A crack image recognition system for top-down subway stations

Abstract

The present invention relates to the technical field of image recognition. The present invention discloses a crack image recognition system for a top-down subway station; it includes a point position adjustment module for adjusting the initial point position to a recognition point position, an image segmentation module for performing regional segmentation on the sampled image to generate a recognition image, a region drawing module for drawing a crack region in the wall tiled image, a crack recognition module for identifying structural cracks from the crack region, and a recognition prompt module for formulating corresponding warning prompt information; the present invention can avoid the phenomenon of blurring at the left and right edge positions of the sampled image due to shooting environmental factors. By means of regional segmentation of the overlapping regions, it ensures that the recognition image can be in a complete and continuous tiled state during subsequent splicing, and combined with an accurate classification of the danger level of the structural cracks, it can improve the image recognition accuracy of the crack region and type on the wall of the top-down subway station.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and more specifically, to a crack image recognition system for a top-down subway station under excavation. Background Art

[0002] Top-down construction is a commonly used construction method in subway station construction. Since top-down construction involves pouring concrete in sections from top to bottom, the lower concrete will shrink due to its own drying shrinkage and temperature shrinkage during the initial setting and final setting, and cracks will occur between the lower concrete and the upper concrete under the action of its own weight, resulting in cracks on the walls inside the subway station. In order to identify the cracks on the walls and facilitate subsequent repair and treatment of the cracks, it is necessary to perform image recognition on the cracks.

[0003] The patent application with the publication number CN116246166A discloses a preprocessing and postprocessing algorithm for building bridge crack image recognition. Through the preprocessing and postprocessing algorithm for building bridge crack image recognition, an original image with any pixel size can be cropped into a standard image with a set pixel size, and then crack recognition is performed on the standard images with the same pixel size, which can effectively improve the accuracy and efficiency of crack recognition. After identifying the cracks, the cropped standard images are stitched and restored, which will not affect the display effect of the crack recognition results;

[0004] When the existing image recognition system performs image recognition on the wall cracks of a subway station, by obtaining wall images at multiple different points and performing visual detection and continuous stitching on the wall images one by one, the image recognition effect of the cracks is achieved. However, due to the interference of the shooting environment such as uneven light or wall reflection inside the subway station when taking wall images, image blurring or information loss will occur in the left and right edge areas of the wall images at adjacent points, and the specific information in the edge areas of the wall images cannot be accurately processed and recognized, making it difficult to ensure the continuity and integrity during subsequent wall image stitching and crack recognition, and reducing the accuracy of the image recognition of the wall cracks in the subway station.

[0005] In view of this, the present invention proposes a crack image recognition system for a top-down subway station to solve the above problems. Summary of the Invention

[0006] In order to overcome the above defects of the prior art and achieve the above object, the present invention provides the following technical solution: A crack image recognition system for a top-down subway station under excavation, which is applied to an image recognition center and includes:

[0007] A point position adjustment module, which pre-sets the initial points for the acquisition terminal to capture the initial images inside the subway station, and based on the point position adjustment criterion, horizontally adjusts the initial points into spaced-apart recognition points;

[0008] The point adjustment criterion is that the area of the overlapping region of two adjacent initial images is 5% of the total area of the initial images.

[0009] An image segmentation module acquires a sampled image of the wall surface at the recognition points, and based on the region segmentation criterion, performs region segmentation on the overlapping region in the sampled image to generate a recognition image.

[0010] The region segmentation criterion is to perform segmentation on the region outside the vertical center line with the vertical center line of the overlapping region as the reference.

[0011] A region drawing module marks the crack points in the recognition image, connects the crack points to enclose local cracks, splices the recognition images into a wall surface tiled image, and draws a crack region in the wall surface tiled image.

[0012] A crack recognition module collects the trend characteristics and span characteristics of the crack region, makes a structural safety determination for the trend characteristics and span characteristics based on the structural safety criterion, and identifies structural cracks from the crack region.

[0013] A recognition prompt module collects the proportion of the extension length and the proportion of the covered area of the structural crack, classifies the structural crack into a high-risk level and a low-risk level, and formulates corresponding warning prompt information.

[0014] Furthermore, the acquisition terminal includes a camera and an electric control trolley.

[0015] The method for setting the initial points is as follows:

[0016] Turn on the camera to keep it running normally, adjust the pitching angle of the camera, and keep the upper edge and the lower edge of the wall surface within the shooting range of the camera.

[0017] Control the electric control trolley to move towards the direction of the wall surface at the nearest position until the upper edge and the lower edge of the shooting range of the camera are horizontally coincident with the upper edge and the lower edge of the wall surface respectively, and then stop moving. Record the position of the moved electric control trolley as the original point.

[0018] Measure the straight-line distance from the original point to the wall surface to obtain a wall point distance value, and draw a point line with a circular structure on the ground of the subway station with a distance of one wall point distance value from the wall surface as the drawing standard.

[0019] Query the corresponding wall width of the shooting range of the camera, and mark the subsequent points distributed at intervals on the point line with a distance less than one wall width as the interval standard. After summarizing the original point and all the subsequent points, A initial points are obtained.

