion process.

6. The method for processing plant growth environment data based on multimodal sensing and deep learning according to claim 1, characterized in that: The specific process of analyzing the growth environment status by fusing plant physiological and electrical data with environmental data is as follows: Acquire enhanced values ​​of plant physiological atmospheric and electrical data, and advance the process of environmental atmospheric and electrical data fusion using environmental data observations; The plant physiological and electrical characteristics are obtained by multiplying the accumulated plant physiological and electrical data enhancement value at time t by its corresponding plant physiological and electrical data fusion weight coefficient. The environmental characteristics are obtained by multiplying the accumulated environmental data enhancement value at time t by its corresponding environmental data fusion weight coefficient. These two types of characteristics are then weighted and combined with the overall fusion weight factor of plant physiological and electrical data factors and conventional environmental data characteristics to obtain the plant growth environment status value.

7. The method for processing plant growth environment data based on multimodal sensing and deep learning according to claim 1, characterized in that: The specific process of the adaptive closed-loop control measure that dynamically adjusts environmental parameters based on the analysis results of the growth environment is as follows: Real-time comparison of plant growth environment status values ​​with the lower and upper thresholds for anomaly detection: When the plant growth environment status value is less than the lower limit threshold for anomaly judgment, the soil moisture of the enhanced value of plant physiological gas and electricity data is extracted, the soil moisture difference is calculated, the irrigation amount is adjusted according to the soil moisture difference, the daily light duration is extended, the gas, temperature and humidity are coordinated and controlled to rise to the growth range, the intervention parameters are recorded in real time, and the process is entered into a deep evaluation process. When the plant growth environment status value is greater than or equal to the lower threshold for anomaly detection and less than or equal to the upper threshold for anomaly detection, maintain the current inspection mode of each control device. When the plant growth environment status value exceeds the upper limit threshold for anomaly judgment, strong light suppression is activated to reduce the power of supplemental lighting equipment, gas balance is regulated through ventilation, and temperature and humidity are reduced and controlled by evaporative cooling curtains. At the same time, an early warning is sent and the process enters the in-depth evaluation process.

8. The method for processing plant growth environment data based on multimodal sensing and deep learning according to claim 1, characterized in that: The abnormal risk assessment is conducted based on the results of the growth environment status analysis. Abnormal plant growth environment status values ​​are obtained, and multimodal comprehensive risk mapping values ​​are obtained by combining historical sample data of plant growth environment status values ​​with nonlinear regression. The abnormal risk value of plant growth is obtained by multiplying the plant growth environment state value by the linear risk weight factor and adding the multimodal comprehensive risk mapping value by the nonlinear risk weight factor.

9. The method for processing plant growth environment data based on multimodal sensing and deep learning according to claim 1, characterized in that: The specific process of intelligently formulating and executing graded response and precise control strategies based on risk levels is as follows: When the risk value of abnormal plant growth is less than the risk threshold: perform routine inspection and data archiving operations, push the current environmental data and plant physiological and electrical data to the next monitoring cycle, and ensure the continuous operation of the monitoring closed loop; When the risk value of abnormal plant growth is greater than or equal to the risk threshold: Enhanced values ​​of plant physiological and electrical data from multiple environmental factors are extracted. Differences in soil moisture content, temperature, air humidity, light intensity, carbon dioxide concentration, ethylene concentration, carbon monoxide concentration, plant microcurrent, and leaf potential are calculated. Targeted controls are then implemented: irrigation and drainage are regulated based on soil moisture content differences; temperature is adjusted based on temperature differences; humidity control is implemented based on air humidity differences; light intensity is adjusted based on light intensity differences; carbon dioxide concentration is adjusted based on carbon dioxide concentration differences; and ventilation and gas purification are regulated based on ethylene and carbon monoxide concentration differences. For plant microcurrent and leaf potential differences, data on soil moisture content, temperature, air humidity, light intensity, carbon dioxide concentration, ethylene, and carbon monoxide concentrations are retrieved. Comparative analysis is used to locate the abnormal causes and activate the control module. Control intensity is allocated according to the proportion of each parameter difference, and an early warning report containing abnormal difference data, cause analysis, and control measures is pushed out. Each anomaly handling process, intervention parameters, their effects, and review records are archived into a multivariate environmental database.

10. The method for processing plant growth environment data based on multimodal sensing and deep learning according to claim 1, characterized in that: The specific process for achieving unified archiving, dynamic visualization, and intelligent traceability management of multi-source environmental data is as follows: The multivariate environmental data, enhanced values ​​of plant physiological and electrical data, plant growth environment status values, plant growth abnormality risk values, early warning reports, and manual review records are archived into the multivariate environmental database according to timestamps, spatial locations, and unique object codes. The system displays real-time observation curves, fluctuation trend reports, in-depth assessment results, intervention execution history, and response effects for each sensor channel, allowing administrators to monitor the entire process from data collection and risk analysis to control execution.

Images