Device for automatically monitoring canopy penetration rainfall of wind-preventing sand-fixing forest

Abstract

This invention discloses a device for automatically monitoring canopy penetration rainfall in windbreak and sand-fixing forests, relating to the field of forestry eco-hydrological monitoring technology. It includes modules for trigger acquisition, signal processing, parameter analysis, evaporation compensation, cross-validation, and storage. The trigger acquisition module responds to a rainfall initiation signal and collects volumetric flow data of penetrating rainwater from a rain collection trough. The signal processing module extracts discontinuous pulse signals from the data and performs noise reduction and smoothing to generate clean flow time-series data. The parameter analysis module analyzes the peak rainfall intensity and duration. The evaporation compensation module combines real-time wind speed and humidity data to perform dynamic calibration of evaporation loss. The cross-validation module verifies the calibration data against the original flip-count sequence of a high-precision tipping bucket rain gauge. The storage module generates and stores the final cumulative measurement value. This invention can eliminate flow signal noise, achieve dynamic compensation for evaporation loss, and improve the accuracy and reliability of canopy penetration rainfall monitoring.

Description

Technical Field

[0001] This invention belongs to the field of forestry ecological hydrological monitoring technology, specifically a device for automatically monitoring the amount of rainfall penetrating the canopy of windbreak and sand-fixing forests. Background Technology

[0002] Conventional monitoring of canopy penetration rainfall in windbreak and sand-fixing forests employs rainwater collection troughs in conjunction with basic flow acquisition components or tipping bucket rain gauges to acquire data. Rainfall values ​​are obtained by directly collecting volumetric flow data of penetrating rainwater and performing simple summation calculations. This method relies on conventional flow signal processing techniques and fixed parameter compensation logic to complete the monitoring process. However, this existing monitoring method is not specifically adapted to the flow characteristics of canopy penetration rainwater. Evaporation loss compensation and data verification both use conventional standardized processing models, and the overall monitoring process follows the basic processing logic of general hydrological monitoring.

[0003] In existing monitoring schemes, the discontinuous pulse signals carried by the throughflow volumetric flow data collected by rainwater troughs cannot be effectively identified and processed. Conventional signal processing methods retain noise interference in the data, failing to generate clean flow time-series data. Rainwater evaporation dynamically changes with real-time environmental wind speed and air humidity in the monitoring area. Existing fixed compensation methods cannot match real-time environmental parameter changes, resulting in numerical deviations in the volumetric flow data due to evaporation. Monitoring data is acquired using only a single acquisition component and is not compared and verified with data recorded by a high-precision tipping bucket rain gauge, lacking a cross-verification step for the measurement results. It is necessary to extract and denoise the discontinuous pulse signals in the throughflow volumetric flow data, perform dynamic calibration of evaporation loss by incorporating real-time environmental wind speed and air humidity, and cross-verify the calibrated data with the original tipping count sequence of the high-precision tipping bucket rain gauge. Summary of the Invention

[0004] This invention aims to solve at least one of the technical problems existing in the prior art;

[0005] Therefore, this invention proposes a device for automatically monitoring the amount of rain penetrating the canopy of windbreak and sand-fixing forests, comprising:

[0006] The acquisition module is triggered in response to a start signal triggered by a rainfall event to acquire the volumetric flow rate of penetrating rainwater collected by the rainwater collection trough.

[0007] The signal processing module extracts the discontinuous pulse signals present in the through-rainwater volumetric flow data, and performs noise reduction and smoothing processing on the discontinuous pulse signals to generate clean flow time series data.

[0008] The parameter parsing module parses the clean flow time series data to obtain the peak rainfall intensity and rainfall duration, and inputs the peak rainfall intensity and rainfall duration into the evaporation compensation calculation model;

[0009] The evaporation compensation module acquires real-time environmental wind speed data and air humidity data of the current monitoring area, and inputs the real-time environmental wind speed data and air humidity data as auxiliary correction parameters into the evaporation compensation calculation model. The evaporation compensation calculation model is run to output the evaporation loss correction coefficient, and the evaporation loss correction coefficient is used to dynamically calibrate the through-rainwater volumetric flow rate data.

[0010] The cross-validation module acquires the original flip count sequence recorded by the high-precision tipping bucket rain gauge and cross-validates the dynamically calibrated through-flow volumetric rainwater data with the original flip count sequence.

[0011] The storage module generates the final cumulative measurement value of canopy penetration rainfall based on the cross-validation results, and stores the final cumulative measurement value in the built-in non-volatile memory.

[0012] Furthermore, acquiring the throughflow volumetric flow rate data collected by the rainwater collection trough includes:

[0013] The pressure sensor array installed at the bottom of the rainwater collection trough is controlled to collect the deformation simulation voltage signal of the pressure surface at a fixed frequency;

[0014] The deformation-simulated voltage signal is differentially amplified by an instrumentation amplifier to obtain an enhanced analog signal;

[0015] A synchronous sample-and-hold operation is performed on the enhanced analog signal, and it is converted into digital sample points by a successive approximation analog-to-digital converter;

[0016] Arrange all digital sampling points according to time series to construct an initial volumetric flow dataset;

[0017] Read the preset geometric dimension parameters of the rainwater collection trough, which include the trough length value, the trough bottom width value, and the side wall height value;

[0018] The effective collection cross-sectional area of ​​the rainwater collection trough is calculated based on the trough length, the trough bottom width, and the sidewall height.

[0019] Divide each digital sampling point in the initial volumetric flow dataset by the effective collection cross-sectional area to obtain the throughflow volumetric flow data per unit area.

[0020] Further, the discontinuous pulse signal is subjected to denoising and smoothing processing, including:

[0021] Identify the signal transition edges in the discontinuous pulse signal and record the timestamp and amplitude of the signal transition edges;

[0022] Calculate the time interval between adjacent signal transition edges and compare the time interval with a preset minimum physical rainfall interval threshold;

[0023] The signal transition edges and their corresponding amplitudes that have a time interval less than the minimum physical rainfall interval threshold are removed to generate a preliminary filtered signal sequence.

[0024] A finite-length impulse response filter based on the Hanning window function is constructed, and the preliminary filtered signal sequence is input as the input sequence to the finite-length impulse response filter;

[0025] Obtain the output sequence of the finite-length impulse response filter, and set the points in the output sequence whose amplitude is lower than a preset noise floor threshold to zero;

[0026] The output sequence after being zeroed out is used as the cleaning flow time series data.

