Long-distance pressure tunnel positioning method based on inner wall feature texture auxiliary identification

Abstract

The invention discloses a long-distance pressure tunnel positioning method based on inner wall feature texture auxiliary identification, and the method is based on the fact that the inner wall of a long-distance pressure tunnel presents different feature textures at different times due to the non-uniform water absorption of the inner wall material. Meanwhile, in combination with the difference of characteristic textures such as pores, gaps, denudation, damage, aging and attachments of concrete structures at different parts of the tunnel and the relevance of the characteristic textures at the same position and at different time, the characteristic textures of the inner wall are associated with the specific position by adopting a mode similar to fingerprint identification based on a panoramic database constructed in a non-pressure period and an emptying period; and finally, comprehensive, accurate and rapid positioning of the underwater non-defect area and the defect area in the'pressure period 'in the long-distance pressure tunnel is achieved.

Description

Technical Field

[0001] This invention relates to the field of tunnel positioning technology, and in particular to a long-distance pressurized tunnel positioning method based on the auxiliary identification of inner wall feature texture. Background Technology

[0002] Due to factors such as high water pressure, erosion from flowing water, and the effects of surrounding rock, pressurized water conveyance tunnels in hydropower stations, pumped storage power stations, and other projects are prone to defects such as concrete erosion, damage, cracks, and leakage after long-term operation. Failure to inspect and repair these defects in a timely manner may seriously affect the safe operation of the project. For long-distance pressurized tunnels (generally over 1 km), due to the difficulty, high cost, and long emptying cycle, which affects project operation and scheduling, drainage is not carried out during most annual tunnel maintenance periods. Underwater maintenance work is primarily conducted using underwater robots.

[0003] When underwater robots are inspecting tunnels, they need to provide the location of the tunnels being inspected. However, due to the limited length of the tunnels (some tunnels are over 10km long), the complex environment such as tunnel bends, intersections of main tunnels and branch tunnels, darkness, and murky water, traditional positioning methods such as GPS are difficult to use in tunnels. Positioning methods such as umbilical cable length, sonar, inertial navigation systems, and Doppler logs are difficult to meet the requirements for high-precision positioning. In addition, because tunnels are subjected to long-term flowing and high-pressure water erosion, it is difficult to place or retain fixed positioning marks on the inner walls of the tunnels. All of these factors have caused great difficulties for underwater robots in positioning within tunnels.

[0004] In the prior art, Chinese patent document CN113252028A discloses a positioning method, electronic device, and storage medium for a robot inside a water conveyance tunnel. This document obtains the position information of these features by combining the structural drawings of the water conveyance tunnel with information about the joints, rivet grooves, monitoring cable manholes, communication optical cable manholes, and other representative features within the tunnel. Its drawback is that, due to the limited information on features in the drawings and the scattered distribution of these features, it is difficult to obtain accurate positioning information based on the features in the drawings, resulting in low underwater positioning accuracy. Summary of the Invention

[0005] To address the existing technical problems, the main objective of this invention is to provide a long-distance pressurized tunnel positioning method based on the assisted identification of inner wall feature textures. This method enables underwater robots to be positioned underwater during tunnel maintenance by utilizing the inner wall feature textures of the tunnel.

[0006] The technical solution adopted in this invention is: a long-distance pressurized tunnel positioning method based on inner wall feature texture-assisted identification, comprising the following steps: S1. Drainage is started inside the tunnel, and the tunnel changes from a pressurized state to a depressurized state, entering the maintenance period. S2. Obtain panoramic image data of the tunnel inner wall during the "pressure-free period" using the first image acquisition device; S3. After the water in the tunnel is drained, panoramic image data of the tunnel wall during the "drainage period" is obtained through the second image acquisition device. S4. Extract the feature texture data of the tunnel inner wall and construct a panoramic database of tunnel inner wall features during the "no-pressure period" and "emptying period"; S5. Merge the panoramic databases of tunnel interior wall features during the "no-pressure period" and "emptying period" to construct a fused feature database; S6. The tunnel is filled with water until it is full, and the tunnel changes from an unpressurized state to a pressurized state, transitioning from the maintenance period to the operation preparation period; afterwards, the tunnel transitions from the operation preparation period to the water-filled operation period, and then to the next maintenance period. S7. Obtain image data of the tunnel inner wall at sampling points during the "pressurized period" using a third image acquisition device; S8. Based on the fusion feature database, locate the sampling points inside the tunnel during the "pressurized period" and the location of the third image acquisition device.

