Asiatic pennywort herb seedling growth abnormity detection system

Abstract

The invention provides a centella asiatica seedling growth anomaly detection system, and relates to the technical field of data processing, and the centella asiatica seedling growth anomaly detection system comprises a parameter acquisition module which is used for establishing a seedbed monitoring grid and dividing the monitoring grid into a plurality of independent monitoring units; for each monitoring unit, extracting characteristic indexes of seedlings in the unit, and collecting environmental parameters in the corresponding unit; and the correction module is used for obtaining unit specificity calibration parameters according to the space coordinates and the historical monitoring data of each monitoring unit, and performing standardized correction on the characteristic indexes and the environmental parameters by using the unit specificity calibration parameters to obtain corrected characteristic indexes and environmental parameters. According to the centella asiatica seedling abnormity detection system and method, through gridding collection, unit specificity correction, model discrimination, type classification and differentiation instruction generation, accurate detection and efficient response to centella asiatica seedling abnormity are integrally achieved, and the seedling culture quality is guaranteed.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a system for detecting abnormal growth of Centella asiatica seedlings. Background Technology

[0002] Currently, some seedling growth monitoring methods based on sensors and image processing have been applied in greenhouse seedbed environments. For example, images of the seedling canopy are collected by deploying cameras at fixed locations, or key parameter data within the seedbed are obtained using temperature, humidity, and light sensors, providing auxiliary support for judging the seedling growth status. However, in practical applications, these methods mostly have certain limitations. On the one hand, because different areas within the seedbed are prone to micro-domain differences in light distribution, air flow, and irrigation uniformity, existing systems mostly use globally unified thresholds or models to make anomaly judgments, failing to fully consider the local micro-domain variability between monitoring units and the differences in basic seedling growth, which may affect the accuracy of the judgment results. On the other hand, even if abnormal growth phenomena can be identified, existing solutions usually cannot effectively distinguish whether the abnormality is caused by disease or by environmental stress (such as uneven local moisture or temperature fluctuations), thus failing to provide targeted control suggestions, and having relatively limited guidance value in actual production. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a detection system for abnormal growth of Centella asiatica seedlings, which can accurately identify the causes of abnormalities and effectively eliminate the influence of differences in the micro-environment inside the seedbed on the monitoring data.

[0004] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: Firstly, a system for detecting abnormal growth in Centella asiatica seedlings includes: The parameter acquisition module is used to establish a seedbed monitoring grid, which is divided into multiple independent monitoring units. For each monitoring unit, the characteristic indicators of the seedlings in the unit are extracted, and the environmental parameters in the corresponding unit are collected. The calibration module is used to obtain unit-specific calibration parameters based on the spatial coordinates and historical monitoring data of each monitoring unit, and to use the unit-specific calibration parameters to standardize and correct the characteristic indicators and environmental parameters, thereby obtaining the corrected characteristic indicators and environmental parameters. The discrimination module is used to input the corrected feature indicators and environmental parameters into the pre-trained growth anomaly recognition model to obtain the anomaly discrimination result; The classification module is used to distinguish between pathogenic abnormalities and environmental stress abnormalities based on the type of abnormality in the abnormality discrimination results. The instruction generation module is used to generate differentiated response instructions based on the exception type.

[0005] In a second aspect, a computing device includes: One or more processors; A storage device for storing one or more programs that, when executed by one or more processors, enable the one or more processors to implement the system.

[0006] Thirdly, a computer-readable storage medium storing a program that, when executed by a processor, implements the system.

[0007] The above-described solution of the present invention has at least the following beneficial effects: By establishing a seedbed monitoring grid, the seedbed is divided into multiple independent monitoring units, enabling precise collection of seedling characteristic indicators and environmental parameters for each unit. Combining the spatial coordinates of the monitoring units with historical data, unit-specific calibration parameters are generated to standardize and correct the collected data. This fully considers the local micro-domain variability and differences in seedling basic growth within each unit, eliminating interference from micro-domain differences at the source of data collection and processing, improving the accuracy of anomaly identification results, and avoiding misjudgments and omissions caused by globally uniform standards. The corrected data is then input into a pre-trained growth anomaly identification model, combined with data such as Centella asiatica tissue culture response, growth stage quality indicators, and the correlation between substrate and physiological response, to accurately identify abnormal characteristic parameters and their corresponding causes. Furthermore, anomalies are clearly categorized into pathogenic anomalies, such as root rot and leaf spot, and environmental stress anomalies, such as uneven watering and strong light exposure, achieving a clear definition of the causes of anomalies and overcoming the limitations of ambiguity in anomaly type identification. Based on the anomaly type, differentiated response plans are generated to address seedling growth anomalies, avoiding problems such as low production efficiency and seedling losses. Attached Figure Description

[0008] Figure 1 This is a schematic diagram of a Centella asiatica seedling abnormality detection system provided in an embodiment of the present invention.

[0009] Figure 2 This is a flowchart illustrating how the corrected feature indicators and environmental parameters are input into a pre-trained growth anomaly recognition model to obtain anomaly discrimination results, as provided in an embodiment of the present invention. Detailed Implementation

[0010] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0011] like Figure 1 As shown, an embodiment of the present invention proposes a system for detecting abnormal growth of Centella asiatica seedlings, comprising: The parameter acquisition module is used to establish a seedbed monitoring grid, which is divided into multiple independent monitoring units. For each monitoring unit, the characteristic indicators of the seedlings in the unit are extracted, and the environmental parameters in the corresponding unit are collected. The calibration module is used to obtain unit-specific calibration parameters based on the spatial coordinates and historical monitoring data of each monitoring unit, and to use the unit-specific calibration parameters to standardize and correct the characteristic indicators and environmental parameters, thereby obtaining the corrected characteristic indicators and environmental parameters. The discrimination module is used to input the corrected feature indicators and environmental parameters into the pre-trained growth anomaly recognition model to obtain the anomaly discrimination result; The classification module is used to distinguish between pathogenic abnormalities and environmental stress abnormalities based on the type of abnormality in the abnormality discrimination results. The instruction generation module is used to generate differentiated response instructions based on the exception type.