[0020] Furthermore, the method for adjusting the recognition points is as follows:

[0021] According to the sequence of the initial point positions, control the electric control trolley to move to A initial point positions in turn, and capture images of the wall through a camera to obtain A initial images;

[0022] Taking the a-th initial image as the reference image, continuously adjust the position of the (a + 1)-th initial point on the point position line, measure the length and width of the overlapping area between the (a + 1)-th initial image and the reference image in real time, and calculate the area of the overlapping area by combining the rectangle area formula;

[0023] When the area of the overlapping area is 5% of the total area of the initial image, stop adjusting the position of the (a + 1)-th initial point. Until the initial image contains all positions of the wall, record the adjusted initial point as the recognition point to obtain B recognition points.

[0024] Furthermore, the method for generating the recognition image is as follows:

[0025] Control the electric control trolley to move to B recognition points in turn, and capture wall images through the camera in turn to obtain B sampling images;

[0026] Draw lines along the upper and lower edges of the overlapping area in the sampling image to obtain the upper boundary line and the lower boundary line respectively;

[0027] Use computer vision technology to identify the midpoints of the upper boundary line and the lower boundary line in the overlapping area one by one. After vertically connecting the two midpoints of the same overlapping area, obtain the vertical center line;

[0028] Mark the area on the left side of the overlapping area in the sampling image and to the left of the vertical center line as the left area, and mark the area on the right side of the overlapping area in the sampling image and to the right of the vertical center line as the right area;

[0029] Separate the left and right areas of the B sampling images respectively, and number the B segmented sampling images in ascending order with 1 as the first number according to the shooting sequence to obtain B recognition images.

[0030] Furthermore, the method for surrounding local cracks is as follows:

[0031] Convert the B recognition images into B grayscale images, mark the brightness values of all pixel points in the B grayscale images one by one, and record the pixel points with brightness values less than the calibrated brightness value as crack points to obtain C crack points;

[0032] Draw circles with the C crack points as the centers and one-third of the length of the upper boundary line as the radius to obtain C pixel circles, and count the total number of pixel points and the number of crack points in the C pixel circles respectively;

[0033] After comparing the number of crack points in each of the C pixel circles with the total number of pixel points respectively, C crack ratios are obtained. The crack points corresponding to the centers of the circles with crack ratios greater than the calibrated crack ratio are marked as boundary points, and D boundary points are obtained.

[0034] The distance values between adjacent two boundary points are measured one by one. The boundary points with distance values less than the calibrated distance value are connected in sequence to form an enclosed area, and the area enclosed by the connection is recorded as a local crack.

[0035] Further, the method for drawing the crack area is as follows:

[0036] The B recognition images are imported into the image processing software in sequence and arranged horizontally to the right, keeping the upper edges and lower edges of the B recognition images horizontally aligned respectively.

[0037] The positions of two adjacent numbered recognition images are adjusted horizontally until the vertical center line on the right side of the b-th recognition image coincides completely with the vertical center line on the left side of the (b + 1)-th recognition image, and then the adjustment stops to obtain a wall tiled image.

[0038] A line is drawn along the position of the boundary points of the local crack to form a closed drawing area, and the closed drawing area is recorded as the crack area, and E crack areas are obtained.

[0039] Further, the method for collecting the trend feature and span feature is as follows:

[0040] A line is drawn along the position of the lower edge of the wall tiled image to obtain the lower boundary line of the wall.

[0041] The point-line distance values from the crack points in the E crack areas to the lower boundary line of the wall are measured one by one. The crack points corresponding to the minimum value and the maximum value of the point-line distance values are marked as the starting point and the ending point respectively. After connecting the starting points and ending points of the E crack areas in sequence, E trend lines are obtained.

[0042] Reference lines are drawn horizontally at the starting points of the E crack areas respectively, and the inclination angles between the E trend lines and the E reference lines are measured one by one to obtain E trend features.

[0043] The distances between the starting points and the ending points in the E crack areas are measured vertically one by one to obtain E span features.

[0044] Further, the structural safety criterion is: the crack areas with crack structural features are recorded as structural cracks.

[0045] The method for identifying structural cracks is:

[0046] The trend features of the E crack areas are compared with the first trend range and the second trend range respectively.

[0047] When the strike feature is within the first structural range or the second structural range, the strike feature is recorded as a crack structural feature;

[0048] Compare the span features of the E crack regions with the span structure value;

[0049] When the span feature is greater than the span structure value, the span feature is recorded as a crack structural feature;

[0050] Count one by one the number of crack structural features in the E crack regions, and record the crack regions with the number of crack structural features being 1 or 2 as structural cracks to obtain F structural cracks.

[0051] Furthermore, the acquisition methods for the extension length ratio and the coverage area ratio are as follows:

[0052] Measure one by one the horizontal distance values between any two crack points in the F structural cracks and the length value of the wall tiled image in the horizontal direction, and compare the maximum value of the horizontal distance values with the length value of the wall tiled image to obtain F extension length ratios;

[0053] Count one by one the number of crack points in the F structural cracks and the total number of pixel points in the wall tiled image, and after comparing the number of crack points with the total number of pixel points in the wall tiled image, obtain F coverage area ratios.

[0054] Furthermore, the classification methods for the high-risk level and the medium-risk level are as follows:

[0055] Compare the extension length ratios and the coverage area ratios of the F structural cracks with the length ratio threshold and the area ratio threshold respectively;

[0056] When the extension length ratio is greater than the length ratio threshold, record the extension length ratio as dangerous data;

[0057] When the coverage area ratio is greater than the area ratio threshold, record the coverage area ratio as dangerous data;

[0058] Count one by one the number of dangerous data in the F structural cracks, classify the structural cracks with the number of dangerous data being 0 as the low-risk level, and classify the structural cracks with the number of dangerous data being 1 or 2 as the high-risk level.

