Integrated optimization layout and high-precision automatic tunnel face three-dimensional reconstruction method and device

Abstract

This invention discloses an integrated optimized layout and high-precision automated three-dimensional reconstruction method and device for tunnel face, comprising the following steps: S1: setting physical attitude constraints for the imaging terminal to control the horizontal tilt deviation of the imaging terminal within a preset threshold range; S2: under the premise of satisfying the physical attitude constraints, using the imaging terminal at at least three equally spaced camera stations in the same cross section in front of the tunnel face, pointing towards the center area of ​​the tunnel face to capture multiple high-resolution original images covering the entire cross section and having overlapping fields of view; S3: using the SFM algorithm to extract and match image feature points from the original images to generate a dense point cloud with high-definition color and texture information; S4: based on the physical attitude constraints, automatically correcting to a world coordinate system with the gravity direction as the normal vector; S5: based on the dense point cloud model and its texture information corrected in step S4, automatically extracting the structural surface traces of the rock mass at the tunnel face and calculating the attitude parameters.

Description

Technical Field

[0001] This invention relates to the fields of tunnel engineering monitoring and computer vision technology, specifically to an integrated optimized layout and high-precision automated three-dimensional reconstruction method and device for tunnel face. Background Technology

[0002] In the New Austrian Tunneling Method (NATM) for tunnel construction, geological logging, over- and under-excavation analysis, and stability evaluation of the tunnel face are crucial for ensuring construction safety and quality. Timely and accurate acquisition of the three-dimensional geometric information and texture features of the tunnel face is of great significance for guiding the blasting design and support parameter adjustments in the next cycle.

[0003] Currently, common methods for obtaining geometric information of the tunnel face mainly include total station surveying, 3D laser scanning, and close-range photogrammetry, but these existing technologies all have certain limitations in practical applications.

[0004] While traditional total station surveying methods offer high accuracy, they typically only collect sparse feature points of the tunnel face outline or cross-section, failing to acquire high-density point cloud data. This makes it difficult to fully reflect the complex, uneven morphology of the tunnel face and lacks texture information, which is detrimental to subsequent geological analysis.

[0005] While 3D laser scanners can acquire high-precision dense point clouds, the equipment is typically expensive and bulky, and the scanning process is time-consuming (usually several minutes to tens of minutes). In the harsh environment of tunnel construction, the transportation and installation of the equipment are extremely inconvenient, and prolonged scanning operations can disrupt the normal construction progress and pose safety hazards.

[0006] While traditional surface photogrammetry (SFM) offers advantages such as lightweight equipment and fast data acquisition, it typically relies on establishing ground control points (GCPs) to address scale ambiguity and coordinate system orientation issues. In the narrow, dark, and dynamically constructed tunnel face, establishing and measuring control points is not only cumbersome and time-consuming but also highly susceptible to construction interference. Furthermore, the dim lighting inside tunnels often requires ordinary cameras to increase ISO to ensure proper exposure, leading to increased image noise and severely impacting feature point matching and point cloud reconstruction quality. Using external strong light sources further complicates the equipment and increases operational difficulty. Summary of the Invention

[0007] The purpose of this invention is to provide an integrated optimized layout and high-precision automated three-dimensional reconstruction method for tunnel face, so as to at least solve the problems of cumbersome operation, bulky equipment, need to set up control points, and poor adaptability to low light environment in the existing technology.

[0008] To address the aforementioned technical problems, in a first aspect, the present invention provides a method and apparatus for integrated optimized layout and high-precision automated three-dimensional reconstruction of tunnel face, comprising the following steps:

[0009] S1: Set the physical attitude constraints of the imaging terminal to control the horizontal tilt deviation of the imaging terminal within a preset threshold range.

[0010] S2: Under the premise of satisfying physical attitude constraints, the imaging terminal is used to take pictures at at least three equally spaced shooting stations in the same cross section in front of the tunnel face, pointing to the center area of ​​the tunnel face respectively, to obtain multiple high-resolution original images covering the entire cross section and having overlapping fields of view.