[0012] In this embodiment of the invention, a seedbed monitoring grid is established, dividing the seedbed into multiple independent monitoring units to achieve precise collection of seedling characteristic indicators and environmental parameters for each unit. Then, by combining the spatial coordinates of the monitoring units with historical data, unit-specific calibration parameters are generated to standardize and correct the collected data. This fully considers the local micro-domain variability of each unit and the differences in basic seedling growth, eliminating interference from micro-domain differences at the source of data collection and processing, improving the accuracy of anomaly identification results, and avoiding misjudgments and omissions caused by globally uniform standards. By inputting the corrected data into a pre-trained growth anomaly identification model, combined with data such as Centella asiatica tissue culture response, growth stage quality indicators, and the correlation between substrate and physiological response, abnormal characteristic parameters and their corresponding causes are accurately identified. Furthermore, anomalies are clearly classified into pathogenic anomalies, such as root rot and leaf spot, and environmental stress anomalies, such as uneven watering and strong light exposure, achieving a clear definition of the causes of anomalies and overcoming the limitations of ambiguity in anomaly type identification. Based on the anomaly type, differentiated response plans are generated to solve seedling growth anomalies and avoid problems such as low production efficiency and seedling loss.

[0013] In a preferred embodiment of the present invention, a seedbed monitoring grid is established, which is divided into multiple independent monitoring units. For each monitoring unit, characteristic indicators of the seedlings within the unit are extracted, and environmental parameters within the corresponding unit are collected, which may include: In this embodiment of the invention, step 100 involves setting a first analysis point, a second analysis point, and a third analysis point at three spatial locations: the left front corner, the right front corner, and the rear center corner of the greenhouse seedling bed. A monitoring grid covering the entire seedling bed is constructed based on these three analysis points. The seedling bed is then divided into multiple independent monitoring units of equal area according to the monitoring grid. Specifically, this includes setting a first analysis point (A, coordinates x1, y1), a second analysis point (B, coordinates x2, y2), and a third analysis point (C, coordinates x3, y3) at the left front corner, the right front corner, and the rear center corner of the greenhouse seedling bed. Each analysis point is positioned using a fixed marker with coordinate markings to ensure no offset. A measuring device is then activated to obtain the actual length and width of the seedling bed. Simultaneously, the coverage of the three points is verified using a triangle area algorithm (coordinate cross product method), i.e., according to the formula... Calculate the area of ​​△ABC. If this area is greater than 1 / 3 of the total area of ​​the seedbed, the three points can be used as reference nodes. If not, fine-tune the coordinates of point C along the front-to-back direction of the seedbed (parallel to the length direction). Each fine-tuning distance is 5% of the seedbed length. After fine-tuning, record the new coordinates of point C (x3', y3') and recalculate the area of ​​△ABC using the coordinate cross product formula. Repeat the fine-tuning and calculation operations until the area of ​​△ABC is greater than 1 / 3 of the total area of ​​the seedbed. Determine the final coordinates of point C and the three reference nodes. Using A, B, and C as references, call the grid drawing system to generate a grid that is horizontally parallel to the length of the seedbed and vertically parallel to the width, ensuring coverage of the entire bed and forming the initial monitoring grid framework. Calculate the total number of monitoring units as: total area of ​​the seedbed ÷ target area of ​​a single unit. Adjust the horizontal and vertical line spacing of the grid according to the total number to evenly divide the seedbed into independent monitoring units of equal area. Each unit corresponds to a unique spatial coordinate.

[0014] Step 101: For each monitoring unit, multispectral image data of the Centella asiatica seedlings within the monitoring unit is acquired using a multispectral imaging device mounted above the seedling bed. Based on the multispectral image data, characteristic indicators of the seedlings within each monitoring unit are extracted, including leaf color characteristics, leaf shape characteristics, and physiological reflectance spectral indicators at different wavelengths. Environmental parameters corresponding to each monitoring unit are also collected, including light intensity, air temperature, air humidity, carbon dioxide concentration, and substrate moisture data. Specifically, a multispectral imaging device is pre-installed on a fixed bracket above the seedling bed. After the device is started... The device measures the vertical distance to the bed surface using its built-in distance sensor, adjusts the lens tilt angle using an angle adjustment motor, and confirms this in real time with image preview until the entire boundary of a single monitoring unit is selected without interference from adjacent seedlings. After calibration, the device sequentially captures images of each unit at preset intervals, such as 30 minutes, automatically switching between visible light (400-760nm) and near-infrared (760-2500nm) bands to acquire seedling image data for each band. This data is then integrated into multispectral image data for the corresponding unit and linked to the unit's spatial coordinates using system coordinates (e.g., binding to "row 2, column 3"). The data is entered into the database. The image analysis system periodically retrieves multispectral image data from each monitoring unit from the database and automatically executes the feature extraction process. Specifically, when extracting leaf color features, the system identifies the leaf pixel region in the image using a color threshold segmentation algorithm, counts the RGB three-color channel values ​​of each pixel in the region, sums the R, G, and B channel values ​​of all pixels respectively, and then divides the sum of the three channel values ​​by the total number of pixels in the leaf pixel region to obtain the average R value, average G value, and average B value of the leaf. When extracting leaf shape features, the system first performs contour edge detection on the leaf pixel region to determine the complete edge of each leaf, measures the maximum straight-line distance between two points on the edge as the leaf length, measures the maximum straight-line distance between two points on the edge perpendicular to the leaf length direction as the leaf width, and calculates the leaf aspect ratio by dividing the leaf length by the leaf width. At the same time, the total number of leaves in the monitoring unit is counted. When extracting physiological reflectance spectral indicators, the system extracts the reflected light intensity values ​​of all pixels in the leaf pixel region for each spectral band of image data and sums them. Then, the sum is divided by the total number of pixels in the leaf pixel region to obtain the average reflected light intensity value of the corresponding band.

[0015] At the center of each monitoring unit, a set of environmental parameter acquisition devices is pre-installed. The light sensor has its photosensitive surface facing upwards to collect light intensity data within the unit; the temperature and humidity sensor's probe is exposed to the air to simultaneously collect air temperature and humidity data within the unit; the carbon dioxide sensor is connected to the air environment within the unit via a duct to collect air carbon dioxide concentration data; and the substrate humidity sensor's probe is inserted into the seedling substrate within the monitoring unit to a depth controlled at 5 cm to collect humidity data within the substrate. Each environmental parameter acquisition sensor and the multispectral imaging device are synchronized via a system clock, collecting data at the same time intervals. After collection, the environmental parameter data is integrated into the corresponding environmental parameter group for the monitoring unit, and then the environmental parameter group is associated and stored with the spatial coordinates of the monitoring unit and the data acquisition timestamp.

