Tunnel earthquake advanced prediction small aperture imaging and automatic interpretation method

Abstract

The application discloses a tunnel earthquake advanced prediction small-aperture imaging and automatic interpretation method, and belongs to the technical field of tunnel engineering geological exploration. Seismic wave data is collected through setting an observation system, and Fresnel body functions are calculated based on the Fresnel condition and a constant-speed medium model after filtering F c , the F c The weight factor is substituted into a Kirchhoff migration imaging formula to complete constraint imaging calculation, and imaging results are output; the imaging color profile is read and converted to an HSV color space, color threshold values are adaptively adjusted based on hue and saturation components, and after multi-scale contour detection fusion and contour grouping optimization, key geological targets are labeled and visualized output; the application realizes F c Fresnel zone correction, solves the problems of small-aperture imaging arc drawing and low resolution; the data-driven automatic interpretation process eliminates the multi-solution nature of manual interpretation, improves prediction accuracy and efficiency, and the imaging results are more in line with actual geological conditions, thereby providing reliable support for tunnel construction safety.

Description

Technical Field

[0001] This invention belongs to the field of tunnel engineering geological exploration technology, specifically involving a method for tunnel seismic advance prediction using small-aperture imaging and automatic interpretation. Background Technology

[0002] Advanced geological prediction for tunnels is a crucial step in ensuring tunnel construction safety and improving construction efficiency and economic benefits. Seismic methods, due to their wide adaptability, ease of operation, long detection range, and high accuracy, have become the mainstream technical means for advanced tunnel prediction. This method involves arranging seismic sources and detector arrays within the tunnel to artificially generate seismic waves and receive reflected and refracted signals. Data processing and imaging are then used to identify geological characteristics and adverse geological formations ahead of the tunnel face.

[0003] However, existing technologies have two major problems in practical applications: First, at the imaging level, the limited width of the tunnel cross section results in an imaging aperture of less than 20m, while the detection depth reaches 100-200m. The lack of large offset data causes obvious arcing in mainstream Kirchhoff migration imaging, which greatly reduces the imaging resolution of geological bodies. Second, at the interpretation level, seismic imaging results rely on manual interpretation by engineers, which is easily affected by cognitive biases of prior geological data, personal experience, and subjective pattern recognition. This leads to multiple interpretations, low efficiency, and no unified standard. Summary of the Invention

[0004] To address the problems mentioned in the background art, this invention provides a method for tunnel earthquake early prediction using small-aperture imaging and automatic interpretation, which solves the problems of arc drawing and low resolution in small-aperture imaging, and also solves the problem of multiple interpretations in manual interpretation.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A method for small-aperture imaging and automatic interpretation in tunnel seismic advance prediction includes the following steps:

[0007] S1: Small-aperture imaging for advanced seismic prediction in tunnels, including the following sub-steps:

[0008] S1.1: Set up a tunnel seismic observation system, arrange seismic sources as the source points and geophone arrays as the receiving points, artificially generate seismic waves, and collect tunnel seismic wave data;

[0009] S1.2: Read the acquired seismic wave data, filter the seismic wave data, and obtain the filtered seismic wave data;

[0010] S1.3: Calculation of Fresnel volume function F based on Fresnel condition and constant velocity medium model c , will F c As a weighting factor for small aperture imaging;

[0011] S1.4: F c Substituting the Kirchhoff migration imaging formula, the filtered seismic wave data was used to implement the F-band imaging. c Constrained Kirchhoff migration imaging calculations;

[0012] S1.5: Output small-aperture imaging results for tunnel seismic advance prediction;

[0013] S2: Image recognition and interpretation for tunnel earthquake early warning, including the following sub-steps:

[0014] S2.1: Read the small aperture imaging results output by S1.5 and obtain the imaging color profile;

[0015] S2.2: Convert the imaging color profile from the RGB color space to the HSV color space to obtain the HSV space imaging profile. The HSV space imaging profile is decomposed into independent components including hue, saturation, and brightness.

