Expressway tunnel inspection method and system

Abstract

The invention relates to the technical field of image recognition, and discloses a highway tunnel inspection method and system, and the method comprises the steps: shooting an image of the inner wall of a tunnel when an inspection vehicle runs along the tunnel; extracting key points from the image of the inner wall of the tunnel, and calculating a transformation matrix according to the key points; performing perspective transformation on the image of the inner wall of the tunnel by using the transformation matrix; inputting the image of the inner wall of the tunnel after perspective transformation into a preset neural network model, and identifying a position with a damage condition from the image of the inner wall of the tunnel by the neural network model; analyzing the level of the damage condition according to a preset strategy; and when the level of the damage condition is higher than a preset level, alarm information is sent to prompt to take measures to deal with the damage condition. According to the method, perspective transformation is carried out on the image to eliminate image distortion caused by the tunnel top curved surface, the damage condition of the tunnel inner wall is identified by using the deep learning technology, and the damage condition of the tunnel inner wall can be efficiently and accurately identified in a complex and changeable tunnel environment.

Description

Technical Field

[0001] The present invention relates to the field of image recognition technology, and more specifically, to a highway tunnel inspection method and system. Background Art

[0002] Highway tunnels are an important part of the modern transportation system. Their safety and reliability directly affect driving safety and road traffic. However, the internal environment of the tunnel is complex, especially the top of the tunnel is usually curved, the lighting conditions are poor, and it is easy to have cracks, peeling and other damage. Traditional tunnel inspection methods mainly rely on manual visual inspection, which is not only time-consuming and labor-intensive, but also greatly affected by human factors, making it difficult to ensure the accuracy and comprehensiveness of the inspection.

[0003] In recent years, with the development of computer vision and artificial intelligence technology, tunnel inspection methods based on image processing and machine learning have gradually attracted attention. These methods use high-definition cameras installed on inspection vehicles to capture images of the inner wall of the tunnel, and use image processing technology to analyze the images to identify damage to the inner wall of the tunnel. However, when processing images of the curved surface at the top of the tunnel, existing methods often have problems with image distortion and uneven lighting, resulting in low recognition accuracy.

[0004] In view of the problems existing in the prior art, the present invention proposes a highway tunnel inspection method and system, which aims to accurately identify the damage of the inner wall of the tunnel and promptly prompt measures to deal with the damage. Summary of the invention

[0005] In order to solve the above-mentioned technical problems, the present application is proposed to provide a highway tunnel inspection method and system, which aims to accurately identify the damage of the inner wall of the tunnel and promptly prompt measures to deal with the damage.

[0006] In a first aspect, the present invention provides a highway tunnel inspection method, comprising: using at least one high-definition camera installed on an inspection vehicle to capture an image of the inner wall of the tunnel while the inspection vehicle is traveling along the tunnel; extracting key points from the image of the inner wall of the tunnel, and calculating a transformation matrix based on the key points; using the transformation matrix to perform a perspective transformation on the image of the inner wall of the tunnel; inputting the image of the inner wall of the tunnel after the perspective transformation into a preset neural network model, and allowing the neural network model to identify locations where damage exists from the image of the inner wall of the tunnel; analyzing the level of the damage according to a preset strategy; and when the level of the damage is higher than a preset level, issuing an alarm message to prompt measures to be taken to deal with the damage.

[0007] Optionally, the above-mentioned highway tunnel inspection method extracts key points from the image of the tunnel inner wall and calculates the transformation matrix according to the key points, including: setting the transformation matrix ; Obtain a preset original image and a target image corresponding to the original image; Take a first set of feature points from the original image, the coordinates of the i-th feature point in the first set of feature points are , take a second set of feature points from the target image, the coordinates of the i-th feature point in the second set of feature points are ; Construct a linear system of equations: , ; Solve the transformation matrix that makes the linear equations valid .

