A tunnel deformation detection method, system, device and storage medium

Abstract

The present invention discloses a tunnel deformation detection method, system, device and storage medium, which relates to the technical field of computer vision processing. The method includes: identifying tunnel targets according to point cloud data and video images; segmenting the tunnel targets to obtain a set of individual targets and storing them in a historical database; wherein the set of individual targets includes several individual targets, and the historical database stores several sets of individual target sets detected at different times; using the left and right delimitation algorithm of the central axis to compare the data of several sets of individual targets in the historical database to obtain the deformation degree of the tunnel targets. The present invention performs segmentation processing on point cloud data, combines historical data for data comparison, and uses the left and right delimitation algorithm of the central axis to determine whether the tunnel has deformed, and at the same time locates the deformed section, which can greatly reduce the workload of track inspection and improve the interval accuracy of deformation detection.

Description

Technical Field

[0001] The present invention relates to the technical field of computer vision processing, and particularly to a tunnel deformation detection method, system, device and storage medium. Background Art

[0002] A tunnel is an engineering structure buried in the ground and is a form of human utilization of underground space. In the operation of railway tunnels, diseases such as water leakage, lining damage, tunnel freezing damage, lining corrosion, extrusion deformation, earthquake damage, fire and air pollution in the tunnel may occur. These diseases have important impacts and threats on the safety, comfort and normal operation of railway tunnels. Therefore, detecting whether a railway tunnel has deformed is directly related to the safety of train operation.

[0003] In order to detect whether a tunnel has deformed, generally a laser scanning device is used to collect the point cloud data of the tunnel. The common methods for collecting tunnel point cloud data include manual handheld scanner collection and total station fixed-point scanning. Among them, manual handheld scanner collection requires a large amount of manpower and material resources and is inefficient, and total station fixed-point scanning requires prior selection of measurement positions and has poor mobility. Therefore, new technical solutions need to be proposed for the disadvantages of collecting tunnel point cloud data by manual handheld scanner collection and total station fixed-point scanning. Summary of the Invention

[0004] The purpose of the present invention is to provide a tunnel deformation detection method, system, device and storage medium to achieve the problem of more convenient and efficient detection of the degree of tunnel deformation.

[0005] To achieve the above purpose, an embodiment of the present invention provides a tunnel deformation detection method, including:

[0006] Identifying tunnel targets according to the point cloud data and video images;

[0007] Segmenting the tunnel targets to obtain a separate target set and storing it in the historical database; wherein the separate target set includes several separate targets, and the historical database stores several groups of separate target sets detected at different times;

[0008] Using the left and right delimiting algorithm of the central axis to perform data comparison on several of the separate target sets in the historical database to obtain the degree of deformation of the tunnel targets.

[0009] Preferably, the identifying tunnel targets according to the point cloud data and video images includes:

[0010] Dividing the point cloud data into blocks to obtain the point cloud data after division;

[0011] Using the view point feature histogram to extract three-dimensional point cloud features from the point cloud data after division to obtain the point cloud features after division;

[0012] Classify and predict the segmented point cloud features, and combine the video images to obtain tunnel targets.

[0013] Preferably, segment the tunnel targets to obtain a set of individual targets and save them in the historical database, including:

[0014] Locate and identify the target image area on the video image;

[0015] Match the video image with the point cloud data in terms of time, and map the target image area to the corresponding area of the point cloud data to obtain the target point cloud area;

[0016] Refine, interpolate or voxelize the target point cloud area to obtain point cloud voxel data;

[0017] Segment the point cloud voxel data at different scales to obtain a set of individual targets and save them in the historical database.

[0018] Preferably, use the left - right boundary determination algorithm based on the central axis to compare the data of several sets of individual targets in the historical database to obtain the deformation degree of the tunnel target, including:

[0019] Register the sets of individual targets detected at different times and perform block processing to obtain a registered data set;

[0020] Use the left - right boundary determination algorithm based on the central axis to calculate the data deviation of the registered data set to obtain the deformation degree of the tunnel target.