[0027] Further, the time-series data of the clean flow is analyzed to obtain the peak rainfall intensity, including:

[0028] A sliding window integration operation is performed on the cleaning flow time series data, wherein the length of the sliding window is set to a preset unit time span;

[0029] Calculate the total integral value of the flow rate within each sliding window, and divide the total integral value of the flow rate by the unit time span to obtain the instantaneous rainfall intensity sequence;

[0030] Traverse all data points in the instantaneous rainfall intensity sequence and find the data point with the largest value as the peak rainfall intensity;

[0031] Record the time point corresponding to the peak rainfall intensity, and mark the time point as the peak occurrence time;

[0032] The number of consecutive data segments with amplitudes greater than zero in the clean flow time series data is counted, and the duration of each consecutive data segment is summed to obtain the rainfall duration.

[0033] Further, the evaporation compensation calculation model is run to output evaporation loss correction coefficients, including:

[0034] The real-time environmental wind speed data collected by the integrated micro weather station is acquired, and the real-time environmental wind speed data is converted into a standard wind speed measurement value;

[0035] The air humidity data collected by the integrated micro weather station is obtained, and the difference between the saturated humidity data and the air humidity data is calculated to obtain the humidity saturation difference;

[0036] Read the current ambient temperature data, and query the preset saturated water vapor pressure temperature lookup table based on the ambient temperature data to obtain the corresponding saturated water vapor pressure value;

[0037] The standard wind speed measurement value, the humidity saturation difference, and the saturated water vapor pressure value are combined into a multi-dimensional input feature vector.

[0038] The multidimensional input feature vector is input into a pre-trained multilayer perceptron regression model, and the multilayer perceptron regression model outputs a predicted estimate of the evaporation rate.

[0039] The evaporation loss ratio is obtained by multiplying the estimated evaporation rate by the rainfall duration and then dividing by the total accumulated amount of the through-rainwater volumetric flow data.

[0040] The evaporation loss ratio is inverted and added to the value to generate an evaporation loss correction coefficient.

[0041] Furthermore, the throughflow volumetric flow rate data is dynamically calibrated using the evaporation loss correction coefficient, including:

[0042] Iterate through each sampling data point in the penetrating rainwater volumetric flow data and obtain the timestamp corresponding to the sampling data point;

[0043] Query the instantaneous flow value corresponding to the cleaning flow time series data based on the timestamp;

[0044] Multiply the instantaneous flow rate value by the evaporation loss correction factor to obtain the calibrated instantaneous flow rate value;

[0045] All calibrated instantaneous flow values ​​are recombined according to the original timestamps to form a new calibrated flow dataset;

[0046] Perform an accumulation and summation operation on the new calibration flow dataset to obtain the calibrated total volumetric flow value.

[0047] Further, acquiring the raw flip count sequence recorded by the high-precision tipping bucket rain gauge includes:

[0048] Monitor the state changes of the reed switch of the high-precision tipping bucket rain gauge, and record a flip event each time the state of the reed switch flips.

[0049] Assign an incrementing event number to each flip event and record the time when the flip event occurs;

[0050] Read the single-tip volume calibration parameters of the high-precision tipping bucket rain gauge, where the single-tip volume calibration parameters represent the fixed rainwater volume corresponding to each tipping.

[0051] The event sequence number, the time, and the single tipping bucket volume calibration parameter are associated and stored to form the original tipping count sequence;

[0052] Obtain the sequence number of the last event in the original flip count sequence and use it as the total number of flips.

[0053] Furthermore, the dynamically calibrated throughflow volumetric flow rate data is cross-validated with the original flip count sequence, including:

[0054] Calculate the total cumulative volume corresponding to the dynamically calibrated through-rainwater volumetric flow rate data, and divide the total cumulative volume by the single tipping bucket volume calibration parameter to obtain the estimated number of tipping times;

[0055] Obtain the total number of flips in the original flip count sequence, and calculate the absolute difference between the estimated number of flips and the total number of flips;

[0056] Determine whether the absolute difference exceeds a preset acceptable error threshold;

[0057] If the absolute difference exceeds the acceptable error threshold, the current monitoring data is determined to be abnormal, and a data abnormality flag is generated.

[0058] If the absolute difference does not exceed the acceptable error threshold, the dynamically calibrated through stormwater volumetric flow data is marked as high-confidence data.

[0059] Furthermore, the generation of the final cumulative measurement of canopy penetration rainfall based on the cross-validation results includes:

[0060] Check if the aforementioned data anomaly flag exists;

[0061] If the data anomaly flag is present, the dynamically calibrated through-flow volumetric flow data is discarded, and the cumulative rainfall is calculated solely based on the original flip count sequence.

[0062] If the data anomaly flag is not present, the calibrated total volumetric flow rate value will be used as the intermediate measurement value of canopy penetration rainfall.

[0063] Obtain the equivalent volume of solid precipitation recorded by the high-precision tipping bucket rain sensor during the current monitoring period;

[0064] The equivalent volume of solid precipitation is algebraically added to the intermediate measurement value to obtain the final cumulative measurement value of canopy penetration rainfall.

[0065] Further, the time interval between adjacent signal transition edges is calculated, and the time interval is compared with a preset minimum physical rainfall interval threshold, including:

[0066] Extract all identified signal transition edges from the discontinuous pulse signal and establish a transition edge index list in chronological order of occurrence.

[0067] Traverse the edge index list, for any two consecutive signal edges, read their corresponding timestamps, and subtract the previous timestamp from the later timestamp to calculate the time interval between adjacent signal edges.

[0068] The preset minimum physical rainfall interval threshold is read from the configuration register. The minimum physical rainfall interval threshold is set according to the typical raindrop falling speed and sensor sampling resolution in the windbreak and sand-fixing forest area.

[0069] Each calculated time interval value is sequentially compared with the minimum physical rainfall interval threshold.

[0070] For signal transition edges whose comparison result is that the time interval value is less than the minimum physical rainfall interval threshold, they are determined to be pseudo transition edges caused by sensor electrical noise or transient interference, and the index identifier of the pseudo transition edge is recorded.

[0071] The index identifiers of all those identified as pseudo-jump edges and their corresponding amplitude data are summarized to generate a list of data to be removed. The discontinuous pulse signals are then cleaned based on the list of data to be removed to generate the preliminary filtered signal sequence.

