Tunnel segment surface image processing method based on linear array camera

Abstract

The present invention discloses a method for processing the apparent image of tunnel segments based on a linear array camera, comprising: continuously collecting the apparent images of tunnel segments by using the linear array camera of a tunnel track detection robot; the forward speed of the tunnel track detection robot is L meters per second, and the line sampling rate of the linear array camera is S kilohertz; converting the apparent images of tunnel segments collected by the linear array camera to obtain the processed photo resolution w n ×h n , performing shift interpolation processing on any photo, including: generating a blank photo with a pixel resolution of w n ×h n , and the gray value of the pixels of the initial photo is 0; using the linear interpolation method to obtain the pixel values after movement of any row until the last row; splicing any photo to complete the processing of the photo. Through the above solution, the present invention has the advantages of strong pertinence, simple logic, low equipment input cost, etc.

Description

Technical Field

[0001] The invention relates to the technical field of tunnel image processing, in particular to a tunnel segment surface image processing method based on a linear array camera. Background Art

[0002] In subway and railway tunnels, due to the influence of external forces, vibrations and other factors, cracks and water seepage may occur in the tunnels. According to the requirements of the operating procedures, regular inspections of the tunnels are required to observe whether the tunnel dome segments are cracked, damaged or leaking, especially in subway tunnels. At present, tunnel inspections in my country mainly use manual inspection methods, which are time-consuming, labor-intensive and inefficient.

[0003] In addition, there are some intelligent equipment in the prior art that are equipped with cameras for automatic collection. However, the tunnel has very high requirements for the accuracy of image collection. For example, the subway tunnel inspection specification requires that the apparent crack width of the inspection segment reach 0.1mm, which puts forward very high requirements for the camera resolution. Therefore, the existing intelligent equipment adopts the method of stacking hardware, carrying multiple (usually 6-8) high-resolution linear array or area array cameras at a time. Each camera is responsible for taking pictures of the tunnel segment in a certain direction, and multiple cameras work simultaneously to achieve the purpose. However, this solution is loaded with more equipment, which is relatively bulky and the cost is also very high. It does not meet the requirements of rapid loading and unloading of rail equipment. For example, the Chinese invention patent with the patent application number "201811228215.3" and the name "A subway tunnel appearance inspection method" and the Chinese invention patent with the patent application number "201910387411.3" and the name "A subway tunnel structure comprehensive inspection vehicle", both technologies are equipped with multiple linear array or area array cameras, and they have the same problem.

[0004] Based on the above situation, the applicant has specially proposed a Chinese invention patent with patent application number "202010138111.4" and name "A tunnel track detection robot". A rotating drive motor, a conductive slip ring, a camera, a light source bracket and a vault camera are arranged on the vault acquisition component. It adopts rotation to collect the image of the tunnel. This technology proposes a full coverage acquisition method of the apparent image of the tunnel segment by a linear array camera. It does not require stacking hardware, but only requires one linear array camera. Through an ingenious rotating mechanism and combined with a movable loading platform, the camera can move in a spiral motion, rotating and moving forward at the same time to complete the full coverage acquisition of the apparent image of the entire tunnel segment.

[0005] However, the photos collected in this way have displacement in each line compared with the previous line because the linear array camera rotates and moves forward during the collection process, so special processing is required to splice them together. Therefore, it is urgent to propose a tunnel segment surface image processing method based on the spiral motion collection of the linear array camera. Summary of the invention

[0006] In view of the above problems, the object of the present invention is to provide a tunnel segment surface image processing method based on a line array camera. The technical solution adopted by the present invention is as follows:

[0007] The tunnel segment surface image processing method based on a linear array camera comprises the following steps:

[0008] The tunnel track detection robot is used to continuously collect the surface image of the tunnel segment by using the linear array camera; the tunnel track detection robot has a forward speed of L meters per second, the linear sampling rate of the linear array camera is S kilohertz; the lateral field of view captured by the linear array camera is V meters;

[0009] Convert the surface image of the tube segment collected by the linear array camera to obtain the processed photo resolution w n ×h n ,in

[0010]

[0011] h n =h c

[0012] Among them, w c h is the original photo resolution width of the tube segment surface image, c represents the original photo resolution height of the tube segment surface image, Δ represents the actual displacement of each row on the photo relative to the previous row, and a represents the actual distance between adjacent pixels in each row on the photo, where Δ=L / (S*1000)m=L / Smm, a=V / w c m=(1000*V) / w c mm;

[0013] Performing shift interpolation processing on any photo includes the following steps:

[0014] Generate a pixel resolution of w n ×h n A blank photo, and the grayscale value of all pixels in the initial photo is 0;

[0015] Use interpolation to obtain the pixel value after the shift of any row until the last row;

[0016] Stitch any photo together to complete your photo processing.