[0021] An embodiment of the present invention also provides a tunnel deformation detection system, including:

[0022] A target recognition module, configured to recognize tunnel targets according to point cloud data and video images;

[0023] A target segmentation module, configured to segment the tunnel targets to obtain a set of individual targets and save them in the historical database; wherein the set of individual targets includes several individual targets, and the historical database stores several sets of individual targets detected at different times;

[0024] A data comparison module, configured to use the left - right boundary determination algorithm based on the central axis to compare the data of several sets of individual targets in the historical database to obtain the deformation degree of the tunnel target.

[0025] Preferably, the target recognition module is further configured to:

[0026] Chunk the point cloud data to obtain the chunked point cloud data; use the viewpoint feature histogram to extract 3D point cloud features from the chunked point cloud data to obtain the chunked point cloud features; classify and predict the chunked point cloud features, and combine with the video image to obtain tunnel targets.

[0027] Preferably, the target segmentation module is further configured to:

[0028] Locate and identify the target image area on the video image; match the video image with the point cloud data in terms of time, and map the target image area to the corresponding area of the point cloud data to obtain the target point cloud area; perform thinning, interpolation or voxelization processing on the target point cloud area to obtain the point cloud voxel data; segment the point cloud voxel data into different scales to obtain a separate target set, and store it in the historical database.

[0029] Preferably, the data comparison module is further configured to:

[0030] Register the separate target sets detected at different times, and perform chunking processing to obtain the registered data set; use the left and right boundary determination algorithm of the central axis to calculate the data deviation of the registered data set to obtain the deformation degree of the tunnel target.

[0031] An embodiment of the present invention further provides a computer terminal device, including one or more processors and a memory. The memory is coupled to the processor and is used to store one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the tunnel deformation detection method as described in any of the above embodiments.

[0032] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the tunnel deformation detection method as described in any of the above embodiments is implemented.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] The tunnel deformation detection method disclosed by the present invention includes: identifying tunnel targets according to point cloud data and video images; segmenting the tunnel targets to obtain a set of individual targets and storing them in a historical database, where the set of individual targets includes several individual targets, and the historical database stores several sets of individual targets detected at different times; using the left and right boundary determination algorithm based on the central axis to compare the data of several sets of individual targets in the historical database to obtain the deformation degree of the tunnel targets. The present invention performs segmentation processing on the point cloud data, combines historical data for data comparison, and can determine whether the tunnel has deformed by using the left and right boundary determination algorithm based on the central axis. At the same time, it locates the deformed section, which can greatly reduce the workload of track inspection and improve the interval accuracy of deformation detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions of the present invention, the drawings required for implementation will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0036] Figure 1 is a flowchart of the tunnel deformation detection method provided by an embodiment of the present invention;

[0037] Figure 2 is a structural diagram of the tunnel deformation detection system provided by an embodiment of the present invention;

[0038] Figure 3 is a structural diagram of a computer terminal device provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0039] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0040] It should be understood that the step numbers used in the text are only for convenient description and do not limit the execution order of the steps.

[0041] It should be understood that the terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include the plural forms.

[0042] The terms "comprising" and "including" indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations.

[0043] The term "and / or" refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0044] Please refer to Figure 1 , Figure 1 which is a schematic flow chart of the tunnel deformation detection method provided by an embodiment of the present invention. In this embodiment, the tunnel deformation detection method includes the following steps:

[0045] S11, identify tunnel targets according to point cloud data and video images.

[0046] Specifically, the acquisition of point cloud data and video images can be synchronously collected using a vehicle-mounted scanning device.

[0047] S12, segment the tunnel targets to obtain a set of individual targets, and save them in the historical database; wherein the set of individual targets includes several individual targets, and the historical database stores several sets of individual target sets detected at different times.

