Bridge deepwater cofferdam construction quality detection method and system based on machine vision

Abstract

The invention relates to the technical field of image data processing, and provides a bridge deepwater cofferdam construction quality detection method and system based on machine vision, and the method comprises the steps: obtaining the scale characteristics of an edge line according to a cofferdam image and a cofferdam scale transformation image; seam edge features are obtained according to the edge lines and the scale features thereof; obtaining a seam probability according to the edge lines and the seam edge features; obtaining a deformation degree according to the edge line; obtaining a temporary deformation degree and a temporary deformation index according to the deformation degree, the water pressure index and the joint probability; according to the temporary deformation index, the temporary deformation degree and the deformation degree, the sealing performance is obtained; and obtaining a bridge deepwater cofferdam construction quality detection result according to the sealing performance. The accurate quality detection result is obtained according to the sealing performance of the edge lines, and the accuracy of the bridge deepwater cofferdam construction quality detection result is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image data processing, and particularly to a method and system for detecting the construction quality of deep-water cofferdams for bridges based on machine vision. Background Art

[0002] A deep-water cofferdam for a bridge is a temporary structure used for underwater construction. By establishing a closed cofferdam area in water, it facilitates construction workers to carry out foundation engineering construction therein. To ensure the tightness and structural stability of the cofferdam and guarantee the safety and engineering quality of bridge foundation construction, construction quality inspection is required. In the underwater construction environment, due to the influence of factors such as water flow and turbid water quality, traditional manual inspection methods are difficult to achieve efficient and accurate quality assessment. Therefore, using machine vision technology through high-precision image acquisition and processing can identify defects, joint tightness, and deformation conditions of the cofferdam structure in real time and accurately, effectively reducing human errors, improving inspection efficiency, and meeting the high standards for construction quality in the deep-water construction environment.

[0003] In the prior art, during the process of using image processing methods to detect the construction quality of cofferdams, when detecting the tightness of cofferdam joints, the underwater construction environment is complex. The impact of water flow not only has a physical effect on the cofferdam structure but may also cause bending, dislocation, or temporary deformation at the joints. The temporary deformation is dynamic and local and changes with the fluctuation of water flow. Traditional image processing methods may not be able to accurately identify the actual state of the joints, resulting in the system being unable to distinguish normal deformation from actual structural cracks, misjudging the temporarily deformed area as a crack or unqualified part, which may lead to incorrect quality assessment and reduce the accuracy of the detection results for the construction quality of deep-water cofferdams for bridges. Summary of the Invention

[0004] The present invention provides a method and system for detecting the construction quality of deep-water cofferdams for bridges based on machine vision to solve the problem of relatively low accuracy of the existing detection results for the construction quality of deep-water cofferdams for bridges. The specific technical solutions adopted are as follows:

[0005] The present invention proposes a method for detecting the construction quality of deep-water cofferdams for bridges based on machine vision. The method includes the following steps:

[0006] Obtain a plurality of cofferdam images and their water pressure indices;

[0007] Obtain a number of cofferdam scale transformation images for each cofferdam image, and obtain a number of edge lines in each cofferdam image and its cofferdam scale transformation images; according to the positional relationship of the edge lines between each cofferdam image and its cofferdam scale transformation images, obtain the scale features of each edge line in each cofferdam image; according to the extension offset relationship and scale features of each edge line in each cofferdam image, obtain the joint edge features of each edge line in each cofferdam image; according to the overall gradient relationship between the edge lines in each cofferdam image and the joint edge features of each edge line, obtain the joint probability of each edge line in each cofferdam image.

[0008] According to the extension bending relationship of each edge line in each cofferdam image, obtain the deformation degree of each edge line in each cofferdam image; according to the correlation relationship between the deformation degree of the edge lines in the cofferdam image and the water pressure index of the cofferdam image, and the joint probability of the edge lines in the cofferdam image, obtain the temporary deformation degree of each edge line in each cofferdam image and the temporary deformation index of each cofferdam image.