[0008] Optionally, the above-mentioned highway tunnel inspection method, taking a first set of feature points from the original image, comprises: constructing a Gaussian pyramid, the bottom layer of the Gaussian pyramid is the original image, and any layer of the Gaussian pyramid except the bottom layer is a low-pass filtered version of the next layer; calculating the invariance score of each point in the original image according to the Gaussian pyramid ,in, is the number of layers of the Gaussian pyramid, , For each point in the original image, Layer, layer's responsiveness; and selecting the first set of feature points from the original image according to the invariance score of each point in the original image.

[0009] Optionally, the aforementioned highway tunnel inspection method, before selecting the first group of feature points from the original image according to the invariance score of each point in the original image, further includes: constructing a rotated image sequence, wherein the first image of the rotated image sequence is the original image, and the kth image of the rotated image sequence is the image after the original image is rotated by k unit angles; and calculating the invariance score correction value of each point in the original image according to the rotated image sequence. ,in, is the number of images in the rotated image sequence, , For each point in the original image, , The responsiveness in the image, is the unit angle; and according to the invariance score correction value of each point in the original image, the invariance score of each point in the original image is corrected.

[0010] Optionally, the aforementioned highway tunnel inspection method further includes: using the transformation matrix Perform a perspective transformation test; analyze the error value of the perspective transformation test result; when the error value of the perspective transformation test result exceeds a preset threshold, re-take the first set of feature points from the original image and the second set of feature points from the target image to recalculate the transformation matrix When the error value of the perspective transformation test result does not exceed a preset threshold, the perspective transformation of the image of the inner wall of the tunnel is performed using the transformation matrix.

[0011] Optionally, the aforementioned highway tunnel inspection method, before inputting the image of the tunnel inner wall after perspective transformation into a preset neural network model, further comprises: detecting the position of each point in the image of the tunnel inner wall before and after perspective transformation; correcting the pixel value of each point in the image of the tunnel inner wall, wherein any point in the image of the tunnel inner wall is taken as the target point, and its positions before and after perspective transformation are respectively and , then the corrected pixel value of the target point is ,in, is the position of the target point before and after the perspective transformation and The distance between is the average pixel value of multiple pixel points centered at the target point before perspective transformation.

[0012] Optionally, the aforementioned highway tunnel inspection method, before inputting the image of the tunnel inner wall after perspective transformation into a preset neural network model, the method further includes: calculating the brightness value of each point in the image of the tunnel inner wall; calculating the brightness distribution of the image of the tunnel inner wall according to the brightness value of each point in the image of the tunnel inner wall; extracting a dark area from the image of the tunnel inner wall according to the brightness distribution of the image of the tunnel inner wall; calculating a dark enhancement factor ,in, is the preset brightness threshold, is a preset enhancement factor; the brightness of each point in the dark area is enhanced using the dark area enhancement factor.

[0013] Optionally, the aforementioned highway tunnel inspection method further includes: when the level of the damage is higher than a preset level, querying the information of vehicles on the road where the tunnel is located; based on the information of the vehicle, querying the information of the mobile terminal of the driver of the vehicle; and sending the alarm information to the mobile terminal of the driver of the vehicle.

[0014] In a second aspect, the present invention provides a highway tunnel inspection system, comprising: a shooting module, which uses at least one high-definition camera installed on an inspection vehicle to capture an image of the inner wall of the tunnel while the inspection vehicle is traveling along the tunnel; a matrix calculation module, which extracts key points from the image of the inner wall of the tunnel and calculates a transformation matrix based on the key points; a perspective transformation module, which uses the transformation matrix to perform a perspective transformation on the image of the inner wall of the tunnel; a damage identification module, which inputs the image of the inner wall of the tunnel after the perspective transformation into a preset neural network model, and the neural network model identifies the location of the damage from the image of the inner wall of the tunnel; a damage analysis module, which analyzes the level of the damage according to a preset strategy; and an alarm processing module, which, when the level of the damage is higher than a preset level, issues an alarm message to prompt measures to deal with the damage.