[0016] This embodiment divides the seedbed into independent monitoring units of equal area by setting analysis points at key locations and constructing a monitoring grid covering the entire seedbed. This eliminates the impact of micro-domain differences caused by varying light, humidity, and airflow distribution within the seedbed on data acquisition. Multispectral imaging equipment is used to selectively collect multispectral image data of seedlings in each monitoring unit. Image analysis is combined to extract detailed characteristic indicators such as leaf color, leaf shape, and multi-band physiological reflectance spectra. Simultaneously, multiple environmental parameters of each monitoring unit are collected, solving the problem of single-dimensional data acquisition in monitoring, which cannot accurately reflect the local seedling growth status.

[0017] In a preferred embodiment of the present invention, unit-specific calibration parameters are obtained based on the spatial coordinates and historical monitoring data of each monitoring unit. These unit-specific calibration parameters are then used to standardize and correct the characteristic indicators and environmental parameters, resulting in corrected characteristic indicators and environmental parameters. This process may include: In this embodiment of the invention, step 200a involves extracting historical environmental parameter benchmark values ​​and historical seedling spectral reflectance benchmark values ​​within the complete growth cycle corresponding to the spatial location attribute from the historical monitoring database, based on the spatial location attribute of the monitoring unit. The historical environmental parameter benchmark values ​​include light intensity benchmark values, air temperature benchmark values, and air humidity benchmark values. Specifically, this includes: determining the spatial location attribute of the monitoring unit based on its unique spatial coordinates within the seedbed monitoring grid. For example, if the coordinates of a monitoring unit correspond to the 2nd row and 4th column of the seedbed, its spatial location attribute is defined as the area in the 2nd row and 4th column of the seedbed. Next, the historical monitoring database is accessed. This database stores long-term monitoring data categorized by different spatial location attributes of the seedbed. For each monitoring unit's spatial location attribute, all complete growth cycle data corresponding to that attribute are filtered out from the database. (A complete growth cycle refers to the entire stage of Centella asiatica seedlings from sowing to meeting transplanting standards, including the emergence period, seedling stage, and mature seedling stage. Meeting transplanting standards specifically means that the seedling height reaches 8-10cm, the total number of leaves is not less than 6, the root length is ≥5cm, and there is no damage from pests or diseases.) Seedlings meeting these specifications can stably adapt to the post-transplant growth environment. From the selected complete growth cycle data, light intensity monitoring data is selected daily from 9:00 AM to 10:00 AM (a fixed time period to avoid the influence of time differences). All light intensity data for this time period within the cycle are summed, and the sum is divided by the total number of data points. The result is the baseline light intensity value for this spatial location attribute. Similarly, air temperature monitoring data for the same time period each day are summed and divided by the total number of data points to obtain the baseline air temperature value. The baseline value of air humidity is obtained by summing the air humidity monitoring data for a period of time and dividing it by the total number of data. From the complete growth cycle data of this spatial location attribute, multispectral image data of seedlings in normal growth state without pests, diseases, or environmental stress, such as uneven watering or abnormal temperature, are selected. Reflectance spectral data of each spectral band, such as visible light and near infrared, are extracted from these images. The reflectance spectral data of all normal growth states under the same band are added together, and the sum is divided by the total number of data. The result is the historical seedling spectral reflectance baseline value of this spatial location attribute.

[0018] Step 200b involves calculating the ratio of the current monitored light intensity value to the light intensity reference value in the environmental parameters of the monitoring unit to obtain the light calibration coefficient; calculating the difference between the current monitored air temperature value and the air temperature reference value in the environmental parameters of the monitoring unit to obtain the temperature calibration coefficient; calculating the difference between the current monitored air humidity value and the air humidity reference value in the environmental parameters of the monitoring unit to obtain the humidity calibration coefficient; and calculating the ratio between the current monitored physiological reflectance spectral index of the monitoring unit and the historical seedling spectral reflectance reference value to obtain the spectral reflectance calibration coefficient. Specifically, this includes acquiring the current environmental parameter monitoring data for each monitoring unit. This data is from the same batch as the environmental parameters collected in step 101. In step 101, environmental parameter acquisition equipment is pre-deployed at the center of each monitoring unit. The light sensor's photosensitive surface faces upwards to collect light intensity data, while the temperature and humidity sensor probes are exposed to the air to synchronously collect air temperature and air humidity data. Each sensor and the multispectral imaging device collect data synchronously and at the same interval according to the system clock. The resulting light intensity, air temperature, and air humidity data represent the current light intensity monitoring value for the current monitoring unit. The system acquires the measured values ​​of current air temperature and current air humidity. Simultaneously, it obtains the current physiological reflectance spectral index of the monitoring unit, which is extracted from the multispectral image data of Centella asiatica seedlings acquired by the multispectral imaging device in step 101. When calculating the light calibration coefficient, the current light intensity monitoring value of the monitoring unit is divided by the corresponding light intensity reference value. Similarly, the temperature calibration coefficient is calculated by subtracting the corresponding air temperature reference value from the current air temperature monitoring value. The humidity calibration coefficient is calculated by subtracting the corresponding air humidity reference value from the current air humidity monitoring value. Finally, the spectral reflectance calibration coefficient is calculated by dividing the current physiological reflectance spectral index monitoring value of the monitoring unit by the corresponding historical seedling spectral reflectance reference value. The light calibration coefficient, temperature calibration coefficient, humidity calibration coefficient, and spectral reflectance calibration coefficient for each monitoring unit together constitute the unit-specific calibration parameters for that unit.