[0011] S3: Use the SFM algorithm to extract and match image feature points from the original image to generate a dense point cloud with high-definition color and texture information;

[0012] S4: Based on physical attitude constraints, a rotation matrix is ​​constructed using prior information about the direction of gravity during shooting. Then, the dense point cloud model generated in step S3 is automatically corrected to the world coordinate system with the direction of gravity as the normal vector using the rotation matrix.

[0013] S5: Based on the dense point cloud model and its texture information corrected in step S4, the structural surface traces of the rock mass at the working face are automatically extracted, and the attitude parameters are calculated.

[0014] Furthermore, the imaging terminal integrates a large-area high-sensitivity image sensor, a large-aperture optical lens, and an attitude sensor; when operating the imaging terminal to take pictures, the horizontal tilt deviation of the imaging terminal is controlled within a preset threshold range through real-time feedback from the attitude sensor.

[0015] Furthermore, the target surface size of the large-target high-sensitivity image sensor is not less than 1 inch, and the single pixel size is not less than 3.0μm; the aperture F-number of the large-aperture optical lens is not greater than 1.8; and the equivalent focal length of the imaging terminal is 24mm to 28mm.

[0016] Furthermore, the physical attitude constraint is that the absolute values ​​of the deviations of the imaging terminal's pitch and roll angles from the absolute horizontal plane are both less than or equal to 1°.

[0017] Furthermore, in step S2, when each camera station takes a picture, the optical axes of the imaging terminal converge at the geometric center region of the working face, and the overlap rate between adjacent images is not less than 90%.

[0018] Further, step S4 includes:

[0019] S41. Extract the camera rotation matrix corresponding to each original image calculated by the SFM algorithm;

[0020] S42. Based on the camera rotation matrix, calculate the gravity direction vector corresponding to each original image in the local coordinate system;

[0021] S43. Calculate the global average vector of all gravity direction vectors and solve for the rotation matrix that rotates the global average vector to be consistent with the gravity axis of the world coordinate system.

[0022] S44. Use a rotation matrix to perform a rigid rotation transformation on the dense point cloud model to complete the gravity direction correction.

[0023] Furthermore, in step S5, the structural surface traces of the rock mass at the tunnel face are identified using a clustering algorithm or an edge detection algorithm.

[0024] In a second aspect, the present invention also provides a tunnel face three-dimensional reconstruction device for implementing the method provided in the first aspect, comprising:

[0025] Imaging terminal, used to acquire high-definition raw images of the tunnel face in low-light environments;

[0026] An attitude sensor, mechanically fixed to the imaging terminal, is used to monitor and feed back the spatial attitude data of the imaging terminal in real time.

[0027] The processing unit is communicatively connected to the imaging terminal and the attitude sensor, and is used to perform steps S3-S5 of the method provided in the first aspect above.

[0028] Furthermore, the imaging terminal is a handheld imaging terminal.

[0029] Furthermore, the attitude sensor is a digital inclinometer, with an attitude angle monitoring accuracy of 0.1°.

[0030] The beneficial effects of this invention are as follows:

[0031] 1. By using physical attitude constraints provided by digital sensors to replace traditional ground control points, the problem of coordinate system deflection in point cloud models without control points is solved, greatly simplifying the on-site operation process; through physical attitude constraints and gravity correction, automated and high-precision reconstruction of the 3D model of the working face and automatic identification of geological occurrence are realized.