[0048] S13, use the left and right delimiting algorithm of the central axis to compare the data of several sets of individual targets in the historical database to obtain the deformation degree of the tunnel targets.

[0049] In an embodiment, step S11, identify tunnel targets according to point cloud data and video images, includes: dividing the point cloud data into blocks to obtain the point cloud data after division; using the viewpoint feature histogram to extract three-dimensional point cloud features from the point cloud data after division to obtain the point cloud features after division; classifying and predicting the point cloud features after division, and combining the video images to obtain tunnel targets. After the above processing, it can be quickly and accurately determined whether the scanned railway scene in the point cloud data contains the key areas (areas including tunnels) for later target segmentation. If this area contains the key areas to be segmented, then this area is further segmented and the deformation is detected.

[0050] Among them, considering the large volume of point cloud data, a large amount of computation is required to identify the entire scene, and there are many interfering objects in the scene. Therefore, the point cloud data is segmented. Classify and predict the features of the segmented point cloud, and combine the video image to obtain tunnel targets, specifically including: establishing a KD tree well-known in the industry based on the viewpoint feature histogram features of different objects, and using the well-known nearest point search algorithm in the industry to classify and predict the features of the segmented point cloud. Finally, count the classification results of all blocks in the scene point cloud, and comprehensively determine whether the scene contains key areas.

[0051] In an embodiment, in step S12, segment the tunnel target to obtain a separate target set and save it in the historical database, including: locating and identifying the target image area on the video image; matching the video image with the point cloud data in time, and mapping the target image area to streamline, interpolate or voxelize the target point cloud area to obtain point cloud voxel data; segment the point cloud voxel data at different scales to obtain a separate target set and save it in the historical database.

[0052] Specifically, the segmentation network includes preliminary segmentation and precise segmentation, where:

[0053] (1) Preliminary segmentation stage: This part is to preliminarily segment the tunnel target area. This target area should be generally the same as the tunnel target area scanned in the previous stage. The purpose of preliminary segmentation is to reduce the interference of irrelevant parts, including:

[0054] a. Use the well-known real-time target detection framework Mask-RCNN in the image field to locate and identify the tunnel target area on the collected railway scene image;

[0055] b. Match the point cloud collected at the same time with the image data, and map the target area identified in the image to the corresponding area in the point cloud;

[0056] This corresponding area is the result of the preliminary segmentation of the railway scene data, and this area excludes the irrelevant points in the railway scene;

[0057] (2) Precise segmentation stage: This part is to use the dedicated point cloud segmentation network PointNet++ to locate and segment a more refined target area for the area preliminarily segmented in the previous step, including:

[0058] a. First, streamline or interpolate the point cloud using a downsampling method to make the point cloud distribution uniform;

[0059] b. Then perform voxelization processing on the point cloud data;

[0060] c. Reconstruct the point cloud segmentation network PointNet++ to quickly segment the regions of tunnel targets with different scales and shapes;

[0061] After the above three-step processing, the tunnel targets can be separately segmented from the important regions for the next step of deformation detection.

[0062] In one embodiment, in step S13, the data comparison of several individual target sets in the historical database is performed using the left and right boundary determination algorithm based on the central axis to obtain the deformation degree of the tunnel target, including: registering the individual target sets detected at different times and performing block processing to obtain the registered data set; calculating the data deviation of the registered data set using the left and right boundary determination algorithm based on the central axis to obtain the deformation degree of the tunnel target.

[0063] (1) The method for detecting the target deformation degree includes:

[0064] a. For the point cloud voxel data of a single tunnel target after segmentation, first register it with the previous historical data using the well-known ICP registration algorithm in the industry;

[0065] b. Since the targets in the railway scene are large, then perform block processing on the registered tunnel point cloud.