[0016] S2.3: Adaptive color threshold adjustment of HSV spatial imaging profile based on independent components, dynamically generating color thresholds according to the color distribution characteristics of the HSV spatial imaging profile itself;

[0017] S2.4: Based on color thresholding, perform multi-scale contour detection and fusion on HSV spatial imaging profiles, extract contours at multiple scaling ratios and integrate the detection results;

[0018] S2.5: Perform contour grouping and optimization on the candidate contours after detection and fusion, remove duplicate contours and select contours corresponding to significant geological targets;

[0019] S2.6: Annotate and visualize the results of the selected key geological targets, and output annotated images and structured data reports.

[0020] Compared with the prior art, the beneficial effects of the present invention are:

[0021] This invention utilizes the Fresnel volume function F c Constraining Kirchhoff migration imaging, using F c By implementing Fresnel zone correction, under the condition of small-aperture tunnel observation, diffraction distortion and arc drawing effects in imaging are effectively reduced, making the imaging results more consistent with actual geological conditions and significantly improving the imaging resolution of geological bodies. At the same time, this invention constructs an automated image interpretation process based on HSV color space conversion, adaptive threshold adjustment, and multi-scale contour detection fusion, completely replacing manual interpretation that relies on personal experience. Through data-driven objective operation, the problem of multiple interpretations is eliminated, and standardized extraction of geological information is achieved, improving interpretation efficiency and prediction accuracy, and providing accurate and reliable technical support for geological prediction in tunnel construction. Attached Figure Description

[0022] Figure 1 A schematic diagram of the Fresnel zone geometry for reflected waves in a homogeneous medium;

[0023] Figure 2 This is a schematic diagram of the imaging results for conventional tunnel earthquake early warning.

[0024] Figure 3 A schematic diagram of small-aperture imaging results for tunnel earthquake early prediction;

[0025] Figure 4 This is a schematic diagram illustrating the results of image recognition and interpretation for tunnel earthquake early prediction. Detailed Implementation

[0026] To facilitate understanding of the technical content of this invention by those skilled in the art, the invention will be further described in detail below with reference to the accompanying drawings and specific examples. It should be understood that the specific examples described herein are merely illustrative and not intended to limit the scope of the invention.

[0027] A method for small-aperture imaging and automatic interpretation in tunnel seismic advance prediction includes the following steps:

[0028] First, set up the observation system, then read the data, perform data filtering, and finally calculate the Fresnel volume function F. c Kirchhoff migration imaging calculations are performed, and the final imaging results are output, specifically:

[0029] For scalar seismic wavefields, the Kirchhoff migration imaging formula is as follows:

[0030] ;

[0031] In the formula, V(m) represents each point in the two-dimensional space that needs to be imaged and its corresponding reflected wave energy, which is a reconstruction of the energy and spatial location of the seismic wave data acquired by the instrument in the tunnel. S t is the travel time of a seismic wave from its source to its reflection point. R This is the travel time of the seismic wave from the reflection point to the detector, corresponding to the time axis of the data acquired by the tunnel seismograph. Vector x represents the spatial distribution of the source and receiver points on the aperture A, which is the area where the source and receiver are located. In the tunnel, it appears as a straight line along the tunnel's direction, but it can also be a curved surface of any shape. To cover the most common case, the diffraction superposition is represented in the form of a surface integral. This refers to taking the time derivative of the input acquired seismic data u, and the weighting function. To ensure the accuracy of amplitude recovery during the backpropagation of the wave field, for elastic wave fields recorded on two or three components, the wave field must be projected onto the polarization direction of any imaging point. In this case, the Kirchhoff migration imaging formula can be expressed as:

[0032] ;

[0033] At this point, the imaging result V(m) consists of a multi-component wavefield that has propagated through reflection, and this wavefield is projected onto the expected polarization direction e of the studied ray path at the receiving point. r Above. Assuming the tunnel geological model is mean, e r It is a unit vector whose direction is determined by the straight ray between the imaging point and the receiving point, and can be calculated by measuring the spatial relationship between the tunnel blast and receiver in the tunnel before data acquisition.

[0034] The above formula reflects the basic principle of Kirchhoff migration imaging. However, in tunnel apertures, the lack of large-aperture data can cause severe image arcing and greatly reduce resolution.

