Lysimeter detection management method and device, electronic equipment and storage medium

Abstract

The invention relates to a lysimeter detection management method and device, electronic equipment and a storage medium, and belongs to the field of equipment detection calibration, and the method comprises the following steps: obtaining weighing data sent by a weighing sensor; low-pass filtering is applied to process weighing data, and high-frequency tiny noise is filtered out; identifying the weighing data by using the trained LSTM classification model, and determining abnormal data; judging whether the abnormal data accords with a first condition: the abnormal duration of the abnormal data in the sliding time window is lower than a preset value; the second condition is that the instantaneous change rate of the abnormal data is greater than a change rate threshold; the third condition is that the abnormal data accords with a symmetric peak form; if the abnormal data meet the first condition, the second condition and the third condition, applying a trained regression model based on machine learning to predict real data corresponding to the abnormal data, and replacing the abnormal data with the real data to obtain filtered weighing data; otherwise, abnormal prompt information is generated. The monitoring precision of the lysimeter is improved.

Description

Technical Field

[0001] The present application relates to the technical field of equipment detection and calibration, and in particular to a lysimeter detection and management method, device, electronic equipment and storage medium. Background Art

[0002] A lysimeter is a standard test device developed and manufactured for measuring evaporation and groundwater-soil water conversion in farmland. By recording and analyzing the soil water cycle, it can be used to study crop growth patterns and the evaporation characteristics of different soils.

[0003] The lysimeter's soil box is filled with soil, exposed to the surface, and planted with crops. A load cell is mounted at the bottom of the box, measuring the overall weight of the box and the soil column within it. The lysimeter also features a sump to collect leaking water and a drainage system to supply water to the soil column. Under conditions simulated by the drainage system, the load cell measures changes in soil water storage. This represents water gain after precipitation or irrigation, or water loss due to transpiration and evaporation, making it an important variable for studying crop and soil evaporation characteristics.

[0004] However, the lysimeter is a high-precision, high-sensitivity instrument. External wind disturbances, movement of personnel outside, surrounding construction vibrations, the presence of large and medium-sized animals, and accidental stepping by personnel will all be detected by the soil lysimeter, leading to abnormal changes in the weighing value, thereby reducing the monitoring accuracy of the lysimeter. Summary of the Invention

[0005] In order to improve the monitoring accuracy of a lysimeter, the present application provides a lysimeter detection management method, device, electronic equipment and storage medium.

[0006] In a first aspect, the present application provides a lysimeter detection and management method, which adopts the following technical solutions: Obtain the time-series continuous weighing data sent by the weighing sensor; Applying low-pass filtering to process the weighing data to filter out high-frequency tiny noise; Applying the trained LSTM classification model to identify the weighing data and determine abnormal data in the weighing data; Determine whether the abnormal data meets the following conditions: First condition: determine whether the abnormal duration of the abnormal data in the sliding time window is lower than a preset value; Second condition: determining that the instantaneous change rate of the abnormal data is greater than the change rate threshold; The third condition: determining that the abnormal data conforms to a symmetrical peak shape; If the abnormal data meets the first condition, the second condition and the third condition, the trained regression model based on machine learning is applied to predict the real data corresponding to the abnormal data, and the real data is used to replace the abnormal data to obtain the filtered weighing data; Otherwise, an exception prompt message is generated.

[0007] By adopting the above technical solution, after the electronic device obtains the weighing data from the weighing sensor, it first filters out high-frequency tiny noise through low-pass filtering technology to eliminate the impact of tiny wind disturbances, the movement of peripheral personnel, and surrounding construction vibrations on the monitoring data. It further identifies abnormal data through LSTM, and after judging under three conditions that the abnormal data is caused by large impacts such as the stay of large and medium-sized animals and accidental stepping by people, it corrects the abnormal data and issues timely warnings for non-impact anomalies, effectively improving the accuracy and reliability of weighing data.

[0008] Furthermore, determining that the abnormal data conforms to a symmetrical peak morphology includes: Extract morphological features from time series data with symmetrical spike morphology; Establishing a morphological model according to each of the morphological features; Describing the morphological features in the morphological model using discrete values ​​to obtain a first feature code; Describing the abnormal data using discrete values ​​to obtain a second feature code; The coincidence rate of the first feature code and the second feature code is determined by comparison. If the coincidence rate reaches a coincidence threshold, it is determined that the abnormal data conforms to a symmetrical spike shape.

[0009] By adopting the above technical solution, the symmetrical spike morphology features are extracted and a model is established. The model and abnormal data are converted into discrete feature codes, and the degree of morphological matching is determined by the overlap rate. This method accurately identifies symmetrical spike morphology, provides a reliable basis for subsequent impact interference correction, and improves the accuracy of anomaly type identification.

[0010] Furthermore, before applying the trained machine learning-based regression model to predict the real data corresponding to the abnormal data, the method further includes: Get historical data sent by the weighing sensor; Filtering the weighing data at the time of impact and the weighing data at N time points before the impact from the historical data, and marking the true value corresponding to the time of impact, wherein the true value includes any one of manually calibrated data, data of weighing sensors at adjacent positions without impact at the same time, and data calculated based on stable data before and after the impact; Establishing multiple training samples, each training sample includes input features consisting of data at N time points before the impact occurs and weighing data at the time the impact occurs, and a target label consisting of a true value corresponding to the time the impact occurs; The training samples are applied to enable the regression model based on machine learning to learn the mapping relationship of the weighing data reflecting the true value, thereby obtaining a trained regression model based on machine learning.

[0011] By employing this technical solution, we construct training samples from historical impact data and annotated true values, allowing the regression model to learn the mapping relationship between weighing data and true values. The resulting trained model can accurately predict the true data at the time of impact, providing reliable model support for subsequent anomaly correction and improving data repair accuracy.

[0012] Furthermore, if the abnormal data does not meet the first condition, the second condition, and the third condition, the method further includes: Obtaining a surface monitoring image at a time corresponding to the abnormal data through a pre-installed lysimeter surface monitoring camera; Obtaining a sensor monitoring image at a time corresponding to the abnormal data through a pre-installed weighing sensor monitoring camera; Using the trained animal recognition model to identify the surface monitoring image, determine whether there is a foreign object in the surface monitoring image; if there is a foreign object, generate a foreign object abnormality prompt message; Applying the trained device detection model to identify the sensor monitoring image, determine whether there is any abnormality in the appearance of the weighing sensor; if there is an abnormality, generate a warning message indicating that the weighing sensor is installed abnormally; If there is no foreign matter in the surface monitoring image and there is no abnormality in the appearance of the weighing sensor, a warning message indicating that the weighing sensor is damaged is generated.

[0013] By employing this technical solution, the system combines dual-camera imagery with AI recognition to identify non-impact anomalies, such as lodged foreign objects and sensor installation anomalies, and accurately provides prompts. If no external cause is apparent, a warning of sensor damage is issued. This intelligently locates the cause of the anomaly, improving troubleshooting efficiency and accuracy.