[0066] For the tunnel point cloud, calculate the deviation between the two-phase data of the tunnel in each block region using the left and right boundary determination algorithm based on the central axis, locate the range of the tunnel deformation region, calculate the envelope volume of the deformation region in the two-phase data of the tunnel, and calculate the maximum distance and average distance of the deviation between the deformation region of the tunnel and the corresponding region in the previous period;

[0067] c. Finally, statistically analyze the overall deformation of the tunnel target. Through the analysis and comparison of multi-phase data, the change trend of the tunnel target deformation can be obtained, and the deformation degree can be quantified.

[0068] (2) The left and right boundary determination algorithm based on the central axis includes:

[0069] a. Obtain the central axis of the tunnel point cloud on the XOY and XOZ planes:

[0070] Intercept the cross-section of the point cloud data along the x direction with Δx = 1 cm, calculate the maximum values y i(max) , z i(max) and minimum values y i(min) , z i(min) of y and z in the cross-section in the x direction. Calculate the average values corresponding to the y and z directions for any point x i (i = 1, 2... n).

[0071]

[0072] Then take x iRespectively with y i , z i Performing quadratic curve fitting, the three-dimensional central axis of the tunnel can be expressed as:

[0073]

[0074] b. Get the x-coordinate center_x of the center point of the cross section:

[0075] Difference the center axis of the two tunnels on the XOY and XOZ planes, and find the extreme point of the difference (i.e., the point where the first derivative is 0 and is within the x range). The x coordinate center_x of this point is the center point of the cross section.

[0076] c. Confirm the left and right boundaries of the deformation area:

[0077] Taking the obtained center_x as the center, calculate the difference between the central axis of the two tunnels on the left and right sides of the x-axis respectively. If the difference does not meet the threshold constraint condition, it is considered that this section belongs to the deformation area. Continuously update the left boundary point bound_l and the right boundary point bound_r. After the update is completed, [bound_l,bound_r] is the detected deformation area. Mark this area with color to highlight it. The setting of the threshold value threshold depends on whether center_x is the maximum value point x_max in the x range or the minimum value point x_min in the x range. When center_x is the maximum value point x_max in the x range, the threshold value threshold is determined by the maximum value num_max and the average value num_mean of the central axis difference of the two tunnels. When center_x is the minimum value point x_min in the x range, the threshold value threshold is determined by the minimum value num_min and the average value num_mean of the central axis difference of the two tunnels. Specifically expressed as:

[0078]

[0079] Among them, α is a decimal in the range of 0 to 1. When center_x is the maximum value point x_max in the range of x, the threshold constraint condition is: the difference between the central axis of the two phases of the tunnel is less than or equal to the threshold. When center_x is the minimum value point x_min in the range of x, the threshold constraint condition is: the difference between the central axis of the two phases of the tunnel is greater than or equal to the threshold. The smaller α is, the smaller the deformation area detected is, and the detected deformation area is the area with severe deformation and needs to be repaired. The larger α is, the larger the deformation area detected is, and it may contain some areas with less severe deformation. For these areas, early warning can be issued to prevent further serious deformation.

[0080] In an embodiment of the present invention, after obtaining on-vehicle point cloud data and video image data, it identifies whether there is a tunnel target in a railway scene. After confirming the existence of a tunnel target, it segments individual tunnel targets in the scene, registers them with the data in the historical database, and performs deformation detection on the paired data of the same tunnel target detected at different time periods using the left-right delimitation algorithm based on the central axis to quantify the degree of tunnel deformation, obtaining the x-axis range of the deformed area, the comparison of the tunnel envelope volume, the maximum and average deviation distances of the tunnel, and summarizing the results for the staff. This embodiment includes the identification of tunnel targets, the segmentation of tunnel targets using the idea of image and point cloud fusion, and the use of the left-right delimitation algorithm based on the central axis to detect the specific area of tunnel deformation. Compared with other deformation detection methods, the left-right delimitation algorithm based on the central axis can not only determine whether the tunnel has deformed, but also accurately locate the deformed section, which can greatly reduce the workload of track inspection and improve the interval accuracy of deformation detection. In addition, the threshold changes dynamically with the change of the difference between the central axes of the tunnel in the previous and current periods. The threshold set in this way has a wider application range compared with a fixed-size threshold and can be applied to tunnels of different sizes, from small subway and road tunnels to large railway tunnels.

