An active layer depth identification method based on envelope fusion algorithm

Abstract

The application discloses a kind of active layer depth identification method based on envelope fusion algorithm, belong to season frozen depth survey technical field, the present application is through field reconnaissance and ground temperature data optimization survey line arrangement and radar frequency, and with ground penetrating radar data acquisition;Data are removed, gain compensation, band-pass filtering and time-depth conversion preprocessing;Then using barycentric method and edge detection method extract bottom interface track, and according to signal definition and continuity adaptive weighted fusion, effectively suppress noise and abnormal value;Finally, with the depth of 0 ℃ isotheral line as benchmark, cross validation and error correction are carried out, and the interface fusion curve of continuous, smooth and centimeter-level precision is output, to calculate the depth of active layer.The method breaks through the traditional interpretation precision limit while maintaining the high-efficiency and non-destructive advantage of GPR, and provides a practical technical solution for the design parameter selection in permafrost engineering investigation.

Description

Technical Field

[0001] This invention relates to the field of seasonal frost depth surveying technology, and in particular to an active layer depth identification method based on an envelope fusion algorithm. Background Technology

[0002] The active layer is the surface soil and rock layer in seasonally frozen soil regions that freezes in the cold season and thaws in the warm season. Its thickness directly determines the scale and distribution of cold-region disasters such as frost heave, thaw settlement, and frost pullout, and is a core parameter for the frost-resistant design of roadbeds, pipelines, pile foundations, and other engineering projects in cold regions. With climate warming and increased engineering activities, the interannual fluctuation of the active layer thickness has intensified. Traditional drilling and point-based ground temperature measurements are insufficient to provide continuous, large-scale, and real-time thickness data, leading to blind spots in exploration, conservative engineering designs, and continuously rising operation and maintenance costs. Therefore, there is an urgent need to develop a rapid, non-destructive, wide-area, and accurate active layer thickness identification technology to support the refined design and risk prevention and control of infrastructure in cold regions.

[0003] Ground-penetrating radar (GPR) technology has been widely applied in the field of permafrost exploration, enabling efficient and continuous kilometer-level profile detection, significantly improving data acquisition efficiency and coverage density. However, current GPR data interpretation and inversion still mainly rely on commercial software integrated into the equipment and manual interpretation, resulting in problems such as strong subjectivity, poor consistency, and low efficiency. There are still significant research gaps and application shortcomings in automated interpretation methods for permafrost radar images. The center of gravity (COG) method identifies interfaces by calculating the weighted average depth of reflected energy, possessing advantages such as strong anti-interference capability and stable extraction results, and is often used for locating strong reflective layers in engineering geophysical exploration. Edge detection captures abrupt interface changes by identifying the first-order difference or gradient extrema of the signal, responding sensitively to subtle interface changes, and has achieved good results in fields such as medical imaging and remote sensing. However, in the field of permafrost radar detection and interpretation, neither the COG method nor the edge detection method has been effectively applied.

[0004] To address this bottleneck, this invention is the first to integrate the centroid method and edge detection method, proposing an active layer depth identification method based on envelope weighted fusion. By combining the advantages of the two algorithms, it utilizes the centroid method's ability to smooth noise and stabilize interface extraction, while leveraging the edge detection method's ability to keenly capture interface abrupt changes. Furthermore, it employs cross-validation and adaptive weight adjustment using geothermal data, effectively overcoming problems such as interface jumps, misalignments, and false reflections that can easily occur with a single algorithm in complex environments. This achieves continuous, accurate, and automatic extraction of the active layer interface, significantly improving the interpretation accuracy and reliability of GPR in cold-region engineering surveys. Summary of the Invention

[0005] The purpose of this invention is to address the technical deficiencies in the existing technology by providing an active layer depth recognition method based on an envelope fusion algorithm.

[0006] The technical solution adopted to achieve the purpose of this invention is:

[0007] An active layer depth recognition method based on envelope fusion algorithm includes the following steps:

[0008] Step 1: Select a study area within the seasonally frozen soil region, conduct on-site reconnaissance of the geological and geophysical information of the study area, obtain favorable acquisition areas for ground-penetrating radar data, delineate survey lines in the acquisition area, and deploy ground-penetrating radar.

