An analysis method for bridge disease detection

Abstract

The application belongs to the field of detection, and discloses an analysis method for bridge disease detection, comprising the following steps: S1, respectively calculating the number of sensitivities of each shooting area on the surface of the bridge; S2, shooting each shooting area according to the number of sensitivities to obtain a set of images of each shooting area; S3, performing multi-frame synthesis processing on the images in the set to obtain a synthesis image of each shooting area; and S4, inputting the synthesis image into a pre-trained neural network model for recognition to obtain a recognition result. The application improves the quality of the obtained synthesis image, thereby being beneficial to improving the accuracy of disease detection on the surface of the bridge.

Description

An analytical method for bridge defect detection Technical Field

[0001] This invention relates to the field of testing, and more particularly to an analytical method for detecting bridge defects. Background Technology

[0002] Traditional bridge defect detection relies on visual inspection, which is inefficient and extremely difficult, especially for inspecting the sides and bottom of bridges. Existing technologies utilize drones to capture images of the bridge surface, then employ image recognition technology to identify defects. However, current techniques capture only a single image of each area of ​​the bridge surface. Since lighting conditions vary across different areas, obtaining a single image may result in poor image quality in some areas, reducing the probability of accurate detection results. Summary of the Invention

[0003] The purpose of this invention is to disclose an analytical method for detecting bridge defects, thereby solving the problem of how to improve the accuracy of detecting defects on the surface of bridges.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] This invention provides an analytical method for detecting bridge defects, comprising:

[0006] S1, calculate the sensitivity of each shooting area on the bridge surface;

[0007] S2, take pictures of each shooting area according to the number of ISO values, and obtain a collection of images of each shooting area;

[0008] S3, perform multi-frame synthesis processing on the images in the collection to obtain a composite image of each shooting area;

[0009] S4. Input the synthesized image into a pre-trained neural network model for recognition and obtain the recognition result.

[0010] Preferably, the process of acquiring the imaging area on the bridge surface is as follows:

[0011] S11, Obtain the 3D model of the bridge;

[0012] S12, Obtain planar images of the 3D model from various perspectives according to preset rules;

[0013] S13, each planar image is divided into multiple shooting areas.

[0014] Preferably, the planar images of the 3D model from various perspectives are obtained according to preset rules, including:

[0015] Obtain the front view, rear view, left view, right view, top view, and bottom view of the 3D model respectively, and project the front view, rear view, left view, right view, top view, and bottom view as planar images of the 3D model onto a 2D plane.

[0016] Preferably, each planar image is divided into multiple shooting areas, including:

[0017] S131, obtain the minimum bounding rectangle of the planar image;

[0018] S132, divide each circumscribed rectangle into multiple sub-regions of the same size;

[0019] S133, determine whether the obtained sub-region contains pixels belonging to the bridge. If not, delete the sub-region and use the area corresponding to the remaining sub-region in the real bridge as the shooting area.

[0020] Preferably, the size of the sub-region is calculated as follows:

[0021] The function for calculating the length of a subregion is:

[0022]

[0023] The function for calculating the width of a subregion is:

[0024]

[0025] Where, silenth and siwidth represent the length and width of the sub-region, respectively; scal represents the scaling ratio of the 3D model; shtfoglen represents the focal length used when capturing the area; eyefoglen represents the standard lens focal length; shtdist represents the distance between the capturing device and the bridge surface when capturing the area; bsdist represents the preset safety distance; subscal represents the preset adjustment ratio; lenth and width represent the length and width of the image captured by the capturing device at a distance of shtdist from the bridge and a focal length of shtfoglen in the world coordinate system, respectively; and α1 and α2 represent the preset weight values.

[0026] Preferably, the method for obtaining the number of photosensitivity values ​​is as follows:

[0027] The sensitivity of the shooting area is calculated using a fixed calculation cycle. Once the sensitivity is calculated, the shooting area is photographed.

[0028] The function for calculating the sensitivity is:

[0029]

[0030] Among them, shocoef h and shocoef h-1 numdef represents the number of photosensitivity values ​​calculated in the h-th and h-1-th calculation cycles, respectively. h-1 and numdef h-2 These represent the number of bridge defects detected in the synthetic images corresponding to the captured area during the (h-1)th and (h-2)th calculation cycles, respectively. std This indicates the preset first quantity.