[0007] The process of opening a drainage system inside the tunnel includes the following steps: S1.1, The water-blocking gate at the beginning of the tunnel is used to block the upstream water. S1.2. Open the drainage facilities inside the tunnel. Under the action of its own weight, the water inside the tunnel is discharged along the opening of the drainage facilities. The water level inside the tunnel gradually decreases and enters the maintenance period.

[0008] The process of acquiring panoramic image data of the tunnel interior wall during the "pressure-free period" using a first image acquisition device includes the following steps: S2.1 When the water level inside the tunnel drops to the first water level line, the drainage facilities inside the tunnel are closed and drainage is stopped. The water level inside the tunnel remains stable and enters the first pressure stabilization period. S2.2. A first image acquisition device is deployed at the tunnel entrance. During the first pressure stabilization period, the first image acquisition device collects panoramic image data of the tunnel wall above the first water level and marks the image location. S2.3. Open the drainage facilities inside the tunnel and continue to drain water. When the water level inside the tunnel drops to the second water level line, close the drainage facilities inside the tunnel and stop draining water. The water level inside the tunnel will stabilize again and enter the second pressure stabilization period. S2.4 During the second pressure stabilization period, the first image acquisition device collects panoramic image data of the tunnel inner wall in the area between the second water level line and the first water level line, and marks the image position; S2.5, repeat S2.3 and S2.4 until panoramic image data of the tunnel interior wall during the "unpressurized period" of the entire tunnel length from the tunnel head to the tunnel tail is obtained.

[0009] After the water inside the tunnel is drained, panoramic image data of the tunnel's inner wall during the "drainage period" is acquired using a second image acquisition device, including the following steps: S3.1 After the water inside the tunnel is drained, it is left to air dry naturally, entering the drying period of the inner wall; S3.2 Open the tunnel end entrance door, deploy the second image acquisition device through the tunnel end entrance, collect image data of the bottom of the inner wall of the tunnel from the tunnel end to the middle of the tunnel, and mark the image position; S3.3. Deploy a second image acquisition device at the tunnel entrance to collect image data of the bottom of the inner wall of the tunnel from the tunnel entrance to the middle of the tunnel, mark the image positions, and ensure that it is combined with the image data of the bottom of the inner wall of the tunnel from the tunnel end to the middle of the tunnel in S3.2 to cover the tunnel inner wall bottom image data during the "drainage period" of the entire tunnel length from the tunnel entrance to the tunnel end. S3.4. Similarly, obtain the top, left, and right image data of the tunnel inner wall during the "drainage period" of the entire tunnel length from the tunnel head to the tunnel tail, and ensure that it is combined with the tunnel inner wall bottom image data in S3.3 to cover the panoramic image data of the tunnel inner wall during the "drainage period" of the entire tunnel length from the tunnel head to the tunnel tail.

[0010] Extracting texture data of the tunnel interior walls and constructing a panoramic database of tunnel interior wall features includes the following steps: S4.1 Obtain relevant parameters for image data acquisition during the "pressure-free period" and "emptying period" respectively. The relevant parameters include the moving speed of the image acquisition device and the video sampling frequency. S4.2 Extract key frame images from the video stream based on the moving speed of the first image acquisition device and the video sampling frequency of the second image acquisition device respectively. The key frame images cover the entire tunnel length from the tunnel head to the tunnel tail and the entire tunnel cross section. S4.3 Based on the keyframe images of the "no-pressure period" and the "emptying period", construct panoramic images of the "no-pressure period" and the "emptying period", extract the edge information of the images, obtain the complete feature texture of the entire tunnel inner wall, and form a panoramic database of tunnel inner wall features during the "no-pressure period" and the "emptying period".