[0059] The technical effects and advantages of a crack image recognition system for a top-down and inverse-construction subway station according to the present invention:

[0060] By setting the recognition points distributed at intervals, the present invention can ensure that there is a certain overlapping area between the left and right edges of adjacent two sampling images, so that there is an overlapping and associated area between adjacent two sampling images, avoiding the phenomenon that the positions of the left and right edges of the sampling images are blurred due to shooting environmental factors. And by means of region segmentation of the overlapping area, the overlapping area in the recognition image can be accurately segmented, ensuring that the recognition image can be in a complete and continuous tiled state during subsequent splicing, providing a good image basis for the subsequent recognition and drawing of the crack area. At the same time, through the structural safety determination of the trend feature and the span feature, the structural features with potential hazards in the subway station can be effectively identified and distinguished, and combined with the accurate classification of the danger degree of the structural cracks, the image recognition accuracy of the crack area and type on the wall of the top-down subway station can be improved, thus providing theoretical support for the subsequent repair operation of the structural cracks. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 It is a schematic structural diagram of a crack image recognition system for a top-down subway station provided in Embodiment 1 of the present invention;

[0062] Figure 2 It is a schematic distribution diagram of the recognition points provided in Embodiment 1 of the present invention;

[0063] Figure 3 It is a schematic flow diagram of a crack image recognition method for a top-down subway station provided in Embodiment 2 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0064] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0065] Embodiment 1: Please refer to Figure 1 and Figure 2 As shown, the crack image recognition system for a top-down subway station in this embodiment is applied to an image recognition center and includes:

[0066] A point position adjustment module, which sets the initial point positions of the acquisition terminals in the subway station, obtains the initial images of the acquisition terminals at the initial point positions, and horizontally adjusts the initial point positions to the recognition points distributed at intervals based on the point position adjustment criteria;

[0067] The initial point refers to the point set in advance in the subway station at a certain distance from the surrounding vertical walls, and serves as the original point for subsequently obtaining the wall image data at various positions and directions inside the subway station. Due to the large internal space of the subway station and the large wall area in each direction, in order to ensure that all positions of the walls in the subway station can be photographed, multiple initial points need to be set at intervals in the subway station.

[0068] The acquisition terminal refers to an intelligent trolley equipped with a high-definition camera and can be wirelessly remotely controlled to move freely, enabling the acquisition terminal to move freely and flexibly to the initial point and capture high-definition images of the wall, and recording the captured high-definition images as initial images;

[0069] Specifically, the acquisition terminal includes a camera and an electric control trolley.

[0070] The method for setting the initial point is as follows:

[0071] Turn on the camera to keep it running normally, adjust the pitching angle of the camera, and keep the upper and lower edges of the wall within the shooting range of the camera;

[0072] Control the electric control trolley to move in the direction of the wall at the nearest position until the upper and lower edges of the shooting range of the camera are horizontally coincident with the upper and lower edges of the wall respectively, and then stop moving, and record the position of the electric control trolley after moving as the original point;

[0073] Measure the straight-line distance from the original point to the wall to obtain the wall point distance value, and use the distance of one wall point distance value from the wall as the drawing standard to draw a circular structure of point lines on the ground of the subway station; the point lines are contour lines used to provide position adjustment limits for the electric control trolley and play a role in limiting the location of the subsequent initial points, ensuring that the distance from each initial point to the wall at the nearest position is the same;

[0074] Query the corresponding wall width of the shooting range of the camera, and use less than one wall width as the interval standard to mark the subsequent points distributed at intervals on the point lines. After summarizing the original point and all subsequent points, A initial points are obtained.

[0075] After setting the initial points, it is necessary to control the acquisition terminal to capture the initial images of the wall at the initial points. At this time, the initial images can only meet the requirement of photographing all positions of the walls in the subway station. Since the distance between adjacent two initial points is different, there is an easy phenomenon that the overlapping area between the initial images at adjacent two initial points is too large, resulting in a large amount of duplicate information between the initial images at adjacent two initial points, reducing the accuracy of subsequent crack identification;

[0076] To ensure that there is not only a reasonable and necessary overlapping area between the initial images at two adjacent initial positions, but also that images can be obtained for all positions on the wall inside the subway station, it is necessary to adjust the positions of the initial positions;

[0077] When adjusting the positions of the initial positions, it is necessary to do so under the limitation of the position adjustment criterion, and record the initial positions after the position adjustment as recognition positions, so that the recognition positions can serve as the correct positions for the acquisition terminal to obtain images of the subway station wall subsequently.

[0078] The position adjustment criterion is: the area of the overlapping region between two adjacent initial images is 5% of the total area of the initial image; this can ensure that there is a certain overlapping region between two adjacent initial images, and it is also possible to perform image shooting and recognition twice on the information at the two side edge positions of the initial image, thereby preventing the phenomenon of blurring at the edge positions of the initial image due to the influence of the shooting environment.