[0072] Compared with the prior art, the beneficial effects of the present invention are:

[0073] By extracting discontinuous pulse signals from the throughflow volumetric flow data and performing denoising and smoothing processing on these signals, interference information mixed in with the flow data can be directly eliminated, and data anomalies caused by non-target signals can be removed. This optimizes the continuity and regularity of the flow time series data, resulting in clean flow time series data that closely reflects the actual flow state of throughflow. Clean flow time series data can reduce the impact of signal fluctuations on parameter analysis, enabling the obtained peak rainfall intensity and rainfall duration to accurately match the actual variation characteristics of canopy throughflow, avoiding interference from noise data in the parameter analysis results, and ensuring the accuracy and stability of rainfall-related parameter analysis.

[0074] Real-time environmental wind speed and air humidity data for the current monitoring area are acquired and used as auxiliary correction parameters input into the evaporation compensation calculation model. The model outputs an evaporation loss correction coefficient, which is then used to dynamically calibrate the throughfall volumetric flow rate data. This allows the evaporation compensation value to change in real time with environmental parameters, closely reflecting the actual loss of rainwater evaporation under different wind speeds and humidity conditions, thus reducing the deviation between the volumetric flow rate data and the actual rainfall value. The dynamically calibrated throughfall volumetric flow rate data is cross-validated with the original flip-count sequence recorded by a high-precision tipping bucket rain gauge. By comparing data from the two acquisition methods, measurement biases caused by a single acquisition method are eliminated, ensuring that the final cumulative measurement value of canopy throughfall rainfall accurately reflects the actual monitoring results. This improves the reliability and accuracy of the monitoring data, making it suitable for practical applications of ecological hydrological monitoring of windbreak and sand-fixing forests. Attached Figure Description

[0075] Figure 1 This is a timing diagram of an automatic monitoring device for monitoring rainfall penetration through the canopy of windbreak and sand-fixing forests according to the present invention.

[0076] Figure 2 A flowchart for performing denoising and smoothing processing on discontinuous pulse signals;

[0077] Figure 3 A graph showing the correction factor for evaporation loss;

[0078] Figure 4 This is a graph comparing the original and calibrated cumulative flow rates.

[0079] Figure 5 This is a diagram showing the flip-count sequence analysis of a tipping bucket rain gauge. Detailed Implementation

[0080] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0081] See Figure 1This invention provides a device for automatically monitoring the penetration rainfall of windbreak and sand-fixing forest canopy. The device includes: a trigger acquisition module, a signal processing module, a parameter analysis module, an evaporation compensation module, a cross-validation module, and a storage module. The trigger acquisition module responds to a start signal triggered by a rainfall event and acquires penetration rainwater volumetric flow data collected by a rainwater collection trough. The signal processing module extracts discontinuous pulse signals present in the penetration rainwater volumetric flow data and performs denoising and smoothing processing on the discontinuous pulse signals to generate clean flow time-series data. The parameter analysis module analyzes the clean flow time-series data to obtain the peak rainfall intensity and rainfall duration, and inputs the peak rainfall intensity and rainfall duration into an evaporation compensation calculation model. The evaporation compensation module acquires real-time environmental wind speed and air humidity data of the current monitoring area, and inputs the real-time environmental wind speed and air humidity data as auxiliary correction parameters into the evaporation compensation calculation model. The evaporation compensation calculation model is run to output an evaporation loss correction coefficient, and the evaporation loss correction coefficient is used to dynamically calibrate the penetration rainwater volumetric flow data. The cross-validation module acquires the original flip count sequence recorded by the high-precision tipping bucket rain gauge and cross-validates the dynamically calibrated throughflow volumetric flow data with the original flip count sequence. The storage module generates the final cumulative measurement of canopy throughflow based on the cross-validation results and stores the final cumulative measurement in the built-in non-volatile memory.

[0082] In one embodiment of the present invention, a trigger acquisition module acquires through-flow volumetric flow data of rainwater collected by a rainwater collection trough. This process involves controlling a pressure sensor array installed at the bottom of the rainwater collection trough to acquire deformation analog voltage signals of the pressure-bearing surface at a fixed frequency. The deformation analog voltage signals are differentially amplified by an instrumentation amplifier to obtain an enhanced analog signal. A synchronous sample-and-hold operation is performed on the enhanced analog signal, and it is converted into digital sampling points using a successive approximation analog-to-digital converter. All digital sampling points are arranged in a time series to construct an initial volumetric flow data set. Preset geometric parameters of the rainwater collection trough, including the trough length, trough bottom width, and sidewall height, are read. The effective collection cross-sectional area of ​​the rainwater collection trough is calculated based on the trough length, trough bottom width, and sidewall height. Each digital sampling point in the initial volumetric flow data set is divided by the effective collection cross-sectional area to obtain the through-flow volumetric flow data per unit area.

[0083] In specific implementation, the process of triggering the acquisition module to obtain the volumetric flow rate data of penetrating rainwater collected by the rainwater collection trough involves controlling the pressure sensor array installed at the bottom of the rainwater collection trough to acquire the deformation analog voltage signal of the pressure surface at a fixed frequency. The deformation analog voltage signal is then differentially amplified by an instrumentation amplifier to obtain an enhanced analog signal. A synchronous sampling and holding operation is performed on the enhanced analog signal, and it is converted into digital sampling points using a successive approximation analog-to-digital converter. All digital sampling points are arranged according to a time series to construct the initial volumetric flow rate dataset. In some embodiments, the fixed frequency is set to a specific Hertz value to match the temporal resolution of the rainfall event, and the conversion process of the digital sampling points uses linear quantization. Preset rainwater collection trough geometric parameters are read, including the trough length, trough bottom width, and sidewall height. The effective collection cross-sectional area of ​​the rainwater collection trough is calculated based on these parameters. In specific implementation, the effective collection cross-sectional area is calculated using the following formula:

[0084]

[0085] in: Indicates the effective collection cross-sectional area. Indicates the length of the tank. This represents the width of the trough bottom. It is understood that the trough geometry parameters are read from non-volatile memory and cached in the processing unit. Each digital sampling point in the initial volumetric flow rate dataset is divided by the effective collection cross-sectional area to obtain the per-unit-area penetrating stormwater volumetric flow rate data, which is organized as a timestamp-indexed sequence. In some embodiments, the pressure sensor array is arranged to cover the entire pressure-bearing surface at the bottom of the trough, and the gain parameter of the differential amplification processing is adjusted via software configuration. Optionally, synchronous sample-and-hold operation is triggered by an external clock signal to ensure timing consistency. Optionally, the resolution of the analog-to-digital converter is set to a fixed number of bits to balance accuracy and power consumption. It is understood that the per-unit-area penetrating stormwater volumetric flow rate data is directly used as input to subsequent signal processing modules to avoid measurement deviations caused by differences in trough dimensions.