[0009] Step 2: Obtain ground temperature data of the collection area and determine the location of the active layer to confirm the operating frequency of the ground penetrating radar. The ground penetrating radar collects data of the collection area along the survey line at the operating frequency. By emitting electromagnetic waves, the electromagnetic waves are reflected when they encounter the interface between the active layer and the unfrozen soil. The reflected electromagnetic waves are received to obtain the raw ground penetrating radar data. Then, the propagation speed of the electromagnetic waves is corrected by drilling and measured data. The raw ground penetrating radar data is preprocessed to obtain the preprocessed radar data signal.

[0010] Step 3: Select the preprocessed radar data signal along a survey line, and use multiple interface extraction methods to extract features from the selected radar data signal, identify the interface between the active layer and the unfrozen soil, and extract the interface location.

[0011] Step 4: Set weight parameters for the interface positions extracted by different interface extraction methods in Step 3, and use a weighted fusion strategy to fuse them to generate fused interface positions;

[0012] Step 5: The depth of the 0℃ isotherm obtained from the geothermal data or the depth of the active layer obtained by drilling is taken as the actual depth. The actual depth is compared with the fusion interface position obtained in Step 4. If the fusion interface position is within the preset error range, the fusion interface position is taken as the calibrated interface position. If the fusion interface position is not within the preset error range, the weighting parameter in Step 4 or the electromagnetic wave propagation speed in Step 2 is adjusted until the fusion interface position is within the preset error range.

[0013] Step 6: Based on the calibrated interface position obtained in Step 5, obtain the fused interface trajectory, and perform secondary processing on the fused interface trajectory to obtain the fused interface curve.

[0014] In the above technical solution, in step 1, the vegetation cover and topography of the research area are investigated to avoid factors that interfere with the ground-penetrating radar signal, so as to obtain a favorable collection area for ground-penetrating radar data based on the surface flatness and access conditions of the research area.

[0015] In the above technical solution, in step 2, the host of the ground penetrating radar equipment is connected to a computer via Ethernet for equipment control and real-time data transmission.

[0016] In the above technical solution, in step 2, the ranging wheel is connected to the ground penetrating radar equipment for communication. The displacement signal generated by the rotation of the ranging wheel is used to trigger the ground penetrating radar data acquisition, so that the ground penetrating radar can collect data along the survey line at equal intervals. The ground penetrating radar equipment is advanced at a constant speed along the survey line, and the original radar data is recorded in real time by a computer.

[0017] In the above technical solution, in step 2, the raw ground-penetrating radar data is preprocessed by background removal, gain compensation, bandpass filtering and time-depth conversion in sequence to enhance the identification effect of the reflection interface in the study area.

[0018] In the above technical solution, step 3, the interface extraction method includes the center of gravity (COG) method and the edge detection method.

[0019] In the above technical solution, the centroid method for interface recognition involves calculating the envelope amplitude by performing envelope analysis on the preprocessed radar data signal. Extract the reflection energy distribution of each A-scan signal from the vertical radar waveform. In each signal, select a target reflection window and calculate the weighted average depth of the energy within the target reflection window as the interface position of that signal. The expression is:

[0020] ;

[0021] in: For the first The envelope amplitude of each sampling point; For the first The depth of a sampling point calculated by converting sampling time and wave velocity; These are the start and end sampling points of the target reflection window; This refers to the interface position extracted using the centroid method.

[0022] Typically, the radar profile signal is obtained from the original radar waveform through the Hilbert Transform. The analytic signal is obtained through the Hilbert transform: Calculate the envelope amplitude ,in, Represents the Hilbert transform. This represents a 90° phase-shifted version of the original signal.

[0023] In the above technical solution, the edge detection method extracts the interface location: Utilizing the characteristic of abrupt changes in reflection intensity at the interface in the preprocessed radar data signal, the first-order difference or gradient image of the A-scan envelope signal is calculated, and the maximum gradient point is extracted as the potential interface location. The edge detection method can sensitively capture abrupt edges at the reflection location, which helps in the precise extraction of the interface. Its mathematical expression is:

[0024] ;

[0025] Or in discrete form:

[0026] ;

[0027] in: For the first The envelope amplitude of each sampling point; For the first The envelope amplitude of each sampling point; This represents the rate of change of the envelope amplitude with depth; The interface position is extracted by the edge detection method; The starting sampling point, This is the termination point for sampling.

[0028] In the above technical solution, the mathematical formula for the weighted fusion strategy in step 4 is:

[0029] ;

[0030] in: , This is a weighting parameter; if the reflected signal is clear and the interface is continuous, then set... If the interface is blurry and severely affected by noise, then set... , This refers to the position of the merged interface.