[0009] According to the temporary deformation index of the cofferdam image, the difference in the temporary deformation degree of the edge lines in the cofferdam image, and the deformation degree of the edge lines, obtain the sealing performance of each edge line in each cofferdam image.

[0010] According to the sealing performance of all the edge lines in all the cofferdam images, obtain the construction quality inspection results of the bridge deep-water cofferdam.

[0011] Further, the specific method for obtaining the scale features of each edge line in each cofferdam image according to the positional relationship of the edge lines between each cofferdam image and its cofferdam scale transformation images is as follows:

[0012] For any edge line in any cofferdam image, the edge lines in all the cofferdam scale transformation images of this cofferdam image that contain positions with an intersection with the position contained in this edge line are denoted as the reference lines of this edge line.

[0013] The calculation method of the scale feature of the b-th edge line in the a-th cofferdam image is:

[0014]

[0015] In the formula, A a,b is the scale feature of the b-th edge line in the a-th cofferdam image; B a,b is the number of reference lines of the b-th edge line in the a-th cofferdam image; C a,b,c is the DTW distance between the b-th edge line in the a-th cofferdam image and its c-th reference line; exp() is the exponential function with the natural constant as the base.

[0016] Further, obtaining the joint edge feature of each edge line in each cofferdam image according to the extension offset relationship and scale feature of each edge line in each cofferdam image includes the following specific method:

[0017] For any edge pixel point on any edge line in any cofferdam image, denote any endpoint of the edge line as the feature endpoint of the edge line, and denote the next adjacent pixel point along the shortest path from the edge pixel point to the feature endpoint along the edge line as the feature adjacent point of the edge pixel point; denote the direction from the edge pixel point to its feature adjacent point as the feature direction of the edge pixel point;

[0018] Denote the minimum included angle between the feature direction of the edge pixel point and the feature direction of its feature adjacent point as the offset angle of the edge pixel point;

[0019] The calculation method of the joint edge feature of the b-th edge line in the a-th cofferdam image is as follows:

[0020]

[0021] In the formula, D a,b is the joint edge feature of the b-th edge line in the a-th cofferdam image; A a,b is the scale feature of the b-th edge line in the a-th cofferdam image; E a,b is the number of edge pixel points on the b-th edge line in the a-th cofferdam image; θ a,b,d is the offset angle of the d-th edge pixel point on the b-th edge line in the a-th cofferdam image; || is the absolute value function; exp() is the exponential function with the natural constant as the base; cos() is the cosine function.

[0022] Further, obtaining the joint probability of each edge line in each cofferdam image according to the overall gradient relationship between the edge lines in each cofferdam image and the joint edge feature of each edge line includes the following specific method:

[0023] Denote any edge line in any cofferdam image as the target edge line; obtain the gradient vector of each edge pixel point on the target edge line; denote the sum vector of the gradient vectors of all edge pixel points on the target edge line as the overall gradient vector of the target edge line;

[0024] The modulus of the difference vector between the overall gradient vector of the target edge line and the overall gradient vector of any edge line other than the target edge line in the cofferdam image is denoted as the degree of difference between the target edge line and this edge line; the inverse proportional normalization result of the mean value of the degrees of difference between the target edge line and all edge lines other than the target edge line in the cofferdam image is denoted as the regularity degree of the target edge line.

[0025] The product of the seam edge feature of the target edge line and the regularity degree is denoted as the seam probability of the target edge line.

[0026] Further, the specific method for obtaining the deformation degree of each edge line in each cofferdam image according to the extension and bending relationship of each edge line includes:

[0027] Obtain the gradient vector of each edge pixel point on each edge line in each cofferdam image;

[0028]

[0029] where, F a,b is the deformation degree of the b-th edge line in the a-th cofferdam image; G a,b is the maximum value of the difference angles between the directions of the gradient vectors of all edge pixel points in the b-th edge line in the a-th cofferdam image; E a,b is the number of edge pixel points on the b-th edge line in the a-th cofferdam image; α a,b,f is the minimum included angle between the direction of the gradient vector of the f-th edge pixel point on the b-th edge line in the a-th cofferdam image and the direction of the gradient vector of its feature adjacent point; norm() is the linear normalization function; cos() is the cosine function.