[0086] In one embodiment of the present invention, the signal processing module performs noise reduction and smoothing processing on discontinuous pulse signals, see reference. Figure 2The process includes identifying signal transition edges in the discontinuous pulse signal and recording the timestamp and amplitude of each transition edge. All identified signal transition edges in the discontinuous pulse signal are extracted, and a transition edge index list is created in chronological order of occurrence. The transition edge index list is traversed, and for any two consecutive signal transition edges, their corresponding timestamps are read, and the time interval between adjacent transition edges is calculated by subtracting the previous timestamp from the later timestamp. A preset minimum physical rainfall interval threshold is read from the configuration register. This minimum physical rainfall interval threshold is set based on the typical raindrop falling speed and sensor sampling resolution in the windbreak and sand-fixing forest area. Each calculated time interval value is compared numerically with the minimum physical rainfall interval threshold. For signal transition edges whose time interval value is less than the minimum physical rainfall interval threshold, they are identified as pseudo-transition edges caused by sensor electrical noise or transient interference, and their index identifiers are recorded. All index identifiers and corresponding amplitude data of pseudo-transition edges are summarized to generate a list of data to be removed. Based on the list of data to be removed, signal transition edges and their corresponding amplitudes with time intervals less than the minimum physical rainfall interval threshold in the discontinuous pulse signals are removed to generate a preliminary filtered signal sequence. A finite-length impulse response (FCR) filter based on the Hanning window function is constructed, and the preliminary filtered signal sequence is input to the FCR filter. The output sequence of the FCR filter is obtained, and points in the output sequence with amplitudes lower than a preset noise floor threshold are set to zero. The zeroed-out output sequence is used as clean flow time-series data.

[0087] In specific implementation, the signal processing module performs denoising and smoothing processing on discontinuous pulse signals. It identifies signal transition edges in the discontinuous pulse signal and records the timestamp and amplitude of each transition edge. It then extracts all identified signal transition edges from the discontinuous pulse signal and establishes a transition edge index list in chronological order of occurrence. In some embodiments, signal transition edge identification is achieved by detecting points where the voltage change rate exceeds a set threshold. The transition edge index list is traversed, and for any two consecutive signal transition edges, their corresponding timestamps are read. The time interval between adjacent signal transition edges is calculated by subtracting the previous timestamp from the subsequent timestamp. In specific implementation, the time interval value... The calculation follows the formula:

[0088]

[0089] in: This represents the calculated time interval value. This represents the timestamp of the i-th signal transition edge in the transition edge index list. This represents the timestamp of the (i+1)th signal transition edge in the transition edge index list. The preset minimum physical rainfall interval threshold is read from the configuration register. This threshold is set based on the typical raindrop falling speed and sensor sampling resolution in the windbreak and sand-fixing forest area. It can be understood that the minimum physical rainfall interval threshold is a pre-calibrated and stored fixed time constant.

[0090] Each calculated time interval value is sequentially compared with the minimum physical rainfall interval threshold. Signal transitions with time interval values ​​less than the minimum physical rainfall interval threshold are identified as pseudo-transitions caused by sensor electrical noise or transient interference, and their index identifiers are recorded. All index identifiers and corresponding amplitude data of pseudo-transitions are compiled to generate a list of data to be removed. Based on this list, signal transitions with time intervals less than the minimum physical rainfall interval threshold and their corresponding amplitudes in discontinuous pulse signals are removed, generating a preliminary filtered signal sequence. Optionally, the removal of pseudo-transitions is accomplished by setting the amplitude data of the corresponding timestamp to zero or removing it from the data sequence. A finite-length impulse response (FLR) filter based on the Hanning window function is constructed, and the preliminary filtered signal sequence is input to the FLR filter. In some embodiments, the order and cutoff frequency of the FLR filter are determined based on the typical spectral characteristics of penetrating rainwater signals. The output sequence of the finite-length impulse response filter is obtained, and points in the output sequence with amplitudes lower than a preset noise floor threshold are set to zero. This noise floor threshold is understood to be derived statistically from the characteristics of the background signal during periods without rainfall. The zeroed-out output sequence is then used as the clean flow time-series data. Optionally, the clean flow time-series data can be stored in memory as an array data structure for subsequent module access.

[0091] In one embodiment of the present invention, the parameter parsing module parses clean flow time-series data to obtain peak rainfall intensity. The process includes performing a sliding window integration operation on the clean flow time-series data, where the length of the sliding window is set to a preset unit time span. The total flow integral value within each sliding window is calculated, and the total flow integral value is divided by the unit time span to obtain an instantaneous rainfall intensity sequence. All data points in the instantaneous rainfall intensity sequence are traversed, and the data point with the largest value is found as the peak rainfall intensity. The time point corresponding to the peak rainfall intensity is recorded and marked as the peak occurrence time. The number of consecutive data segments with amplitudes greater than zero in the clean flow time-series data is counted, and the duration of each consecutive data segment is accumulated to obtain the rainfall duration.

[0092] In practical implementation, the parameter parsing module parses the clean flow time series data to obtain the peak rainfall intensity. A sliding window integration operation is performed on the clean flow time series data, with the length of the sliding window set to a preset unit time span. The total flow integral value within each sliding window is calculated, and this total value is divided by the unit time span to obtain the instantaneous rainfall intensity sequence. In practical implementation, the calculation of each data point in the instantaneous rainfall intensity sequence follows the formula:

[0093]

[0094] in: This represents the instantaneous rainfall intensity corresponding to the j-th sliding window. This represents the total flow integral value within the current sliding window. This represents the preset unit time span. It can be understood that the unit time span is a configurable fixed value. In some embodiments, the sliding window's movement step size is set to a data point interval, thereby generating a high temporal resolution sequence of instantaneous rainfall intensity. Referring to Table 1, the process of calculating the sliding window integral over a set of clean flow time-series data segments is illustrated.