[0017] Furthermore, the method of obtaining the pixel value of any row after the movement by using the interpolation method comprises the following steps:

[0018] Taking the first row of the photo as the reference, find the right-shifted pixels of the kth row of the photo. The expression is:

[0019]

[0020] Where k is greater than 1 and less than or equal to h c The natural number of

[0021] When offset is an integer, the kth row of the photo is shifted right by offset pixels, and the grayscale value of the left pixel of the kth row of the photo after the shift is set to 0;

[0022] When offset is a decimal, linear interpolation is performed on the k rows of the converted photo. The specific steps are as follows:

[0023] Round the offset down to get the integer O f , round up the offset to get the integer O c , the integer O c =O f +1;

[0024] Set the first O of the kth row of the converted photo f The pixel gray value corresponding to the point is 0;

[0025] Assign the gray value of the first point in the kth row of the original photo to the gray value of the oth point in the kth row of the converted photo. c points;

[0026] From the kth row, c +1 point to start calculating grayscale interpolation, the Oth point of the kth row c +i point gray value is d·G i+1 +(1-d)·G i ; The G i is the gray value of the i-th point in the k-th row of the original photo; i+1 is the gray value of the i+1th point in the kth row of the original image; d is the decimal part of offset, that is, d=offset-O f i is greater than or equal to 1 and less than w c A natural number.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] The present invention cleverly converts any photo to be suitable for image stitching that is synchronized with the movement of the tunnel track detection robot and the rotation of the linear array camera, providing a guarantee for the stitched image; in addition, the present invention uses linear interpolation to interpolate any row of any photo, and obtains images synchronized with the movement of the tunnel track detection robot and the rotation of the linear array camera, so as to solve the technical problem of large investment in traditional tunnel detection stacking camera detection equipment; in this way, the present invention can realize the detection and stitching of tunnel segment surface images with one camera. In summary, the present invention has the advantages of simple logic and low equipment investment cost, and has high practical value and promotion value in the field of tunnel image processing technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope of protection. For those skilled in the art, other related drawings can be obtained based on these drawings without creative work.

[0030] Figure 1 It is a schematic diagram for comparing the photos before and after the conversion of the present invention. DETAILED DESCRIPTION

[0031] In order to make the purpose, technical scheme and advantages of this application clearer, the present invention is further described below in conjunction with the accompanying drawings and embodiments, and the embodiments of the present invention include but are not limited to the following embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0032] Example

[0033] like Figure 1 As shown, this embodiment provides a tunnel segment surface image processing method based on a line array camera; in this embodiment, the actual width corresponding to the field of view of the line array camera is V meters (V is greater than 0), the line sampling rate of the line array camera is S kilohertz, the forward speed of the mobile platform is L meters per second (L is greater than 0), and the line array camera is set to sample h c A photo is generated after the line, that is, the photo pixel height is h c , the resolution of each photo is w c ×h cIn this embodiment, each photo is converted and interpolated, and finally stitched into a large picture. Since the mobile platform moves forward every time the linear array camera samples a line, each row on the photo should have an actual displacement relative to the previous row. Let this displacement be Δ, then Δ=L / (S*1000)m=L / Smm. On the imaging plane of the camera, each row actually needs to be shifted to the right by a corresponding distance relative to the previous row. This distance can be expressed in pixels. The actual distance between adjacent pixels in each row on the photo is a, a=V / w c m=(1000*V) / w c mm, then theoretically, each row on the phase plane needs to be moved right by Δ / a pixels relative to the previous row. In practical applications, Δ is usually smaller than a, so the relative movement of each row is usually less than one pixel, and each row of pixels needs to be interpolated.

[0034] In this embodiment, the specific steps of performing conversion interpolation processing on each photo are as follows:

[0035] The first step is to calculate the resolution of the processed photo, assuming the original photo resolution is w c ×h c , the resolution of the processed photo is w n ×h n ,but ,h n =h c , that is, the photo height remains unchanged, the width is widened, and the last row of the photo needs to be moved right by h relative to the first row c -1 times, so the width needs to be increased by the above pixels and rounded up to ensure complete inclusion.

[0036] The second step is to generate a pixel resolution of w n ×h n For a blank photo, the grayscale value of all pixels is initially set to 0;

[0037] The second step is to calculate the new pixel value after each row shift, starting from the second row. Assume that the current row is the kth row (k>1 and k≤h c ), theoretically, the kth row needs to be shifted right relative to the first row pixels. This number is usually a decimal rather than an integer. When it is an integer, each pixel is directly shifted to the right without interpolation. The brightness of the pixels on the left after the right shift can be set to 0. When the calculated offset is a decimal, the brightness of each pixel in the kth row of the processed image needs to be interpolated. The specific interpolation method is as follows:

[0038] (1) The offset is rounded up and down respectively to obtain two integers: f and O c , where Of is the value rounded down, O c is the value rounded up. Then O c =O f +1;

[0039] (2) On the processed photo, add the O f The gray value corresponding to each point is set to 0;

[0040] (3) Assign the gray value of the first point in the row of the original image to the gray value of the Oth point in the row of the processed image. c points (the first point on the left is not interpolated);

[0041] (4) From the processed image c +1 point to start calculating grayscale interpolation, and set the current calculation point to be the Oth point of the processed image c +i points, where i≥1 and i<w c , let the gray value of the i-th point of the original image be G i , then the processed image O c +i pixel gray value is d·G i+1 +(1-d)·G i , where d is the fractional part of offset, i.e. d = offset - O f .