[0032] 2. By adopting a handheld shooting method and combining it with a "shoot and go" operation mode, the on-site data collection time can be reduced to less than 1 minute, without interfering with the normal tunnel construction. Through the hardware combination of a large target surface sensor and a large aperture lens, the amount of light entering the camera is significantly improved. Low-noise, high-definition images can be obtained without complex supplementary lighting in the dim environment of the tunnel, ensuring the accuracy of subsequent point cloud reconstruction. Attached Figure Description

[0033] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, use the same reference numerals to denote the same or similar parts. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0034] Figure 1 This is a schematic diagram of the overall implementation process of this method;

[0035] Figure 2 This is a side view showing the image capture position and angle along the tunnel excavation direction in this embodiment, illustrating the relative position of the camera and the tunnel face and the application of the inclinometer;

[0036] Figure 3 This is a schematic diagram of the field of view overlap acquired through multi-field coverage in this embodiment;

[0037] Figure 4 This diagram illustrates the coordinate correction principle based on gravity constraints, showing the process of correcting a point cloud from a skewed state to a horizontal state.

[0038] Figure 5 Here is an example of the original image set of the tunnel face obtained by this method; where: Figure 5 (a) is the image of the tunnel face at the intermediate measuring point. Figure 5 (b) is the image of the working face at the left measuring point. Figure 5 (c) is the tunnel face image of the measuring point on the right;

[0039] Figure 6 This is a schematic diagram of the three-dimensional point cloud model of the tunnel face reconstructed based on this method;

[0040] Figure 7 This is an automatic identification map of structural surfaces based on the 3D point cloud of the tunnel face. Detailed Implementation

[0041] like Figure 1 The integrated optimized layout and high-precision automated 3D reconstruction method for tunnel face, as shown, includes the following steps:

[0042] S1: Set the physical attitude constraints of the imaging terminal to control the horizontal tilt deviation of the imaging terminal within a preset threshold range.

[0043] S2: Under the premise of satisfying physical attitude constraints, use the imaging terminal to take pictures at at least three equally spaced shooting stations in the same cross section in front of the tunnel face, pointing to the center area of ​​the tunnel face respectively, to obtain multiple (at least three) high-resolution original images covering the entire cross section and having overlapping fields of view.

[0044] S3: Use the existing Structure for Motion Recovery (SFM) algorithm to extract and match image feature points from the original image to generate a dense point cloud with high-definition color and texture information;

[0045] S4: Based on physical attitude constraints, a rotation matrix is ​​constructed using prior information about the direction of gravity during shooting. Then, the dense point cloud model generated in step S3 is automatically corrected to the world coordinate system with the direction of gravity as the normal vector using the rotation matrix.

[0046] S5: Based on the dense point cloud model and its texture information corrected in step S4, the structural surface traces of the rock mass at the working face are automatically extracted, and the attitude parameters are calculated.

[0047] This method uses physical attitude constraints provided by digital sensors to replace traditional ground control points, solving the problem of coordinate system deflection in point cloud models without control points and greatly simplifying the on-site operation process. Through physical attitude constraints and gravity correction, it realizes automated, high-precision reconstruction of the three-dimensional model of the working face and automatic identification of geological occurrence.

[0048] According to one embodiment of this application, the imaging terminal integrates a large-area high-sensitivity image sensor, a large-aperture optical lens, and an attitude sensor. When the imaging terminal is used for shooting, the horizontal tilt deviation of the imaging terminal is controlled within a preset threshold range through real-time feedback from the attitude sensor. Considering that the illumination in tunnels is usually below 50 Lux, this method abandons the traditional flash lighting scheme and adopts a passive high-sensitivity scheme of "large sensor + large aperture".

[0049] According to one embodiment of this application, the target surface size of the large-area high-sensitivity image sensor is not less than 1 inch, and the single pixel size is not less than 3.0 μm; the aperture F-number of the large-aperture optical lens is not greater than 1.8; and the equivalent focal length of the imaging terminal is 24mm to 28mm. This embodiment, through the combination of a large-area high-sensitivity image sensor and a large-aperture optical lens, significantly improves the light intake and signal-to-noise ratio in low-light environments, ensuring that high-definition, low-noise images can be acquired without additional light sources in dim tunnel environments, thus guaranteeing the accuracy and reliability of subsequent SFM algorithm feature matching and point cloud reconstruction from the source.