[0019] Step 201: Using the unit-specific calibration parameters corresponding to each monitoring unit, standardize and correct the seedling characteristic indicators and environmental parameters of the corresponding monitoring unit to obtain the corrected seedling characteristic indicators and corrected environmental parameters for each monitoring unit. Specifically, this includes: identifying the types of seedling characteristic indicators and environmental parameters that need correction; the seedling characteristic indicators include leaf color characteristics (average R value, average G value, average B value), leaf morphology characteristics (leaf length-to-width ratio, total number of leaves), and physiological reflectance spectral indicators extracted in step 101; the environmental parameters include light intensity, air temperature, air humidity, carbon dioxide concentration, and substrate moisture collected in step 101; and correcting the seedling characteristic indicators, i.e. The spectral reflectance calibration coefficient ranges from 0.8 to 1.2. For the average R value in leaf color characteristics, the corrected average R value is obtained by multiplying the average R value by the spectral reflectance calibration coefficient for that unit. Similarly, the corrected average G value is obtained by multiplying the average G value by the spectral reflectance calibration coefficient, and the corrected average B value is obtained by multiplying the average B value by the spectral reflectance calibration coefficient. For the leaf aspect ratio in leaf morphology characteristics, the corrected leaf aspect ratio is obtained by multiplying the leaf aspect ratio by the spectral reflectance calibration coefficient. The total number of leaves does not require coefficient correction and is directly used as the corrected total number of leaves. For physiological reflectance spectral indices, the corrected physiological reflectance spectral indices are obtained by multiplying the current physiological reflectance spectral index by the spectral reflectance calibration coefficient. The system calculates the reflectance spectral index; it standardizes environmental parameters, with the illumination calibration coefficient ranging from 0.7 to 1.3. For illumination intensity, the current illumination intensity is multiplied by the unit's illumination calibration coefficient to obtain the corrected illumination intensity. The temperature calibration coefficient ranges from -4℃ to +4℃. For air temperature, the current air temperature is subtracted from the unit's temperature calibration coefficient. If the coefficient is +3℃, it indicates the current temperature is 3℃ higher than the baseline value; subtracting the coefficient returns the temperature to the baseline level. If the coefficient is -2℃, it indicates the current temperature is 2℃ lower than the baseline value; subtracting the coefficient increases the temperature to the baseline level, thus obtaining the corrected air temperature. For air humidity, the current air humidity is calculated by subtracting the unit's temperature calibration coefficient. The humidity calibration coefficient of the unit is used to obtain the corrected air humidity. The carbon dioxide concentration and substrate humidity are first compared with the normal growth cycle benchmark values ​​of the corresponding spatial location attributes in the historical monitoring database. The same difference correction logic as that for air temperature is used. The corrected carbon dioxide concentration and corrected substrate humidity are obtained by subtracting the corresponding concentration / humidity benchmark difference from the current monitoring value (which is calculated from the historical benchmark value). All corrected leaf color characteristics, leaf shape characteristics, and physiological reflectance spectral indicators are integrated into the corrected seedling characteristic indicators of the monitoring unit. All corrected light intensity, air temperature, air humidity, carbon dioxide concentration, and substrate humidity are integrated into the corrected environmental parameters of the monitoring unit.

[0020] This embodiment extracts historical benchmark values ​​by combining the spatial location attributes of monitoring units, ensuring that the extracted data matches the local micro-domain characteristics of the unit and avoiding calibration deviations caused by using a globally unified benchmark value. Unit-specific calibration parameters are obtained through targeted ratio and difference calculations, enabling the calibration coefficient of each monitoring unit to accurately reflect the difference between the current data and the historical benchmark, avoiding the problem that a unified calibration standard cannot adapt to the micro-domain differences of different units. By correcting seedling characteristic indicators and environmental parameters using unit-specific calibration parameters, data deviations caused by micro-domain differences and different historical growth bases in different monitoring units are eliminated, making the corrected data comparable across units and solving the problem of data accuracy being affected by micro-domain differences.

[0021] like Figure 2 As shown, in another preferred embodiment of the present invention, inputting the corrected feature indicators and environmental parameters into a pre-trained growth anomaly recognition model to obtain anomaly discrimination results may include: In this embodiment of the invention, step 300 involves fusing the corrected seedling characteristic indicators and corrected environmental parameters of the same monitoring unit to generate a multi-dimensional feature vector representing the overall growth status of the monitoring unit. Specifically, this includes: clarifying the source of the data items to be fused, i.e., both the corrected seedling characteristic indicators and the corrected environmental parameters are data standardized and corrected in step 201. The corrected seedling characteristic indicators include the corrected average R value, average G value, average B value, leaf aspect ratio, total number of leaves, and average reflected light intensity values ​​for each wavelength band; the corrected environmental parameters include the corrected light intensity, air temperature, air humidity, carbon dioxide concentration, and substrate humidity; and unifying the data format for all data items. All data are converted to a numerical format of the same magnitude, for example, the unit of light intensity is unified to lux, the unit of air temperature is unified to degrees Celsius, and the values ​​of each characteristic index are retained to two decimal places. The processed data are sorted according to a preset characteristic order, which is: corrected average R value, average G value, average B value, leaf aspect ratio, total number of leaves, average reflected light intensity value of each band, light intensity, air temperature, air humidity, carbon dioxide concentration, and substrate humidity. All sorted data items are arranged in sequence to form a vector containing multiple numerical elements. This vector is the multidimensional feature vector representing the overall growth status of the monitoring unit. Each element corresponds to a feature parameter, and the number of dimensions of the vector is consistent with the total number of data items.

[0022] Step 301 involves inputting the multidimensional feature vector into a pre-trained growth anomaly identification model. Each feature parameter in the multidimensional feature vector is compared one by one with the predefined threshold range of the corresponding feature parameter under normal growth conditions of Centella asiatica. Abnormal feature parameters that are greater than or equal to the upper limit of the threshold range or less than or equal to the lower limit of the threshold range are identified, and the deviation of the abnormal feature parameter from the corresponding threshold boundary value is calculated. Specifically, this includes constructing and training the growth anomaly identification model. First, historical data of Centella asiatica under normal growth conditions (no pests or diseases, no environmental stress) is collected, covering three growth stages: emergence, seedling stage, and mature seedling stage. At least 300 sets of data are collected for each stage. Each set of data includes seedling characteristic indicators (average R value, average G value, average B value, leaf length-to-width ratio, total number of leaves, average reflected light intensity value of each band) and environmental parameters (light intensity, air temperature, air humidity, carbon dioxide concentration, substrate moisture) for that stage. All data undergoes standardization correction processing as described in step 201. The data interface reads two types of data: first, the aforementioned historical normal data (including growth stage labels, feature parameter names, and corresponding values); second, the data to be judged (the multi-dimensional feature vector generated in step 300, including the feature parameter sequence index, and the corrected feature indicators and environmental parameters corresponding to each monitoring unit). After automatically parsing the data structure, it is stored in a two-level directory of growth stage-feature parameter. For example, the seedling stage-average R value directory only stores the corrected average R value normal data for that stage, and the seedling stage-air temperature directory only stores the corrected air temperature normal data for that stage. For model training, the first step is to divide the collected and classified historical normal data into a training set and a validation set in a 7:3 ratio. The second step is to retrieve all data under that category in the training set for a certain feature parameter of the same growth stage, first calculate the average value (the sum of all data ÷ the total number of data), then calculate the deviation of each data from the average value, sort all deviations in ascending order, and take the deviation at the 2.5% position as the lower limit deviation, and the 97th position as the lower limit deviation.The deviation at the 5% mark is taken as the upper limit deviation. The parameter value range corresponding to this deviation interval (average + lower limit deviation to average + upper limit deviation) is the initial normal threshold range for this parameter at this growth stage. For example, if the average R value of the training set data at the seedling stage is 180, the lower limit deviation is -10, and the upper limit deviation is +10, the initial threshold range is [170, 190]. The third step is to verify the accuracy of the initial threshold range using validation set data. The number of data in the validation set whose values ​​are within the initial threshold range is counted, and this number is divided by the total number of data in the validation set to obtain the percentage. If the percentage is less than 95%, that is, the percentage of normal data exceeding the threshold range exceeds 5%, the upper and lower limits of the deviation interval are readjusted, such as expanding it to the deviation at the 1% to 99% mark, and the threshold is recalculated. The process involves several steps: First, the threshold range is calculated and verified until 95% of the data in the validation set falls within the threshold range, thus determining the final normal threshold range for that parameter at that growth stage. Second, following the steps in steps two and three, the threshold range calculation and verification for all feature parameters at each growth stage are completed. Simultaneously, the core functional logic of the model is determined (reading parameter values ​​and their corresponding growth stages from the data to be judged, retrieving the corresponding final threshold range; marking an anomaly if the parameter value is ≤ the lower threshold or ≥ the upper threshold, otherwise marking it as normal; calculating deviation, i.e., if an abnormal parameter is ≥ the upper threshold, calculating (parameter value - upper threshold) ÷ upper threshold; if an abnormal parameter is ≤ the lower threshold, calculating (lower threshold - parameter value) ÷ lower threshold, with the result rounded to three decimal places), thus completing model training.