[0042] Repeat step 3 for each row until the last row.

[0043] In this embodiment, the system is used in a subway tunnel in Tianjin. The radius of the subway tunnel is 2.77 meters. A DALSA 4K linear array camera is selected and equipped with a suitable telephoto lens so that the camera's field of view width at 2.77 meters is 0.8 meters. Assuming that the moving platform's forward direction is the x-axis, the camera's shooting accuracy in the x-direction is 0.2 mm when the lateral resolution is 4000. The cross-sectional circumference of the subway tunnel is 2×π×2.77≈17.404m=17404mm. In order to achieve an accuracy of sampling every 0.2 mm in the direction of rotation around the x-axis, the linear array camera needs to sample 17404÷0.2≈87022 times for one rotation. In order to ensure full coverage acquisition, the platform's forward distance should be less than 0.8 meters when the linear array camera rotates one circle. In order to ensure a certain overlap rate, the platform is set to move forward 0.6 meters. The maximum sampling rate of a 4k line array camera is 80kHz. Under stable conditions, it is allowed to operate at a sampling rate of 70kHz. At this time, the time required for the entire rotating system to rotate one circle is: 87022 / 70000≈1.243 seconds, and the forward speed of the entire mobile platform is 60 / 1.243=49.5cm / s=0.495m / s.

[0044] Then we have L = 0.495, S = 70, V = 0.8. Assume that the camera generates a photo every 1000 lines sampled, that is, h c =1000, then the resolution of the photo is 4000×1000. The actual displacement of each row of the photo relative to the previous row is Δ=L / S=0.495 / 70≈0.00707 mm, and the actual distance corresponding to adjacent pixels is a=(1000*0.8) / 4000=0.2 mm, so the pixel offset of each row of the photo relative to the previous row is Δ / a=0.00707 / 0.2=0.03535, which is a very small decimal, so linear interpolation is necessary.

[0045] The above embodiments are only preferred embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any changes that adopt the design principles of the present invention and are made through non-creative labor on this basis should fall within the protection scope of the present invention.

Claims

1. A tunnel segment surface image processing method based on a linear array camera, characterized in that: The following steps are involved: The tunnel track detection robot uses a linear array camera to continuously collect the surface image of the tunnel segment; the tunnel track detection robot travels at a speed of L meters per second; the linear sampling rate of the linear array camera is S kilohertz; the lateral field of view captured by the linear array camera is V meters; Convert the surface image of the tube segment collected by the linear array camera to obtain the processed photo resolution w n ×h n ,in h n =h c Among them, w c h is the original photo resolution width of the tube segment surface image, c represents the original photo resolution height of the tube sheet surface image, Δ represents the actual displacement of each row on the photo relative to the previous row, and a represents the actual distance between adjacent pixels in each row on the photo, where Δ=L(S*1000)m=LSmm, a=Vw c m=(1000*V)w c mm; Performing shift interpolation processing on any photo includes the following steps: Generate a pixel resolution of w n ×h n A blank photo, and the grayscale value of all pixels in the initial photo is 0; Use interpolation to obtain the pixel value after the shift of any row until the last row; Stitch any photo together to complete the photo processing; The method of obtaining the pixel value of any row after the movement by using the interpolation method comprises the following steps: Taking the first row of the photo as the reference, find the right-shifted pixels of the kth row of the photo. The expression is: Where k is greater than 1 and less than or equal to h c The natural number of When offset is an integer, the kth row of the photo is shifted right by offset pixels, and the grayscale value of the left pixel of the kth row of the photo after the shift is set to 0; When offset is a decimal, linear interpolation is performed on the k rows of the converted photo. The specific steps are as follows: Round the offset down to get the integer O f , round up the offset to get the integer O c , the integer O c =O f +1; Set the first O of the kth row of the converted photo f The pixel gray value corresponding to the point is 0; Assign the gray value of the first point in the kth row of the original photo to the gray value of the oth point in the kth row of the converted photo. c points; From the kth row, c +1 point to start calculating grayscale interpolation, the Oth point of the kth row c +i point gray value is d·G i+1 +(1-d)·G i ; said G i is the gray value of the i-th point in the k-th row of the original photo; i+1 is the gray value of the i+1th point in the kth row of the original image; d is the decimal part of offset, that is, d=offset-O f , i is greater than or equal to 1 and less than w c A natural number.

Images