[0031] In the above technical solution, in step 6, the secondary processing is curve smoothing, missing value smoothing and interpolation.

[0032] Compared with the prior art, the beneficial effects of the present invention are:

[0033] This invention's identification method integrates the advantages of the centroid method and edge detection method. Through the joint extraction and weighted processing of multiple feature information, it achieves automatic identification of the depth of the active layer of seasonally frozen soil. Compared with traditional single algorithms, it exhibits superior performance in terms of accuracy, stability, and adaptability. By introducing preprocessing techniques such as background removal, gain compensation, bandpass filtering, and temporal depth, it effectively reduces the impact of noise interference and false reflections in radar data, improving the continuity and reliability of the interpretation results. Furthermore, this method supports the processing of radar data at different frequencies and is suitable for frozen interface identification tasks under various soil hydrothermal conditions.

[0034] This invention improves the quality of collected data from the source by optimizing the layout of survey lines and radar frequencies through field reconnaissance and geothermal data. Background removal, gain compensation, bandpass filtering, and time-depth conversion preprocessing are performed on the raw ground-penetrating radar data to significantly enhance the reflection signal at the freeze-thaw interface. Subsequently, the centroid method and edge detection method are used to extract the bottom interface trajectory, and adaptive weighted fusion is performed based on signal clarity and continuity to effectively suppress noise and outliers. Finally, cross-validation and error correction are performed using the 0℃ isotherm depth as a benchmark to output a continuous, smooth interface fusion curve with centimeter-level accuracy, and to calculate the depth of the active layer. This method maintains the high efficiency and non-destructive nature of ground-penetrating radar (GPR) while overcoming the limitations of traditional interpretation accuracy, providing a practical technical solution for engineering surveys in seasonally frozen soil regions that can be directly used to determine design parameters. Attached Figure Description

[0035] Figure 1 This is a schematic diagram of the ground-penetrating radar detection principle of the present invention.

[0036] Figure 2 This is an image of the active layer detected by the ground-penetrating radar of the present invention.

[0037] Figure 3 This is a schematic diagram of the identification method of the present invention.

[0038] Figure 4 These are radar image cross-sections and active layer bottom interface recognition results of the present invention. Detailed Implementation

[0039] The present invention will be further described in detail below with reference to specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0040] An active layer depth recognition method based on envelope fusion algorithm includes the following steps:

[0041] like Figures 1-3As shown, in step 1, a study area is selected in the seasonally frozen soil region, and the geological and geophysical information of the study area is surveyed on-site to obtain favorable collection areas for ground-penetrating radar data. The collection area is divided into survey lines and ground-penetrating radar is deployed, and the survey line division is optimized through ground temperature data.

[0042] Furthermore, the study analyzes the vegetation cover and environmental conditions of the research area, avoiding terrain features such as water accumulation, snow accumulation, dense vegetation areas, and electromagnetic interference sources that may interfere with ground-penetrating radar signals, in order to reduce conditions affecting radar signal propagation. Priority is given to selecting areas with relatively flat surfaces, moderate hardness, and conditions conducive to the stable advancement of the ground-penetrating radar system, such as alpine grasslands, areas with frequent freeze-thaw cycles, or areas with obvious freeze-thaw interface changes, as favorable acquisition areas for ground-penetrating radar data. These acquisition areas are conducive to obtaining high-quality ground-penetrating radar data, ensuring clear identification of the interface between the active layer and unfrozen soil, and minimizing signal reflection interference.

[0043] Furthermore, appropriate survey line layout areas are delineated within the favorable acquisition areas of ground-penetrating radar data to ensure that the survey lines cover key areas of the acquisition area. Within the survey line layout area, several parallel or intersecting survey lines are laid out according to the lateral variation characteristics of seasonal permafrost distribution to ensure the coverage and resolution of subsequent data acquisition.

[0044] Preferably, the layout of survey lines follows the principle of "covering representative areas while taking into account the characteristics of frost depth changes," prioritizing areas with different landform types and significant frost depth changes for focused deployment. Survey lines should be as perpendicular as possible to the direction of frost depth changes to enhance the clarity of the radar reflector.