[0050] According to one embodiment of this application, the imaging terminal may be a camera equipped with a 1-inch CMOS sensor, with a single pixel size of 3.2μm, which significantly improves the light intake compared to ordinary mobile phone sensors; the lens is a fixed-focus lens with a large aperture of f / 1.8 and an equivalent focal length of 24mm.

[0051] According to one embodiment of this application, the physical attitude constraint is that the absolute values ​​of the deviations of the imaging terminal's pitch and roll angles relative to the absolute horizontal plane are both less than or equal to 1°. This embodiment strictly controls the horizontal tilt angle deviation within ±1°, meaning that the camera's attitude satisfies the following physical constraint:

[0052] (1) Roll angle ( ) The camera is not tilted left or right.

[0053] (2) Pitch angle ( ) The camera's optical axis is horizontal, with no upward or downward view.

[0054] Ideally, the camera coordinate system (Camera coordinate system) definition: The axis is the direction of the camera's optical axis. The axis is perpendicular to the image sensor. The axis is the horizontal direction of the image sensor and the world coordinate system. (definition: The axis is parallel to the direction of gravity (vertically upward). Rotational relationships between planes (horizontal plane) This can be simplified. Typically, the rotation matrix is ​​constructed using Euler angles:

[0055]

[0056] Due to physical constraints Then the vertical axis of the camera coordinate system Theoretically, it should be aligned with the gravity axis of the world coordinate system. (or (Depending on the definition) Parallel or orthogonal. In this embodiment (assuming the camera is held horizontally for shooting), when the tiltmeter reading is 0, the camera's imaging plane is perpendicular to the horizontal plane. At this time, the gravity vector in the world coordinate system... Projection in the camera coordinate system It should correspond to the vertical axis direction of the camera.

[0057] This method provides high-precision physical prior information for calculating the direction of gravity by setting physical attitude constraints. This strict constraint makes it possible to use the average gravity vector of multiple images to correct the point cloud coordinate system, directly eliminating model tilt error. It is the key to eliminating the need for control point layout and realizing automatic horizontal correction of the coordinate system.

[0058] According to one embodiment of this application, in step S2, when each camera station takes pictures, the optical axes of the imaging terminals converge at the geometric center region of the working face, and the overlap rate between adjacent images is not less than 90%. This embodiment ensures the consistency of the coverage area and the stability of the geometric relationship of the multi-view images by making the optical axes of the imaging terminals converge at the center when the camera station takes pictures; and the high overlap rate of not less than 90% greatly increases the number of common feature points between images, significantly improving the success rate and robustness of feature point matching in the SFM algorithm, thereby generating a more complete and accurate dense point cloud.

[0059] According to one embodiment of this application, step S4 includes:

[0060] S41. Extract the camera rotation matrix corresponding to each original image calculated by the SFM algorithm; this step specifically includes: after performing bundle adjustment on N original images using the SFM algorithm, the camera extrinsic parameters of each image in an arbitrarily initialized local coordinate system (L) will be output, including the rotation matrix. , ( =1,2,...,N);

[0061] S42. Based on the camera rotation matrix, calculate the gravity direction vector corresponding to each original image in the local coordinate system; under the ideal "horizontal shooting" physical constraint, the Y-axis of the camera coordinate system (C) (usually corresponding to the vertical direction of the image sensor and pointing "downward") should be aligned with the gravity direction (i.e., the Z-axis or -Z-axis) of the world coordinate system (W); therefore, for the first... The image shows the direction vector of gravity in the local coordinate system L. It can be obtained by applying the inverse (or transpose) of its camera rotation matrix to the "downward" direction in the camera coordinate system, as expressed by the formula:

[0062]

[0063] Where: It is assumed here that the Y-axis of the image coordinate system is downward; Represents the "down" vector in the camera coordinate system;