[0014] Furthermore, the lysimeter includes a plurality of weighing sensors, and if the weighing data has no abnormal data in a continuous time period, the method further includes: Obtaining weighing data sent by each weighing sensor; Comparing the similarity between any two weighing data; If there is a similarity lower than the similarity threshold, the corresponding weighing sensor is determined to be an abnormal weighing sensor; Analyze the historical weighing data of the abnormal weighing sensor to determine whether there is a data mutation point; If there is a data mutation point, the weighing data corresponding to the abnormal weighing sensor is temporarily eliminated, and the remaining weighing data is used as the test data; If there is no data mutation point, the mean of each weighing data is taken as the test data.

[0015] By adopting the above technical solution, when the data is continuous and normal, the abnormal sensor is located by comparing the similarity of multi-sensor weighing data; combined with the sudden change of historical data, the outliers are eliminated or the average is taken to ensure the reliability of the test data and improve the measurement stability.

[0016] Furthermore, the comparing the similarity between any two weighing data includes: Establish a coordinate system about time and weighing value; Marking the two weighing data on the coordinate system; Determine a weighing peak value of each weighing data and a peak time corresponding to the weighing peak value; In chronological order, the two weighing peak values ​​with the closest peak time in the two weighing data are used as a group of comparison data, and the next weighing peak value of each of the two weighing data is combined as a second group of comparison data until the weighing data time ends, and multiple groups of comparison data are determined in sequence; Determine a first difference between two peak times and a second difference between two weighing peaks in each set of comparison data; In each set of comparison data, if the first difference is lower than a first threshold, the first mark value is determined to be 1; if the first difference is not lower than the first threshold, the first mark value is determined to be 0; if the second difference is lower than a second threshold, the second mark value is determined to be 1; if the second difference is not lower than the second threshold, the second mark value is determined to be 0; Adding the first label value and the second label value to obtain a label value corresponding to each group of comparison data; The ratio of the number of the marked values ​​2 in each group of comparison data to the total number of comparison data groups is determined as the similarity between the two weighing data.

[0017] By adopting the above technical solution, through the dual-dimensional comparison of peak time and weighing value, the similarity is quantified into the proportion of marked values ​​meeting the standard, and the consistency of the two weighing data is accurately measured, providing a reliable basis for abnormal sensor identification and improving the scientific nature of data comparison.

[0018] Furthermore, the method further comprises: Establish a data display interface; The weighing data sent by the weighing sensor, the weighing data after abnormal data is marked, and the weighing data after filtering are synchronously displayed on the data display interface; Abnormal prompt information is displayed on the data display interface.

[0019] By adopting the above technical solution, the data display interface simultaneously displays the original weighing data, abnormality-marked data, and filtered data, and displays abnormality prompts. This achieves full data process visualization, facilitates intuitive tracking of data changes and abnormality handling results, and improves data monitoring and analysis efficiency.

[0020] In a second aspect, the present application provides a lysimeter detection and management device, which adopts the following technical solution: The weighing data acquisition module is used to obtain the time-series continuous weighing data sent by the weighing sensor; A low-pass filter processing module is used to apply low-pass filtering to the weighing data to filter out high-frequency tiny noise; An abnormal data identification module is used to apply the trained LSTM classification model to identify the weighing data and determine abnormal data in the weighing data; The condition judgment module is used to judge whether the abnormal data meets the following conditions: First condition: determine whether the abnormal duration of the abnormal data in the sliding time window is lower than a preset value; Second condition: determining that the instantaneous change rate of the abnormal data is greater than a threshold; The third condition: determining that the abnormal data conforms to a symmetrical peak shape; a data filtering module, configured to, when the abnormal data all meet the first condition, the second condition, and the third condition, apply a trained regression model based on machine learning to predict real data corresponding to the abnormal data, and replace the abnormal data with the real data to obtain filtered weighing data; The abnormality prompt module is used to generate abnormality prompt information when the abnormal data does not meet the first condition, the second condition and the third condition.

[0021] By adopting the above technical solution: after the weighing data acquisition module obtains the weighing data of the weighing sensor, the low-pass filtering processing module first filters out high-frequency tiny noise through low-pass filtering technology to eliminate the impact of tiny wind disturbances, peripheral personnel walking, surrounding construction vibrations, etc. on the monitoring data. The abnormal data identification module further identifies abnormal data through LSTM, and after the condition judgment module triple-conditionally judges that the abnormal data is a data anomaly caused by large impacts such as the stay of large and medium-sized animals and accidental stepping by people, the data filtering module corrects the abnormal data, and the abnormal prompt module promptly warns of non-impact anomalies, effectively improving the accuracy and reliability of weighing data.

[0022] In a third aspect, the present application provides an electronic device, which adopts the following technical solution: An electronic device, comprising: at least one processor; Memory; At least one computer program, wherein the at least one computer program is stored in the memory and configured to be executed by the at least one processor, and the at least one computer program is configured to: execute a lysimeter detection and management method as described in any one of the first aspects.

[0023] By adopting the above technical solution, the processor executes the computer program in the memory, obtains the weighing data of the weighing sensor, and first uses low-pass filtering technology to filter out high-frequency tiny noise, eliminate the impact of tiny wind disturbances, peripheral personnel walking, surrounding construction vibrations, etc. on the monitoring data, and further uses LSTM to identify abnormal data. After judging under three conditions that the abnormal data is a data anomaly caused by large impacts such as the stay of large and medium-sized animals and accidental stepping by people, the abnormal data is corrected, and timely warnings are issued for non-impact anomalies, effectively improving the accuracy and reliability of weighing data.

[0024] In a fourth aspect, the present application provides a computer-readable storage medium, which adopts the following technical solution: A computer-readable storage medium stores a computer program that can be loaded by a processor and executes a lysimeter detection and management method as described in any one of the first aspects.

[0025] By adopting the above technical solution, the processor executes the computer program in the computer-readable storage medium, obtains the weighing data of the weighing sensor, and first filters out high-frequency tiny noise through low-pass filtering technology to eliminate the impact of tiny wind disturbances, the movement of peripheral personnel, and surrounding construction vibrations on the monitoring data. It further identifies abnormal data through LSTM, and after judging under three conditions that the abnormal data is a data anomaly caused by large impacts such as the stay of large and medium-sized animals and accidental stepping by people, it corrects the abnormal data and issues timely warnings for non-impact anomalies, effectively improving the accuracy and reliability of weighing data.

[0026] In summary, this application includes at least one of the following beneficial technical effects: 1. After the electronic device obtains the weighing data from the weighing sensor, it first uses low-pass filtering technology to filter out high-frequency noise and eliminate the impact of small wind disturbances, the movement of people outside, and surrounding construction vibrations on the monitoring data. It further uses LSTM to identify abnormal data. After determining under three conditions that the abnormal data is caused by large impacts such as the presence of large or medium-sized animals or accidental stepping on by people, it corrects the abnormal data and issues timely warnings for non-impact anomalies, effectively improving the accuracy and reliability of weighing data. 2. For non-impact anomalies, the system combines dual-camera images with AI recognition to identify causes such as lodged foreign objects and sensor installation anomalies, providing accurate prompts. If there are no obvious external causes, it will warn of sensor damage. This intelligently locates the cause of the anomaly, improving troubleshooting efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 2 is a cross-sectional view of the lysimeter in the embodiment of the present application.