[0045] Step 2: Obtain ground temperature data of the collection area and determine the location of the active layer. Based on this, confirm the measurement of the ground penetrating radar antenna with operating frequencies of 250 MHz and 500 MHz to balance detection depth and resolution. The ground penetrating radar collects image data of the collection area at equal intervals along the survey line. The ground penetrating radar emits electromagnetic waves, which are reflected when they encounter the interface between the active layer and the unfrozen soil. The reflected electromagnetic waves are received to obtain the raw ground penetrating radar data. Then, the electromagnetic wave propagation speed is corrected by drilling and measured data. The raw ground penetrating radar data is preprocessed to obtain the preprocessed radar data signal.

[0046] Furthermore, the ground-penetrating radar includes a main unit, transmitting and receiving antennas, and the main unit is connected to a computer via Ethernet to establish a real-time data transmission channel. The main unit of the ground-penetrating radar is also communicatively connected (via cable) to a ranging wheel. During the data acquisition phase, the computer software controls the initialization, parameter settings, and antenna configuration of the ground-penetrating radar. The displacement signal generated by the rotation of the ranging wheel triggers the ground-penetrating radar data acquisition. Preferably, the ranging wheel and the main unit of the ground-penetrating radar are synchronized. When the ground-penetrating radar moves along the survey line, it ensures that the antenna is in close contact with the ground and maintains good coupling with the ground surface. Data acquisition is triggered once every set distance the ranging wheel advances, ensuring uniform data acquisition intervals and accurate spatial positioning. The ground-penetrating radar automatically records raw (radar profile) data such as radar waveforms, sampling time, and location coordinates. The raw data is transmitted in real time and stored in the computer. The signal quality is monitored through the real-time waveform display function on the computer software, and the acquired signals are initially interpreted. If severe signal interference or abnormal echoes are detected, the ground-penetrating radar acquisition parameters are adjusted promptly or acquisition is repeated to ensure data integrity and validity.

[0047] Furthermore, ground-penetrating radar (GPR) data processing software was used to preprocess the raw GPR data sequentially, including background removal, gain compensation, bandpass filtering, and time-depth transformation. Specifically, this enhanced the identification of reflective interfaces in the study area, reduced random noise in the data signal, and improved the signal-to-noise ratio. Therefore, the preprocessed data signal showed less noise than the unprocessed data, and was also clearer than the original data because the filtering process helped highlight structural features. Energy gain compensation compensated for the energy attenuation of the radar data signal, making the energy distribution of the preprocessed radar data signal more uniform and improving signal consistency.

[0048] The background removal process aims to remove background noise and highlight localized abnormal reflections. The removal method is the horizontal averaging method, and the formula is as follows:

[0049] ;

[0050] in, The signal after background removal. For the first The road in the The raw ground-penetrating radar data signals at each time sampling point, where N is the total number of channels. Average background value (column-direction average) for each time sampling point.

[0051] The gain compensation is an artificial amplification process applied to signals that gradually weaken over time (corresponding to detection depth). Essentially, it reduces the gain of early (shallow) signals and increases the gain of later (deep) signals, thus visually "balancing" the entire radar profile and facilitating the identification of deep targets. Bandpass filtering is a crucial step, aiming to remove unwanted frequency components and retain effective signal components.

[0052] like Figure 1 As shown, ground-penetrating radar transmits incident waves into the ground, generating reflected waves at the interface between the active layer and unfrozen soil. t2 represents the total time it takes for the electromagnetic wave to propagate to another interface after passing through the interface between the active layer and unfrozen soil, while t1 represents the time it takes for the electromagnetic wave to travel from transmission to being captured by the receiving antenna, i.e., the "two-way travel time." The active layer depth = (electromagnetic wave propagation speed × two-way travel time) / 2. Time-depth conversion transforms the visually displayed time axis (unit: nanoseconds, ns) on the radar profile into a practically understandable depth axis (unit: meters, m). The propagation speed v of electromagnetic waves in a medium (unit: meters / nanoseconds, m / ns) is not constant and is corrected using the common-center method or based on the dielectric properties of the medium obtained through drilling.

[0053] like Figure 1 In the ground-penetrating radar profile, mV and -mV represent the amplitude and polarity of the received electromagnetic wave signal in the radar image. If the echo is positive polarity (mV), it indicates that the electromagnetic wave has entered the unfrozen soil (high ε) from the active layer (low ε).

[0054] Step 3: Select the preprocessed radar data signal along a survey line, and use multiple interface extraction methods to extract features from the selected radar data signal to identify the interface between the active layer and the unfrozen soil, and extract the interface location. The interface extraction methods include the center of gravity (COG) method and the edge detection method.