[0079] The adjustment method for the recognition positions is as follows:

[0080] According to the sequence of the initial position markings, control the electric control trolley to move to A initial positions in turn, and capture images of the wall through the camera to obtain A initial images;

[0081] Taking the a-th initial image as the reference image, continuously adjust the position of the (a + 1)-th initial position on the position line, measure the length and width of the overlapping region between the (a + 1)-th initial image and the reference image in real time, and calculate the area of the overlapping region in combination with the rectangle area formula, where a = 1, 2,..., A;

[0082] When the area of the overlapping region is 5% of the total area of the initial image, stop adjusting the position of the (a + 1)-th initial position;

[0083] Until all positions on the wall are included in the initial image, record the adjusted initial positions as recognition positions to obtain B recognition positions.

[0084] It should be noted that there will also be a certain proportion of overlapping regions in the captured images of the first recognition position and the last recognition position to ensure that all captured images can follow the position adjustment criterion and improve the processing accuracy of the subsequent captured images.

[0085] Based on the above adjustment method for the recognition positions, obtain the recognition positions, and combine with Figure 2 to represent the positions and distributions of the recognition positions, the position line, the wall, and the acquisition terminal.

[0086] The image segmentation module acquires the sampled images of the wall surface at the recognition points, and based on the region segmentation criterion, performs region segmentation on the overlapping regions in the sampled images to generate recognition images;

[0087] The sampled image refers to the real-time image of the wall surface taken by the acquisition terminal at the recognition points. Due to the influence of the point position adjustment criterion, there will be a certain overlapping region between two adjacent sampled images at this time. In order to ensure the accurate recognition of wall cracks in the follow-up, it is necessary to perform region segmentation processing on the overlapping regions of two adjacent sampled images, so that the wall positions corresponding to two adjacent sampled images can maintain continuity and uniqueness, and the sampled images after region segmentation are recorded as recognition images;

[0088] When performing region segmentation on two adjacent sampled images, it is necessary to perform separate segmentation on the overlapping regions of the two sampled images based on the actual wall position corresponding to the overlapping region of the sampled images. Therefore, it is necessary to perform region segmentation processing under the limitation of the region segmentation criterion.

[0089] The region segmentation criterion is: taking the vertical center line of the overlapping region as the benchmark, segment the regions outside the vertical center line; this can ensure that the regions in the overlapping region of the sampled image that coincide with the adjacent sampled image are segmented off, avoiding the negative interference effects brought by the regions at the two side edge positions of the sampled image due to the shooting environment, thereby improving the clarity of each sampled image.

[0090] The generation method of the recognition image is:

[0091] Control the electric control trolley to move to B recognition points in turn, and sequentially capture the wall surface images through the camera to obtain B sampled images;

[0092] Draw lines along the upper and lower edges of the overlapping region in the sampled image to obtain the upper boundary line and the lower boundary line respectively;

[0093] Use computer vision technology to identify the midpoints of the upper boundary line and the lower boundary line in the overlapping region one by one, and after vertically connecting the two midpoints of the same overlapping region, obtain the vertical center line;

[0094] Mark the region on the left side of the overlapping region in the sampled image and to the left of the vertical center line as the left region, and mark the region on the right side of the overlapping region in the sampled image and to the right of the vertical center line as the right region; the specific positions of the left region and the right region in the sampled image are the left and right edges of the sampled image respectively, so that the left region and the right region can be used as the objects for subsequent region segmentation of the overlapping region;

[0095] Separate the left regions and right regions of the B sampled images respectively, and in the order of shooting, starting with 1 as the first number, sequentially number the B sampled images after segmentation in ascending order to obtain B recognition images.

[0096] It should be noted that the recognized images obtained by region segmentation can replace all the sampled images one by one, that is, all positions on the wall of the subway station can be displayed. After obtaining the recognized images, by numbering each recognized image, the recognized images can be distinguished, which is convenient for accurately marking the recognized images with crack phenomena in the follow-up.

[0097] The region drawing module marks the crack points in the recognized images, connects the crack points to enclose a local crack, and splices the recognized images into a wall tiled image, and draws the crack region in the wall tiled image;

[0098] The crack point refers to the pixel point corresponding to the crack position in the recognized image, that is, the pixel points at the crack position and the pixel points at non-crack positions can be accurately distinguished.

[0099] After marking the crack points, the crack points are in a discrete state at this time, resulting in the crack position in the recognition region not being able to maintain a relatively complete and closed state. Therefore, it is necessary to summarize and enclose the crack points at different positions in the recognized image, and record the corresponding positions of the summarized and enclosed crack points as local cracks, so that the local cracks can independently represent the cracks on each recognized image.

[0100] The method for enclosing the local crack is as follows:

[0101] Convert B recognized images into B grayscale images, and mark the brightness values of all pixel points in the B grayscale images one by one;

[0102] Compare the brightness value of the pixel point with the calibrated brightness value, and mark the pixel points with brightness values less than the calibrated brightness value as crack points to obtain C crack points; the calibrated brightness value refers to the maximum value of the brightness values of the pixel points recognized as crack points. Since the brightness value in the grayscale image determines the brightness degree of the pixel point position, the smaller the brightness value of the pixel point, the lower the brightness and the darker the picture. In the actual grayscale image of the wall crack, the crack is usually a black background with low brightness. Therefore, the calibrated brightness value can limit the brightness value of the pixel points at the crack position;

[0103] Draw circles with the C crack points as the centers and one-third of the length of the upper boundary line as the radius respectively to obtain C pixel circles, and respectively count the total number of pixel points and the number of crack points in the C pixel circles;

[0104] After comparing the number of crack points in C pixel circles with the total number of pixel points, C crack ratios are obtained. The crack point corresponding to the center of the circle with a crack ratio greater than the calibrated crack ratio is recorded as a boundary point, and D boundary points are obtained. The calibrated crack ratio refers to the minimum crack ratio of the pixel circle where the crack point recorded as a boundary point is located, which provides a numerical limit for the identification of boundary points.