[0028] Figure 2 yes Figure 1 Enlarged schematic diagram of part A.

[0029] Figure 3 1 is a diagram of the electrical control structure of the lysimeter in the embodiment of the present application.

[0030] Figure 4 It is a flow chart of the lysimeter detection management method in the embodiment of the present application.

[0031] Figure 5 It is a structural block diagram of the lysimeter detection and management device in an embodiment of the present application.

[0032] Figure 6 It is a structural block diagram of the electronic device in the embodiment of the present application.

[0033] Figure numerals: 1. Basic platform; 11. Sump; 12. Drainage ditch; 13. Sump pump; 14. Sump level sensor; 2. Protective shell; 21. Cylinder; 22. Stairwell; 23. Top cover; 3. Soil box; 31. Filter layer; 4. Weighing platform; 41. Weighing base; 42. Weighing sensor; 43. Equal-diameter tee; 5. Soil moisture sensor; 6. Controller; 7. Communication module; 8. Cloud platform; 9. Pressure measuring tube; 10. Water level control equipment; 1001. First water tank; 1002. Second water tank; 1003. Water supply tank; 1004. First solenoid valve; 1005. Water volume sensor; 1006. Second water tank water level sensor; 1007. Water supply pump; 15. Surface monitoring camera; 16. Weighing sensor monitoring camera; 300. Electronic equipment. DETAILED DESCRIPTION

[0034] To make the purpose, technical solutions, and advantages of the embodiments of this application more clear, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0035] In this document, the term "and / or" simply describes a relationship between related objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, or B exists alone. Furthermore, the character " / " in this document, unless otherwise specified, generally indicates an "or" relationship between the related objects.

[0036] The present application embodiment provides a lysimeter, referring to Figure 1 and Figure 3 The lysimeter includes a basic platform 1, a protective shell 2, a soil box 3, a weighing platform 4 and an electronic device 300.

[0037] First, the basic platform 1 is used to support the protective shell 2, the soil box 3 and the weighing platform 4, and the basic platform 1 is pre-set underground.

[0038] Protective shell 2, used to protect soil box 3, is placed on foundation platform 1 and is surrounded by soil. Protective shell 2 comprises a cylindrical body 21 and a stairwell 22, which are connected and internally communicated with each other. The top of stairwell 22 serves as the entrance to protective shell 2, allowing personnel to enter cylindrical body 21 through stairwell 22.

[0039] The soil box 3 and the weighing platform 4 are arranged in the cylinder 21. The soil box 3 is a cylindrical barrel with an open top. The soil box 3 is made of 6mm carbon steel plate on all sides and 8mm carbon steel plate on the bottom. The soil is filled in the soil box 3 according to the experimental requirements, and plants are planted in the soil, and the upper end of the soil box 3 is exposed to the ground. Among them, the soil can be filled with different materials according to the experimental requirements, and the plants planted can be crops or ordinary vegetation according to the research and development requirements. The protective shell 2 also includes a circular top cover 23, the outer edge of the top cover 23 is connected to the top edge of the cylinder 21, and the inner edge is connected to the upper end of the soil box 3. The inner edge of the top cover 23 is higher than the outer edge, so the top cover 23 is an arc-shaped slope with a high middle and low sides, which effectively drains the accumulated water around the soil box 3.

[0040] The weighing platform 4 includes a weighing base 41, which can be cast from concrete and fixed to the base platform 1. The weighing platform 4 also includes three load cells 42 mounted on the weighing base 41. The top ends of the load cells 42 are fixed to the bottom end surface of the soil box 3, and the three load cells 42 are centrally symmetrical about the center of the bottom end surface of the soil box 3. The load cells 42 measure the total weight of the soil box 3 and the soil inside.

[0041] To monitor soil moisture, multiple soil moisture sensors 5 are arranged along the axial line from top to bottom within the soil box 3. These sensors are suitable for measuring soil moisture, offering high accuracy, fast response, and stable output. They are minimally affected by soil salinity and are suitable for a wide range of soil types. They can be buried in the soil for extended periods, are resistant to long-term electrolysis and corrosion, and are vacuum-sealed and completely waterproof. Furthermore, because soil temperature and humidity near the atmosphere fluctuate frequently, the spacing between the soil moisture sensors 5 from top to bottom is increasingly larger to improve detection sensitivity, resulting in a denser arrangement of sensors near the surface and a sparser arrangement farther from the surface.

[0042] A sensor mounting hole is provided on the side wall of the soil box 3, and the soil moisture and temperature sensor is inserted into the soil through the sensor mounting hole. A connector is installed outside the sensor mounting hole. A high waterproof rubber ring is fixed on the side of the connector close to the side wall of the soil box 3. The connector is fixed to the soil box 3 by screws. The wires of the soil moisture, temperature and humidity sensor pass through the connector and are arranged in the wire groove on the outer wall of the soil box 3. Therefore, the soil moisture and temperature sensor is stably installed on the soil box 3.

[0043] Reference Figure 1 and Figure 3 A control box is disposed within the protective housing 2, and a controller 6 is disposed within the control box. Controller 6 is connected to each soil moisture and temperature sensor, and is also connected to a communication module 7. Communication module 7 transmits soil moisture and temperature data from the soil moisture and temperature sensors to a cloud platform 8. Similarly, controller 6 is connected to each weighing sensor 42 and uploads weighing data to the cloud platform 8. The electronic device 300 can then obtain soil moisture data, soil temperature data, and weighing data via the cloud platform 8.

[0044] Reference Figure 1 and Figure 2A filter layer 31 is provided at the inner bottom of the soil box 3. The filter layer 31 can separate water leaking from the soil. A sump 11 and a drainage ditch 12 are provided on the foundation platform 1. A sump pump 13 and a sump level sensor 14 are provided in the sump 11. The drainage ditch 12 is connected to the outside world, and the discharge port of the sump pump 13 is connected to the drainage ditch 12. An outlet pipe is provided on the side wall of the soil box 3 at the lower end of the filter layer 31. The outlet pipe is connected to an equal-diameter tee 43. One end of the outlet pipe is connected to a pressure measuring tube 9. The pressure measuring tube 9 can be a transparent tube and is vertically upward. The groundwater level in the soil box 3 can be intuitively reflected through the pressure measuring tube 9. The other end is connected to a hose and is connected to the water level control device 10. The hose connection is smooth and does not affect the weighing of the weighing sensor 42. The sump pump 13 and the sump level sensor 14 are connected to the controller 6. The sump pump 13 and the sump liquid level sensor 14 are connected to the controller 6. When the liquid level in the sump 11 is higher than the first liquid level threshold, the sump pump 13 is turned on to drain the water in the sump 11 to the drain ditch 12. When the liquid level is lower than the second liquid level threshold, the sump pump 13 is turned off.