[0015] The above one or more technical solutions of the present invention have at least one or more of the following beneficial effects:

[0016] According to the technical solution of the present invention, the image is perspective transformed to eliminate the image distortion caused by the curved surface of the tunnel top, and the damage to the inner wall of the tunnel is identified using deep learning technology. This can efficiently and accurately identify the damage to the inner wall of the tunnel in a complex and changeable tunnel environment, provide strong support for the maintenance and management of the tunnel, and ensure the safety and normal operation of the tunnel. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0018] Figure 1 is a flow chart of a highway tunnel inspection method according to an embodiment of the present application;

[0019] Figure 2 is a partial flow chart of a highway tunnel inspection method according to an embodiment of the present application;

[0020] Figure 3 is another layout flow chart of the highway tunnel inspection method according to an embodiment of the present application;

[0021] Figure 4 is another partial flow chart of the highway tunnel inspection method according to an embodiment of the present application;

[0022] Figure 5 4 is a block diagram of a highway tunnel inspection system according to an embodiment of the present application. DETAILED DESCRIPTION

[0023] Some embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0024] like Figure 1 As shown, in one embodiment of the present invention, a highway tunnel inspection method is provided, comprising:

[0025] Step S110, using at least one high-definition camera installed on the inspection vehicle to capture images of the inner wall of the tunnel while the inspection vehicle is traveling along the tunnel.

[0026] Step S120, extracting key points from the image of the inner wall of the tunnel, and calculating a transformation matrix based on the key points.

[0027] Step S130: Perform perspective transformation on the image of the inner wall of the tunnel using a transformation matrix.

[0028] In this embodiment, by extracting key points from the image of the inner wall of the tunnel and calculating the perspective transformation matrix, the image is perspective transformed, thereby eliminating the image distortion caused by the curved surface of the tunnel top and improving the geometric accuracy of the image.

[0029] Step S140, inputting the image of the inner wall of the tunnel after perspective transformation into a preset neural network model, and using the neural network model to identify locations where damage exists from the image of the inner wall of the tunnel.

[0030] In this embodiment, the image after perspective transformation and image enhancement is input into a preset neural network model, and the damage of the inner wall of the tunnel is identified using deep learning technology, thereby improving the accuracy and robustness of the identification.

[0031] Step S150: analyzing the damage level according to a preset strategy.

[0032] Step S160: When the level of the damage is higher than a preset level, an alarm message is issued to prompt taking measures to deal with the damage.

[0033] In this embodiment, when the level of the damage is higher than a preset level, an alarm message is immediately issued to prompt relevant personnel to take measures to deal with the damage and ensure the safe operation of the tunnel.

[0034] Specifically, the method may further include: (1) when the level of damage is higher than a preset level, querying the information of vehicles on the road where the tunnel is located; (2) querying the information of the mobile terminal of the driver of the vehicle based on the information of the vehicle; and (3) sending alarm information to the mobile terminal of the driver of the vehicle.

[0035] In this embodiment, when the damage to the inner wall of the tunnel reaches a preset level, the system will immediately query the vehicle information on the road where the tunnel is located and send an alarm message to the mobile terminal of the relevant vehicle driver. After receiving the alarm, the driver can quickly take measures such as slowing down, detouring or parking to avoid potential safety accidents.

[0036] According to the technical solution of this embodiment, the image is perspective transformed to eliminate the image distortion caused by the curved surface of the tunnel top, and the damage to the inner wall of the tunnel is identified using deep learning technology. This can efficiently and accurately identify the damage to the inner wall of the tunnel in a complex and changeable tunnel environment, provide strong support for the maintenance and management of the tunnel, and ensure the safety and normal operation of the tunnel.

[0037] like Figure 2 As shown, another embodiment of the present invention provides a highway tunnel inspection method. Compared with the above-mentioned embodiment, the highway tunnel inspection method of this embodiment, step S120, includes:

[0038] Step S210, set the transformation matrix .