[0031] Preferably, images are taken of each shooting area according to the sensitivity level, resulting in a set of images for each shooting area, including:

[0032] S21 controls the shooting equipment to fly to the shooting area;

[0033] S22, controls the main optical axis of the lens carried by the shooting equipment to be perpendicular to the center of the shooting area;

[0034] S23 acquires the ambient brightness value and automatic ISO value of the shooting area through the shooting device;

[0035] S24, the set of sensitivity is calculated by the shooting device based on the ambient brightness value, the automatic ISO value and the number of sensitivity values;

[0036] S24, the shooting device takes pictures of the shooting area using each sensitivity in the sensitivity set, and obtains a set of images of the shooting area.

[0037] Preferably, the shooting device calculates a set of ISO sensitivity based on ambient brightness, automatic ISO value, and the number of ISO values, including:

[0038] Get the number of elements in the sensitivity set:

[0039]

[0040] Where numfele represents the number of elements in the sensitivity set, shocoef represents the number of sensitivity values ​​obtained in S1, envbri represents the ambient brightness value, stdbri represents the preset ambient brightness baseline value, and shocoef ste This indicates the preset second quantity;

[0041] If numfele is even, the elements in the sensitivity set are calculated using the following function:

[0042]

[0043] If numfele is odd, the elements in the sensitivity set are calculated using the following function:

[0044]

[0045] Among them, iso k This represents the k-th element in the sensitivity set, where k ∈ [1, numfele].

[0046] This invention first obtains the photosensitivity of the shooting area, and then controls the number of images to be captured in that area based on the photosensitivity, thus avoiding capturing the same number of images for all shooting areas. The photosensitivity can be determined according to the actual situation of the shooting area and the previous damage conditions, making the photosensitivity adaptable to the actual situation of the shooting area. This ensures that different shooting areas have a sufficient number of images for multi-frame synthesis processing, improving the quality of the resulting synthesized images and thus improving the accuracy of surface damage detection on bridges. Attached Figure Description

[0047] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 is a schematic diagram of an analytical method for detecting bridge defects according to the present invention.

[0049] Figure 2 is a schematic diagram of the process of acquiring the imaging area on the bridge surface according to the present invention. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] As shown in Figure 1, in one embodiment, the present invention provides an analytical method for detecting bridge defects, including...

[0052] S1, calculate the sensitivity of each shooting area on the bridge surface;

[0053] S2, take pictures of each shooting area according to the number of ISO values, and obtain a collection of images of each shooting area;

[0054] S3, perform multi-frame synthesis processing on the images in the collection to obtain a composite image of each shooting area;

[0055] S4. Input the synthesized image into a pre-trained neural network model for recognition and obtain the recognition result.

[0056] This invention first obtains the photosensitivity of the shooting area, and then controls the number of images to be captured in that area based on the photosensitivity, thus avoiding capturing the same number of images for all shooting areas. The photosensitivity can be determined according to the actual situation of the shooting area and the previous damage conditions, making the photosensitivity adaptable to the actual situation of the shooting area. This ensures that different shooting areas have a sufficient number of images for multi-frame synthesis processing, improving the quality of the resulting synthesized images and thus improving the accuracy of surface damage detection on bridges.

[0057] Preferably, as shown in Figure 2, the process of acquiring the imaging area on the bridge surface is as follows:

[0058] S11, Obtain the 3D model of the bridge;

[0059] S12, Obtain planar images of the 3D model from various perspectives according to preset rules;

[0060] S13, each planar image is divided into multiple shooting areas.

[0061] Specifically, the 3D model of the bridge is created by acquiring point cloud data according to a pre-defined method, and the resulting 3D model can adaptively change size according to the scaling ratio.

[0062] The scaling ratio is calculated using the following function:

[0063]

[0064] zomrat represents the scaling ratio, aimlength represents the length of the 3D model, and reallength represents the actual length of the bridge.

[0065] For example, the scaling ratio could be That is, the unit length in the 3D model is one ten-thousandth of the actual length of the bridge.

[0066] Preferably, the planar images of the 3D model from various perspectives are obtained according to preset rules, including:

[0067] Obtain the front view, rear view, left view, right view, top view, and bottom view of the 3D model respectively, and project the front view, rear view, left view, right view, top view, and bottom view as planar images of the 3D model onto a 2D plane.