[0045] The water level control device 10 includes a first water tank 1001, a second water tank 1002, and a water supply tank 1003. The first water tank 1001 is connected to a water pipe that feeds into the sump 11, and a first solenoid valve 1004 and a water level sensor 1005 are installed on the water pipe. A second water tank water level sensor 1006 is installed in the second water tank 1002. A water supply pump 1007 is installed on the pipe connecting the water supply tank 1003 and the second water tank 1002. The water supply pump 1007 opens the water supply tank 1003 to supply water to the second water tank 1002. The first solenoid valve 1004, water level sensor 1005, and water supply pump 1007 are all connected to the controller 6.

[0046] The controller 6 controls the water level control device 10 according to the actual groundwater level, and needs to make the liquid level in the soil box 3 consistent with the actual groundwater level.

[0047] After determining the actual groundwater level through technical means, controller 6 adjusts the water supply pump 1007 or the first solenoid valve 1004 to adjust the water supply, aligning the water level threshold in first water tank 1001 with the actual groundwater level. If the actual groundwater level remains unchanged and the second water tank water level sensor 1006 detects that the liquid level is greater than the water level threshold, controller 6 opens first solenoid valve 1004. The water discharged into sump 11 represents seepage from the soil column within soil box 3. The water volume sensor 1005 records the amount of seepage, which controller 6 then transmits to cloud platform 8. If the second water tank water level sensor 1006 detects that the liquid level is below the water level threshold, controller 6 opens water supply pump 1007 to restore the liquid level to the water level threshold.

[0048] When the groundwater level rises, the water level threshold is updated, and the controller 6 turns on the water supply pump 1007 to supply water to the second water tank 1002; when the groundwater level drops, the water level threshold is updated, and the controller 6 opens the first solenoid valve 1004 to make the water levels in the first water tank 1001, the second water tank 1002 and the soil box 3 consistent.

[0049] A lysimeter surface monitoring camera 15 is also installed on the surface of the lysimeter to photograph the surface and transmit the surface monitoring image to the controller 6. A weighing sensor 42 monitoring camera is also installed on the weighing platform 4 to photograph the weighing sensor 42 and transmit the sensor monitoring image to the controller 6. The controller 6 uploads the obtained image to the cloud platform 8.

[0050] For the soil column measured by the soil lysimeter, the water balance equation is: △S=P+I+Q-△R-ET; Where △S is the change in soil water storage, P is precipitation, I is irrigation, Q is groundwater flow, △R is net surface runoff, and ET is evaporation.

[0051] For the lysimeter system, △R can generally be ignored and the equation can be changed to ET=P+I+Q-△S; Precipitation (P) and irrigation (I) can be measured directly using rain gauges and water meters. The change in soil water storage (ΔS) represents the increase in water after precipitation or irrigation, or measures water loss through evaporation and transpiration. These are more difficult to measure, so high-precision weighing systems are used to measure ΔS.

[0052] Groundwater flow Q represents the amount of water supplied to and removed from the soil column by the lysimeter water level control device 10. When the groundwater level remains constant, the lysimeter does not require additional water flow adjustment. In this case, the amount of water supplied to the soil column by the lysimeter (Q) equals the amount of groundwater recharged to the soil (Eg), i.e., Eg = Q (equivalent to the amount of groundwater actively recharging the soil under natural conditions). The amount of water removed from the soil column by the lysimeter (Q) equals the amount of soil water recharged to the groundwater (Rg), i.e., Rg = Q (equivalent to the amount of excess soil water that infiltrates and recharges the groundwater).

[0053] When the actual groundwater level rises or falls, the water level control device 10 needs to "add or remove water" to keep the water level in the soil box 3 consistent with the actual water level (to ensure that the experimental conditions are consistent with nature). At this time, Eg and Rg need to be calculated in combination with the water level change (△H) and the coefficients (a, b): If the actual groundwater level rises by △H: Eg=Qa·△H; Q is the water supply of the water level control device 10, and a·△H is the amount of water that naturally infiltrates into the soil column due to rising water levels; If the actual groundwater level drops by △H: Rg=b·△HQ; Q is the drainage volume of the water level control device 10, that is, the leakage volume, and b·△H is the volume of water lost due to natural infiltration in the soil column as the water level drops.

[0054] a and b are coefficients determined from water absorption and dehydration experiments. In practice, a and b can be calculated using ΔQ, which is the change in water content when the groundwater level rises or falls, and can be measured using a moisture meter.

[0055] Therefore, staff can conduct experiments based on the lysimeter to obtain research variables such as transpiration, groundwater recharge to soil water, and soil water recharge to groundwater.

[0056] The present application embodiment discloses a lysimeter detection and management method. Figure 4 , executed by an electronic device, which can be a server or a terminal device. The server can be a standalone physical server, a server cluster or distributed system consisting of multiple physical servers, or a cloud server providing cloud computing services. The terminal device can be, but is not limited to, a smartphone, tablet computer, desktop computer, etc., including (steps S101 to S106): Step S101: Acquire the time-series continuous weighing data sent by the weighing sensor.

[0057] Specifically, the electronic device obtains the weighing data sent by the weighing sensor through the cloud platform, wherein the weighing data is time-series continuous and can reflect the changes in the weighing data including time characteristics.

[0058] Step S102: Apply low-pass filtering to process the weighing data to filter out high-frequency tiny noise.

[0059] Specifically, electronic interference from the load cell itself, environmental vibration, or electromagnetic interference can cause high-frequency, subtle noise in the weighing data. This noise doesn't affect the overall data trend, but it can make the data appear cluttered. Therefore, low-pass filtering removes this high-frequency noise, leaving only the true, low-frequency weight change information, resulting in smoother data.

[0060] In addition to these high-frequency, subtle noises, large external impacts can also affect weighing data. Examples include large or medium-sized animals lingering on the lysimeter, strong wind disturbances, and people accidentally stepping on them. These impacts typically cause the weighing value to increase or decrease abnormally, then change direction and become smaller or larger. To filter out the noise caused by these large impacts, an intelligent recognition algorithm is further employed.

[0061] Step S103: Apply the trained LSTM classification model to identify the weighing data and determine abnormal data in the weighing data.

[0062] Specifically, the electronic device trains the LSTM classification model in advance. During training, multiple sets of weighing data sets containing normal and abnormal labels are obtained, and the data sets are divided into training sets and validation sets. An LSTM model is constructed, trained with the training set, and parameters are adjusted with the validation set until the classification accuracy meets the standard, thereby obtaining a trained LSTM classification model.

[0063] When the weighing data is input into the LSTM classification model, abnormal data in the weighing data can be identified.

[0064] Step S104: Determine whether the abnormal data meets the following conditions: The first condition is to ensure that the duration of abnormal data within the sliding time window is less than the preset value. Second condition: determine whether the instantaneous change rate of abnormal data is greater than the change rate threshold; The third condition: Make sure that the abnormal data conforms to a symmetrical spike shape.

[0065] Specifically, the abnormal data caused by large shocks are characterized by short duration, steep fluctuations, and symmetrical spike shapes.