[0039] Step S220, obtaining a preset original image and a target image corresponding to the original image.

[0040] Step S230, taking a first set of feature points from the original image, the coordinates of the i-th feature point in the first set of feature points are , take the second set of feature points from the target image, the coordinates of the i-th feature point in the second set of feature points are .

[0041] In this embodiment, feature points are extracted from the original image and the target image, and a linear equation group is constructed through these feature points, thereby ensuring the accuracy and stability of feature point matching.

[0042] Step S240, constructing a linear equation system: , .

[0043] In this embodiment, the transformation matrix is ​​determined by solving a group of linear equations, thereby ensuring high accuracy of image registration and reducing errors caused by image deformation.

[0044] Step S250, solving the transformation matrix that makes the linear equations valid .

[0045] In this embodiment, the image registration problem is converted into a mathematical problem by setting a transformation matrix and constructing a linear equation group, thereby ensuring the calculation accuracy of the transformation matrix.

[0046] Step S260, using the transformation matrix Perform a perspective transform test.

[0047] Step S270, analyzing the error value of the perspective transformation test result.

[0048] Step S280: When the error value of the perspective transformation test result exceeds the preset threshold, the first set of feature points is taken from the original image and the second set of feature points is taken from the target image to calculate the transformation matrix again. .

[0049] In this embodiment, when the error value of the perspective transformation test result exceeds the preset threshold, the feature points are re-extracted and the transformation matrix is ​​calculated to ensure the high accuracy of the final transformation matrix. Through the feedback mechanism of the error value, the extraction of feature points and the calculation of the transformation matrix are dynamically adjusted, thereby improving the adaptive ability and robustness of the system.

[0050] Step S290, when the error value of the perspective transformation test result does not exceed the preset threshold, executing perspective transformation of the image of the inner wall of the tunnel using the transformation matrix.

[0051] According to the technical solution of this embodiment, during highway tunnel inspection, the high precision and robustness of image registration are ensured by accurately calculating the transformation matrix and dynamically adjusting the feature point extraction, thereby improving the accuracy of identifying damage to the tunnel inner wall, providing a more reliable guarantee for tunnel maintenance and management.

[0052] like Figure 3 As shown, another embodiment of the present invention provides a highway tunnel inspection method. Compared with the above-mentioned embodiment, the highway tunnel inspection method of this embodiment, step S220 includes:

[0053] Step S310: construct a Gaussian pyramid, where the bottom layer of the Gaussian pyramid is the original image, and any layer in the Gaussian pyramid except the bottom layer is a low-pass filtered version of the next layer.

[0054] In this embodiment, a Gaussian pyramid is constructed to perform multi-scale analysis on the original image, and the response change of each point at different scales is calculated, thereby ensuring the stability of the feature points at different scales.

[0055] Step S320, calculating the invariance score of each point in the original image according to the Gaussian pyramid ,in, is the number of layers of the Gaussian pyramid, , For each point in the original image, Layer, The responsiveness of the layer.

[0056] In this embodiment, by calculating the scale invariance score of each point, feature points with smaller changes in responsiveness at different scales are screened out, thereby improving the scale invariance of the feature points.

[0057] Step S330, constructing a rotated image sequence, wherein the first image in the rotated image sequence is the original image, and the kth image in the rotated image sequence is the image after the original image is rotated by k unit angles.

[0058] In this embodiment, by constructing a rotation image sequence, multi-angle analysis is performed on the original image, and the response change of each point at different rotation angles is calculated, thereby ensuring the stability of the feature points at different rotation angles.

[0059] Step S340, calculating the invariance score correction value of each point in the original image according to the rotated image sequence ,in, is the number of images in the rotated image sequence, , For each point in the original image, , The responsiveness in the image, The unit angle is .

[0060] Step S350, correcting the invariance score of each point in the original image according to the invariance score correction value of each point in the original image.