[0011] The panoramic databases of tunnel interior wall features during the "unpressurized period" and "emptying period" are merged, including the following steps: S5.1 Standardize the panoramic data of tunnel interior wall features during the "no-pressure period" and "emptying period"; S5.2 Bind the characteristic data of "pressure-free period" and "emptying period" to the same physical coordinate system; S5.3 Feature layer fusion, retaining feature textures, including water film textures during the pressureless period and drying textures during the drainage period; S5.4 Remove noise and invalid information from the fused data.

[0012] The edge information of the extracted image is performed using the Canny operator. In the frame image after edge extraction, the pixel value is calculated pixel by pixel to count the percentage of edge information in the whole image. By adjusting the threshold of the Canny operator, the percentage of pixels that reflect edge information in each edge image is within the threshold range.

[0013] Image data of the tunnel inner wall at sampling points during the "pressurized period" is acquired using a third image acquisition device, including the following steps: S7.1 Deploy a third image acquisition device at the tunnel entrance at the first end; S7.2. Remotely control the third image acquisition device to the sampling point to acquire the video stream.

[0014] Based on a panoramic database of tunnel wall features, the location of sampling points within the tunnel during the "pressurized period" and the position of the third image acquisition device are determined, including the following steps: S8.1 Compare the video stream sampled in real time along the tunnel by the third image acquisition device with the fusion feature database. Combine the moving speed and direction of the third image acquisition device with the distance and direction between it and the sampling point. When one frame in the video stream successfully matches the tunnel inner wall feature texture in the fusion feature database, record the current sampling point position, which is the non-defect area that has not changed. S8.2 When a defect occurs inside the tunnel, feature texture matching is performed on the area surrounding the defect using the steps in S8.1. The location of the sampling points in the surrounding area is recorded, and then the location of the defect area is calculated. S8.3. By converting the obtained locations of surrounding sampling points with the distance and direction of the third image acquisition device, the location of the third image acquisition device inside the tunnel is determined.

[0015] The steps for comparing the real-time sampled video stream with the tunnel interior wall feature texture data in the fused feature database include: 1) Decompose the real-time video stream into frame images; 2) Convert the frame image to a grayscale image; 3) Extract the edge information of each grayscale image and compare it with the tunnel inner wall feature texture data in the fusion feature database.

[0016] Compared with the prior art, the present invention has the following beneficial effects: This invention divides long-distance pressurized tunnels into three periods based on their operational conditions: the "unpressurized period" where water flows freely across the tunnel surface and the inner walls are moist; the "drainage period" where there is no water flow and the inner walls are dry; and the "pressurized period" where water flows from the tunnel's head to its tail, filling the entire tunnel length. Utilizing valuable data collected during annual drainage maintenance, the invention identifies different texture characteristics of the inner walls at different times due to uneven water absorption. It also considers the differences in texture characteristics of concrete structures in different parts of the tunnel, such as pores, cracks, erosion, damage, aging, and deposits, as well as the correlation between texture characteristics at the same location at different times. Based on a fusion feature database constructed from the "unpressurized period" and "drainage period," a "fingerprint recognition" method is used to associate the texture characteristics of the inner walls with specific locations. Ultimately, this achieves comprehensive, accurate, and rapid underwater positioning of both non-defective and defective areas within the "pressurized period" of long-distance pressurized tunnels. Attached Figure Description

[0017] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the positioning method of the present invention.

[0019] Figure 2 This is a schematic diagram of the water level status inside the tunnel during the "pressure-free period" of this invention. Detailed Implementation

[0020] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] The tunnel's inner wall is made of concrete, which contains sand particles that are not evenly distributed. When the tunnel is dry, the inner wall appears the same, but when it is wet, the water absorption varies greatly due to the different pore sizes. By using a method similar to "fingerprint recognition," the image of the tunnel's inner wall can be identified, and its location within the tunnel can be determined based on the differences in the image.