[0066] Therefore, in the first condition, the sliding time window is set to match the maximum duration of the shock, for example, 2 minutes, and the corresponding preset value is also set to 2 minutes. If the real-time monitoring of the weighing signal changes from normal to abnormal and then back to normal within the sliding time window within 2 minutes, it meets the short-term characteristics.

[0067] In the second condition, impact can cause a significant spike in the rate of change. For example, a person accidentally stepping on something can cause the weighing value to change by ±5-10mm within 1 second, with a rate of change of ±5-10mm / s. Therefore, the instantaneous rate of change of the weighing data can be used to determine whether there is a sharp fluctuation. When calculating the instantaneous rate of change, the difference between the weighing values ​​of two adjacent sampling points is divided by the sampling interval. The rate of change threshold is set and adaptively adjusted based on 3-5 times the standard deviation of historical normal data to avoid the poor adaptability of fixed thresholds.

[0068] In the third condition, the spike caused by the impact is symmetrical in the vertical direction and / or in the horizontal direction, and first deviates in one direction and then recovers in the opposite direction. To compare and determine whether the third condition is met, the electronic device performs the following method (steps S11 to S15): Step S11: extracting morphological features from time series data with symmetrical spike morphology.

[0069] Specifically, the electronic device obtains time series data with symmetrical spike morphology from historical data, and then extracts morphological features, which include: peak height of the spike (the difference between the highest point and the baseline), duration (the time from the start of fluctuation to the recovery of the baseline), left-right symmetry (whether the slopes of the rising segment and the falling segment are close), baseline position (stable value before fluctuation), etc.

[0070] Step S12: establishing a morphological model according to various morphological features.

[0071] Specifically, the electronic device integrates the extracted multiple morphological features to form a standard model that can "define the symmetrical spike morphology". The model is essentially a "feature set" that clearly defines the feature range that the symmetrical spike should meet.

[0072] Step S13: using discrete values ​​to describe the morphological features in the morphological model to obtain a first feature code.

[0073] Specifically, the features of the morphological model are “digitized and discretized” and converted into a string of comparable “codes” to facilitate the subsequent calculation of the overlap rate.

[0074] For example: discretize model features: Symmetry: "Rise / fall time difference ≤ 0.1 second" is recorded as "1"; Peak value: "8-12 kg" is recorded as "2"; Duration: "0.8-1.2 seconds" is recorded as "3"; The first feature code is "1-2-3", that is, each number corresponds to a discrete result of a feature.

[0075] Step S14: Use discrete values ​​to describe the abnormal data to obtain a second feature code.

[0076] Specifically, the electronic device first extracts the morphological features of the abnormal data that needs to be judged, and then converts it into a "second feature code" using the same discrete rules.

[0077] The characteristics of some abnormal data are "rise time 0.4 seconds, fall time 0.5 seconds (difference 0.1 seconds), peak value 9 kg, duration 1.0 second". After discretization according to the rules, the second characteristic code is "1-2-3".

[0078] Step S15: comparing and determining the coincidence rate of the first feature code and the second feature code; if the coincidence rate reaches a coincidence threshold, it is determined that the abnormal data conforms to the upper and lower symmetrical peak shape.

[0079] Specifically, the proportion of discrete values ​​at the same position that are consistent. For example, the first feature code "1-2-3" and the second feature code "1-2-3" have a 100% coincidence rate; if the second feature code is "1-4-3", the coincidence rate is 2 / 3≈67%.

[0080] The overlap threshold is set according to the accuracy requirements, such as 80%. If the overlap rate ≥ the overlap threshold, the abnormal data is judged to be a "symmetrical spike shape".

[0081] Determine whether the abnormal data meets the first condition, the second condition and the third condition. If so, execute step S105: apply the trained machine learning-based regression model to predict the real data corresponding to the abnormal data, and replace the abnormal data with the real data to obtain the filtered weighing data; otherwise, execute step S106: generate abnormal prompt information.

[0082] First, the electronic device pre-trains a regression model based on machine learning. The method includes (steps S21 to S24): Step S21: Acquire historical data sent by the weighing sensor.

[0083] Specifically, the electronic device can obtain historical data from the data stored in the cloud platform.

[0084] Step S22: Filter out the weighing data at the time of impact and the weighing data at N time points before the impact from the historical data, and mark the true value corresponding to the time of impact. The true value includes any one of the following: manually calibrated data, data of weighing sensors at adjacent positions without impact at the same time, and data calculated based on stable data before and after the impact.

[0085] Specifically, assuming an impact occurs at a certain moment (such as 10:00:20) and the sensor measurement value is abnormal (such as 210 kg), the data at that moment and the data at N time points before the impact (such as N=3, that is, normal data at 10:00:10, 10:00:00, and 09:59:50) are filtered. These "pre-impact data" can reflect the stable state before the impact and assist the model in determining the true value.

[0086] The measured value at the moment of impact (210 kg) is disturbed and needs to be marked as the true weight. The true value can be obtained by manually weighing the true weight at that moment. Other sensors in the same soil box that were not impacted at the same time can also use their measured values ​​as the true value. Estimated data: Based on the data before impact (e.g. 200.1 kg at 10:00:10) and the data after recovery to stability (e.g. 200 kg at 10:00:30), the actual value at the time of impact is estimated to be approximately 200 kg.

[0087] Step S23: Establish multiple training samples, each training sample includes input features consisting of data at N time points before the impact and weighing data at the time of impact, and a target label consisting of the true value corresponding to the time of impact.

[0088] Step S24: applying the training samples to enable the regression model based on machine learning to learn the mapping relationship from the weighing data to the true value, thereby obtaining a trained regression model based on machine learning.

[0089] Specifically, through a large number of samples, the model will learn rules such as "the more stable the data before the impact, the higher the measured value at the time of impact, and the closer the true value is to the stable value before the impact."

[0090] After training is completed, when a new shock occurs, the model can output the corrected true value by inputting "N data before the shock + current shock measurement value".

[0091] When the electronic device obtains the real data, it replaces the abnormal data and obtains the processed weighing data.

[0092] In another possible implementation, if the abnormal data is not a fluctuation caused by a large impact, that is, the abnormal data does not meet the first, second, and third conditions, it may be that an object has remained on the lysimeter for a long time or the weighing sensor itself has a fault. In order to further eliminate the abnormality, the method further includes (steps S31 to S35): Step S31: obtaining a surface monitoring image at a time corresponding to the abnormal data through a pre-installed lysimeter surface monitoring camera.

[0093] Step S32: obtaining the sensor monitoring image at the time corresponding to the abnormal data through the pre-installed weighing sensor monitoring camera.

[0094] Specifically, the electronic device obtains surface monitoring images and sensor monitoring images through the cloud, and then obtains the surface monitoring images and sensor monitoring images corresponding to the abnormal data moment.

[0095] Step S33: Using the trained animal recognition model to identify the surface monitoring image, determine whether there is a foreign object in the surface monitoring image; if there is a foreign object, generate a foreign object abnormality prompt message.