[0081] Please refer to Figure 2 , Figure 2 which is a schematic structural diagram of a tunnel deformation detection system provided by an embodiment of the present invention. In this embodiment, the tunnel deformation detection system is characterized by including:

[0082] A target recognition module 21, configured to recognize tunnel targets according to point cloud data and video images.

[0083] Specifically, the acquisition of point cloud data and video images can be synchronously collected using on-vehicle scanning equipment.

[0084] A target segmentation module 22, configured to segment the tunnel targets to obtain a set of individual targets and save them in the historical database; wherein the set of individual targets includes several individual targets, and the historical database stores several sets of individual target sets detected at different times.

[0085] A data comparison module 23, configured to perform data comparison on several sets of individual target sets in the historical database using the left-right delimitation algorithm based on the central axis to obtain the degree of deformation of the tunnel target.

[0086] In one embodiment, the target recognition module 21 is further configured to: divide the point cloud data into blocks to obtain the block-divided point cloud data; extract three-dimensional point cloud features from the block-divided point cloud data by using a viewpoint feature histogram to obtain block-divided point cloud features; classify and predict the block-divided point cloud features, and combine the video image to obtain tunnel targets. After the above processing, it can be quickly and accurately determined whether the scanned railway scene in the point cloud data contains key areas (areas containing tunnels) for later target segmentation. If this area contains key areas to be segmented, further segmentation and deformation detection are performed on this area.

[0087] Among them, considering that the volume of point cloud data is large, a large amount of computation is required to recognize the entire scene, and there are many interfering objects in the scene, so the point cloud data is divided into blocks. Classifying and predicting the block-divided point cloud features and combining the video image to obtain tunnel targets specifically includes: establishing a KD tree well-known in the industry based on the viewpoint feature histogram features of different objects, and using the well-known nearest point search algorithm in the industry to classify and predict the block-divided point cloud features. Finally, the classification results of all blocks in the scene point cloud are statistically analyzed to comprehensively determine whether the scene contains key areas.

[0088] In one embodiment, the target segmentation module 22 is further configured to: locate and identify a target image area on the video image; match the video image with the point cloud data in terms of time, and map the target image area to the corresponding area of the point cloud data to obtain a target point cloud area; perform thinning, interpolation, or voxelization processing on the target point cloud area to obtain point cloud voxel data; divide the point cloud voxel data into different scales to obtain a separate target set and store it in the historical database.

[0089] Specifically, the segmentation network includes preliminary segmentation and precise segmentation, where:

[0090] (1) Preliminary segmentation stage: This part is for preliminary segmentation of the tunnel target area. This target area should be generally the same as the tunnel target area scanned in the previous stage. The purpose of preliminary segmentation is to reduce the interference of irrelevant parts, including:

[0091] a. Use the well-known real-time target detection framework Mask-RCNN in the image field to locate and identify the tunnel target area on the collected railway scene image;

[0092] b. Match the point cloud and image data collected at the same time, and map the target area identified in the image to the corresponding area in the point cloud;

[0093] This corresponding area is the result of preliminary segmentation of the railway scene data, and this area excludes irrelevant points in the railway scene;

[0094] (2) Precise segmentation stage: This part is to use the dedicated point cloud segmentation network PointNet++ to locate and segment more refined target areas for the areas preliminarily segmented in the previous step, including:

[0095] a. First, use the downsampling method to streamline or interpolate the point cloud to make the point cloud distribution uniform;

[0096] b. Then, perform voxelization processing on the point cloud data;

[0097] c. Then, build the point cloud segmentation network PointNet++ to quickly segment the areas of tunnel targets with different scales and shapes;

[0098] After the above three-step processing, the tunnel targets can be separately segmented from the important areas for the next deformation detection.