[0030] Further, the specific method for obtaining the temporary deformation degree of each edge line in each cofferdam image is:

[0031] Denote any cofferdam image as the main image; denote any edge line in the main image as the main edge line; obtain the edge lines in all edge lines of any cofferdam image other than the main image, where the number of pixel points overlapping with the pixel points of the main edge line is greater than the preset overlap threshold, and denote them as the matching lines of the main edge line;

[0032] Denote the sequence formed by the deformation degrees of the main edge line and its matching lines as the deformation sequence of the main edge line;

[0033] According to the order of the edge lines corresponding to the deformation degrees in the deformation sequence of the main edge line, denote the sequence formed by the water pressure indices of the cofferdam images where the main edge line and its matching lines are located as the water pressure sequence of the main edge line.

[0034] The Pearson correlation coefficient between the deformation sequence of the main edge line and the water pressure sequence is denoted as the water pressure influence index of the main edge line.

[0035] The product of the joint probability of the main edge line and the water pressure influence index is denoted as the temporary deformation degree of the main edge line.

[0036] Furthermore, the specific method for obtaining the temporary deformation index of each cofferdam image is as follows:

[0037] The mean value of the temporary deformation degrees of all edge lines in the main image is denoted as the temporary deformation index of the main image.

[0038] Furthermore, the specific method for obtaining the sealing performance of each edge line in each cofferdam image according to the temporary deformation index of the cofferdam image, the difference in the temporary deformation degrees of the edge lines in the cofferdam image, and the deformation degree of the edge line includes:

[0039] For any edge line in any cofferdam image, the linearly normalized result of the difference between the temporary deformation degree of the edge line and the temporary deformation index of the cofferdam image is denoted as the water flow action index of the edge line;

[0040] The product of the inversely normalized result of the deformation degree of the edge line and the water flow action index is denoted as the sealing performance of the edge line.

[0041] Furthermore, the specific method for obtaining the construction quality inspection result of the bridge deep - water cofferdam according to the sealing performance of all edge lines in all cofferdam images includes:

[0042] If there are edge lines with sealing performance less than or equal to the sealing threshold in all cofferdam images, the construction quality of the cofferdam is unqualified; if the sealing performance of all edge lines in all cofferdam images is greater than the sealing threshold, the construction quality of the cofferdam is qualified.

[0043] The present invention also proposes a construction quality inspection system for bridge deep - water cofferdams based on machine vision. The system includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the steps of the above - mentioned method are implemented.

[0044] The beneficial effects of the present invention are as follows: When detecting the construction quality of a deep-water cofferdam for a bridge, since it is easy to be affected by the underwater environment when photographing the deep-water cofferdam of the bridge, resulting in inaccurate acquisition of the joint edge, the present invention obtains the joint probability of each edge line by the positional relationship of the edge lines in images of different scales, the extension and offset relationship of the edge lines, and combines the overall gradient relationship between the edge lines to judge the possibility of each edge line being a joint edge; Since the water flow impact will cause deformation in the joint area of the deep-water cofferdam of the bridge, and the deformation types are divided into temporary deformation and permanent deformation, it is necessary to distinguish between temporary deformation and permanent deformation. The present invention obtains the deformation degree of each edge line, and then obtains the temporary deformation degree of each edge line in each cofferdam image and the temporary deformation index of each cofferdam image according to the correlation between the deformation degree of the edge lines in the cofferdam image and the water pressure index of the cofferdam image, so as to judge the degree of influence of the edge lines by the water flow; When detecting the construction quality of a deep-water cofferdam for a bridge, the tightness at each joint is mainly judged. The present invention obtains the tightness of each edge line in each cofferdam image through the temporary deformation index of the cofferdam image, the difference in the temporary deformation degree of the edge lines in the cofferdam image, and the deformation degree of the edge lines. Thus, the present invention obtains an accurate detection result of the construction quality of the deep-water cofferdam of the bridge through the tightness of all edge lines in all cofferdam images. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following-described drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0046] Figure 1 It is a schematic flow chart of a method for detecting the construction quality of a deep-water cofferdam for a bridge based on machine vision provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0048] Please refer to Figure 1 , which shows a flow chart of a method for detecting the construction quality of a deep-water cofferdam for a bridge based on machine vision provided by an embodiment of the present invention. The method includes the following steps:

[0049] Step S001: Obtain a number of cofferdam images and their water pressure indices.