[0061] Step S360: Select a first group of feature points from the original image according to the invariance score of each point in the original image.

[0062] According to the technical solution of this embodiment, by comprehensively considering the scale invariance score and the rotation invariance score correction value, the invariance score of each point in the original image is corrected to ensure the comprehensive stability of the feature point. According to the corrected invariance score, the point with a higher score is selected as the feature point to ensure the accuracy of the selection of the feature point.

[0063] like Figure 4 As shown, another embodiment of the present invention provides a highway tunnel inspection method. Compared with the above-mentioned embodiment, the highway tunnel inspection method of this embodiment, before step S140, includes:

[0064] Step S410, detecting the position of each point in the image of the tunnel inner wall before and after the perspective transformation.

[0065] Step S420, correcting the pixel value of each point in the image of the inner wall of the tunnel, wherein any point in the image of the inner wall of the tunnel is taken as the target point, and its positions before and after the perspective transformation are respectively and , then the corrected pixel value of the target point is ,in, is the position of the target point before and after the perspective transformation and The distance between It is the average pixel value of multiple pixels centered at the target point before perspective transformation.

[0066] In this embodiment, according to the distance The mean If weighted adjustment is performed, the distance The larger the value, the corrected pixel value The closer to the mean , thereby reducing the image distortion caused by perspective transformation.

[0067] According to the technical solution of this embodiment, the accuracy of the position information is ensured by detecting the specific position of each point in the tunnel inner wall image before and after the perspective transformation. The quality and accuracy of the image after the perspective transformation are ensured by correcting the pixel value, thereby improving the recognition accuracy of the tunnel inner wall damage and providing a more reliable guarantee for the maintenance and management of the tunnel.

[0068] Another embodiment of the present invention provides a highway tunnel inspection method. Compared with the above-mentioned embodiment, the highway tunnel inspection method of this embodiment, before step S140, further includes:

[0069] (1) Calculate the brightness value of each point in the image of the inner wall of the tunnel.

[0070] (2) Calculating the brightness distribution of the image of the inner wall of the tunnel according to the brightness value of each point in the image of the inner wall of the tunnel.

[0071] (3) Extracting a dark area from the image of the tunnel inner wall according to the brightness distribution of the image of the tunnel inner wall.

[0072] (4) Calculate the dark enhancement factor ,in, is the preset brightness threshold, is the preset enhancement factor.

[0073] In this embodiment, according to the formula, the enhancement factor gradually increases with the change of the brightness value, which is beneficial to improving the enhancement effect of the dark area.

[0074] (5) Using the dark area enhancement factor to enhance the brightness of each point in the dark area.

[0075] According to the technical solution of this embodiment, the enhancement of the dark area makes it easier to detect damage such as fine cracks and peeling, thereby improving the accuracy of detection.

[0076] like Figure 5 As shown, in one embodiment of the present invention, a highway tunnel inspection system is provided, comprising:

[0077] The shooting module 510 uses at least one high-definition camera installed on the inspection vehicle to capture images of the inner wall of the tunnel while the inspection vehicle is traveling along the tunnel.

[0078] The matrix calculation module 520 extracts key points from the image of the inner wall of the tunnel and calculates the transformation matrix based on the key points.

[0079] The perspective transformation module 530 performs perspective transformation on the image of the inner wall of the tunnel using a transformation matrix.

[0080] In this embodiment, by extracting key points from the image of the inner wall of the tunnel and calculating the perspective transformation matrix, the image is perspective transformed, thereby eliminating the image distortion caused by the curved surface of the tunnel top and improving the geometric accuracy of the image.

[0081] The damage identification module 540 inputs the image of the tunnel inner wall after perspective transformation into a preset neural network model, and the neural network model identifies the location where damage exists from the image of the tunnel inner wall.