[0099] In one embodiment, the data comparison module 23 is further configured to: register the individual target sets detected at different times, perform block processing, and obtain a registered data set; calculate the data deviation of the registered data set using the left and right boundary determination algorithm of the central axis to obtain the deformation degree of the tunnel target.

[0100] (1) The method for detecting the deformation degree of the target includes:

[0101] a. For the voxel data of the point cloud of a single tunnel target after segmentation, first use the well-known ICP registration algorithm in the industry to register it with the previous historical data;

[0102] b. Since the targets in the railway scene are large, then perform block processing on the registered tunnel point cloud.

[0103] For the tunnel point cloud, in each block area, use the left and right boundary determination algorithm of the central axis to calculate the deviation between the two-phase data of the tunnel before and after, locate the range of the tunnel deformation area, calculate the envelope volume of the deformation area in the two-phase data of the tunnel before and after, and calculate the maximum distance and average distance of the deviation of the tunnel deformation area from the corresponding area in the previous period;

[0104] c. Finally, statistically analyze the overall deformation of the tunnel target. Through the analysis and comparison of multi-phase data, the change trend of the tunnel target deformation can be obtained, and the deformation degree can be quantified.

[0105] (2) The left and right boundary determination algorithm of the central axis includes:

[0106] a. Obtain the central axis of the tunnel point cloud on the XOY and XOZ planes: [[ID=X]]

[0107] Intercept the cross-section of the point cloud data along the x direction with Δx = 1 cm, and calculate the maximum values y i(max) , z i(max) and minimum values yi(min) , z i(min) . Calculate any point x i (i = 1, 2…n) respectively correspond to the average values in the y and z directions.

[0108]

[0109] Then take x i and respectively compare it with y i , z i to perform quadratic curve fitting. The obtained three-dimensional central axis of the tunnel can be expressed as:

[0110]

[0111] b. Obtain the x coordinate center_x of the center point of the cross-section:

[0112] Take the difference between the central axes of the tunnels in the two periods on the XOY and XOZ planes, and take the extreme value points of the difference (that is, find the points where the first derivative is 0 and within the x range). The x coordinate center_x of this point is the center point of the cross-section.

[0113] c. Confirm the left and right boundaries of the deformation area:

[0114] Taking the obtained center_x as the center, respectively calculate the difference between the central axes of the tunnels in the two periods on the left and right sides of the x-axis. If the difference does not meet the threshold limit condition, then this cross-section is considered to belong to the deformation area. Continuously update the left boundary point bound_l and the right boundary point bound_r. The [bound_l, bound_r] after the update is the detected deformation area. Mark the color of this area to highlight it. The setting of the threshold threshold depends on whether center_x is the maximum value point x_max or the minimum value point x_min within the x range. When center_x is the maximum value point x_max within the x range, the threshold threshold is determined by the maximum value num_max and the average value num_mean of the differences between the central axes of the tunnels in the two periods. When center_x is the minimum value point x_min within the x range, the threshold threshold is determined by the minimum value num_min and the average value num_mean of the differences between the central axes of the tunnels in the two periods. Specifically expressed as:

[0115]

[0116] Among them, α is a decimal within the range of 0 to 1. When center_x is the maximum value point x_max within the range of x, the threshold limit condition is that the difference between the central axes of the tunnels in two consecutive periods is less than or equal to the threshold. When center_x is the minimum value point x_min within the range of x, the threshold limit condition is that the difference between the central axes of the tunnels in two consecutive periods is greater than or equal to the threshold. The smaller α is, the smaller the detected deformation area is, and the detected deformation area is the area with serious deformation that needs to be repaired with emphasis. The larger α is, the larger the detected deformation area is, and it may contain some areas with less serious deformation. For this part of the area, early warning can be given to prevent the deformation from becoming more serious.