[0055] Furthermore, the interface position is extracted using the centroid method: envelope analysis is performed on the preprocessed radar data signal to calculate the envelope amplitude. The reflected energy distribution of each vertical A-scan signal is extracted. Within each signal, a target reflection window is selected, and the weighted average depth of the energy within that window is calculated as the interface position for that signal. The centroid method has strong anti-interference capabilities and can smoothly extract the position of high-energy reflection areas. The mathematical expression for the centroid method is:

[0056] ;

[0057] in: For the first The envelope amplitude of each sampling point; For the first The depth of a sampling point calculated by converting sampling time and wave velocity; These are the start and end sampling points of the target reflection window; This refers to the interface position extracted using the centroid method. Typically, the radar profile signal is obtained from the original radar waveform through the Hilbert Transform. The analytic signal is obtained through the Hilbert transform: Calculate the envelope amplitude ,in, Represents the Hilbert transform. This represents a 90° phase-shifted version of the original signal.

[0058] Edge detection method for extracting abrupt interface locations: Utilizing the abrupt changes in reflection intensity at the interface in the preprocessed radar data signal, the first-order difference or gradient image of the A-scan envelope signal is calculated, and the point of maximum gradient is identified as the potential interface. Edge detection can sensitively capture abrupt edges at reflection locations, which is helpful for precise interface localization and extraction. Its mathematical expression is:

[0059] ;

[0060] Or in discrete form:

[0061] ;

[0062] in: For the first The envelope amplitude of each sampling point; For the first The envelope amplitude of each sampling point; This represents the rate of change of the envelope amplitude with depth; The interface position is extracted by the edge detection method; The starting sampling point, This is the termination point for sampling.

[0063] Step 4: Set weight parameters for the interface positions extracted by different interface extraction methods in Step 3, and use a weighted fusion strategy to fuse them to generate fused interface positions. The weight parameters are adjusted based on the interface positions measured during drilling.

[0064] The mathematical formula for the weighted fusion strategy is:

[0065] ;

[0066] in: , The centroid method weight parameters, These are the weight parameters for the edge detection method. If the reflected signal is clear and the interface is continuous, then set... If the interface is blurry and severely affected by noise, then set... , To integrate the interface positions.

[0067] Step 5: The depth of the 0℃ isotherm obtained from the geothermal data or the depth of the active layer obtained by drilling is taken as the actual depth. The actual depth is compared with the fusion interface position obtained in Step 4. If the fusion interface position is within the preset error range (±0.10 m), the fusion interface position is taken as the calibrated interface position. If the fusion interface position is not within the preset error range, the weighting parameter in Step 4 or the electromagnetic wave propagation speed in Step 2 is adjusted until the fusion interface position is within the preset error range.

[0068] Step 6: Based on the calibrated interface position obtained in Step 5, obtain the fused interface trajectory, perform curve smoothing on the fused interface trajectory, smooth out missing values ​​and interpolate to obtain the fused interface curve.

[0069] Furthermore, in the interface identification process, the active layer interface is defined as a reflection anomaly layer with a significant envelope amplitude response in the radar image. An envelope calculation and waveform feature tracking algorithm based on Hilbert transform, combined with multi-channel smoothing, is used to extract continuous reflection interfaces. Simultaneously, based on ground temperature data measured by field-deployed ground temperature sensors, a ground temperature variation curve with depth is plotted, and the 0℃ isotherm of this curve is used as the criterion for the boundary between the active layer of frozen soil and the unfrozen soil. By comparing this boundary with the reflection anomaly layer in the radar image profile, comprehensive automatic identification of the frost depth interface can be achieved. To ensure interpretation accuracy, manual drilling or other geophysical methods can also be introduced for cross-validation.

[0070] like Figure 4 As shown, in order to demonstrate that the method of the present invention has good recognition accuracy and engineering applicability, the interface positions extracted by the centroid method and the interface positions extracted by the edge detection method are processed twice to obtain the interface curves of the centroid method and the interface curves of the edge method. Then, a straight line (fitting curve) is fitted to the fused interface curves obtained in step 4 according to their direction. The trend of the interface between the active layer and the unfrozen soil is intuitively perceived through the fitted straight line. Finally, the above lines are visualized and superimposed to output an image, which is then compared with the original radar image profile.