[0023] Based on the trained growth anomaly identification model, a multidimensional feature vector is input into the model. The model first extracts each feature parameter from the multidimensional feature vector (corresponding to the data items integrated in step 300), and then compares each feature parameter with the normal threshold range for the growth stage to which the parameter belongs. If the feature parameter value is greater than or equal to the upper limit of the corresponding threshold range, or if the feature parameter value is less than or equal to the lower limit of the corresponding threshold range, the feature parameter is marked as an abnormal feature parameter; if the feature parameter value is between the upper and lower limits of the threshold range, the feature parameter is marked as a normal feature parameter. After marking the abnormal feature parameters, the model further calculates the relationship between the abnormal feature parameters and the corresponding threshold boundary values. Deviation is calculated as follows: if the abnormal feature parameter is greater than or equal to the upper threshold, the deviation is calculated as: Deviation = (Abnormal feature parameter value - Upper threshold value) ÷ Upper threshold value; if the abnormal feature parameter is less than or equal to the lower threshold, the deviation is calculated as: Deviation = (Lower threshold value - Abnormal feature parameter value) ÷ Lower threshold value. For example, if the normal upper threshold of a feature parameter is 200, and the value of the feature parameter input into the model is 220, the deviation calculated by the model is: (220 - 200) ÷ 200 = 0.1; if the normal lower threshold of a feature parameter is 23℃, and the value of the feature parameter input into the model is 21℃, the deviation calculated by the model is: (23 - 21) ÷ 23 ≈ 0.087.

[0024] Step 302a: Based on the type and corresponding deviation of the abnormal feature parameters, establish an abnormal feature parameter combination pattern. Specifically, this includes: statistically analyzing the types of all abnormal feature parameters within the same monitoring unit (abnormal feature parameters are those marked in step 301 that deviate from the normal threshold range; types include corrected average R value, average G value, and other corrected indicators and parameters determined in step 201). For example, the abnormal feature parameter type may be the corrected average G value, matrix humidity, or near-infrared band average reflected light intensity value; recording the deviation corresponding to each abnormal feature parameter, associating the abnormal feature parameter type with the corresponding deviation to form basic data in key-value pair form, for example, the corrected average G value corresponds to a deviation of 0.08, matrix humidity corresponds to a deviation of 0.12, and the near-infrared band average reflected light intensity value corresponds to a deviation of 0.05; and arranging the above-associated abnormal feature parameter types and deviation data sequentially according to the sorting order of the feature parameters in the multidimensional feature vector (the sorting order preset in step 300) to form a set of structured data. This structured data is the abnormal feature parameter combination pattern, which contains information on the types of abnormal features and the severity of each abnormality.

[0025] Step 302b involves matching the abnormal feature parameter combination pattern with a pre-stored library of typical abnormal feature parameter patterns for Centella asiatica. This library includes feature parameter patterns for root rot, leaf spot, uneven water distribution, strong light exposure, and drastic temperature fluctuations. Specifically, the library contains five patterns: root rot, leaf spot, uneven water distribution, strong light exposure, and drastic temperature fluctuations. Each pattern clearly defines the types of abnormal feature parameters and their corresponding deviation ranges. For example, the root rot pattern includes three parameters: substrate moisture, average leaf B-value, and average near-infrared reflected light intensity, with deviation ranges of 0.1 to 0.3, 0.05 to 0.2, and 0.08 to 0.25, respectively. The leaf spot disease model includes three parameters: average leaf R value, average leaf G value, and total number of leaves, with deviation ranges of 0.06 to 0.22, 0.07 to 0.23, and 0.09 to 0.3, respectively. The uneven water distribution, strong light exposure, and drastic temperature fluctuation models all specify the corresponding abnormal characteristic parameter types and deviation ranges in the same format. During comparison, the abnormal characteristic parameter combination model established in step 302a is compared with the five typical models in the library one by one. For each comparison, the number of abnormal characteristic parameter types shared by both models is first counted, such as the number of parameter types shared by the combination model and the root rot disease model. Then, the number of common types in the combination model that fall within the deviation range of the typical model is calculated. For example, if the substrate moisture deviation of 0.15 in the combination model falls within the range of 0.1-0.3 in the root rot disease model, it is counted as 1. The number of common parameter types and the number that meet the deviation range obtained from the comparison with each typical model are recorded.