[0035] like Figure 1 As shown, the gray area represents the Fresnel zone below the isochronous surface (the constant two-way travel time line of the scattering point M). The angle γ represents the angle between the measured polarization direction (i.e., the wave propagation direction for P-waves) and the theoretical polarization direction (defined by the wave propagation direction originating from point D on the isochronous surface).

[0036] For any seismic wave ray between two points, the Fresnel zone can be represented as follows: Figure 1 As shown, consider a direct seismic ray from the source point S to the receiver point R, with a length of l. The scattering point D lies within the neighborhood of the ray. The length of the ray from the source point to D is s, and the length of the scattered ray from D to the receiver point R is r. The Fresnel condition is as follows: where λ is the wavelength of the seismic wave, obtained from the measured seismic wave velocity at a given frequency, and n is an integer representing the nth order Fresnel body.

[0037] ;

[0038] In the constant velocity medium model, the nth-order Fresnel volume expression for ray SMR is:

[0039] ;

[0040] ;

[0041] Among them, H1, H2 and H n It's a numerical threshold determined by geometry and wavelength. H1 represents the threshold of the first Fresnel boundary, used to determine whether to completely preserve the data. H2 represents the threshold of the second Fresnel boundary, used to determine whether to enter the linear decay region or discard the data directly. n The discrimination threshold corresponding to the nth order Fresnel volume.

[0042] At this point, Kirchhoff's offset formula becomes the following equation, F c As a weighting factor in tunnel small-aperture imaging, it takes into account the influence of Nell band correction in cases of insufficient coverage, thus greatly reducing the impact of arc drawing:

[0043] ;

[0044] Image recognition and interpretation for tunnel earthquake early warning includes the following steps:

[0045] The imaging color profile is read and converted to the HSV color space. The color threshold is adaptively adjusted based on hue and saturation components. After multi-scale contour detection and fusion, and contour grouping optimization, key geological targets are labeled and visualized. Specifically:

[0046] First, the input offset imaging color profile is converted from the RGB color space to the HSV color space. The HSV space decomposes color information into three independent components: hue, saturation, and lightness, which is closer to human visual perception and facilitates effective separation of different geological responses based on color features. Then, adaptive color threshold adjustment is performed. The core of this step is to dynamically determine the hue and saturation threshold ranges for extracting target areas (such as strongly reflective interfaces) based on the color distribution characteristics of each image, thereby avoiding over-segmentation or under-segmentation problems caused by fixed thresholds. Specifically: first, an initial screening threshold (e.g., saturation > 40) is applied to the saturation component to highlight potential target areas with vivid colors; based on this, the histogram of the hue component within this area is statistically analyzed to identify its dominant peak, which reflects the dominant geological reflection color tendency in the image; then, the upper and lower offsets of the hue threshold are dynamically calculated based on the peak position—if the peak is located in the red region (hue value near 0° or 180°), the red threshold range is expanded to cover its cross-boundary characteristics; if the peak is located in the blue region (hue value approximately 120°), the blue threshold is adjusted accordingly. Meanwhile, the saturation offset is adaptively constrained based on the statistical dispersion (such as standard deviation) of the image saturation components, thereby ensuring that the threshold setting is adapted to the overall color contrast of the image. This adaptive process is ultimately manifested as a dynamic threshold generation mechanism for different color ranges.

[0047] Building upon this foundation, multi-scale contour detection and fusion are performed. Since the scale of geological targets in seismic profiles varies with depth and lithology, single-scale detection is prone to missing small-scale targets or failing to completely segment large-scale structures. Therefore, this method performs color segmentation and contour extraction in parallel at multiple scaling ratios (e.g., original scale 1.0, reduced scale 0.5, enlarged scale 1.25). At each scale, a color mask is generated using the aforementioned adaptive threshold to extract candidate contours; the coordinates of all detected contours are uniformly mapped back to the original image size, achieving integration of cross-scale detection results. The extracted contours are further grouped and optimized. Detection at different scales may generate multiple overlapping or adjacent contours for the same geological body. Therefore, this method performs clustering based on the intersection-union ratio (IU) of the contour bounding boxes, merging duplicate contours belonging to the same geological entity. Then, the contours are sorted according to their pixel area, and the most significant targets (e.g., the top 3 with the largest areas) are selected as key interpretation objects, thereby improving interpretation efficiency while ensuring the representativeness of the results.