[0096] Step S34: Apply the trained device detection model to identify the sensor monitoring image and determine whether there is any abnormality in the appearance of the weighing sensor; if there is an abnormality, generate a warning message indicating that the weighing sensor is installed abnormally.

[0097] Specifically, the AI ​​model is used to analyze the surface image. If foreign objects such as animals and discarded items are identified on the surface or soil box of the lysimeter, it means that the abnormal data may be caused by the pressure of these foreign objects, and a foreign object retention abnormality prompt message is issued.

[0098] Analyze sensor images and check for any issues with the sensor's appearance and installation. If any abnormalities are detected, such as a loose sensor, tilted sensor, or a surface clogged with mud, the abnormal data may be caused by an installation problem, and an installation abnormality alert will be issued.

[0099] Step S35: If there is no foreign matter in the surface monitoring image and there is no abnormality in the appearance of the weighing sensor, a warning message indicating that the weighing sensor is damaged is generated.

[0100] Specifically, if there is no foreign matter on the surface, the appearance and installation of the sensor are normal, but the weighing data is still abnormal, it may indicate an internal fault in the sensor, such as circuit damage or sensitivity failure, and a sensor damage prompt message will be issued.

[0101] Furthermore, theoretically, the data of the three weighing sensors are equal, but in reality there are slight differences. If there is no abnormal data in the weighing data within a continuous time period, it can be further ruled out whether there is any abnormality in the weighing data itself without the influence of external force impact. The method further includes (steps S41 to S46): Step S41: Acquire the weighing data sent by each weighing sensor.

[0102] Step S42: Compare the similarity between any two weighing data.

[0103] Specifically, the weight measurement data sent in real time by all load cells in the lysimeter are collected.

[0104] When calculating the similarity between two weighing data, a coordinate system about time and weighing value is established; the two weighing data are marked on the coordinate system; the weighing peak value of each weighing data and the peak time corresponding to the weighing peak value are determined; in chronological order, the two weighing peak values ​​with the closest peak time in the two weighing data are used as a group of comparison data, and the next weighing peak value of each of the two weighing data is combined as a second group of comparison data until the weighing data time ends, and multiple groups of comparison data are determined in sequence; a first difference between the two peak times and a second difference between the two weighing peak values ​​in each group of comparison data are determined; in each group of comparison data, if the first difference is lower than a first threshold value, the first label value is determined to be 1, and if the first difference is not lower than the first threshold value, the first label value is determined to be 0; if the second difference is lower than the second threshold value, the second label value is determined to be 1, and if the second difference is not lower than the second threshold value, the second label value is determined to be 0; the first label value and the second label value are added to obtain the label value corresponding to each group of comparison data; the ratio of the number of label values ​​2 in each group of comparison data to the total number of comparison data groups is determined as the similarity between the two weighing data.

[0105] For example, mark all the data points of sensor A and sensor B in the coordinate system, connect them to form two curves, and determine the peak values: A: (0,100), (2,105), (4,110), (6,105), (8,100); B: (0,102), (2.5,110), (4,112), (6,107), (8,102); The first peak of A (0, 100) and the first peak of B (0, 102) are the first set of comparison data, (2, 105) and (2.5, 108) are the second set of comparison data, and so on, 5 sets of comparison data are obtained.

[0106] Taking the first set of comparison data as an example, the first difference (time difference) is: |4s-4s|=0s; the second difference (weighing value difference) is: |100kg-102kg|=2kg.

[0107] Assuming the first threshold = 1 second and the second threshold = 5 kg, then the first difference (0 s) < the first threshold (1 s) → the first mark value = 1; the second difference (2 kg) < the second threshold (5 kg) → the second mark value = 1; the mark value = 1 + 1 = 2.

[0108] Since there are 5 sets of comparison data in total, and there are 4 sets with a mark value of 2, the similarity = 4 / 5 = 80%, and the similarity between the current two weighing data is 0.8.

[0109] Step S43: If there is a similarity lower than the similarity threshold, the corresponding weighing sensor is determined to be an abnormal weighing sensor.

[0110] Specifically, taking the similarity threshold set to 60% as an example, if the similarity between A and B is 80% (higher than the preset value of 60%), and the similarity between A and C is 30% (lower than 60%), then C is preliminarily determined to be an abnormal sensor.

[0111] Step S44: Analyze the historical weighing data of the abnormal weighing sensor to determine whether there is a data mutation point.

[0112] Specifically, retrieve the abnormal sensor's past measurement data to check for any sudden jumps, such as a sudden jump from 100 kg to 200 kg, or a change from stable fluctuations to chaotic fluctuations. Sudden changes often indicate temporary sensor interference, such as impact, poor contact, or a sudden malfunction. The absence of sudden changes may indicate long-term sensor drift, such as a slow decrease in accuracy.

[0113] Step S45: If there is a data mutation point, the weighing data corresponding to the abnormal weighing sensor is temporarily eliminated, and the remaining weighing data is used as the test data.

[0114] Step S46: If there is no data mutation point, the mean of each weighing data is used as the test data.

[0115] Specifically, if the abnormal sensor has a sudden change in data, its data is temporarily not used, and the data of other normal sensors are used as valid data. If the abnormal sensor has no sudden change point, it may be a slight error, so all sensor data are retained and their average is taken as the final data.

[0116] Furthermore, in order to intuitively display the weighing data and data anomalies, the above method further includes: Establish a data display interface; display the weighing data sent by the weighing sensor, the weighing data after abnormal data is marked, and the weighing data after filtering in a time-synchronous manner on the data display interface; display abnormal prompt information on the data display interface.

[0117] In order to better implement the above method, the present application embodiment also provides a lysimeter detection management device, referring to Figure 5 The lysimeter detection and management device 200 includes: The weighing data acquisition module 201 is used to obtain the time-series continuous weighing data sent by the weighing sensor; A low-pass filter processing module 202 is used to apply low-pass filtering to the weighing data to filter out high-frequency small noise; Abnormal data identification module 203, used to identify weighing data using the trained LSTM classification model and determine abnormal data in the weighing data; The condition judgment module 204 is used to judge whether the abnormal data meets the following conditions: The first condition is to ensure that the duration of abnormal data within the sliding time window is less than the preset value. Second condition: determine whether the instantaneous change rate of abnormal data is greater than the change rate threshold; The third condition: make sure that the abnormal data conforms to the symmetrical spike shape; The data filtering module 205 is used to apply the trained regression model based on machine learning to predict the real data corresponding to the abnormal data when the abnormal data meets the first condition, the second condition and the third condition, and replace the abnormal data with the real data to obtain the filtered weighing data; The abnormality prompt module 206 is used to generate abnormality prompt information when the abnormal data does not meet the first condition, the second condition and the third condition.

[0118] Furthermore, when the condition judgment module 204 determines that the abnormal data conforms to a symmetrical peak shape, it is specifically configured to: Extract morphological features from time series data with symmetrical spike morphology; Establish a morphological model based on various morphological features; Using discrete values ​​to describe the morphological features in the morphological model, and obtaining a first feature code; Use discrete values ​​to describe abnormal data and obtain the second feature code; The coincidence rate of the first characteristic code and the second characteristic code is determined by comparison. If the coincidence rate reaches a coincidence threshold, it is determined that the abnormal data conforms to the symmetrical spike shape.