[0068] Specifically, the front view can be an image from any side of the bridge, while the other side is the rear view. The top view is an image from the road surface above the bridge, and the bottom view is an image from the bottom of the bridge. The left and right views are images from the left and right sides of the bridge, respectively.

[0069] Since this invention only detects the surface, only a planar image is needed.

[0070] Preferably, each planar image is divided into multiple shooting areas, including:

[0071] S131, obtain the minimum bounding rectangle of the planar image;

[0072] S132, divide each circumscribed rectangle into multiple sub-regions of the same size;

[0073] S133, determine whether the obtained sub-region contains pixels belonging to the bridge. If not, delete the sub-region and use the area corresponding to the remaining sub-region in the real bridge as the shooting area.

[0074] Specifically, after obtaining the sub-region, the corresponding region in the 3D model can be transformed from the coordinate system of the 3D model to the coordinate system of the real world, thereby obtaining the shooting area in the bridge.

[0075] By deleting sub-regions that do not contain pixels of the bridge, invalid shooting areas can be avoided, thereby reducing the probability of obtaining invalid images.

[0076] Preferably, the size of the sub-region is calculated as follows:

[0077] The function for calculating the length of a subregion is:

[0078]

[0079] The function for calculating the width of a subregion is:

[0080]

[0081] Where, silenth and siwidth represent the length and width of the sub-region, respectively; scal represents the scaling ratio of the 3D model; shtfoglen represents the focal length used when capturing the area; eyefoglen represents the standard lens focal length; shtdist represents the distance between the capturing device and the bridge surface when capturing the area; bsdist represents the preset safety distance; subscal represents the preset adjustment ratio; lenth and width represent the length and width of the image captured by the capturing device at a distance of shtdist from the bridge and a focal length of shtfoglen in the world coordinate system, respectively; and α1 and α2 represent the preset weight values.

[0082] In this invention, the size of the sub-region is not preset. Presetting the sub-region size would limit the scaling ratio, the parameters of the shooting device, and the shooting distance, making it difficult to adjust according to actual needs. Therefore, this invention comprehensively determines the size of the sub-region by considering three variables: the scaling ratio, the focal length used when shooting the area, and the distance between the shooting device and the bridge surface. This greatly improves the adaptability of the invention during implementation, allowing the focal length and the distance between the shooting device and the bridge surface to be set according to actual conditions.

[0083] Specifically, the world coordinate system is the coordinate system in the real world. When the shooting distance and angle are determined, the size of the image captured by the lens in the real world is also determined.

[0084] Specifically, the filming equipment could be a drone equipped with a camera.

[0085] Preferably, the method for obtaining the number of photosensitivity values ​​is as follows:

[0086] The sensitivity of the shooting area is calculated using a fixed calculation cycle. Once the sensitivity is calculated, the shooting area is photographed.

[0087] The function for calculating the sensitivity is:

[0088]

[0089] Among them, shocoef h and shocoef h-1 numdef represents the number of photosensitivity values ​​calculated in the h-th and h-1-th calculation cycles, respectively. h-1 and numdef h-2 These represent the number of bridge defects detected in the synthetic images corresponding to the captured area during the (h-1)th and (h-2)th calculation cycles, respectively.std This indicates the preset first quantity.

[0090] Specifically, when h is less than or equal to 2, the sensitivity is a preset value.

[0091] In this invention, the sensitivity is calculated based on the relationship between the number of bridge defects present in the shooting area during the first two calculation cycles. If numdef h-1 Less than numdef h-2 This indicates that the number of defects in the bridge has decreased, thus reducing the amount of light sensitivity; if numdef h-1 Equal to numdef h-2 If the sensitivity is reduced, the amount remains unchanged; if numdef h-1 Greater than numdef h-2 This indicates that the number of defects in the bridge has increased, thereby increasing the amount of light sensitivity to increase the number of images in the corresponding shooting area and obtain a higher quality composite image.

[0092] Specifically, a calculation cycle refers to the time period from when the sensitivity value is calculated in this calculation to when the sensitivity value is calculated again.

[0093] In this invention, after calculating the number of photosensitivity values, steps S2-S4 are performed sequentially. After step S4 is completed, the current calculation cycle is waited for to end.