[0048] Finally, the results are labeled and visualized. For the selected key geological targets, bounding boxes are drawn and their attribute information, such as inferred geological type, salience ranking, and pixel size, is labeled. Simultaneously, a semi-transparent color mask is generated and overlaid on the original image to visually present the target areas identified by the algorithm. The final output includes labeled images and a structured data report.

[0049] This application utilizes Fresnel migration aperture-constrained small-aperture Kirchhoff migration imaging, which results in less diffraction distortion in the image under small-aperture conditions, leading to imaging results that better reflect actual geological conditions. Furthermore, this application employs an intelligent interpretation method for tunnel seismic images based on adaptive color segmentation, multi-scale contour fusion, and target optimization. This method replaces manual interpretation relying on personal experience with a data-driven, objective process, enabling the automatic and standardized extraction of key geological information from images.

[0050] The following conclusions are drawn using the processed data from this invention:

[0051] like Figure 2 , Figure 3 and Figure 4 As shown in the comparison, it was found that Figure 3 The reflective interface in the medium image is significantly straighter, while Figure 2 The intermediate reflection interface exhibits an arc-shaped pattern, resulting in significant diffraction. This indicates that the proposed method can better reduce the impact of the arc and improve the resolution of reflections in the leading direction; through Figure 4 The image recognition interpretation results can clearly identify the two strongest reflective interfaces from the three profiles. By combining the extracted reflective interfaces with geological quality, the geological anomalies can be identified more accurately, reducing the ambiguity of manual identification and improving accuracy.

Claims

1. A method for early prediction of tunnel earthquakes using small-aperture imaging and automatic interpretation, characterized in that, Includes the following steps: S1: Small-aperture imaging for advanced seismic prediction in tunnels, including the following sub-steps: S1.1: Set up a tunnel seismic observation system, arrange seismic sources as the source points and geophone arrays as the receiving points, artificially generate seismic waves, and collect tunnel seismic wave data; S1.2: Read the acquired seismic wave data, filter the seismic wave data, and obtain the filtered seismic wave data; S1.3: Calculation of Fresnel volume function F based on Fresnel condition and constant velocity medium model c , will F c As a weighting factor for small aperture imaging; S1.4: F c Substituting the Kirchhoff migration imaging formula, the filtered seismic wave data was used to implement the F-band imaging. c Constrained Kirchhoff migration imaging calculations; S1.5: Output small-aperture imaging results for tunnel seismic advance prediction; S2: Image recognition and interpretation for tunnel earthquake early warning, including the following sub-steps: S2.1: Read the small aperture imaging results output by S1.5 and obtain the imaging color profile; S2.2: Convert the imaging color profile from the RGB color space to the HSV color space to obtain the HSV space imaging profile. The HSV space imaging profile is decomposed into independent components including hue, saturation, and brightness. S2.3: Adaptive color threshold adjustment of HSV spatial imaging profile based on independent components, dynamically generating color thresholds according to the color distribution characteristics of the HSV spatial imaging profile itself; The adaptive color threshold adjustment performed on the HSV spatial imaging profile is as follows: First, apply a preliminary screening threshold with a saturation greater than a preset value X1 to the saturation component of the HSV spatial imaging profile. Then, statistically analyze the histogram of the hue component in the region after preliminary screening and identify the main peak. Based on the position of the main peak, dynamically calculate the upper and lower offset of the hue threshold. At the same time, adaptively constrain the saturation offset according to the statistical dispersion of the saturation component to generate a dynamic color threshold. The vertical offset of the hue threshold is dynamically calculated based on the position of the main peak, specifically as follows: If the peak value is located in the red region, expand the red threshold range to cover its cross-boundary characteristics; if the peak value is located in the blue region, adjust the blue threshold range accordingly. S2.4: Based on color thresholding, perform multi-scale contour detection and fusion on HSV spatial imaging profiles, extract contours at multiple scaling ratios and integrate the detection results; S2.5: Perform contour grouping and optimization on the candidate contours after detection and fusion, remove duplicate contours and select contours corresponding to significant geological targets; S2.6: Annotate and visualize the results of the selected key geological targets, and output annotated images and structured data reports.