[0022] Based on the working conditions of the tunnel, the tunnel is divided into three periods: the "unpressurized period" in which water flows freely on the surface and the inner wall is moist; the "drainage period" in which there is no water flow and the inner wall is dry; and the "pressurized period" in which water flows throughout the entire length of the tunnel from the beginning to the end.

[0023] See Figure 1 This embodiment discloses a long-distance pressurized tunnel positioning method based on inner wall feature texture-assisted identification, including the following steps: S1. Drainage is started inside the tunnel, and the tunnel changes from a pressurized state to a depressurized state, entering the maintenance period. S2. Obtain panoramic image data of the tunnel inner wall during the "pressure-free period" using the first image acquisition device; S3. After the water in the tunnel is drained, panoramic image data of the tunnel wall during the "drainage period" is obtained through the second image acquisition device. S4. Extract the feature texture data of the tunnel inner wall and construct a panoramic database of tunnel inner wall features during the "no-pressure period" and "emptying period"; S5. Merge the panoramic databases of tunnel interior wall features during the "no-pressure period" and "emptying period" to construct a fused feature database; S6. The tunnel is filled with water until it is full, and the tunnel changes from an unpressurized state to a pressurized state, transitioning from the maintenance period to the operation preparation period; afterwards, the tunnel transitions from the operation preparation period to the water-filled operation period, and then to the next maintenance period. S7. Obtain image data of the tunnel inner wall at sampling points during the "pressurized period" using a third image acquisition device; S8. Based on the fusion feature database, locate the sampling points inside the tunnel during the "pressurized period" and the location of the third image acquisition device.

[0024] This invention is based on the fact that the inner wall of a tunnel exhibits different texture characteristics at different times due to the uneven water absorption of the inner wall material. It adopts a "fingerprint recognition" method to associate the texture characteristics of the inner wall with specific locations, and finally achieves positioning within the tunnel.

[0025] In this invention, the first image acquisition device can be a remotely controlled camera boat; the second image acquisition device can be a vehicle-mounted image acquisition device or a drone; and the third image acquisition device can be an underwater image acquisition device, such as a submersible.

[0026] See Figure 2 In S1, the process of opening the drainage system inside the tunnel includes the following steps: S1.1, The water-blocking gate at the beginning of the tunnel is used to block the upstream water. S1.2. Open the drainage facilities inside the tunnel. Under the action of its own weight, the water inside the tunnel is discharged along the opening of the drainage facilities. The water level inside the tunnel gradually decreases and enters the maintenance period.

[0027] In S2, panoramic image data of the tunnel interior wall during the "pressure-free period" is acquired using a first image acquisition device, including the following steps: S2.1 When the water level inside the tunnel drops to the first water level line, the drainage facilities inside the tunnel are closed and drainage is stopped. The water level inside the tunnel remains stable and enters the first pressure stabilization period. S2.2. Deploy a first image acquisition device through the tunnel entrance (such as the water inlet vent). During the first pressure stabilization period, the first image acquisition device collects panoramic image data of the tunnel inner wall in the area above the first water level and marks the image position. S2.3. Open the drainage facilities inside the tunnel and continue to drain water. When the water level inside the tunnel drops to the second water level line, close the drainage facilities inside the tunnel and stop draining water. The water level inside the tunnel will stabilize again and enter the second pressure stabilization period. S2.4 During the second pressure stabilization period, the first image acquisition device collects panoramic image data of the tunnel inner wall in the area between the second water level line and the first water level line, and marks the image position; S2.5, repeat S2.3 and S2.4 until panoramic image data of the tunnel interior wall during the "unpressurized period" of the entire tunnel length from the tunnel head to the tunnel tail is obtained.

[0028] By employing a cyclical pattern of lowering water level, stabilizing pressure, acquiring data, and then lowering water level again, images of the inner wall between different water levels are collected in different areas. This avoids image blurring caused by unstable water levels, such as reflections from water surface fluctuations. The primary image acquisition device can be a remotely controlled camera boat that can move in the residual water during the depressurization period, eliminating the need for manual entry into the depths of the tunnel, reducing personnel safety risks, and allowing for flexible adjustment of the acquisition angle to obtain panoramic data.