[0117] Please refer to Figure 3 , an embodiment of the present invention provides a computer terminal device, including one or more processors and a memory. The memory is coupled to the processor and is used to store one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the tunnel deformation detection method in any of the above embodiments.

[0118] The processor is used to control the overall operation of the computer terminal device to complete all or part of the steps of the above tunnel deformation detection method. The memory is used to store various types of data to support the operation of the computer terminal device. These data may include, for example, instructions for any application program or method operating on the computer terminal device, and application program-related data. The memory can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, abbreviated as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, abbreviated as EEPROM), erasable programmable read-only memory (Erasable Programmable Read-Only Memory, abbreviated as EPROM), programmable read-only memory (Programmable Read-Only Memory, abbreviated as PROM), read-only memory (Read-Only Memory, abbreviated as ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0119] In an exemplary embodiment, the computer terminal device may be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components, and is used to execute the above-mentioned tunnel deformation detection method and achieve the same technical effects as those of the above method.

[0120] In another exemplary embodiment, there is also provided a computer-readable storage medium including a computer program. When the computer program is executed by a processor, it implements the steps of the tunnel deformation detection method in any one of the above embodiments. For example, the computer-readable storage medium may be the above-mentioned memory including the computer program, and the above computer program may be executed by the processor of the computer terminal device to complete the above-mentioned tunnel deformation detection method and achieve the same technical effects as those of the above method.

[0121] The above is the preferred embodiment of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements are also regarded as the protection scope of the present invention.

Claims

1. A tunnel deformation detection method, characterized in that, Including: Identifying tunnel targets based on point cloud data and video images; Segmenting the tunnel targets to obtain a set of individual targets and storing them in a historical database; wherein the set of individual targets includes a number of individual targets, and the historical database stores several sets of individual targets detected at different times; Using the left - right delimiting algorithm based on the central axis to compare data of several sets of individual targets in the historical database to obtain the deformation degree of the tunnel targets; The step of using the left - right delimiting algorithm based on the central axis to compare data of several sets of individual targets in the historical database to obtain the deformation degree of the tunnel targets includes: Registering sets of individual targets detected at different times and performing block processing to obtain a registered data set; Using the left - right delimiting algorithm based on the central axis to calculate the data deviation of the registered data set to obtain the deformation degree of the tunnel targets; The step of using the left - right delimiting algorithm based on the central axis to calculate the data deviation of the registered data set to obtain the deformation degree of the tunnel targets includes: Obtaining a first central axis of the tunnel point cloud data in the XOY plane and a second central axis in the XOZ plane in the registered data set; Based on the first central axis and the second central axis, obtaining the x - axis coordinate of the cross - section center point; Taking the x - axis coordinate of the cross - section center point as the center, calculating the difference between the central axes of the tunnel in the previous and current periods respectively to the left and right sides of the x - axis; Judging whether the difference between the central axes meets the threshold limit condition. If not, determining the cross - section as a deformation area; The threshold limit condition is: Wherein, threshold represents the threshold, center_x represents the x - axis coordinate of the cross - section center point, x_max represents the maximum value within the x range, x_min represents the minimum value within the x range, α represents a decimal within the range of 0 to 1, num_max represents the maximum value of the difference between the central axes of the tunnel in the previous and current periods, num_min represents the minimum value of the difference between the central axes of the tunnel in the previous and current periods, and num_mean represents the average value of the difference between the central axes of the tunnel in the previous and current periods; When the center_x is the maximum value point x_max within the x range, the threshold limit condition is: the difference between the central axes of the tunnel in the previous and current periods is less than or equal to the threshold; When the center_x is the minimum value point x_min within the x range, the threshold limit condition is: the difference between the central axes of the tunnel in the previous and current periods is greater than or equal to the threshold.