[0119] Furthermore, the lysimeter detection management device 200 further includes: Historical data acquisition module, used to obtain historical data sent by the weighing sensor; A weighing data screening module is used to screen out the weighing data at the time of impact and the weighing data at N time points before the impact from the historical data, and mark the true value corresponding to the time of impact. The true value includes any one of the following: manually calibrated data, data of weighing sensors at adjacent positions without impact at the same time, and data calculated based on stable data before and after the impact; A training sample establishment module is used to establish multiple training samples, each of which includes input features consisting of data from N time points before the impact and weighing data at the time of impact, and a target label consisting of the true value corresponding to the time of impact; The training module is used to apply training samples to enable the regression model based on machine learning to learn the mapping relationship between the weighing data and the true value, so as to obtain a trained regression model based on machine learning.

[0120] Furthermore, the lysimeter detection management device 200 further includes: A surface monitoring image acquisition module is used to acquire a surface monitoring image at a time corresponding to abnormal data through a pre-installed lysimeter surface monitoring camera; A sensor monitoring image acquisition module is used to acquire the sensor monitoring image at the time corresponding to the abnormal data through a pre-installed weighing sensor monitoring camera; The surface monitoring image judgment module is used to apply the trained animal recognition model to identify the surface monitoring image and determine whether there is a foreign object in the surface monitoring image; if a foreign object is present, a foreign object abnormality prompt message is generated; The sensor monitoring image judgment module is used to apply the trained equipment detection model to identify the sensor monitoring image and determine whether there is any abnormality in the appearance of the weighing sensor; if there is an abnormality, a warning message indicating that the weighing sensor is installed abnormally is generated; The prompt module is used to generate a warning message indicating that the weighing sensor is damaged when there is no foreign matter in the surface monitoring image and there is no abnormality in the appearance of the weighing sensor.

[0121] Furthermore, the lysimeter detection management device 200 also includes: A first weighing data acquisition module is used to acquire the weighing data sent by each weighing sensor; A similarity determination module is used to compare the similarity between any two weighing data; an abnormal weighing sensor determination module, configured to determine that the corresponding weighing sensor is an abnormal weighing sensor if there is a similarity lower than a similarity threshold; Data mutation analysis module, used to analyze the historical weighing data of abnormal weighing sensors and determine whether there is a data mutation point; The processing module is used to temporarily eliminate the weighing data corresponding to the abnormal weighing sensor when there is a data mutation point, and use the remaining weighing data as test data; When there is no data mutation point, the mean of each weighing data is taken as the test data.

[0122] Furthermore, the similarity determination module is specifically configured to: Establish a coordinate system about time and weighing value; Mark the two weighing data on the coordinate system; Determine the weighing peak value of each weighing data and the peak time corresponding to the weighing peak value; In chronological order, the two weighing peaks with the closest peak time in the two weighing data are taken as a group of comparison data, and the next weighing peak value of each of the two weighing data is combined as the second group of comparison data until the weighing data time ends, and multiple groups of comparison data are determined in sequence; Determine a first difference between two peak times and a second difference between two weighing peaks in each set of comparison data; In each set of comparison data, if the first difference is lower than the first threshold, the first mark value is determined to be 1; if the first difference is not lower than the first threshold, the first mark value is determined to be 0; if the second difference is lower than the second threshold, the second mark value is determined to be 1; if the second difference is not lower than the second threshold, the second mark value is determined to be 0; Adding the first label value and the second label value to obtain a label value corresponding to each set of comparison data; The ratio of the number of marked values ​​2 in each group of comparison data to the total number of comparison data groups is determined as the similarity between the two weighing data.

[0123] Furthermore, the lysimeter detection management device 200 also includes: A data display interface establishment module is used to establish a data display interface; The display module is used to synchronously display the weighing data sent by the weighing sensor, the weighing data after abnormal data is marked, and the weighing data after filtering on the data display interface; The prompt information display module is used to display abnormal prompt information on the data display interface.

[0124] The various variations and specific examples of the methods in the aforementioned embodiments are also applicable to the lysimeter detection and management device of this embodiment. Through the aforementioned detailed description of the lysimeter detection and management method, those skilled in the art can clearly understand the implementation method of the lysimeter detection and management device in this embodiment. Therefore, for the sake of brevity of the specification, they will not be described in detail here.

[0125] In order to better implement the above method, the embodiment of the present application provides an electronic device, referring to Figure 6 , electronic device 300 includes: a processor 301, a memory 303, and a display screen 305. The memory 303 and the display screen 305 are both connected to the processor 301, for example, via a bus 302. Optionally, the electronic device 300 may further include a transceiver 304. It should be noted that in actual applications, there is not limited to one transceiver 304, and the structure of the electronic device 300 does not constitute a limitation on the embodiments of the present application.

[0126] Processor 301 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 301 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, or a combination of a DSP and a microprocessor.

[0127] Bus 302 may include a path for transmitting information between the aforementioned components. Bus 302 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. Bus 302 may be divided into an address bus, a data bus, and a control bus.

[0128] The memory 303 may be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto.

[0129] The memory 303 is used to store application code for executing the solution of the present application, and the execution is controlled by the processor 301. The processor 301 is used to execute the application code stored in the memory 303 to implement the content shown in the above method embodiment.

[0130] Figure 6 The electronic device 300 shown is merely an example and should not limit the functions and scope of use of the embodiments of the present application.

[0131] An embodiment of the present application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the lysimeter detection and management method provided in the above embodiment. The processor executes the computer program in the computer-readable storage medium, obtains the weighing data of the weighing sensor, and first uses low-pass filtering technology to filter out high-frequency tiny noise, eliminates the impact of tiny wind disturbances, peripheral personnel walking, surrounding construction vibrations, etc. on the monitoring data, and further identifies abnormal data through LSTM. After judging under three conditions that the abnormal data is a data anomaly caused by a large impact such as the stay of large and medium-sized animals and accidental stepping by people, the abnormal data is corrected, and timely warnings are issued for non-impact anomalies, effectively improving the accuracy and reliability of the weighing data.

[0132] In this embodiment, a computer-readable storage medium may be a tangible device that retains and stores instructions used by an instruction execution device. The computer-readable storage medium may be, but is not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any combination thereof. Specifically, the computer-readable storage medium may be a portable computer disk, a hard disk, a USB flash drive, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disc (DVD), a memory stick, a floppy disk, an optical disc, a magnetic disk, a mechanical encoding device, or any combination thereof.

[0133] The computer program in this embodiment includes program code for executing all of the aforementioned methods. The program code may include instructions corresponding to the steps of the methods provided in the aforementioned embodiments. The computer program can be downloaded from a computer-readable storage medium to various computing / processing devices, or downloaded to an external computer or external storage device via a network (e.g., the Internet, a local area network, a wide area network, and / or a wireless network). The computer program can be executed entirely on a user's computer or as a standalone software package.