[0029] In S3, after the water inside the tunnel is drained, panoramic image data of the tunnel wall during the "drainage period" is acquired using a second image acquisition device, including the following steps: S3.1 After the water inside the tunnel is drained, it is left to air dry naturally, entering the drying period of the inner wall; S3.2 Open the tunnel end entrance access door (such as the volute access door), deploy the second image acquisition device through the tunnel end entrance, collect image data of the bottom of the inner wall of the tunnel from the tunnel end to the middle of the tunnel, and mark the image position; S3.3. Deploy a second image acquisition device at the tunnel entrance to collect image data of the bottom of the inner wall of the tunnel from the tunnel entrance to the middle of the tunnel, mark the image positions, and ensure that it is combined with the image data of the bottom of the inner wall of the tunnel from the tunnel end to the middle of the tunnel in S3.2 to cover the tunnel inner wall bottom image data during the "drainage period" of the entire tunnel length from the tunnel entrance to the tunnel end. S3.4. Similarly, obtain the top, left, and right image data of the tunnel inner wall during the "drainage period" of the entire tunnel length from the tunnel head to the tunnel tail, and ensure that it is combined with the tunnel inner wall bottom image data in S3.3 to cover the panoramic image data of the tunnel inner wall during the "drainage period" of the entire tunnel length from the tunnel head to the tunnel tail.

[0030] During the emptying period, the tunnel interior walls are dry, exposing the bottom area covered by water during the pressureless period and details obscured by the water film, such as micro-cracks and pinholes. This creates a wet-dry contrast with the pressureless period data, improving the integrity of feature textures. Image data of the tunnel interior walls' bottom, top, left, and right sides during the "emptying period" are acquired for the entire tunnel length from the tunnel's head to its tail, ensuring full cross-section coverage without blind spots. Combined verification is used to avoid splicing errors.

[0031] In S4, the feature texture data of the tunnel inner wall is extracted to construct a panoramic database of tunnel inner wall features, including the following steps: S4.1 Obtain relevant parameters for image data acquisition during the "pressure-free period" and "emptying period" respectively. The relevant parameters include the moving speed of the image acquisition device and the video sampling frequency. S4.2 Extract key frame images from the video stream based on the moving speed of the first image acquisition device and the video sampling frequency of the second image acquisition device respectively. The key frame images cover the entire tunnel length from the tunnel head to the tunnel tail and the entire tunnel cross section. S4.3 Based on the keyframe images of the "no-pressure period" and the "emptying period", construct panoramic images of the "no-pressure period" and the "emptying period", extract the edge information of the images, obtain the complete feature texture of the entire tunnel inner wall, and form a panoramic database of tunnel inner wall features during the "no-pressure period" and the "emptying period".

[0032] By combining equipment movement speed and video sampling frequency to calculate keyframes, data redundancy caused by too many frames is avoided while ensuring keyframe coverage of the entire tunnel, balancing data volume and effectiveness. A database is constructed categorized into "no-pressure period" and "emptying period" to facilitate hierarchical processing during subsequent merging, and features are bound to location.

[0033] In S5, the panoramic databases of tunnel interior wall features during the "no-pressure period" and "emptying period" are merged, including the following steps: S5.1 Standardize the panoramic data of tunnel interior wall features during the "no-pressure period" and "emptying period".

[0034] Eliminate the underlying differences between the "no-pressure period" and the "empty period" databases to provide standardized data formats and feature dimensions for subsequent merging, and avoid merging failures caused by format confusion.

[0035] S5.2 Bind the characteristic data of "pressure-free period" and "emptying period" to the same physical coordinate system; S5.3 Feature layer fusion, retaining feature textures, including water film textures during the pressureless period and drying textures during the drainage period; S5.4 Remove noise and invalid information from the fused data.

[0036] By unifying data formats and feature dimensions, feature incompatibility caused by different acquisition devices and periods is avoided, ensuring a consistent merging basis. Features are bound to the same physical coordinate system to resolve misalignment issues of different features at the same location, providing a unified benchmark for subsequent positioning. Water film textures during the depressurization period (reflecting differences in material water absorption) and crack features during the drainage period (key information on structural defects) are selectively retained to achieve feature complementarity. Redundant and abnormal data (such as blurred images and discontinuous features) are removed to reduce the computational burden of subsequent matching, improving the database's "lightweight" and "reliability."