[0095] Table 1: Sliding Window Integral Calculation Table

[0096] Window start time point (index j) Total flow rate (ml) Calculated instantaneous rainfall intensity (ml / min) 1 15.2 1.52 2 18.7 1.87 3 22.4 2.24 4 20.1 2.01 5 16.8 1.68

[0097] The process iterates through all data points in the instantaneous rainfall intensity sequence, finding the data point with the largest value as the peak rainfall intensity. The time point corresponding to the peak rainfall intensity is recorded and marked as the peak occurrence time. In some embodiments, the traversal process uses a sequential comparison algorithm, initializing a temporary variable to store the currently found maximum value and its index. Optionally, if multiple data points with equal and identical values ​​exist in the instantaneous rainfall intensity sequence, the time point corresponding to the first occurrence of this data point is marked as the peak occurrence time. The number of consecutive data segments with amplitudes greater than zero in the clean flow time series data is counted, and the duration of each consecutive data segment is accumulated to obtain the rainfall duration. In a specific implementation, a consecutive data segment with an amplitude greater than zero refers to a sequence of several consecutive flow values ​​greater than zero in the clean flow time series data. The duration of each consecutive data segment is calculated by multiplying the number of data points in that segment by the time interval between individual data points. Optionally, for noise with amplitudes fluctuating around zero, a small threshold slightly greater than zero can be set to determine whether the data point effectively participates in the rainfall duration statistics. It is understood that the statistical results of the rainfall duration are used as input for the subsequent evaporation compensation calculation model.

[0098] See Figure 3This is a curve showing the evaporation loss correction coefficient, fully illustrating the dynamic changes of the correction coefficient during rainfall. The curve exhibits a symmetrical V-shaped change, initially decreasing and then increasing. In the initial stage of rainfall (1-6 minutes), the coefficient continuously decreases from 1.00, reaching its lowest value of 0.91 at the 6th minute. In the later stage of rainfall (6-12 minutes), the coefficient gradually increases from 0.91, eventually returning to 1.00. The correction coefficient is used to compensate for rainwater evaporation loss caused by environmental factors (wind speed, humidity, temperature) during rainfall. A coefficient <1 indicates that the original flow data needs to be amplified and calibrated to offset the measurement deviation caused by evaporation. The curve visually presents the dynamic correction effect of the evaporation compensation calculation model, verifying the model's accurate perception of evaporation loss during rainfall. The higher the rainfall intensity, the lower the correction coefficient and the greater the compensation magnitude, consistent with physical laws.

[0099] In one embodiment of the present invention, the evaporation compensation module runs an evaporation compensation calculation model to output an evaporation loss correction coefficient. The process includes acquiring real-time environmental wind speed data collected by an integrated micro-weather station and converting the real-time environmental wind speed data into a standard wind speed measurement value. It also acquires air humidity data collected by the integrated micro-weather station and calculates the difference between the saturated humidity data and the air humidity data to obtain a humidity saturation difference. The module reads the current ambient temperature data and queries a preset saturated vapor pressure-temperature lookup table based on the ambient temperature data to obtain the corresponding saturated vapor pressure value. The standard wind speed measurement value, the humidity saturation difference, and the saturated vapor pressure value are combined into a multi-dimensional input feature vector. This multi-dimensional input feature vector is input into a pre-trained multilayer perceptron regression model, which outputs a predicted evaporation rate estimate. The evaporation rate estimate is multiplied by the rainfall duration and then divided by the total accumulated amount of the through-rainwater volumetric flow rate data to obtain the evaporation loss ratio. Finally, the evaporation loss ratio is inverted and added to a single value to generate an evaporation loss correction coefficient. The process of dynamically calibrating the throughflow volumetric flow rate data using the evaporation loss correction coefficient involves iterating through each sampling data point in the throughflow volumetric flow rate data and obtaining the timestamp corresponding to each sampling data point. Based on the timestamp, the corresponding instantaneous flow rate value in the clean flow time series data is retrieved. This instantaneous flow rate value is multiplied by the evaporation loss correction coefficient to obtain the calibrated instantaneous flow rate value. All calibrated instantaneous flow rate values ​​are then recombined according to their original timestamps to form a new calibration flow rate dataset. Finally, an accumulation and summation operation is performed on the new calibration flow rate dataset to obtain the calibrated total volumetric flow rate value.

[0100] In practical implementation, the evaporation compensation module runs an evaporation compensation calculation model to output an evaporation loss correction coefficient. It acquires real-time environmental wind speed data collected by the integrated micro-weather station and converts this data into a standard wind speed measurement value, with the unit being meters per second. It acquires air humidity data collected by the integrated micro-weather station and calculates the difference between the saturated humidity data and the actual air humidity data to obtain the humidity saturation difference, with the unit being grams per cubic meter. It reads the current ambient temperature data and, based on the ambient temperature data, queries a preset saturated vapor pressure-temperature lookup table to obtain the corresponding saturated vapor pressure value, with the unit being kilopascals. In practical implementation, the calculation of the evaporation rate estimate follows the formula:

[0101]

[0102] in: This represents an estimated evaporation rate, expressed in millimeters per hour. This represents the standard wind speed measurement value. Indicates the difference in humidity saturation. This represents the saturated vapor pressure value. The calibration coefficient k is a composite dimension that makes the dimensions of both sides of the equation consistent, and it is obtained experimentally. It is understood that the value of the calibration coefficient k is related to local climate conditions and the installation environment of the device. In specific implementations, the evaporation compensation calculation model will read the preset value of the calibration coefficient k from the configuration file before running the above formula. In some embodiments, the standard wind speed measurement value, humidity saturation difference, saturated vapor pressure value, and the calculated estimated evaporation rate value can be temporarily stored during processing, as shown in Table 2.