[0105] The calculation formula for the crack ratio is: ;

[0106] Where, is the crack ratio of the c-th pixel circle, c=1,2,...,c, is the number of crack points in the c-th pixel circle, is the total number of pixels in the c-th pixel circle;

[0107] The spacing between two adjacent boundary points is measured one by one. Boundary points with spacing values less than the calibrated spacing value are connected one by one, and the area enclosed by the connected lines is recorded as a local crack. The calibrated spacing value is the maximum spacing value between two adjacent boundary points in the same local crack. This provides a basis for connecting boundary points belonging to the same local crack, thus preventing the incorrect connection of boundary points in different local cracks.

[0108] When obtaining local cracks, the local cracks at this time can only represent the individual crack positions on each recognition image, so that all local cracks are in a discrete and discontinuous state. In order to fully and continuously represent the wall crack phenomenon of the subway station, it is necessary to tile and splice the recognition images so that the discrete recognition images can form a complete wall tile image.

[0109] After the identified images are stitched together into a tiled wall image, the location of the local cracks in each identified image needs to be identified so that the local cracks at consecutive positions on the wall can be combined in an orderly manner and eventually form multiple complete crack areas.

[0110] The method for drawing the crack area is:

[0111] Import B identification images into the image processing software in sequence and arrange them to the right, keeping the upper and lower edges of the B identification images horizontally aligned;

[0112] Adjust the positions of two adjacent numbered recognition images in the horizontal direction until the vertical center line on the right side of the b-th recognition image completely coincides with the vertical center line on the left side of the b+1-th recognition image, and then stop adjusting to obtain a wall tile image, where b = 1, 2, ..., B;

[0113] Draw a line along the position of the boundary points of the local cracks to form a closed drawing area, and denote the closed drawing area as the crack area, obtaining E crack areas.

[0114] It should be noted that when the recognition images are tiled and spliced, the local cracks in each recognition image will also be tiled and seamlessly spliced, so that the local cracks that are associated and in a continuous position on the wall can be spliced and combined one by one, achieving the splicing accuracy of the local cracks corresponding to the same crack position on the wall.

[0115] The crack recognition module collects the trend characteristics and span characteristics of the crack area, makes a structural safety determination for the trend characteristics and span characteristics based on the structural safety criteria, and identifies the structural cracks from the crack area;

[0116] After obtaining the crack areas in the tiled state, the cracks existing on the wall of the subway station will remain in a continuous and complete state. At this time, there will be differences in the size, shape, and position of each crack area. Due to the reason of top-down construction, the lower-layer concrete will shrink and cracks will occur between the upper and lower concrete under the action of its own weight, and this crack will affect the structural safety of the wall of the subway station. Therefore, the crack area corresponding to this crack is denoted as the structural crack.

[0117] The trend characteristic is used to represent the inclination amplitude of the crack trend on the tiled image of the wall, and the span characteristic is used to represent the size of the crack span of the crack area;

[0118] The acquisition methods of the trend characteristic and the span characteristic are as follows:

[0119] Draw a line along the position of the lower edge of the tiled image of the wall to obtain the lower boundary line of the wall;

[0120] Measure one by one the point-line distance values from the crack points in the E crack areas to the lower boundary line of the wall, and denote the crack points corresponding to the minimum value and the maximum value of the point-line distance values as the starting point and the ending point respectively. After connecting the starting points and the ending points of the E crack areas in sequence, obtain E trend lines;

[0121] Draw a reference line horizontally along the starting point of each of the E crack areas, and measure one by one the inclination angles between the E trend lines and the E reference lines to obtain E trend characteristics;

[0122] Measure one by one the distances between the starting point and the ending point in the E crack areas along the vertical direction to obtain E span characteristics.

[0123] After collecting the strike characteristics and span characteristics, it is necessary to conduct structural safety assessment on the strike characteristics and span characteristics of the crack area. That is, based on the size of the strike characteristics and span characteristics of each crack area, it is determined whether the crack area will affect the structural safety of the wall, and the crack area that has a negative impact on the structural safety of the wall will be recorded as a structural crack.

[0124] When identifying structural cracks within the crack area, it is necessary to perform structural safety assessment on the strike characteristics and span characteristics under the constraints of the structural safety criteria to ensure the accuracy of structural crack identification.

[0125] The structural safety criterion is: crack areas with crack structural characteristics are recorded as structural cracks; cracks that appear in the wall structure and have a negative impact on the safety of the wall structure can be accurately identified.