[0094] The calculation cycle is set because bridge defects do not need to be detected frequently; they only need to be detected periodically. In this invention, the time interval between two consecutive bridge defects detections is approximately equal to one calculation cycle.

[0095] Preferably, images are taken of each shooting area according to the sensitivity level, resulting in a set of images for each shooting area, including:

[0096] S21 controls the shooting equipment to fly to the shooting area;

[0097] S22, controls the main optical axis of the lens carried by the shooting equipment to be perpendicular to the center of the shooting area;

[0098] S23 acquires the ambient brightness value and automatic ISO value of the shooting area through the shooting device;

[0099] S24, the set of sensitivity is calculated by the shooting device based on the ambient brightness value, the automatic ISO value and the number of sensitivity values;

[0100] S24, the shooting device takes pictures of the shooting area using each sensitivity in the sensitivity set, and obtains a set of images of the shooting area.

[0101] In this invention, the shooting device can be a drone. When shooting, the drone first flies to a position perpendicular to the center of the shooting area, maintaining a distance from the bridge surface corresponding to the shooting area. After acquiring the ambient brightness value and the automatic ISO value, this invention can calculate a set of ISO values ​​by combining the number of ISO values. The shooting area is then photographed using elements from this set of ISO values, with one image captured for each ISO value. During shooting, all parameters related to the shooting process remain unchanged except for the ISO value, thus obtaining a set of images of the shooting area.

[0102] Preferably, the shooting device calculates a set of ISO sensitivity based on ambient brightness, automatic ISO value, and the number of ISO values, including:

[0103] Get the number of elements in the sensitivity set:

[0104]

[0105] Where numfele represents the number of elements in the sensitivity set, shocoef represents the number of sensitivity values ​​obtained in S1, envbri represents the ambient brightness value, stdbri represents the preset ambient brightness baseline value, and shocoef ste This indicates the preset second quantity;

[0106] If numfele is even, the elements in the sensitivity set are calculated using the following function:

[0107]

[0108] If numfele is odd, the elements in the sensitivity set are calculated using the following function:

[0109]

[0110] Among them, iso k This represents the k-th element in the ISO sensitivity set, where k ∈ [1, numfele], and iso std This indicates the amount of change in the preset photosensitivity.

[0111] Specifically, the number of elements in the sensitivity set is related to the ambient brightness and the number of sensitivity values ​​obtained in S1. When the number of sensitivity values ​​obtained in S1 remains constant, the higher the ambient brightness, the smaller the number of elements in the sensitivity set, indicating that the lighting conditions in the shooting area are sufficient and a large number of images are not needed for multi-frame synthesis to improve the quality of the synthesized image. Conversely, a larger number of images are acquired to improve the quality of the synthesized image. This setting avoids taking the same number of images for all shooting areas, thus effectively ensuring the quality of the synthesized image obtained in each shooting area while reducing the total number of images obtained from shooting all shooting areas. This reduces the computational pressure in S4 and facilitates faster recognition results.

[0112] Preferably, the identification results are the types of bridge defects contained in the synthetic image, the location of each bridge defect, and the number of each type of bridge defect.

[0113] Preferably, the types of bridge defects include cracks, voids, pitting, honeycombing, exposed reinforcement, and potholes.

[0114] Preferably, the images in the set are subjected to multi-frame synthesis processing to obtain a synthesized image of each shooting area, including:

[0115] Calculate the image score for each image in the set;

[0116] Multi-frame synthesis is performed based on image scores to obtain a synthesized image.

[0117] Preferably, the image score can be obtained using algorithms such as the BRISQUE algorithm or the NIMA algorithm.

[0118] Preferably, multi-frame synthesis is performed based on image scores to obtain a synthesized image, including:

[0119] Use the following function to perform multi-frame synthesis:

[0120]

[0121] Where compimg represents the synthesized image, scor img The image score is represented by `img`, `imgset` represents the set obtained in S2, and `img` represents an element in `imgset`.

[0122] Specifically, in this invention, the images are not superimposed in the same proportion during image synthesis. Instead, they are superimposed based on the image scores. The higher the image score, the greater the proportion of the image referenced in the synthesized image. This results in a larger proportion of information from high-quality images in the synthesized image, effectively improving the quality of the synthesized image.