[0103] Table 2: Input and Output Table of Evaporation Compensation Model

[0104] Input / output parameters symbol Example values unit Standard wind speed measurement value 1.5 meters per second Humidity saturation difference 2.3 g / m³ Saturated water vapor pressure 2.8 kPa Calibration coefficient 0.15 Composite dimensional units Evaporation rate estimate 0.185 mm / hour

[0105] The evaporation rate estimate is multiplied by the rainfall duration and then divided by the total accumulated amount of through-flow volumetric flow data to obtain the evaporation loss ratio, a dimensionless value between 0 and 1. The rainfall duration and total accumulated amount are obtained from the outputs of the parameter parsing module and the trigger acquisition module. The evaporation loss ratio is inverted and added to the value to generate an evaporation loss correction coefficient, which is also a dimensionless value. In some embodiments, when the calculated evaporation loss correction coefficient exceeds a preset upper limit, the evaporation loss correction coefficient is clamped to that upper limit.

[0106] The process of dynamically calibrating throughflow volumetric flow data using an evaporation loss correction factor involves iterating through each sampling data point in the throughflow volumetric flow data and obtaining the corresponding timestamp. Based on the timestamp, the corresponding instantaneous flow value in the clean flow time series data is then retrieved. In practice, this query is performed by matching identical or nearest-neighbor timestamps. The instantaneous flow value is multiplied by the evaporation loss correction factor to obtain the calibrated instantaneous flow value, which has the same physical units and dimensions as the original instantaneous flow value. All calibrated instantaneous flow values ​​are then recombined according to their original timestamps to form a new calibration flow dataset. The new calibration flow dataset maintains the same data structure as the original throughflow volumetric flow data. Optionally, for timestamps in the clean flow time series data where the instantaneous flow value is zero, the corresponding calibrated instantaneous flow value is also set to zero. An accumulation and summation operation is performed on the new calibration flow dataset to obtain the calibrated total volumetric flow value. This accumulation and summation operation iterates through all data points in the new calibration flow dataset, and the dimension of the calibrated total volumetric flow value is length. In some embodiments, the calibrated total volumetric flow rate value is stored in an intermediate variable for use by the cross-validation module.

[0107] See Figure 4 This is a graph comparing the original and calibrated cumulative flow, fully demonstrating the difference in cumulative flow before and after evaporation compensation. During the first 0-1 hours, rainfall intensity is low, evaporation loss is minimal, the correction coefficient is close to 1.00, and the calibration effect is not significant. From 1-3 hours, rainfall intensity increases, ambient wind speed is high, and humidity is low, exacerbating evaporation loss, causing the correction coefficient to continuously decrease, and the calibration range to gradually increase. From 3-5 hours, rainfall ends, evaporation loss stops, the correction coefficient rises back to 1.00, calibration is complete, and the cumulative flow no longer changes. This visually quantifies the actual calibration effect of the evaporation compensation calculation model, clearly demonstrating the impact of evaporation loss on canopy penetration rainfall measurement. It provides visual support for the reliability of monitoring data, proving that the system can effectively offset measurement errors caused by environmental factors and improve the accuracy of canopy penetration rainfall measurement.

[0108] In one embodiment of the present invention, the cross-validation module acquires the original flip count sequence recorded by a high-precision tipping bucket rain gauge. This process involves monitoring the state changes of the reed switch of the high-precision tipping bucket rain gauge, recording a flip event each time the reed switch state flips. Each flip event is assigned an incrementing event number, and the time of occurrence is recorded. The single-tip volume calibration parameter of the high-precision tipping bucket rain gauge is read, representing the fixed rainwater volume corresponding to each flip. The event number, the time, and the single-tip volume calibration parameter are associated and stored to form the original flip count sequence. The last event number in the original flip count sequence is obtained and used as the total number of flips. The process of cross-validating the dynamically calibrated penetrating rainwater volumetric flow data with the original flip count sequence involves calculating the total cumulative volume corresponding to the dynamically calibrated penetrating rainwater volumetric flow data, and dividing the total cumulative volume by the single-tip volume calibration parameter to obtain the estimated number of flips. The total number of flips in the original flip count sequence is obtained, and the absolute difference between the estimated flip count and the total number of flips is calculated. It is determined whether the absolute difference exceeds a preset acceptable error threshold. If the absolute difference exceeds the acceptable error threshold, the current monitoring data is deemed abnormal, and a data anomaly flag is generated. If the absolute difference does not exceed the acceptable error threshold, the dynamically calibrated throughflow volumetric flow data is marked as high-confidence data. The storage module generates the final cumulative measurement value of canopy throughflow rainfall based on the cross-validation results, by checking for the existence of the data anomaly flag. If the data anomaly flag exists, the dynamically calibrated throughflow volumetric flow data is discarded, and the cumulative rainfall is calculated only based on the original flip count sequence. If the data anomaly flag does not exist, the calibrated total volumetric flow value is used as the intermediate measurement value of canopy throughflow rainfall. The equivalent volume of solid precipitation recorded by the high-precision tipping bucket rain gauge during the current monitoring period is obtained. The equivalent volume of solid precipitation is algebraically added to the intermediate measurement value to obtain the final cumulative measurement value of canopy penetration rainfall.

[0109] In specific implementation, the cross-validation module acquires the original flip count sequence recorded by the high-precision tipping bucket rain gauge. This process involves monitoring the state changes of the reed switch of the high-precision tipping bucket rain gauge. Each time the reed switch flips, a flip event is recorded, and each flip event is assigned an incrementing event number, along with the time of occurrence. The single-tip volume calibration parameter of the high-precision tipping bucket rain gauge is read. This parameter represents the fixed volume of rainwater corresponding to each flip. The event number, time, and single-tip volume calibration parameter are associated and stored to form the original flip count sequence. The last event number in the original flip count sequence is obtained and used as the total number of flips. In some embodiments, changes in the reed switch state are captured through an interrupt service routine to ensure the accuracy of the recorded time. The process of cross-validating the dynamically calibrated throughflow volumetric flow data with the original tipping count sequence involves calculating the total cumulative volume corresponding to the dynamically calibrated throughflow volumetric flow data, and dividing the total cumulative volume by the single tipping bucket volume calibration parameter to obtain the estimated tipping count. In practice, the calculation of the estimated tipping count follows the formula:

[0110]

[0111] in: This indicates the estimated number of flips. This represents the total accumulated volume corresponding to the dynamically calibrated throughflow stormwater volumetric flow rate data. This represents the single-tilting-bucket volume calibration parameter of a high-precision tipping-bucket rain gauge. The total number of tippings in the original tipping count sequence is obtained, and the absolute difference between the estimated tipping count and the total number of tippings is calculated. It is then determined whether the absolute difference exceeds a preset acceptable error threshold. This acceptable error threshold is pre-set based on sensor accuracy and measurement requirements. If the absolute difference exceeds the acceptable error threshold, the current monitoring data is considered abnormal, and a data anomaly flag is generated. If the absolute difference does not exceed the acceptable error threshold, the dynamically calibrated through-flow data is marked as high-confidence data. In some embodiments, the acceptable error threshold is set as a fixed percentage of the total number of tippings.