[0126] The identification method of structural cracks is:

[0127] Compare the strike characteristics of the E fracture areas with the first strike range and the second strike range respectively;

[0128] When the strike feature is within the first structural range or the second structural range, and the crack inclination amplitude of the crack region is within the inclination amplitude range corresponding to the structural crack, the strike feature of the crack region cannot pass the structural safety assessment and the strike feature is recorded as a crack structural feature; the first strike range is a small inclination range level for the crack inclination amplitude, and the second strike range is a large inclination range level for the crack inclination amplitude; for example, the first strike range is 0 degrees to 15 degrees, and the second strike range is 45 degrees to 60 degrees;

[0129] Compare the span characteristics of the E crack regions with the span structure value; the span structure value is the minimum span amplitude of the crack region identified as a structural crack; for example, the span structure value is 0.4 mm;

[0130] When the span feature is greater than the span structure value, the crack span amplitude of the crack area is within the crack span amplitude range corresponding to the structural crack. In this case, the span feature of the crack area cannot pass the structural safety assessment and the span feature is recorded as a crack structure feature.

[0131] The number of crack structural features in the E crack regions is counted one by one, and the crack regions with 1 or 2 crack structural features are recorded as structural cracks, thereby obtaining F structural cracks.

[0132] The identification and warning module collects crack distribution parameters of structural cracks, classifies structural cracks into high-risk and low-risk levels, and formulates early warning information corresponding to the high-risk and low-risk levels;

[0133] The crack distribution parameters are used to diversely represent the specific morphological distribution of structural cracks on the wall surface, and can provide a data basis for identifying the level of the dangerous impact of structural cracks on the wall surface, and provide a basis for the subsequent identification of structural cracks of high and medium risk levels.

[0134] The crack distribution parameters include the extension length ratio and the coverage area ratio;

[0135] The extension length ratio refers to the ratio between the distribution length of the structural crack and the length of the flat-laid image of the wall surface, which can represent the crack length of the structural crack. When the extension length ratio is larger, the crack length is also larger, and the crack danger index is also larger;

[0136] The coverage area ratio refers to the ratio between the area of the structural crack and the area of the flat-laid image of the wall surface, which can represent the crack area of the structural crack. When the coverage area ratio is larger, the crack area is also larger, and the crack danger index is also larger;

[0137] The acquisition methods of the extension length ratio and the coverage area ratio are as follows:

[0138] Measure the horizontal distance value between any two crack points in F structural cracks and the length value of the flat-laid image of the wall surface one by one along the horizontal direction, and compare the maximum value of the horizontal distance value with the length value of the flat-laid image of the wall surface to obtain F extension length ratios;

[0139] The calculation formula of the extension length ratio is: ;

[0140] In the formula, is the extension length ratio of the f-th structural crack, f = 1, 2,..., F, is the maximum value of the horizontal distance value of the f-th structural crack, is the length value of the flat-laid image of the wall surface;

[0141] Count the number of crack points in F structural cracks and the total number of pixel points in the flat-laid image of the wall surface one by one, and compare the number of crack points with the total number of pixel points in the flat-laid image of the wall surface to obtain F coverage area ratios;

[0142] The calculation formula of the coverage area ratio is: ;

[0143] In the formula, is the coverage area ratio of the f-th structural crack, is the number of crack points of the f-th structural crack, is the total number of pixel points in the flat-laid image of the wall surface.

[0144] High-risk level and low-risk level are used to represent the severity of the potential hazards brought by structural cracks to the wall surface, and are used as the result of distinguishing the risk level of structural cracks. Specifically, the high-risk level means that the risk level of the structural crack is relatively high, and the low-risk level means that the risk level of the structural crack is relatively low.

[0145] The methods for dividing the high-risk level and the medium-risk level are as follows:

[0146] Compare the extension length ratio and the coverage area ratio of F structural cracks with the length ratio threshold and the area ratio threshold respectively; the length ratio threshold is the maximum value of the extension length ratio when it is not recorded as dangerous data, so as to judge whether the extension length of the structural crack will pose a danger to the wall surface; the area ratio threshold is the maximum value of the coverage area ratio when it is not recorded as dangerous data, so as to judge whether the coverage area of the structural crack will pose a danger to the wall surface.

[0147] When the extension length ratio is greater than the length ratio threshold, it indicates that the crack length of the structural crack will cause serious harm to the wall surface, and then record the extension length ratio as dangerous data.

[0148] When the coverage area ratio is greater than the area ratio threshold, it indicates that the crack coverage area of the structural crack will cause serious harm to the wall surface, and then record the coverage area ratio as dangerous data.

[0149] Count the number of dangerous data in F structural cracks one by one. Classify the structural cracks with the number of dangerous data being 0 as the low-risk level, and classify the structural cracks with the number of dangerous data being 1 or 2 as the high-risk level.

[0150] After classifying into the high-risk level and the low-risk level, it is necessary to formulate corresponding prompt information for the identification results of structural cracks at different levels, that is: early warning prompt information, so as to help maintenance personnel quickly and accurately master the identification results of each structural crack on the wall surface.

[0151] The early warning prompt information includes the information of "high risk, please repair immediately" and the information of "low risk, can be repaired timely".

[0152] Specifically, when the structural crack is at the high-risk level, at this time, the structural crack needs to be repaired immediately, and then formulate the information of "high risk, please repair immediately".

[0153] When the structural crack is at the low-risk level, at this time, the structural crack does not need to be repaired immediately, and then formulate the information of "low risk, can be repaired timely".