[0026] Step 302c: When the matching degree between the abnormal feature parameter combination pattern and any typical abnormal feature parameter pattern is greater than the first threshold, it is determined that the Centella asiatica seedlings in the monitoring unit have a corresponding type of abnormal state, and an abnormal state discrimination result containing the abnormal type identifier and the degree of abnormality is obtained. Specifically, it includes: matching degree = (number of common abnormal feature parameter types that meet the deviation range) ÷ (total number of all abnormal feature parameter types in the typical abnormal pattern); for example, the root rot pattern contains 3 parameters, and the combination pattern has 2 parameters that it shares and both meet the deviation range, then the matching degree = 2 ÷ 3 ≈ 0.67; set the first threshold, which is set to 0.6. This threshold is determined by analyzing more than 100 sets of historical abnormal matching data. The specific calculation method is to use different thresholds for each set of historical data, such as 0.5, 0.55, 0.6, and 0. Anomaly type discrimination is performed at level 65. The number of discrimination results that match the actual anomaly type at each threshold is counted. The discrimination accuracy is then calculated by dividing the number of matching results by the total number of historical data sets. When the threshold is set to 0.6, the calculated discrimination accuracy reaches 95%, so this value is determined as the first threshold. If the matching degree of a typical pattern is >0.6, the monitoring unit is judged to have this type of anomaly. If the matching degree is ≤0.6, other patterns are compared until all patterns are compared. For the judged anomaly type, the average deviation of all corresponding anomaly feature parameters is calculated (total deviation ÷ number of parameters). The level is divided according to the average value, i.e., the average value <0.1 is mild anomaly, the average value between 0.1 and 0.2 is moderate anomaly, and the average value >0.2 is severe anomaly. The anomaly type, such as root rot, is integrated with the severity level to form the final discrimination result.

[0027] This embodiment combines multi-dimensional data such as the hormone response of Centella asiatica tissue culture, quality during the growth period, and physiological response of the propagation substrate to provide more comprehensive background support for the analysis of abnormal characteristic parameters, reducing misjudgments based on a single data dimension. Simultaneously, by matching with a pre-stored library of typical abnormal patterns, it further anchors the abnormality type, ensuring more accurate discrimination results. By establishing abnormal characteristic parameter combination patterns and calculating deviation, it can not only locate the abnormality type, such as root rot or uneven water distribution, but also classify mild, moderate, and severe abnormalities based on the average deviation, avoiding blind treatment. Standardized combination patterns, pattern library matching, and threshold determination processes can quickly output abnormality discrimination results. Furthermore, the clear identification of abnormality types and degrees can guide managers to precisely adjust the environment, such as controlling light and humidity, or taking preventative measures, such as controlling root rot.

[0028] In a preferred embodiment of the present invention, classifying abnormalities into pathogenic abnormalities or environmental stress abnormalities based on the abnormality type in the abnormality discrimination result may include: In this embodiment of the invention, step 400, when the abnormality type identifier in the abnormality status discrimination result is root rot or leaf spot, the abnormality type is classified as pathogenic abnormality; when the abnormality type identifier in the abnormality status discrimination result is uneven water distribution, strong light exposure, or drastic temperature fluctuation, the abnormality type is classified as environmental stress abnormality. Specifically, this includes: first obtaining the abnormality status discrimination result output in step 302c, extracting the abnormality type identifier in the result, classifying it according to the specific category of the abnormality type identifier, and if the abnormality type identifier is root rot or leaf spot, classifying the abnormality type as pathogenic abnormality, wherein root rot and leaf spot are both pathogen-related abnormality types defined in the typical abnormality feature parameter pattern library in step 302b; If the anomaly type is identified as uneven moisture distribution, strong light exposure, or drastic temperature fluctuation, the anomaly type is classified as an environmental stress anomaly. Uneven moisture distribution, strong light exposure, and drastic temperature fluctuation are all environmental factor-related anomaly types defined in the typical anomaly feature parameter pattern library in step 302b. After classification, a structured classification result containing "anomaly type identifier - anomaly classification (pathogenic anomaly / environmental stress anomaly) - anomaly degree" is obtained.

[0029] This embodiment accurately classifies anomalies into pathogenic anomalies (root rot, leaf spot) and environmental stress anomalies (uneven water distribution, strong light, temperature fluctuations), clearly distinguishing between biotic pathogenic factors and abiotic environmental factors. After classification, two differentiated response logics can be directly applied, such as pathogenic anomalies corresponding to disease control and environmental stress anomalies corresponding to environmental optimization or equipment maintenance. This eliminates the need for repeated analysis of the anomaly's nature, reduces intermediate decision-making steps, and accelerates the conversion speed from judgment results to execution instructions.

[0030] In a preferred embodiment of the present invention, generating a differentiated response instruction based on the anomaly type may include: In this embodiment of the invention, step 500 involves determining the specific environmental stress factor based on the anomaly type identifier in the anomaly state discrimination result when the anomaly type is an environmental stress anomaly, and generating a corresponding environmental parameter optimization scheme based on the anomaly degree quantification value. When the anomaly type identifier is uneven water distribution and the anomaly degree quantification value exceeds a second threshold, the system analyzes and determines whether it is caused by irrigation equipment failure. If it is determined to be equipment failure, an equipment maintenance instruction is generated. Specifically, this includes: first, obtaining the environmental stress anomaly classification result output in step 400, and extracting the anomaly type identifier and the anomaly degree quantification value (i.e., the average deviation calculated in step 302c); determining the specific environmental stress factor based on the anomaly type identifier; if the anomaly type identifier is uneven water distribution, then... The environmental stressor is abnormal substrate moisture; if it is strong light exposure, the specific environmental stressor is abnormal light intensity; if it is drastic temperature fluctuation, the specific environmental stressor is abnormal air temperature. Combining the specific environmental stressor and the quantitative value of the degree of abnormality, a corresponding environmental parameter optimization scheme is generated. For example, if the abnormality type is strong light exposure and the average deviation is 0.15 (moderate abnormality), the current light intensity is 8000 lux, and the normal threshold range of light intensity for seedlings is 3000-5000 lux, then an optimization scheme is generated to reduce the light intensity from the current 8000 lux to the normal threshold range for seedlings [3000 lux, 5000 lux], with a single adjustment not exceeding 1000 lux, and the adjustment is completed within 2 hours. If the anomaly type is drastic temperature fluctuation with an average deviation of 0.09 (mild anomaly), and the current air temperature is 29℃, while the normal threshold range for air temperature during the seedling stage is 23-27℃, then an optimized solution is generated to adjust the air temperature from the current 29℃ to the normal threshold range for the seedling stage [23℃, 27℃], gradually lowering the temperature through the ventilation system, monitoring the temperature every 30 minutes, and ensuring stability within the target range for 1 hour. Simultaneously, if the anomaly type is uneven moisture distribution and the quantified anomaly exceeds the second threshold, the second threshold is determined by collecting 500 sets of abnormal substrate moisture data from the past 12 months, from which 120 sets of abnormal data caused by irrigation equipment malfunctions (sprinkler blockage, pipe leakage) are selected. These 120 sets... First, extract the deviation of the abnormal feature parameter corresponding to each group of data, that is, the matrix moisture deviation calculated in step 301. Then, add up the 120 deviation values ​​to get the sum. For example, if the sum is 28.8, divide the sum by the number of data groups 120 to get the average deviation of these 120 data groups, which is 0.24. To ensure coverage of abnormal situations caused by most equipment failures, round up the average deviation of 0.24 to one decimal place to determine 0.25 as the second threshold. When the abnormality quantification value exceeds 0.25, retrieve the irrigation equipment operation log of the monitoring unit for the past 24 hours (including water pressure records, normal range 0.2 to 0.4 MPa; irrigation duration records, normal single irrigation 15 to 20 minutes). If the log shows that the water pressure is continuously lower than 0.2MPa or irrigation duration fluctuations exceeding 100%. If the time is less than a minute, it is determined to be a device malfunction, and a maintenance command is generated to perform nozzle unclogging (rinsing the nozzles with 50℃ clean water) and pipeline sealing test on the irrigation equipment of the XX monitoring unit (focusing on checking the joints). After completion, the water pressure is tested to the range of 0.2 to 0.4 MPa.