[0112] The storage module generates the final cumulative measurement value of canopy penetration rainfall based on the cross-validation results. This process involves checking for data anomaly indicators. If an anomaly indicator is present, the dynamically calibrated penetration volumetric flow data is discarded, and the cumulative rainfall is calculated solely based on the original flip count sequence. In practice, the method for calculating the cumulative rainfall solely based on the original flip count sequence is to multiply the total number of flips by the single-flip bucket volume calibration parameter. It can be understood that when a data anomaly indicator exists, the system assumes that the volumetric measurement subsystem composed of the rainwater collection trough and pressure sensor array may have a temporary fault or calibration drift. If no data anomaly indicator is present, the calibrated total volumetric flow value is used as the intermediate measurement value of canopy penetration rainfall. The equivalent volume of solid precipitation recorded by the high-precision flip bucket rain gauge within the current monitoring period is obtained. The equivalent volume of solid precipitation is algebraically added to the intermediate measurement value to obtain the final cumulative measurement value of canopy penetration rainfall. Optionally, the equivalent volume of solid precipitation is measured and recorded after melting the solid precipitation using a thermal heating device inside the high-precision flip bucket rain gauge. Optionally, the final cumulative measurement value, along with its corresponding timestamp and data quality identifier, is stored in the built-in non-volatile memory.

[0113] See Figure 5 This is an analysis chart of the flip count sequence of a tipping bucket rain gauge, demonstrating the flip count and event distribution patterns of a high-precision tipping bucket rain gauge. The overall trend is a uniform, step-like increase without significant fluctuations, reflecting a stable rainfall process without sudden interruptions during the monitoring period. The time span is positively correlated with the number of flips; at fixed time intervals, the count sequence increases by one step, corresponding to one tipping bucket flip. At the 100th minute, the cumulative number of flips reaches 14, the highest value within the monitoring period. The smoothness of the step-like increase indicates stable rainfall intensity in the windbreak and sand-fixing forest area during the monitoring period, without short-term heavy rainfall or interruptions, meeting the requirements for continuous field monitoring. The uniformity of the time intervals reflects the stability of the sensor's sampling and triggering mechanism, without mechanical jamming or signal delay issues. By analyzing the correspondence between the number of flips and time, rainfall intensity can be calculated, and sensor sensitivity can be evaluated.

[0114] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A device for automatically monitoring rainfall penetration through the canopy of windbreak and sand-fixing forests, characterized in that, include: The acquisition module is triggered in response to a start signal triggered by a rainfall event to acquire the volumetric flow rate of penetrating rainwater collected by the rainwater collection trough. The signal processing module extracts the discontinuous pulse signals present in the through-rainwater volumetric flow data, and performs noise reduction and smoothing processing on the discontinuous pulse signals to generate clean flow time series data. The parameter parsing module parses the clean flow time series data to obtain the peak rainfall intensity and rainfall duration, and inputs the peak rainfall intensity and rainfall duration into the evaporation compensation calculation model; The evaporation compensation module acquires real-time environmental wind speed data and air humidity data of the current monitoring area, and inputs the real-time environmental wind speed data and air humidity data as auxiliary correction parameters into the evaporation compensation calculation model. The evaporation compensation calculation model is run to output the evaporation loss correction coefficient, and the evaporation loss correction coefficient is used to dynamically calibrate the through-rainwater volumetric flow rate data. The cross-validation module acquires the original flip count sequence recorded by the high-precision tipping bucket rain gauge and cross-validates the dynamically calibrated through-flow volumetric rainwater data with the original flip count sequence. The storage module generates the final cumulative measurement value of canopy penetration rainfall based on the cross-validation results, and stores the final cumulative measurement value in the built-in non-volatile memory.

2. The device for automatically monitoring rainfall penetration through the canopy of windbreak and sand-fixing forests according to claim 1, characterized in that, The acquisition of through-flow volumetric flow data collected by the rainwater collection trough includes: The pressure sensor array installed at the bottom of the rainwater collection trough is controlled to collect the deformation simulation voltage signal of the pressure surface at a fixed frequency; The deformation-simulated voltage signal is differentially amplified by an instrumentation amplifier to obtain an enhanced analog signal; A synchronous sample-and-hold operation is performed on the enhanced analog signal, and it is converted into digital sample points by a successive approximation analog-to-digital converter; Arrange all digital sampling points according to time series to construct an initial volumetric flow dataset; Read the preset geometric dimension parameters of the rainwater collection trough, which include the trough length value, the trough bottom width value, and the side wall height value; The effective collection cross-sectional area of ​​the rainwater collection trough is calculated based on the trough length, the trough bottom width, and the sidewall height. Divide each digital sampling point in the initial volumetric flow dataset by the effective collection cross-sectional area to obtain the throughflow volumetric flow data per unit area.

3. The device for automatically monitoring rainfall penetration through the canopy of windbreak and sand-fixing forests according to claim 2, characterized in that, Performing denoising and smoothing processing on the discontinuous pulse signal includes: Identify the signal transition edges in the discontinuous pulse signal and record the timestamp and amplitude of the signal transition edges; Calculate the time interval between adjacent signal transition edges and compare the time interval with a preset minimum physical rainfall interval threshold; The signal transition edges and their corresponding amplitudes that have a time interval less than the minimum physical rainfall interval threshold are removed to generate a preliminary filtered signal sequence. A finite-length impulse response filter based on the Hanning window function is constructed, and the preliminary filtered signal sequence is input as the input sequence to the finite-length impulse response filter; Obtain the output sequence of the finite-length impulse response filter, and set the points in the output sequence whose amplitude is lower than a preset noise floor threshold to zero; The output sequence after being zeroed out is used as the cleaning flow time series data.