[0082] In this embodiment, the image after perspective transformation and image enhancement is input into a preset neural network model, and the damage of the inner wall of the tunnel is identified using deep learning technology, thereby improving the accuracy and robustness of the identification.

[0083] The damage analysis module 550 analyzes the level of damage according to a preset strategy.

[0084] The alarm processing module 560 issues an alarm message to prompt taking measures to deal with the damage situation when the level of the damage situation is higher than a preset level.

[0085] In this embodiment, when the level of the damage is higher than a preset level, an alarm message is immediately issued to prompt relevant personnel to take measures to deal with the damage and ensure the safe operation of the tunnel.

[0086] Specifically, the method may further include: (1) when the level of damage is higher than a preset level, querying the information of vehicles on the road where the tunnel is located; (2) querying the information of the mobile terminal of the driver of the vehicle based on the information of the vehicle; and (3) sending alarm information to the mobile terminal of the driver of the vehicle.

[0087] In this embodiment, when the damage to the inner wall of the tunnel reaches a preset level, the system will immediately query the vehicle information on the road where the tunnel is located and send an alarm message to the mobile terminal of the relevant vehicle driver. After receiving the alarm, the driver can quickly take measures such as slowing down, detouring or parking to avoid potential safety accidents.

[0088] According to the technical solution of this embodiment, the image is perspective transformed to eliminate the image distortion caused by the curved surface of the tunnel top, and the damage to the inner wall of the tunnel is identified using deep learning technology. This can efficiently and accurately identify the damage to the inner wall of the tunnel in a complex and changeable tunnel environment, provide strong support for the maintenance and management of the tunnel, and ensure the safety and normal operation of the tunnel.

[0089] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details disclosed above are only for the purpose of illustration and ease of understanding, not for limitation, and the above details do not limit the present application to being implemented by adopting the above specific details.

[0090] The block diagrams of the devices, apparatuses, equipment, and systems involved in this application are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagram. As will be appreciated by those skilled in the art, these devices, apparatuses, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including", "comprising", "having", etc. are open words, referring to "including but not limited to", and can be used interchangeably with them. The words "or" and "and" used here refer to the words "and / or" and can be used interchangeably with them, unless the context clearly indicates otherwise. The words "such as" used here refer to the phrase "such as but not limited to", and can be used interchangeably with them.

[0091] It should also be noted that in the apparatus, device and method of the present application, each component or each step can be decomposed and / or recombined. Such decomposition and / or recombination should be regarded as equivalent solutions of the present application.

[0092] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the aspects shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

[0093] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. A highway tunnel inspection method, characterized in that: include: Using at least one high-definition camera installed on the inspection vehicle, taking an image of the inner wall of the tunnel while the inspection vehicle is traveling along the tunnel; Extract key points from the image of the inner wall of the tunnel, and calculate a transformation matrix based on the key points; Performing perspective transformation on the image of the inner wall of the tunnel using the transformation matrix; Inputting the image of the inner wall of the tunnel after perspective transformation into a preset neural network model, and allowing the neural network model to identify locations where damage exists from the image of the inner wall of the tunnel; Analyzing the level of the damage according to a preset strategy; When the level of the damage situation is higher than a preset level, an alarm message is issued to prompt taking measures to deal with the damage situation.

2. The highway tunnel inspection method according to claim 1, characterized in that: Extracting key points from the image of the inner wall of the tunnel and calculating a transformation matrix according to the key points include: Let the transformation matrix ; Acquire a preset original image and a target image corresponding to the original image; Take a first set of feature points from the original image, and the coordinates of the i-th feature point in the first set of feature points are , take a second set of feature points from the target image, the coordinates of the i-th feature point in the second set of feature points are ; Construct a system of linear equations: ， ， Solve the transformation matrix that makes the linear equations valid .