[0071] The above description is only a preferred embodiment of the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for active layer depth recognition based on envelope fusion algorithm, characterized in that, Includes the following steps: Step 1: Select a study area within the seasonally frozen soil region, conduct on-site reconnaissance of the geological and geophysical information of the study area, obtain favorable acquisition areas for ground-penetrating radar data, delineate survey lines in the acquisition area, and deploy ground-penetrating radar. Step 2: Obtain ground temperature data of the collection area and estimate the location of the active layer to confirm the operating frequency of the ground penetrating radar. The ground penetrating radar collects data of the collection area along the survey line at the operating frequency. By emitting electromagnetic waves, the electromagnetic waves are reflected when they encounter the interface between the active layer and the unfrozen soil. The reflected electromagnetic waves are received to obtain the raw data of the ground penetrating radar. The raw data of the ground penetrating radar is preprocessed to obtain the preprocessed radar data signal. Step 3: Select the preprocessed radar data signal on a survey line, and use multiple interface extraction methods to extract features from the selected radar data signal, identify the interface between the active layer and the unfrozen soil, and extract the interface position. The interface extraction methods include the centroid method and the edge detection method. The centroid method for interface recognition involves performing envelope analysis on the preprocessed radar data signal and calculating the envelope amplitude. The reflection energy distribution of each A-scan signal of the vertical radar waveform is extracted. In each signal, a target reflection window is selected, and the weighted average depth of the energy within the target reflection window is calculated as the interface position of that signal. The edge detection method extracts the interface position by using the feature of abrupt change in reflection intensity at the interface of the preprocessed radar data signal, calculating the first-order difference or gradient image of the A-scan envelope signal, and extracting the maximum gradient point as the potential interface position. Step 4: Set weight parameters for the interface positions extracted by different interface extraction methods in Step 3, and use a weighted fusion strategy to fuse them to generate fused interface positions; Step 5: Use the depth of the 0℃ isotherm obtained from the geothermal data or the depth of the active layer obtained by drilling as the actual depth. Compare the actual depth with the fusion interface position obtained in Step 4. If the fusion interface position is within the preset error range, use the fusion interface position as the calibrated interface position. If the fusion interface position is not within the preset error range, return to Step 4 to adjust the weight parameters in Step 4. Step 6: Based on the calibrated interface position obtained in Step 5, obtain the fused interface trajectory. Perform secondary processing on the fused interface trajectory to obtain the fused interface curve. Calculate the active layer depth based on the fused interface curve.

2. The active layer depth recognition method according to claim 1, characterized in that, In step 1, the vegetation cover and topography of the research area are investigated to avoid factors that interfere with the ground-penetrating radar signal, so as to obtain a favorable collection area for ground-penetrating radar data based on the surface flatness and access conditions of the research area; at the same time, the survey line division is optimized through ground temperature data.

3. The active layer depth recognition method according to claim 1, characterized in that, In step 2, the host of the ground penetrating radar device is connected to a computer via Ethernet for device control and real-time data transmission.

4. The active layer depth recognition method according to claim 3, characterized in that, In step 2, the ranging wheel is connected to the ground-penetrating radar device. The displacement signal generated by the rotation of the ranging wheel triggers the ground-penetrating radar to collect data, so that the ground-penetrating radar can collect data along the survey line at equal intervals. The ground-penetrating radar device is moved at a constant speed along the survey line, and the original radar data is recorded in real time by a computer.

5. The active layer depth recognition method according to claim 1, characterized in that, In step 2, the raw ground-penetrating radar data is preprocessed sequentially by background removal, gain compensation, bandpass filtering, and time-depth conversion.

6. The active layer depth recognition method according to claim 1, characterized in that, The mathematical expression for the centroid method is: ； in: For the first i The envelope amplitude of each sampling point; For the first The depth of a sampling point calculated by converting sampling time and wave velocity; , These are the start and end sampling points of the target reflection window; This refers to the interface position extracted using the centroid method.

7. The active layer depth recognition method according to claim 1, characterized in that, The mathematical expression for the edge detection method is: ； Or in discrete form: ； in: For the first The envelope amplitude of each sampling point; For the first The envelope amplitude of each sampling point; This represents the rate of change of the envelope amplitude with depth; The interface position is extracted by the edge detection method; The starting sampling point, This is the termination point for sampling.

8. The active layer depth recognition method according to claim 1, characterized in that, In step 4, the mathematical formula for the weighted fusion strategy is: ； in: , For weight parameters, This refers to the position of the merged interface.

9. The active layer depth recognition method according to claim 1, characterized in that, In step 6, the secondary processing includes curve smoothing, missing value smoothing, and interpolation.

Images