4. The device for automatically monitoring rainfall penetration through the canopy of windbreak and sand-fixing forests according to claim 3, characterized in that, Parsing the clean flow time-series data to obtain the peak rainfall intensity includes: A sliding window integration operation is performed on the cleaning flow time series data, wherein the length of the sliding window is set to a preset unit time span; Calculate the total integral value of the flow rate within each sliding window, and divide the total integral value of the flow rate by the unit time span to obtain the instantaneous rainfall intensity sequence; Traverse all data points in the instantaneous rainfall intensity sequence and find the data point with the largest value as the peak rainfall intensity; Record the time point corresponding to the peak rainfall intensity, and mark the time point as the peak occurrence time; The number of consecutive data segments with amplitudes greater than zero in the clean flow time series data is counted, and the duration of each consecutive data segment is summed to obtain the rainfall duration.

5. The device for automatically monitoring rainfall penetration through the canopy of windbreak and sand-fixing forests according to claim 4, characterized in that, Running the evaporation compensation calculation model to output evaporation loss correction coefficients includes: The real-time environmental wind speed data collected by the integrated micro weather station is acquired, and the real-time environmental wind speed data is converted into a standard wind speed measurement value; The air humidity data collected by the integrated micro weather station is obtained, and the difference between the saturated humidity data and the air humidity data is calculated to obtain the humidity saturation difference; Read the current ambient temperature data, and query the preset saturated water vapor pressure temperature lookup table based on the ambient temperature data to obtain the corresponding saturated water vapor pressure value; The standard wind speed measurement value, the humidity saturation difference, and the saturated water vapor pressure value are combined into a multi-dimensional input feature vector. The multidimensional input feature vector is input into a pre-trained multilayer perceptron regression model, and the multilayer perceptron regression model outputs a predicted estimate of the evaporation rate. The evaporation loss ratio is obtained by multiplying the estimated evaporation rate by the rainfall duration and then dividing by the total accumulated amount of the through-rainwater volumetric flow data. The evaporation loss ratio is inverted and added to the value to generate an evaporation loss correction coefficient.

6. The device for automatically monitoring rainfall penetration through the canopy of windbreak and sand-fixing forests according to claim 5, characterized in that, Dynamic calibration of the throughflow volumetric flow rate data using the evaporation loss correction coefficient includes: Iterate through each sampling data point in the penetrating rainwater volumetric flow data and obtain the timestamp corresponding to the sampling data point; Query the instantaneous flow value corresponding to the cleaning flow time series data based on the timestamp; Multiply the instantaneous flow rate value by the evaporation loss correction factor to obtain the calibrated instantaneous flow rate value; All calibrated instantaneous flow values ​​are recombined according to the original timestamps to form a new calibrated flow dataset; Perform an accumulation and summation operation on the new calibration flow dataset to obtain the calibrated total volumetric flow value.

7. The device for automatically monitoring rainfall penetration through the canopy of windbreak and sand-fixing forests according to claim 6, characterized in that, The acquisition of the raw flip count sequence recorded by the high-precision tipping bucket rain gauge includes: Monitor the state changes of the reed switch of the high-precision tipping bucket rain gauge, and record a flip event each time the state of the reed switch flips. Assign an incrementing event number to each flip event and record the time when the flip event occurs; Read the single-tip volume calibration parameters of the high-precision tipping bucket rain gauge, where the single-tip volume calibration parameters represent the fixed rainwater volume corresponding to each tipping. The event sequence number, the time, and the single tipping bucket volume calibration parameter are associated and stored to form the original tipping count sequence; Obtain the sequence number of the last event in the original flip count sequence and use it as the total number of flips.

8. The device for automatically monitoring rainfall penetration through the canopy of windbreak and sand-fixing forests according to claim 7, characterized in that, Cross-validation is performed between the dynamically calibrated throughflow stormwater volumetric flow data and the original flip count sequence, including: Calculate the total cumulative volume corresponding to the dynamically calibrated through-rainwater volumetric flow rate data, and divide the total cumulative volume by the single tipping bucket volume calibration parameter to obtain the estimated number of tipping times; Obtain the total number of flips in the original flip count sequence, and calculate the absolute difference between the estimated number of flips and the total number of flips; Determine whether the absolute difference exceeds a preset acceptable error threshold; If the absolute difference exceeds the acceptable error threshold, the current monitoring data is determined to be abnormal, and a data abnormality flag is generated. If the absolute difference does not exceed the acceptable error threshold, the dynamically calibrated through stormwater volumetric flow data is marked as high-confidence data.

9. The device for automatically monitoring rainfall penetration through the canopy of windbreak and sand-fixing forests according to claim 8, characterized in that, The generation of the final cumulative measurement of canopy penetration rainfall based on cross-validation results includes: Check if the aforementioned data anomaly flag exists; If the data anomaly flag is present, the dynamically calibrated through-flow volumetric flow data is discarded, and the cumulative rainfall is calculated solely based on the original flip count sequence. If the data anomaly flag is not present, the calibrated total volumetric flow rate value will be used as the intermediate measurement value of canopy penetration rainfall. Obtain the equivalent volume of solid precipitation recorded by the high-precision tipping bucket rain sensor during the current monitoring period; The equivalent volume of solid precipitation is algebraically added to the intermediate measurement value to obtain the final cumulative measurement value of canopy penetration rainfall.

10. The device for automatically monitoring rainfall penetration through the canopy of windbreak and sand-fixing forests according to claim 9, characterized in that, Calculate the time interval between adjacent signal transition edges and compare the time interval with a preset minimum physical rainfall interval threshold, including: Extract all identified signal transition edges from the discontinuous pulse signal and establish a transition edge index list in chronological order of occurrence. Traverse the edge index list, for any two consecutive signal edges, read their corresponding timestamps, and subtract the previous timestamp from the later timestamp to calculate the time interval between adjacent signal edges. The preset minimum physical rainfall interval threshold is read from the configuration register. The minimum physical rainfall interval threshold is set according to the typical raindrop falling speed and sensor sampling resolution in the windbreak and sand-fixing forest area. Each calculated time interval value is sequentially compared with the minimum physical rainfall interval threshold. For signal transition edges whose comparison result is that the time interval value is less than the minimum physical rainfall interval threshold, they are determined to be pseudo transition edges caused by sensor electrical noise or transient interference, and the index identifier of the pseudo transition edge is recorded. The index identifiers of all those identified as pseudo-jump edges and their corresponding amplitude data are summarized to generate a list of data to be removed. The discontinuous pulse signals are then cleaned based on the list of data to be removed to generate the preliminary filtered signal sequence.

Images