3. The highway tunnel inspection method according to claim 2, characterized in that: Taking a first set of feature points from the original image, including: Constructing a Gaussian pyramid, wherein the bottom layer of the Gaussian pyramid is the original image, and any layer except the bottom layer of the Gaussian pyramid is a low-pass filtered version of the next layer; According to the Gaussian pyramid, the invariance score of each point in the original image is calculated. ,in, is the number of layers of the Gaussian pyramid, , For each point in the original image, Layer, The responsiveness of the layer; The first group of feature points is selected from the original image according to the invariance score of each point in the original image.

4. The highway tunnel inspection method according to claim 3, characterized in that: Before selecting the first set of feature points from the original image according to the invariance score of each point in the original image, the method further includes: Constructing a rotation image sequence, wherein the first image of the rotation image sequence is the original image, and the kth image of the rotation image sequence is the image after the original image is rotated by k unit angles; Calculate the invariance score correction value of each point in the original image according to the rotated image sequence ,in, is the number of images in the rotated image sequence, , For each point in the original image, , The responsiveness in the image, is the unit angle; The invariance score of each point in the original image is corrected according to the invariance score correction value of each point in the original image.

5. The highway tunnel inspection method according to claim 2, characterized in that: Also includes: Using the transformation matrix Perform perspective transformation test; Analyzing the error value of the perspective transformation test result; When the error value of the perspective transformation test result exceeds a preset threshold, the first set of feature points is taken from the original image and the second set of feature points is taken from the target image to calculate the transformation matrix again. ; When the error value of the perspective transformation test result does not exceed a preset threshold, the perspective transformation of the image of the inner wall of the tunnel is performed using the transformation matrix.

6. The highway tunnel inspection method according to claim 1, characterized in that: Before inputting the perspective-transformed image of the inner wall of the tunnel into a preset neural network model, the method further includes: Detecting the position of each point in the image of the tunnel inner wall before and after the perspective transformation; The pixel value of each point in the image of the inner wall of the tunnel is corrected, wherein any point in the image of the inner wall of the tunnel is taken as the target point, and the positions of the target point before and after the perspective transformation are respectively and , then the corrected pixel value of the target point is ,in, is the position of the target point before and after the perspective transformation and The distance between is the average pixel value of multiple pixel points centered at the target point before perspective transformation.

7. The highway tunnel inspection method according to claim 1, characterized in that: Before inputting the perspective-transformed image of the inner wall of the tunnel into a preset neural network model, the method further includes: Calculating the brightness value of each point in the image of the inner wall of the tunnel; Calculating the brightness distribution of the image of the inner wall of the tunnel according to the brightness value of each point in the image of the inner wall of the tunnel; Extracting a dark area from the image of the inner wall of the tunnel according to the brightness distribution of the image of the inner wall of the tunnel; Calculate dark enhancement factor ,in, is the preset brightness threshold, is the preset enhancement factor; The brightness of each point in the dark area is enhanced using the dark area enhancement factor.

8. The highway tunnel inspection method according to claim 1, characterized in that: Also includes: When the level of the damage condition is higher than a preset level, querying information of vehicles on the road where the tunnel is located; According to the information of the vehicle, query the information of the mobile terminal of the driver of the vehicle; The alarm information is sent to a mobile terminal of the driver of the vehicle.

9. Highway tunnel inspection system, characterized in that: include: A shooting module uses at least one high-definition camera installed on the inspection vehicle to take images of the inner wall of the tunnel while the inspection vehicle is traveling along the tunnel; A matrix calculation module extracts key points from the image of the inner wall of the tunnel and calculates a transformation matrix based on the key points; A perspective transformation module, using the transformation matrix to perform perspective transformation on the image of the inner wall of the tunnel; A damage identification module, inputting the image of the tunnel inner wall after perspective transformation into a preset neural network model, and allowing the neural network model to identify the location where damage exists from the image of the tunnel inner wall; A damage analysis module, analyzing the level of the damage according to a preset strategy; The alarm processing module sends out an alarm message to prompt taking measures to deal with the damage situation when the level of the damage situation is higher than a preset level.

Images