[0037] Specifically, the edge information of the image is extracted using the Canny operator. In the frame image after edge extraction, the pixel value is calculated pixel by pixel to count the percentage of edge information in the whole image. By adjusting the threshold of the Canny operator, the percentage of pixels that reflect edge information in each edge image is within the threshold range.

[0038] In S6, filling the tunnel with water involves the following steps: S6.1. Clean up all maintenance materials inside the tunnel to ensure that there are no foreign objects inside the tunnel and that it is ready for water filling. S6.2 Open the water filling valve of the water-blocking gate at the beginning of the tunnel, and the tunnel begins to be filled with water; S6.3 The water level inside the tunnel gradually rises until the pressure is equalized upstream and downstream of the water-retaining gate at the tunnel's head. Then, the water-filling valve of the water-retaining gate at the tunnel's head is closed, and the maintenance period transitions into the operational preparation period. First, foreign objects are removed to ensure water-filling conditions. Then, water is slowly filled by opening the water-filling valve, equalizing the pressure, and closing the valve to avoid structural impact or cavitation caused by a sudden rise in water level.

[0039] The transition from the operational preparation period to the water-passing operation period within the tunnel, and then to the next maintenance period, includes the following steps: S6.4 Raise the water-blocking gate at the beginning of the tunnel to ensure that the tunnel interior has the conditions for flow. S6.5. The unit starts up and transitions from the preparation period to the water-flow operation period; S6.6 The unit is shut down, transitioning from the water-passing operation period to the next maintenance period.

[0040] In S7, image data of the tunnel inner wall at sampling points during the "pressurized period" is acquired through a third image acquisition device, including the following steps: S7.1 Deploy a third image acquisition device at the tunnel entrance at the first end; S7.2. Remotely control the third image acquisition device to the sampling point to acquire the video stream.

[0041] Using an underwater submersible as a third image acquisition device allows for stable operation under pressurized water conditions, overcoming the limitation that close-range acquisition of inner wall images is impossible during pressurized periods.

[0042] In S8, based on the panoramic database of tunnel wall features, the location of sampling points within the tunnel during the "pressurized period" and the position of the third image acquisition device are determined, including the following steps: S8.1 Compare the video stream sampled in real time along the tunnel by the third image acquisition device with the fusion feature database. Combine the moving speed and direction of the third image acquisition device with the distance and direction between it and the sampling point. When one frame in the video stream successfully matches the tunnel inner wall feature texture in the fusion feature database, record the current sampling point position, which is the non-defect area that has not changed. S8.2 When a defect occurs inside the tunnel, feature texture matching is performed on the area surrounding the defect using the steps in S8.1. The location of the sampling points in the surrounding area is recorded, and then the location of the defect area is calculated. S8.3. By converting the obtained locations of surrounding sampling points with the distance and direction of the third image acquisition device, the location of the third image acquisition device inside the tunnel is determined.

[0043] The core idea of ​​matching real-time video streams with a fusion database, combined with device motion parameters (speed and direction), is applicable to both direct matching of non-defect areas and location of defects by matching around the defect. When a defect area cannot be directly matched, the device position is calculated by using known positions, distances, and directions in the surrounding area, thus solving the localization problem in scenarios with missing features.

[0044] Specifically, the steps for comparing the real-time sampled video stream with the tunnel interior wall feature texture data in the fused feature database include: 1) Decompose the real-time video stream into frame images; 2) Convert the frame image to a grayscale image; 3) Extract the edge information of each grayscale image and compare it with the tunnel inner wall feature texture data in the fusion feature database.

[0045] Of course, the present invention may have many other embodiments. Without departing from the spirit and essence of the present invention, those skilled in the art can make various corresponding changes and modifications according to the present invention, but these corresponding changes and modifications should all fall within the protection scope of the appended claims.