[0154] It should be noted that when formulating high-risk, please inspect immediately information, maintenance personnel can convert the specific position of the high-risk structural cracks in the wall tile image to the corresponding position on the wall of the subway station, thereby locating the exact position of the high-risk structural cracks, making it convenient for maintenance personnel to carry out subsequent filling, pouring and other operations on the high-risk structural cracks. When formulating low-risk, timely maintenance information, maintenance personnel can also locate its exact position and carry out filling, pouring and other operations on it at the appropriate time.

[0155] In this embodiment, by setting identification points distributed at intervals, it is possible to ensure that the left and right edges of two adjacent sampling images have a certain overlapping area, so that there is an overlapping associated area between the two adjacent sampling images, avoiding the phenomenon that the left and right edge positions of the sampling images are blurred due to shooting environment factors, and by segmenting the overlapping area, the overlapping area in the identification image can be accurately segmented, ensuring that the identification image can be in a complete and continuous tiled state during subsequent splicing, providing a good image basis for the subsequent identification and drawing of crack areas, and at the same time, through the structural safety identification of the strike characteristics and span characteristics, the structural features with hidden dangers in the subway station can be effectively identified and distinguished, and combined with the accurate classification of the danger level of the structural cracks, the image recognition accuracy of the crack area and type on the wall of the cover-cut reverse subway station can be improved, thereby providing theoretical support for the subsequent structural crack repair operation.

[0156] Example 2: Please refer to Figure 3 As shown, for parts not described in detail in this embodiment, please refer to the description of the first embodiment. A crack image recognition method for a cover-excavation reverse construction subway station is provided, which is applied to an image recognition center and is implemented based on a crack image recognition system for a cover-excavation reverse construction subway station, including:

[0157] S1: Initial points for capturing initial images by the acquisition terminal are pre-set in the subway station. Based on the point adjustment criteria, the initial points are horizontally adjusted to recognition points with interval distribution.

[0158] S2: Obtain a sample image of the wall at the recognition point, perform region segmentation on the overlapping areas in the sample image based on the region segmentation criterion, and generate a recognition image;

[0159] S3: Mark the crack points in the recognition image, enclose the crack points into a local crack by connecting the lines, and splice the recognition image into a wall tile image, and draw the crack area in the wall tile image;

[0160] S4: Collect the strike and span characteristics of the crack area, perform structural safety assessment on the strike and span characteristics based on the structural safety criteria, and identify structural cracks in the crack area;

[0161] S5: Collect the proportion of the extension length and the proportion of the coverage area of the structural cracks, classify the structural cracks into high-risk levels and low-risk levels, and formulate corresponding warning prompt information.

[0162] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention.

Claims

1. A crack image recognition system for top-down inverse construction subway stations, which is applied to an image recognition center, is characterized in that, Including: A point position adjustment module, which pre-sets an initial point position for the acquisition terminal to capture an initial image within a subway station, and horizontally adjusts the initial point position to recognition point positions distributed at intervals based on the point position adjustment criterion. The point position adjustment criterion is: the area of the overlapping region of two adjacent initial images is 5% of the total area of the initial image. An image segmentation module, which acquires a sampled image of the wall surface at the recognition point position, and performs region segmentation on the overlapping region in the sampled image based on the region segmentation criterion to generate a recognition image. The region segmentation criterion is: taking the vertical center line of the overlapping region as a reference, segmenting the region outside the vertical center line. The method for generating the recognition image is as follows: Controlling the electric control trolley to move to B recognition point positions in sequence, and successively capturing wall surface images through a camera to obtain B sampled images. Drawing lines along the upper edge and the lower edge positions of the overlapping region in the sampled image to respectively obtain an upper boundary line and a lower boundary line. Using computer vision technology to identify the midpoints of the upper boundary line and the lower boundary line in the overlapping region one by one, and after vertically connecting the two midpoints of the same overlapping region, obtaining a vertical center line. Denoting the region on the left side of the overlapping region in the sampled image and to the left of the vertical center line as the left region, and denoting the region on the right side of the overlapping region in the sampled image and to the right of the vertical center line as the right region. Respectively segmenting the left region and the right region of the B sampled images, and sequentially numbering the B segmented sampled images in ascending order with 1 as the first number according to the shooting sequence to obtain B recognition images. A region drawing module, which marks the crack points in the recognition image, connects the crack points to enclose a local crack, and splices the recognition images into a wall surface tiled image, and draws a crack region in the wall surface tiled image. The method for enclosing the local crack is as follows: Converting the B recognition images into B grayscale images, and respectively marking the brightness values of all pixel points in the B grayscale images, and denoting the pixel points with brightness values less than the calibrated brightness value as crack points to obtain C crack points. Respectively drawing circles with the C crack points as the centers and one-third of the length of the upper boundary line as the radius to obtain C pixel circles, and respectively counting the total number of pixel points and the number of crack points in the C pixel circles. After comparing the number of crack points in the C pixel circles with the total number of pixel points respectively, obtaining C crack ratios, and denoting the crack points corresponding to the centers with crack ratios greater than the calibrated crack ratio as boundary points to obtain D boundary points. Measuring the distance values between adjacent two boundary points one by one, connecting the boundary points with distance values less than the calibrated distance value to enclose, and denoting the enclosed region as a local crack. A crack recognition module, which acquires the trend characteristics and span characteristics of the crack region, conducts structural safety determination on the trend characteristics and span characteristics based on the structural safety criterion, and identifies structural cracks from the crack region. An identification prompt module, which acquires the extension length ratio and the coverage area ratio of the structural crack, classifies the structural crack into a high-risk level and a low-risk level, and formulates corresponding warning prompt information.