[0064] S43. Calculate the global average vector of all gravity direction vectors and solve for the rotation matrix that rotates this global average vector to align with the gravity axis of the world coordinate system. Since the gravity direction of all cameras is physically consistent (i.e., the true gravity direction in the local coordinate system L), it will be slightly dispersed due to SFM calculation errors and minor attitude deviations. By calculating these N... The average value can yield a more robust, globally averaged gravity vector that represents the direction of gravity in the local coordinate system L. To correct the point cloud to the world coordinate system (Z-axis upwards), we need to solve for a rotation matrix. , so that:

[0065]

[0066] Align the direction of gravity in the local coordinate system with the vertical downward axis of the world coordinate system; thus, through Rotating the entire point cloud corrects it to an absolute coordinate system based on the direction of gravity; this rotation matrix... The solution can be obtained using the Rodriguez formula or quaternion interpolation. Rotate to The required axis angle pair.

[0067] S44. Perform a rigid rotation transformation on the dense point cloud model using a rotation matrix to correct the gravity direction; the transformation formula is expressed as:

[0068]

[0069] in: It is a scale factor (calibrated using the known width of the face or a reference object). The translation vector is used. By utilizing the ±1° physical constraint during shooting, the tilt error of the point cloud model is directly eliminated, making the output model naturally have the horizontal reference system required for engineering, thus saving the tedious process of setting up control points.

[0070] According to one embodiment of this application, in step S5, structural surface traces of the rock mass at the tunnel face are identified using a clustering algorithm or an edge detection algorithm. Mature clustering algorithms (such as region growing) or edge detection algorithms can automatically and objectively extract the spatial distribution (traces) of structural surfaces such as faults and joints in the rock mass, providing a basis for subsequent occurrence parameter calculations. This replaces traditional manual identification and improves the efficiency and accuracy of geological logging.

[0071] The attitude parameters include the strike, dip, and dip angle of the structural plane; this method achieves this by identifying abrupt changes in the normal vector of the point cloud or color differences in the image information. Secondly, the present invention also provides a three-dimensional reconstruction device for a tunnel face for implementing the method provided in the first aspect, comprising:

[0072] Imaging terminal, used to acquire high-definition raw images of the tunnel face in low-light environments;

[0073] An attitude sensor, mechanically fixed to the imaging terminal, is used to monitor and feed back the spatial attitude data of the imaging terminal in real time.

[0074] The processing unit is communicatively connected to the imaging terminal and the attitude sensor, and is used to perform steps S3-S5 of the method provided in the first aspect above.

[0075] This device integrates hardware selection, data acquisition, and intelligent processing, enabling a single person to quickly complete the entire process from acquisition to analysis on-site, thus realizing the productization and engineering application of the technology.

[0076] According to one embodiment of this application, the attitude sensor is a digital inclinometer with an attitude angle monitoring accuracy of 0.1°. The use of a high-precision digital inclinometer with an accuracy of 0.1° ensures the reliability of the physical attitude measurement data. This high precision is the fundamental guarantee for the effective execution of ±1° attitude constraints and ultimately for the recovery of the absolute spatial attitude (i.e., gravity direction) of the point cloud under control-point-free conditions. It is one of the core components for the device to achieve its expected technical specifications.

[0077] According to one embodiment of this application, the imaging terminal is a handheld imaging terminal. By adopting a handheld shooting method and combining it with a "shoot and go" operation mode, the on-site data acquisition time can be reduced to less than 1 minute, without interfering with the normal tunnel construction.