[0031] Step 501: When the anomaly type is pathogenic anomaly, determine the specific disease type based on the anomaly type identifier in the anomaly status discrimination result, determine the warning level based on the anomaly degree quantification value, and generate a disease control alarm based on the disease type and warning level. Specifically, this includes: obtaining the pathogenic anomaly classification result obtained in step 400, extracting the anomaly type identifier and the anomaly degree quantification value (i.e., the average deviation calculated in step 302c); determining the specific disease type based on the anomaly type identifier; if the anomaly type identifier is root rot, the specific disease type is a root fungal disease; if it is leaf spot, the specific disease type is a leaf bacterial disease; and determining the warning level based on the anomaly degree quantification value; if the average deviation is <0.1 (mild anomaly), the warning level is set to level one. Level 1: If the average deviation is between 0.1 and 0.2 (moderate abnormality), the warning level is set to Level 2; if the average deviation is > 0.2 (severe abnormality), the warning level is set to Level 3; a disease control alarm is generated according to the specific disease type and warning level. For example, if the abnormality type is root rot and the warning level is Level 2 (moderate abnormality), a control alarm is generated indicating that the Centella asiatica seedlings in monitoring unit XX have moderate root rot, and it is recommended to spray with XX concentration of hymexazol fungicide within 24 hours and control the substrate humidity to near the lower limit of the normal threshold; if the abnormality type is leaf spot and the warning level is Level 3 (severe abnormality), a control alarm is generated indicating that the Centella asiatica seedlings in monitoring unit XX have severe leaf spot, and it is necessary to immediately remove the diseased plants and carry out preventive spraying on the surrounding seedlings, while strengthening ventilation to reduce air humidity.

[0032] This embodiment identifies environmental stress factors (substrate humidity, light, and temperature) and generates concrete optimization schemes based on average deviations, such as the adjustment range and timing of light and temperature, avoiding growth fluctuations caused by blind regulation. Simultaneously, a second threshold is determined based on historical data to accurately identify water anomalies caused by irrigation equipment malfunctions, generating targeted maintenance instructions to reduce the continuous impact of equipment failures on Centella asiatica seedling growth and improve environmental control efficiency and equipment maintenance accuracy. Early warning levels are divided according to average deviations, and differentiated prevention and control alarms are generated based on disease types, with tiered measures corresponding to mild to severe anomalies. This avoids resource waste caused by over-control and can curb disease spread through timely intervention, reducing the harm of diseases to the seedling population and ensuring the quality of Centella asiatica seedling cultivation. Response logic is designed separately for environmental stress and pathogenic anomalies: environmental anomalies focus on parameter optimization and equipment maintenance, while pathogenic anomalies focus on graded early warning and disease control. Furthermore, the instructions include specific operations, reducing manual decision-making costs and improving the accuracy of Centella asiatica seedling anomaly management.

[0033] Embodiments of the present invention also provide a computing device, including: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the system as described above. All implementations in the above system embodiments are applicable to this embodiment and can achieve the same technical effects.

[0034] Embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the system as described above. All implementations in the above system embodiments are applicable to this embodiment and can achieve the same technical effects.

[0035] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A system for detecting abnormal growth in Centella asiatica seedlings, characterized in that, include: The parameter acquisition module is used to establish a seedbed monitoring grid, which is divided into multiple independent monitoring units; For each monitoring unit, characteristic indicators of seedlings within the unit are extracted, and environmental parameters within the corresponding unit are collected. The calibration module is used to obtain unit-specific calibration parameters based on the spatial coordinates and historical monitoring data of each monitoring unit, and to use the unit-specific calibration parameters to standardize and correct the characteristic indicators and environmental parameters, thereby obtaining the corrected characteristic indicators and environmental parameters. The discrimination module is used to input the corrected feature indicators and environmental parameters into the pre-trained growth anomaly recognition model to obtain the anomaly discrimination result; The classification module is used to distinguish between pathogenic abnormalities and environmental stress abnormalities based on the type of abnormality in the abnormality discrimination results. The instruction generation module is used to generate differentiated response instructions based on the exception type.

2. The Centella asiatica seedling growth abnormality detection system according to claim 1, characterized in that, Establish a seedbed monitoring grid, and divide the monitoring grid into multiple independent monitoring units; For each monitoring unit, characteristic indicators of seedlings within the unit were extracted, and environmental parameters within the corresponding unit were collected, including: The first analysis point, the second analysis point, and the third analysis point were set at three spatial points: the left front corner, the right front corner, and the rear middle corner of the greenhouse seedling bed. Based on the first analysis point, the second analysis point, and the third analysis point, a monitoring grid covering the entire seedling bed was constructed. The seedling bed was then divided into multiple independent monitoring units of equal area according to the monitoring grid. For each monitoring unit, multispectral image data of Centella asiatica seedlings within the monitoring unit is collected using a multispectral imaging device installed above the seedling bed. Based on the multispectral image data, characteristic indicators of the seedlings within each monitoring unit are extracted, including leaf color characteristics, leaf shape characteristics, and physiological reflectance spectral indicators at different wavelengths. Environmental parameters corresponding to each monitoring unit are also collected, including light intensity, air temperature, air humidity, carbon dioxide concentration, and substrate moisture data.