Claims

1. A long-distance pressurized tunnel positioning method based on inner wall feature texture auxiliary recognition, characterized in that, The method comprises the following steps: S1, opening drainage in the tunnel, the tunnel is changed from a pressure state to a non-pressure state, and enters a maintenance period; S2, acquiring panoramic image data of the inner wall of the tunnel in the "non-pressure period" by a first image acquisition device; S3, after the water flow in the tunnel is emptied, acquiring panoramic image data of the inner wall of the tunnel in the "emptying period" by a second image acquisition device; S4, extracting feature texture data of the inner wall of the tunnel, and constructing a "non-pressure period" and "emptying period" inner wall feature panoramic database of the tunnel; S5, merging the "non-pressure period" and "emptying period" inner wall feature panoramic database of the tunnel, and constructing a fusion feature database; S6, filling water in the tunnel to full water, the tunnel is changed from a non-pressure state to a pressure state, from a maintenance period to a running preparation period; then the tunnel is changed from a running preparation period to a water running period, and enters a next maintenance period; S7, acquiring image data of the inner wall of the tunnel at a sampling point in the "pressure period" by a third image acquisition device; S8, based on the fusion feature database, positioning the position of the sampling point in the "pressure period" and the third image acquisition device in the tunnel.

2. The long-range pressurized tunnel positioning method based on the inner wall feature texture auxiliary identification according to claim 1, characterized in that, The step of opening drainage in the tunnel comprises the following steps: S1.1, closing a water blocking gate at the head of the tunnel to block the upstream water; S1.2, opening drainage facilities in the tunnel, the water flow in the tunnel is discharged along the drainage facility orifice under the action of gravity, the water level in the tunnel is gradually lowered, and the tunnel enters a maintenance period.

3. The long-range pressurized tunnel positioning method based on the inner wall feature texture auxiliary identification according to claim 1, characterized in that, The step of acquiring panoramic image data of the inner wall of the tunnel in the "non-pressure period" by the first image acquisition device comprises the following steps: S2.1, when the water level in the tunnel is lowered to a first water level line, the drainage facilities in the tunnel are closed, the drainage is stopped, the water level in the tunnel is kept stable, and a first stable pressure period is entered; S2.2, deploying the first image acquisition device through the entrance at the head of the tunnel, during the first stable pressure period, the first image acquisition device collects panoramic image data of the inner wall of the tunnel in the region above the first water level line, and marks the image position; S2.3, opening the drainage facilities in the tunnel, continuing to drain, when the water level in the tunnel is lowered to a second water level line, the drainage facilities in the tunnel are closed, the drainage is stopped, the water level in the tunnel is kept stable again, and a second stable pressure period is entered; S2.4, during the second stable pressure period, the first image acquisition device collects panoramic image data of the inner wall of the tunnel in the region between the second water level line and the first water level line, and marks the image position; S2.5, repeating S2.3 and S2.4 until the "non-pressure period" panoramic image data of the inner wall of the tunnel from the head of the tunnel to the end of the tunnel is acquired.

4. The long-range pressurized tunnel positioning method based on inner wall feature texture aided recognition according to claim 1, characterized in that, The step of acquiring panoramic image data of the inner wall of the tunnel in the "emptying period" by the second image acquisition device after the water flow in the tunnel is emptied comprises the following steps: S3.1, after the water flow in the tunnel is emptied, waiting for natural drying, and entering an inner wall drying period; S3.2, opening an entrance door at the end of the tunnel, deploying the second image acquisition device through the entrance at the end of the tunnel, collecting image data of the inner wall bottom from the end of the tunnel to the middle of the tunnel, and marking the image position; S3.3, placing the second image acquisition device through the entrance of the tunnel head, collecting image data of the inner wall bottom from the tunnel head to the middle of the tunnel, marking the image position, and ensuring that the image data of the inner wall bottom from the tunnel tail to the middle of the tunnel in S3.2 is combined to cover the "emptying period" image data of the inner wall bottom of the entire tunnel length from the tunnel head to the tunnel tail; S3.4, in the same way, acquiring the "emptying period" image data of the inner wall top, left side and right side of the entire tunnel length from the tunnel head to the tunnel tail, and ensuring that the image data of the inner wall top, left side and right side of the entire tunnel length from the tunnel head to the tunnel tail is combined with the image data of the inner wall bottom in S3.3 to cover the "emptying period" panoramic image data of the entire tunnel length from the tunnel head to the tunnel tail.