2. The tunnel deformation detection method according to claim 1, wherein The step of identifying tunnel targets based on point cloud data and video images includes: Segmenting the point cloud data to obtain segmented point cloud data; Using the view - point feature histogram to extract three - dimensional point cloud features from the segmented point cloud data to obtain segmented point cloud features; Classifying and predicting the segmented point cloud features and combining with the video images to obtain tunnel targets.

3. The tunnel deformation detection method according to claim 1, characterized in that The step of segmenting the tunnel targets to obtain a set of individual targets and storing them in a historical database includes: Locating and identifying a target image area on the video image; Matching the time of the video image with the point cloud data and mapping the target image area to the corresponding area of the point cloud data to obtain a target point cloud area; After thinning or interpolating the target point cloud region, further perform voxelization processing to obtain point cloud voxel data; Segment the point cloud voxel data into different scales to obtain a separate target set, and save it in the historical database.

4. A tunnel deformation detection system, characterized in that, Including: A target recognition module for recognizing tunnel targets based on point cloud data and video images; A target segmentation module for segmenting the tunnel targets to obtain a separate target set and saving it in the historical database; wherein the separate target set includes several individual targets, and the historical database stores several sets of separate target sets detected at different times; A data comparison module for using the left and right boundary determination algorithm of the central axis to compare the data of several separate target sets in the historical database to obtain the deformation degree of the tunnel target; The data comparison module is further used for: Registering the separate target sets detected at different times and performing block processing to obtain a registered data set; using the left and right boundary determination algorithm of the central axis to calculate the data deviation of the registered data set to obtain the deformation degree of the tunnel target; The calculating the data deviation of the registered data set by using the left and right boundary determination algorithm of the central axis to obtain the deformation degree of the tunnel target includes: Obtain the first central axis of the tunnel point cloud data in the XOY plane and the second central axis in the XOZ plane in the registered data set; Based on the first central axis and the second central axis, obtain the x-axis coordinate of the cross-section center point; Taking the x-axis coordinate of the cross-section center point as the center, calculate the difference between the central axes of the tunnel in the previous and current periods respectively to the left and right sides of the x-axis; Judge whether the difference between the central axes meets the threshold limit condition. If not, determine the cross-section as a deformation area; The threshold limit condition is: Wherein, threshold represents the threshold, center_x represents the x-axis coordinate of the cross-section center point, x_max represents the maximum value within the x range, x_min represents the minimum value within the x range, α represents a decimal within the range of 0 to 1, num_max represents the maximum value of the difference between the central axes of the tunnel in the previous and current periods, num_min represents the minimum value of the difference between the central axes of the tunnel in the previous and current periods, and num_mean represents the average value of the difference between the central axes of the tunnel in the previous and current periods; When the center_x is the maximum value point x_max within the x range, the threshold limit condition is: the difference between the central axes of the tunnel in the previous and current periods is less than or equal to the threshold; When the center_x is the minimum value point x_min within the x range, the threshold limit condition is: the difference between the central axes of the tunnel in the previous and current periods is greater than or equal to the threshold.

5. The tunnel deformation detection system according to claim 4, characterized in that, The target recognition module is further used for: Segment the point cloud data to obtain the segmented point cloud data; use the view point feature histogram to extract the three-dimensional point cloud features of the segmented point cloud data to obtain the segmented point cloud features; Classify and predict the segmented point cloud features, and combine the video images to obtain tunnel targets.

6. The tunnel deformation detection system according to claim 4, wherein The target segmentation module is further used for: Locate and identify the target image area on the video image; perform time matching between the video image and the point cloud data, and map the target image area to the corresponding area of the point cloud data to obtain the target point cloud area; after thinning or interpolating the target point cloud area, further perform voxelization processing to obtain point cloud voxel data; segment the point cloud voxel data into different scales to obtain a separate target set and store it in the historical database.

7. A computer terminal device, characterized in that, Comprising: One or more processors; A memory coupled to the processor for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the tunnel deformation detection method according to any one of claims 1 to 3.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the tunnel deformation detection method according to any one of claims 1 to 3.

Images