2. The crack image recognition system for a top-down inverse construction subway station according to claim 1, characterized in that, The acquisition terminal includes a camera and an electric control trolley. The method for setting the initial point position is as follows: Turn on the camera to keep it running normally, adjust the pitch angle of the camera, and keep the upper edge and lower edge of the wall within the camera's shooting range; Control the electric control trolley to move towards the direction of the wall at the nearest position until the upper edge and lower edge of the camera's shooting range are horizontally coincident with the upper edge and lower edge of the wall respectively, and then stop moving. Record the position of the electric control trolley after movement as the original point; Measure the straight-line distance from the original point to the wall to obtain the wall point distance value. Taking a distance equal to one wall point distance value from the wall as the drawing standard, draw a point line of a circular structure on the ground of the subway station; Query the corresponding wall width of the camera's shooting range, and mark subsequent points distributed at intervals on the point line with a spacing less than one wall width. After summarizing the original point and all subsequent points, obtain A initial points; 3. The crack image recognition system for a top-down and inverse construction subway station according to claim 2, wherein, The method for identifying points is as follows: According to the order of marking of the initial points, control the electric control trolley to move to the A initial points in sequence, and obtain A initial images by shooting the wall with the camera; Taking the a-th initial image as the reference image, continuously adjust the position of the (a + 1)-th initial point on the point line, measure the length and width of the overlapping area between the (a + 1)-th initial image and the reference image in real time, and calculate the area of the overlapping area in combination with the rectangle area formula; When the area of the overlapping area is 5% of the total area of the initial image, stop adjusting the position of the (a + 1)-th initial point. Until the initial image contains all positions of the wall, record the adjusted initial point as the identified point, and obtain B identified points; 4. The crack image recognition system for the top-down inverse construction subway station according to claim 3, characterized in that, The method for drawing the crack area is as follows: Import the B identified images into the image processing software in sequence and arrange them horizontally to the right, keeping the upper edge and lower edge of the B identified images horizontally aligned respectively; Adjust the positions of two adjacent identified images in the horizontal direction until the vertical center line on the right side of the b-th identified image coincides completely with the vertical center line on the left side of the (b + 1)-th identified image, and then stop adjusting to obtain a tiled wall image; Draw a line along the position of the boundary points of the local crack to form a closed drawing area, and record the closed drawing area as the crack area to obtain E crack areas; 5. The crack image recognition system for a top-down subway station according to claim 4, characterized in that, The method for collecting the trend feature and span feature is as follows: Draw a line along the lower edge position of the tiled wall image to obtain the lower boundary line of the wall; Measure one by one the point-line distance values from the crack points in the E crack areas to the lower boundary line of the wall. Respectively record the crack points corresponding to the minimum value and the maximum value of the point-line distance values as the starting point and the ending point, and connect the starting points and ending points of the E crack areas in sequence to obtain E trend lines; Draw reference lines horizontally at the starting points of the E crack areas respectively, and measure the inclination angles between the E trend lines and the E reference lines one by one to obtain E trend features; Measure the distances between the starting points and ending points in the E crack areas vertically one by one to obtain E span features; 6. The crack image recognition system for a top-down and reverse construction subway station according to claim 5, wherein The structural safety criterion is: Record the crack area with the crack structure feature as the structural crack; The method for identifying the structural crack is as follows: Compare the trend features of the E crack areas with the first trend range and the second trend range respectively; When the strike feature is within the first structural range or the second structural range, the strike feature is recorded as a crack structural feature; Compare the span features of the E crack regions with the span structure value; When the span feature is greater than the span structure value, the span feature is recorded as a crack structural feature; Count one by one the number of crack structural features in the E crack regions, and record the crack regions with the number of crack structural features being 1 or 2 as structural cracks to obtain F structural cracks.

7. The crack image recognition system for a top-down and inverse construction subway station according to claim 6, wherein, The acquisition methods for the extension length ratio and the coverage area ratio are as follows: Measure one by one the horizontal distance values between any two crack points in the F structural cracks and the length value of the wall tiled image in the horizontal direction, and compare the maximum value of the horizontal distance values with the length value of the wall tiled image to obtain F extension length ratios; Count one by one the number of crack points in the F structural cracks and the total number of pixel points in the wall tiled image, and obtain F coverage area ratios after comparing the number of crack points with the total number of pixel points in the wall tiled image.

8. A crack image recognition system for a top-down and reverse construction subway station according to claim 7, characterized in that, The classification methods for the high-risk level and the medium-risk level are as follows: Compare the extension length ratios and the coverage area ratios of the F structural cracks with the length ratio threshold and the area ratio threshold respectively; When the extension length ratio is greater than the length ratio threshold, record the extension length ratio as dangerous data; When the coverage area ratio is greater than the area ratio threshold, record the coverage area ratio as dangerous data; Count one by one the number of dangerous data in the F structural cracks, classify the structural cracks with the number of dangerous data being 0 as the low-risk level, and classify the structural cracks with the number of dangerous data being 1 or 2 as the high-risk level.

Images