3. The Centella asiatica seedling growth abnormality detection system according to claim 2, characterized in that, Based on the spatial coordinates and historical monitoring data of each monitoring unit, unit-specific calibration parameters are obtained. These parameters are then used to standardize and correct the characteristic indicators and environmental parameters, resulting in the corrected characteristic indicators and environmental parameters, including: Based on the spatial coordinates of each monitoring unit in the seedbed, the spatial location attributes of the monitoring unit are determined. Combined with the long-term environmental monitoring data and historical seedling growth data corresponding to the spatial location attributes of each monitoring unit in the historical monitoring database, unit-specific calibration parameters including light calibration coefficient, temperature calibration coefficient, humidity calibration coefficient and spectral reflectance calibration coefficient are determined for each monitoring unit. Using the unit-specific calibration parameters corresponding to each monitoring unit, the seedling characteristic indicators and environmental parameters of the corresponding monitoring unit are standardized and corrected respectively to obtain the corrected seedling characteristic indicators and corrected environmental parameters of each monitoring unit.

4. The Centella asiatica seedling growth abnormality detection system according to claim 3, characterized in that, Based on the spatial coordinates of each monitoring unit in the seedbed, the spatial location attributes of the monitoring units are determined. Combining this with long-term environmental monitoring data and historical seedling growth data corresponding to the spatial location attributes of each monitoring unit in the historical monitoring database, unit-specific calibration parameters are determined for each monitoring unit, including light calibration coefficient, temperature calibration coefficient, humidity calibration coefficient, and spectral reflectance calibration coefficient. Based on the spatial location attributes of the monitoring unit, historical environmental parameter benchmark values ​​and historical seedling spectral reflectance benchmark values ​​within the complete growth cycle corresponding to the corresponding spatial location attributes are extracted from the historical monitoring database. The historical environmental parameter benchmark values ​​include light intensity benchmark values, air temperature benchmark values, and air humidity benchmark values. The light intensity monitoring value and the light intensity reference value are compared to the current monitoring values ​​of the environmental parameters of the monitoring unit to obtain the light calibration coefficient; the air temperature monitoring value and the air temperature reference value are compared to the current monitoring values ​​of the environmental parameters of the monitoring unit to obtain the temperature calibration coefficient; the air humidity monitoring value and the air humidity reference value are compared to the current monitoring values ​​of the environmental parameters of the monitoring unit to obtain the humidity calibration coefficient; and the spectral reflectance calibration coefficient is obtained by comparing the current monitoring value of the physiological reflectance spectral index of the monitoring unit with the historical seedling spectral reflectance reference value.

5. The Centella asiatica seedling growth abnormality detection system according to claim 4, characterized in that, The corrected feature indicators and environmental parameters are input into a pre-trained growth anomaly recognition model to obtain anomaly discrimination results, including: The corrected seedling characteristic indicators and corrected environmental parameters of the same monitoring unit are fused together to generate a multi-dimensional feature vector representing the overall growth status of the monitoring unit. The multidimensional feature vector is input into the pre-trained growth anomaly recognition model. By comparing each feature parameter in the multidimensional feature vector with the threshold range of the corresponding feature parameter under the predefined normal growth state of Centella asiatica, abnormal feature parameters that are greater than or equal to the upper limit of the threshold range or less than or equal to the lower limit of the threshold range are identified, and the deviation of the abnormal feature parameter from the corresponding threshold boundary value is calculated. Based on the tissue culture response characteristics of Centella asiatica under different hormone ratios, the quality index data of each growth stage, and the correlation data between substrate type and seedling physiological response during division propagation, the abnormal characteristic parameters and corresponding deviations were comprehensively analyzed to determine the factors causing the abnormality of the characteristic parameters and obtain the abnormal status discrimination results of Centella asiatica seedlings in the monitoring unit.

6. The Centella asiatica seedling growth abnormality detection system according to claim 5, characterized in that, Based on tissue culture response characteristics of *Centella asiatica* under different hormone ratios, quality index data at each growth stage, and correlation data between substrate type and seedling physiological response during division propagation, a comprehensive analysis was conducted on abnormal characteristic parameters and their corresponding deviations to identify the factors contributing to the abnormality of characteristic parameters. This yielded the results of abnormal state identification for *Centella asiatica* seedlings within the monitoring unit, including: Establish an abnormal feature parameter combination pattern based on the type and corresponding deviation of the abnormal feature parameters; The abnormal feature parameter combination pattern was matched and compared with the pre-stored typical abnormal feature parameter pattern library of Centella asiatica. The typical abnormal feature parameter pattern library of Centella asiatica includes root rot disease feature parameter pattern, leaf spot disease feature parameter pattern, uneven water distribution feature parameter pattern, strong light irradiation feature parameter pattern, and drastic temperature fluctuation feature parameter pattern. When the matching degree between the abnormal feature parameter combination pattern and any typical abnormal feature parameter pattern is greater than the first threshold, it is determined that there is a corresponding type of abnormal state in the Centella asiatica seedlings in the monitoring unit, and an abnormal state discrimination result containing the abnormal type identifier and the degree of abnormality is obtained.

7. The Centella asiatica seedling growth abnormality detection system according to claim 6, characterized in that, Based on the type of abnormality in the abnormality discrimination results, abnormalities are classified into pathogenic abnormalities or environmental stress abnormalities, including: When the abnormality type in the abnormality status identification result is root rot or leaf spot, the abnormality type is classified as pathogenic abnormality; when the abnormality type in the abnormality status identification result is uneven water distribution, strong light exposure, or drastic temperature fluctuation, the abnormality type is classified as environmental stress abnormality.

8. The Centella asiatica seedling growth abnormality detection system according to claim 7, characterized in that, Generate differentiated response instructions based on the anomaly type, including: When the anomaly type is environmental stress anomaly, the specific environmental stress factor is determined according to the anomaly type identifier in the anomaly state discrimination result, and the corresponding environmental parameter optimization scheme is generated in combination with the anomaly degree quantification value; when the anomaly type identifier is uneven water distribution and the anomaly degree quantification value exceeds the second threshold, it is analyzed and judged whether it is caused by irrigation equipment failure. If it is determined to be equipment failure, an equipment maintenance instruction is generated. When the anomaly type is pathogenic, the specific disease type is determined based on the anomaly type identifier in the anomaly status discrimination result, the warning level is determined by combining the anomaly degree quantification value, and a disease prevention and control alarm is generated based on the disease type and warning level.

9. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the system as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, implements the system as described in any one of claims 1 to 8.

Images