2. The method for tunnel seismic advance prediction with small-aperture imaging and automatic interpretation according to claim 1, characterized in that, Calculate the Fresnel body function F in S1.3 c The Fresnel condition formula used is as follows: ，n=1,2,…； Where l is the length of the direct seismic ray from the source point to the receiver point, s is the length of the ray from the source point to the scattering point, r is the length of the scattered ray from the scattering point to the receiver point, λ is the wavelength of the seismic wave obtained from the measured seismic wave velocity and frequency, and n is an integer representing the nth order Fresnel body.

3. The method for tunnel seismic advance prediction using small-aperture imaging and automatic interpretation according to claim 2, characterized in that, Calculate the Fresnel body function F in S1.3 c The expression for an nth-order Fresnel body under the constant velocity medium model is as follows: ； ； In the formula, γ is the angle between the measured polarization direction and the theoretical polarization direction, where H1, H2 and H n It is a numerical threshold determined by geometry and wavelength. H1 represents the threshold of the first Fresnel boundary, used to determine whether to retain it completely, and H2 represents the threshold of the second Fresnel boundary, used to determine whether to enter the linear decay region or discard it directly. n The discrimination threshold corresponding to the nth order Fresnel volume.

4. The method for tunnel seismic advance prediction with small-aperture imaging and automatic interpretation according to claim 3, characterized in that, The basic formula for calculating Kirchhoff migration imaging in S1.4 is as follows: ； In the formula, V(m) represents the two-dimensional spatial point of the image and the corresponding reflected wave energy, and t S t is the travel time of a seismic wave from its source to its reflection point. R The vector x represents the travel time of the seismic wave from the reflection point to the detector, and represents the spatial distribution of the source and receiver points on the aperture A. To perform time derivative calculation on the filtered seismic wave data u, A weighting function is used to ensure accurate recovery of the wave field amplitude during backpropagation.

5. The method for tunnel seismic advance prediction with small-aperture imaging and automatic interpretation according to claim 4, characterized in that, S1.1 acquires seismic wave data in the elastic wave field. The basic formula for Kirchhoff migration imaging calculation is: ； Before performing imaging calculations, the elastic wave field is projected onto the polarization direction of any imaging point, where the polarization direction is determined by the unit vector e, which is a straight ray between the imaging point and the receiving point. r It means, e r It was calculated by measuring the spatial relationship between the tunnel blast and the receiver.

6. The method for tunnel seismic advance prediction using small-aperture imaging and automatic interpretation according to claim 5, characterized in that, In S1.4, the Fresnel body function F c The constraint formula after substituting Kirchhoff's migration imaging formula is: ； The filtered seismic wave data obtained from S1.2 is processed through the Fresnel body function F. c Fresnel band correction was applied to Kirchhoff migration imaging.

7. The method for tunnel seismic advance prediction using small-aperture imaging and automatic interpretation according to claim 1, characterized in that, The multi-scale contour detection and fusion performed on the threshold-adjusted HSV spatial imaging profile in S2.4 are as follows: Under multiple scaling ratios of the preset original scale Y1, reduced scale Y2, and enlarged scale Y3, color segmentation and contour extraction are performed in parallel on the HSV spatial imaging profile after threshold adjustment. The coordinates of the detected contours at each scale are uniformly mapped back to the original image size, realizing the integration of cross-scale detection results and obtaining candidate contours.

8. The method for tunnel earthquake early prediction with small-aperture imaging and automatic interpretation according to claim 7, characterized in that, The grouping and optimization of candidate contours in S2.5 are as follows: Candidate contours are clustered and grouped based on the intersection-union ratio of contour bounding boxes. Duplicate contours belonging to the same geological entity are merged. Contours are sorted according to the pixel area of ​​each independent contour and the contours corresponding to significant targets are selected as key geological target contours. The results of the annotation and visualization output of the key geological target outlines in S2.6 are as follows: Boundary boxes are drawn for the outlines of key geological targets, and attribute information such as inferred geological type, salience ranking, and pixel size is labeled. A semi-transparent color mask is generated and overlaid on the original color profile of the small aperture imaging for tunnel seismic advance prediction. Finally, labeled images and structured data reports containing geological target attributes are output.

Images