5. The long-range pressurized tunnel positioning method based on inner wall feature texture aided recognition according to claim 1, characterized in that, Extracting the feature texture data of the inner wall of the tunnel, constructing the feature panoramic database of the inner wall of the tunnel, including the following steps: S4.1, acquiring the relevant parameters during the acquisition of "non-pressure period" and "emptying period" image data respectively, the relevant parameters including the moving speed of the image acquisition device and the video sampling frequency; S4.2, extracting the key frame images in the video stream according to the moving speed of the first image acquisition device and the second image acquisition device and the video sampling frequency respectively, the key frame images covering the entire tunnel length from the tunnel head to the tunnel tail and the entire tunnel section; S4.3, constructing the "non-pressure period" and "emptying period" panoramic images according to the "non-pressure period" and "emptying period" key frame images respectively, extracting the edge information of the images, obtaining the complete feature texture of the entire tunnel inner wall, and forming the "non-pressure period" and "emptying period" inner wall feature panoramic database of the tunnel.

6. The long-range pressurized tunnel positioning method based on inner wall feature texture aided recognition according to claim 1, characterized in that, Merging the "non-pressure period" and "emptying period" inner wall feature panoramic database of the tunnel, including the following steps: S5.1, standardizing the "non-pressure period" and "emptying period" inner wall feature panoramic data; S5.2, unifying the feature data of the "non-pressure period" and "emptying period" to the same physical coordinate system; S5.3, feature layer fusion, retaining feature texture, feature texture including water film texture of non-pressure period and dry texture of emptying period; S5.4, eliminating noise and invalid information after fusion.

7. The long-range pressurized tunnel positioning method based on inner wall feature texture aided recognition according to claim 5, characterized in that, The edge information of the image is extracted by Canny operator, and the percentage of edge information in the whole image is calculated by calculating the pixel value of each pixel in the frame image after edge extraction. By adjusting the threshold value of Canny operator, the percentage of pixel points representing edge information in the whole image in each edge image is within the threshold interval.

8. The long-range pressurized tunnel positioning method based on inner wall feature texture aided recognition of claim 1, wherein, Acquiring image data of the inner wall of the tunnel at the "pressure period" sampling point by the third image acquisition device, including the following steps: S7.1, placing the third image acquisition device at the entrance of the tunnel head; S7.2, remotely controlling the third image acquisition device to the sampling point to acquire the video stream.

9. The long-range pressurized tunnel positioning method based on inner wall feature texture aided recognition according to claim 1, characterized in that, Based on the fused feature database, positioning the position of the "pressure period" sampling point and the third image acquisition device in the tunnel, including the following steps: S8.1, compare the video stream of the third image acquisition device real-time sampling along the tunnel with the fusion feature database, combined with the moving speed, direction of the third image acquisition device and the distance, direction between the sampling point, when one frame in the video stream matches the tunnel wall feature texture in the fusion feature database successfully, record the current sampling point position, which is the non-defect area without change; S8.2, when the defect appears in the tunnel, the feature texture matching in step S8.1 is applied to the surrounding area of the defect to record the sampling point position of the surrounding area, and then the positioning of the defect area is calculated; S8.3, the distance and direction of the third image acquisition device are converted to determine the position of the third image acquisition device in the tunnel.

10. The long-range pressurized tunnel positioning method based on inner wall feature texture aided recognition according to claim 9, characterized in that, The steps of comparing the real-time sampling video stream with the tunnel wall feature texture data in the fusion feature database include: 1) decompose the real-time shooting video stream into frame images; 2) convert the frame image to a gray space image; 3) extract the edge information of each gray space image and compare it with the tunnel wall feature texture data in the fusion feature database.

Images