[0050] It should be noted that the purpose of this embodiment is to perform construction quality inspection on deep - water cofferdams of bridges based on machine vision. Therefore, it is first necessary to collect cofferdam images. Since the deep - water cofferdam of a bridge needs to surround the entire construction location, it is necessary to collect images of the cofferdam from multiple angles.

[0051] Specifically, arrange industrial underwater cameras underwater so that any complete surface of the cofferdam is included in the lens of the industrial underwater camera; arrange LED underwater lights so that the light of the LED underwater lights evenly covers the surface of the cofferdam; arrange a pressure sensor at the center position of the complete surface in the lens of the industrial underwater camera;

[0052] Use the industrial underwater camera to collect the initial RGB images of the cofferdam once every 1 second at each moment, and at the same time of collecting the initial RGB images of the cofferdam, use the pressure sensor to collect the water pressure values. The collection of the initial RGB images of the cofferdam and the water pressure values lasts for one minute.

[0053] It should be noted that since the deep - water cofferdam of a bridge surrounds the entire construction location, when performing construction quality inspection on the cofferdam, it is necessary to collect multiple complete surfaces of the cofferdam to ensure that each complete surface of the cofferdam is reliable and stable. This embodiment analyzes all the initial RGB images of the cofferdam obtained from one complete surface as an example.

[0054] It should be noted that due to more noise interference in the underwater environment, in order to make the features of the cofferdam clearer, it is necessary to further enhance the initial RGB images of the cofferdam using an image enhancement algorithm.

[0055] Specifically, perform grayscale processing on all the initial RGB images of the cofferdam to obtain a number of cofferdam grayscale images; perform median filtering processing and contrast enhancement processing on all the cofferdam grayscale images to obtain a number of cofferdam images; among them, grayscale processing, median filtering, and contrast enhancement are well - known technologies, and the specific methods are not introduced here;

[0056] For any cofferdam image, record the water pressure value collected at the same moment as the initial RGB image corresponding to this cofferdam image as the water pressure index of this cofferdam image.

[0057] Step S002: Obtain a number of cofferdam scale transformation images for each cofferdam image, and obtain a number of edge lines in each cofferdam image and its cofferdam scale transformation images; according to the positional relationship of the edge lines between each cofferdam image and its cofferdam scale transformation images, obtain the scale characteristics of each edge line in each cofferdam image; according to the extension offset relationship and scale characteristics of each edge line in each cofferdam image, obtain the joint edge characteristics of each edge line in each cofferdam image; according to the overall gradient relationship between the edge lines in each cofferdam image and the joint edge characteristics of each edge line, obtain the joint probability of each edge line in each cofferdam image.

[0058] It should be noted that during the construction of a deep-water cofferdam for a bridge, a foundation support structure, i.e., the pile foundation of the cofferdam, is established in the water. After transporting each part of the cofferdam and connecting them through joints, methods such as steel plate welding and bolt connection are usually used. The main function of the cofferdam is to isolate the water body and provide a dry environment for bridge construction. The joint is a weak link in the cofferdam structure. If the seal is not good, it may cause displacement, misalignment, or damage at the joint under the action of water pressure, thus affecting the stability of the cofferdam and even possibly causing the cofferdam to fail and affecting the construction safety of the bridge. Therefore, it is necessary to detect the tightness of the cofferdam joint. So it is necessary to obtain the joint area from the cofferdam images collected in the underwater environment.

[0059] It should be further noted that when photographing a deep-water cofferdam for a bridge in the underwater environment, it is easy to be affected by the environment, resulting in inaccurate acquisition of the joint edge. Therefore, multi-scale edge detection is used to obtain the scale characteristics of the edge lines in the cofferdam image.