[0134] The above are all preferred embodiments of the present application, and are not intended to limit the scope of protection of the present application. Therefore, any equivalent changes made based on the structure, shape, and principle of the present application should be included in the scope of protection of the present application.

[0135] In addition, it should be understood that relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. The terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes 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.

Claims

1. A lysimeter detection and management method, characterized in that: The method is performed by an electronic device and includes: Obtain the time-series continuous weighing data sent by the weighing sensor; Applying low-pass filtering to process the weighing data to filter out high-frequency tiny noise; Applying the trained LSTM classification model to identify the weighing data and determine abnormal data in the weighing data; Determine whether the abnormal data meets the following conditions: First condition: determine whether the abnormal duration of the abnormal data in the sliding time window is lower than a preset value; Second condition: determining that the instantaneous change rate of the abnormal data is greater than the change rate threshold; The third condition: determining that the abnormal data conforms to a symmetrical peak shape; If the abnormal data meets the first condition, the second condition and the third condition, the trained regression model based on machine learning is applied to predict the real data corresponding to the abnormal data, and the real data is used to replace the abnormal data to obtain the filtered weighing data; Otherwise, an exception prompt message is generated.

2. The lysimeter detection and management method according to claim 1, characterized in that: Determining that the abnormal data conforms to a symmetrical peak shape includes: Extract morphological features from time series data with symmetrical spike morphology; Establishing a morphological model according to each of the morphological features; Describing the morphological features in the morphological model using discrete values ​​to obtain a first feature code; Describing the abnormal data using discrete values ​​to obtain a second feature code; The coincidence rate of the first feature code and the second feature code is determined by comparison. If the coincidence rate reaches a coincidence threshold, it is determined that the abnormal data conforms to a symmetrical spike shape.

3. The lysimeter detection and management method according to claim 1, characterized in that: Before applying the trained machine learning-based regression model to predict real data corresponding to the abnormal data, the method further includes: Get historical data sent by the weighing sensor; Filtering the weighing data at the time of impact and the weighing data at N time points before the impact from the historical data, and marking the true value corresponding to the time of impact, wherein the true value includes any one of manually calibrated data, data of weighing sensors at adjacent positions without impact at the same time, and data calculated based on stable data before and after the impact; Establishing multiple training samples, each training sample includes input features consisting of data at N time points before the impact occurs and weighing data at the impact occurs, and a target label consisting of a true value corresponding to the impact occurs; The training samples are applied to enable the regression model based on machine learning to learn the mapping relationship of the weighing data reflecting the true value, thereby obtaining a trained regression model based on machine learning.

4. The lysimeter detection and management method according to claim 1, characterized in that: If the abnormal data does not meet the first condition, the second condition, and the third condition, the method further includes: Obtaining a surface monitoring image at a time corresponding to the abnormal data through a pre-installed lysimeter surface monitoring camera; Obtaining a sensor monitoring image at a time corresponding to the abnormal data through a pre-installed weighing sensor monitoring camera; Using the trained animal recognition model to identify the surface monitoring image, determine whether there is a foreign object in the surface monitoring image; if there is a foreign object, generate a foreign object abnormality prompt message; Applying the trained device detection model to identify the sensor monitoring image, determine whether there is any abnormality in the appearance of the weighing sensor; if there is an abnormality, generate a warning message indicating that the weighing sensor is installed abnormally; If there is no foreign matter in the surface monitoring image and there is no abnormality in the appearance of the weighing sensor, a warning message indicating that the weighing sensor is damaged is generated.

5. The lysimeter detection and management method according to claim 1, characterized in that: The lysimeter includes a plurality of weighing sensors, and if the weighing data does not contain abnormal data in a continuous time period, the method further includes: Obtaining weighing data sent by each weighing sensor; Comparing the similarity between any two weighing data; If there is a similarity lower than the similarity threshold, the corresponding weighing sensor is determined to be an abnormal weighing sensor; Analyze the historical weighing data of the abnormal weighing sensor to determine whether there is a data mutation point; If there is a data mutation point, the weighing data corresponding to the abnormal weighing sensor is temporarily eliminated, and the remaining weighing data is used as the test data; If there is no data mutation point, the mean of each weighing data is taken as the test data.

6. The lysimeter detection and management method according to claim 5, characterized in that: The comparing the similarity between any two weighing data includes: Establish a coordinate system about time and weighing value; Marking the two weighing data on the coordinate system; Determine a weighing peak value of each weighing data and a peak time corresponding to the weighing peak value; In chronological order, the two weighing peak values ​​with the closest peak time in the two weighing data are used as a group of comparison data, and the next weighing peak value of each of the two weighing data is combined as a second group of comparison data until the weighing data time ends, and multiple groups of comparison data are determined in sequence; Determine a first difference between two peak times and a second difference between two weighing peaks in each set of comparison data; In each set of comparison data, if the first difference is lower than a first threshold, the first mark value is determined to be 1; if the first difference is not lower than the first threshold, the first mark value is determined to be 0; if the second difference is lower than a second threshold, the second mark value is determined to be 1; if the second difference is not lower than the second threshold, the second mark value is determined to be 0; Adding the first label value and the second label value to obtain a label value corresponding to each group of comparison data; The ratio of the number of the marked values ​​2 in each group of comparison data to the total number of comparison data groups is determined as the similarity between the two weighing data.

7. The lysimeter detection and management method according to claim 1, characterized in that: The method further comprises: Establish a data display interface; The weighing data sent by the weighing sensor, the weighing data after abnormal data is marked, and the weighing data after filtering are synchronously displayed on the data display interface; Abnormal prompt information is displayed on the data display interface.

8. A lysimeter detection and management device, characterized in that: include: The weighing data acquisition module is used to obtain the time-series continuous weighing data sent by the weighing sensor; A low-pass filter processing module is used to apply low-pass filtering to the weighing data to filter out high-frequency tiny noise; An abnormal data identification module is used to apply the trained LSTM classification model to identify the weighing data and determine abnormal data in the weighing data; The condition judgment module is used to judge whether the abnormal data meets the following conditions: First condition: determine whether the abnormal duration of the abnormal data in the sliding time window is lower than a preset value; Second condition: determining that the instantaneous change rate of the abnormal data is greater than the change rate threshold; The third condition: determining that the abnormal data conforms to a symmetrical peak shape; a data filtering module, configured to, when the abnormal data all meet the first condition, the second condition, and the third condition, apply a trained regression model based on machine learning to predict real data corresponding to the abnormal data, and replace the abnormal data with the real data to obtain filtered weighing data; The abnormality prompt module is used to generate abnormality prompt information when the abnormal data does not meet the first condition, the second condition and the third condition.

9. An electronic device, characterized in that: include: at least one processor; Memory; At least one computer program, wherein the at least one computer program is stored in the memory and configured to be executed by the at least one processor, and the at least one computer program is configured to: execute a lysimeter detection and management method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The invention stores a computer program that can be loaded by a processor and executes a lysimeter detection and management method according to any one of claims 1 to 7.

Images