[0078] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions 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 solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for integrating optimized layout and high-precision automated three-dimensional reconstruction of tunnel face, characterized in that, Includes the following steps: S1: Set the physical attitude constraint of the imaging terminal so that the horizontal tilt angle deviation of the imaging terminal is controlled within a preset threshold range; S2: Under the premise of satisfying the physical attitude constraints, the imaging terminal is used to take pictures at at least three equally spaced camera stations in the same cross section in front of the tunnel face, pointing to the center area of ​​the tunnel face respectively, to obtain multiple high-resolution original images covering the entire cross section and having overlapping fields of view. S3: Use the SFM algorithm to extract and match image feature points from the original image to generate a dense point cloud with high-definition color and texture information; S4: Based on the physical attitude constraints, a rotation matrix is ​​constructed using the prior information of the gravity direction during shooting. Then, the dense point cloud model generated in step S3 is automatically corrected to the world coordinate system with the gravity direction as the normal vector using the rotation matrix. S5: Based on the dense point cloud model and its texture information corrected in step S4, the structural surface traces of the rock mass at the working face are automatically extracted, and the attitude parameters are calculated.

2. The integrated optimized layout and high-precision automated three-dimensional reconstruction method for tunnel face according to claim 1, characterized in that, The imaging terminal integrates a large-area high-sensitivity image sensor, a large-aperture optical lens, and an attitude sensor; when the imaging terminal is operated to take pictures, the horizontal tilt angle deviation of the imaging terminal is controlled within a preset threshold range through real-time feedback from the attitude sensor.

3. The integrated optimized layout and high-precision automated three-dimensional reconstruction method for tunnel face according to claim 2, characterized in that, The target surface size of the large-target high-sensitivity image sensor is not less than 1 inch, and the single pixel size is not less than 3.0μm; the aperture F-number of the large-aperture optical lens is not greater than 1.8; and the equivalent focal length of the imaging terminal is 24mm to 28mm.

4. The integrated optimized layout and high-precision automated three-dimensional reconstruction method for tunnel face according to any one of claims 1-3, characterized in that, The physical attitude constraint is that the absolute values ​​of the deviations of the pitch angle and roll angle of the imaging terminal relative to the absolute horizontal plane are both less than or equal to 1°.

5. The integrated optimized layout and high-precision automated three-dimensional reconstruction method for tunnel face according to claim 1, characterized in that, In step S2, when each camera station takes pictures, the optical axes of the imaging terminals converge at the geometric center region of the working face, and the overlap rate between adjacent images is not less than 90%.

6. The integrated optimized layout and high-precision automated three-dimensional reconstruction method for tunnel face according to claim 1, characterized in that, Step S4 includes: S41. Extract the camera rotation matrix corresponding to each original image calculated by the SFM algorithm; S42. Based on the camera rotation matrix, calculate the gravity direction vector corresponding to each original image in the local coordinate system; S43. Calculate the global average vector of all gravity direction vectors and solve for the rotation matrix that rotates the global average vector to be consistent with the gravity axis of the world coordinate system. S44. Use the rotation matrix to perform a rigid rotation transformation on the dense point cloud model to complete the gravity direction correction.

7. The integrated optimized layout and high-precision automated three-dimensional reconstruction method for tunnel face according to claim 1, characterized in that, In step S5, the structural surface traces of the rock mass at the tunnel face are identified by a clustering algorithm or an edge detection algorithm.

8. A three-dimensional reconstruction device for a tunnel face for implementing the method of any one of claims 1-7, characterized in that, include: Imaging terminal, used to acquire high-definition raw images of the tunnel face in low-light environments; An attitude sensor is mechanically fixed to the imaging terminal and is used to monitor and feed back the spatial attitude data of the imaging terminal in real time. The processing unit is communicatively connected to the imaging terminal and the attitude sensor, and is used to execute steps S3-S5 as described in any one of claims 1-8.

9. The tunnel face three-dimensional reconstruction device according to claim 8, characterized in that, The attitude sensor is a digital tiltmeter, and its attitude angle monitoring accuracy reaches 0.1°.

10. The tunnel face three-dimensional reconstruction device according to claim 9, characterized in that, The imaging terminal is a handheld imaging terminal.

Images