[0123] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. An analytical method for detecting bridge defects, characterized in that, include: S1, calculate the sensitivity of each shooting area on the bridge surface; the sensitivity is obtained by calculating the sensitivity of the shooting area using a fixed calculation cycle, and then taking a picture of the shooting area after the sensitivity is calculated; the calculation function for the sensitivity is: Among them, shocoef h and shocoef h-1 numdef represents the number of photosensitivity values ​​calculated in the h-th and h-1-th calculation cycles, respectively. h-1 and numdef h-2 These represent the number of bridge defects detected in the synthetic images corresponding to the captured area during the (h-1)th and (h-2)th calculation cycles, respectively. std S2, based on the preset first quantity, takes pictures of each shooting area according to the sensitivity, obtaining a set of images for each shooting area; S3, performs multi-frame synthesis processing on the images in the set to obtain a synthesized image for each shooting area, including: calculating the image score of each image in the set; performing multi-frame synthesis based on the image scores to obtain a synthesized image, including: using the following function for multi-frame synthesis: Where compimg represents the synthesized image, scor img S2 represents the image score of img, imgset represents the set obtained in S2, and img represents the element in imgset; S4, the synthesized image is input into a pre-trained neural network model for recognition to obtain the recognition result.

2. The analytical method for detecting bridge defects according to claim 1, characterized in that, The process of acquiring the shooting area on the bridge surface is as follows: S11, acquire the three-dimensional model of the bridge; S12, acquire planar images of the three-dimensional model from various perspectives according to preset rules; S13, divide each planar image into multiple shooting areas.

3. The analytical method for detecting bridge defects according to claim 2, characterized in that, According to preset rules, obtain planar images of the 3D model from various perspectives, including: obtaining the front view, rear view, left view, right view, top view and bottom view of the 3D model respectively, and projecting the front view, rear view, left view, right view, top view and bottom view as planar images of the 3D model onto a 2D plane.

4. The analytical method for bridge defect detection according to claim 3, characterized in that, Each planar image is divided into multiple shooting areas, including: S131, obtaining the minimum bounding rectangle of the planar image; S132, dividing each bounding rectangle into multiple sub-regions of the same size; S133, determining whether the obtained sub-regions contain pixels belonging to the bridge, if not, deleting the sub-region, and using the area corresponding to the remaining sub-region in the real bridge as the shooting area.

5. The analytical method for detecting bridge defects according to claim 4, characterized in that, The size of a subregion is calculated as follows: The length of a subregion is calculated using the following function: The function for calculating the width of a subregion is: Where, silenth and siwidth represent the length and width of the sub-region, respectively; scal represents the scaling ratio of the 3D model; shtfoglen represents the focal length used when capturing the area; eyefoglen represents the standard lens focal length; shtdist represents the distance between the capturing device and the bridge surface when capturing the area; bsdist represents the preset safety distance; subscal represents the preset adjustment ratio; lenth and width represent the length and width of the image captured by the capturing device at a distance of shtdist from the bridge and a focal length of shtfoglen in the world coordinate system, respectively; and α1 and α2 represent the preset weight values.

6. The analytical method for detecting bridge defects according to claim 1, characterized in that, The process involves capturing images of each shooting area based on the sensitivity level, resulting in a set of images for each shooting area. This includes: S21, controlling the shooting device to fly towards the shooting area; S22, controlling the main optical axis of the lens carried by the shooting device to be perpendicular to the center of the shooting area; S23, acquiring the ambient brightness value and auto ISO value of the shooting area through the shooting device; S24, calculating the sensitivity set by the shooting device based on the ambient brightness value, auto ISO value, and sensitivity level; and S25, capturing images of the shooting area using each sensitivity level in the sensitivity set to obtain a set of images of the shooting area.

7. The analytical method for detecting bridge defects according to claim 1, characterized in that, The shooting device calculates the ISO set based on the ambient brightness value, the auto ISO value, and the number of ISO values, including: obtaining the number of elements in the ISO set: Where numfele represents the number of elements in the sensitivity set, shocoef represents the number of sensitivity values ​​obtained in S1, envbri represents the ambient brightness value, stdbri represents the preset ambient brightness baseline value, and shocoef ste This represents the preset second quantity; if numfele is even, the elements in the sensitivity set are calculated using the following function: If numfele is odd, the elements in the sensitivity set are calculated using the following function: Among them, iso k This represents the k-th element in the sensitivity set, where k ∈ [1, numfele].

Images