[0060] Specifically, for any cofferdam image, use the Gaussian pyramid to obtain a number of scale images of this cofferdam image; use the bilinear interpolation algorithm to restore all scale images to the same size as this cofferdam image to obtain a number of cofferdam scale transformation images of this cofferdam image; perform Canny edge detection on this cofferdam image and all its cofferdam scale transformation images respectively to obtain a number of edge lines in this cofferdam image and each of its cofferdam scale transformation images; among them, the processes of Canny edge detection and obtaining scale images using the Gaussian pyramid are well-known technologies, and the specific methods will not be introduced here.

[0061] It should be noted that in addition to the joint edge of the cofferdam, the edge lines in the cofferdam image will also have interference edges caused by the underwater environment and the surface texture of the cofferdam. Compared with the joint edge of the cofferdam, the interference edges are more blurred in the image, that is, in the cofferdam scale transformation image, due to different scales, the interference edges may disappear.

[0062] Specifically, for any edge line in any cofferdam image, the edge lines in all the scale-transformed images of the cofferdam image that contain positions intersecting with the positions contained in the edge line are recorded as the reference lines of the edge line; among them, each cofferdam image contains multiple edge lines, and each edge line corresponds to multiple reference lines.

[0063] The calculation method of the scale feature of the b-th edge line in the a-th cofferdam image is as follows:

[0064]

[0065] In the formula, A a,b is the scale feature of the b-th edge line in the a-th cofferdam image; B a,b is the number of reference lines of the b-th edge line in the a-th cofferdam image; C a,b,c is the DTW distance (the DTW distance between the positions of the edge line and the reference line) between the b-th edge line in the a-th cofferdam image and its c-th reference line; exp() is the exponential function with the natural constant as the base.

[0066] It should be noted that the larger B a,b is, the clearer the b-th edge line in the a-th cofferdam image is, and the more likely it belongs to the seam edge. The larger it is, the higher the correlation between the b-th edge line in the a-th cofferdam image and its reference lines, and the more likely the reference lines belong to the same edge before and after scale transformation.

[0067] It should be noted that since the seams during cofferdam splicing are approximately straight lines and the extension directions of all seam edges are relatively unified, the seam edge features of the edge lines are judged according to the extension rules of the edge lines in the cofferdam image.

[0068] Specifically, for any edge pixel point in any edge line in any cofferdam image, any one end point of the edge line is recorded as the feature end point of the edge line (only one end point of the edge line is used as the feature end point, and other end points are no longer used as feature end points, and subsequent analyses are also carried out for other edge pixel points in the same way). The next adjacent pixel point on the shortest path from the edge pixel point along the edge line to the feature end point is recorded as the feature adjacent point of the edge pixel point; the direction from the edge pixel point to its feature adjacent point is recorded as the feature direction of the edge pixel point; it should be noted that the feature end point of the edge line has no feature adjacent point and feature direction.

[0069] The minimum included angle between the feature direction of the edge pixel point and the feature direction of its feature adjacent point is recorded as the offset angle of the edge pixel point; it should be noted that the offset angle of the feature end point of the edge line is recorded as 0.

[0070] The calculation method of the seam edge feature of the b-th edge line in the a-th cofferdam image is as follows:

[0071]

[0072] In the formula, D a,b is the seam edge feature of the b-th edge line in the a-th cofferdam image; A a,b is the scale feature of the b-th edge line in the a-th cofferdam image; E a,b is the number of upper edge pixels of the b-th edge line in the a-th cofferdam image; θ a,b,d is the offset angle of the d-th edge pixel on the b-th edge line in the a-th cofferdam image; || is the absolute value function; exp() is the exponential function with the natural constant as the base; cos() is the cosine function.

[0073] It should be noted that the larger it is, the smaller the bending degree of the b-th edge line in the a-th cofferdam image, the less the offset occurs along the edge line, and the more likely it belongs to the seam edge.

[0074] It should be noted that the seams of the cofferdam are parallel, so the possibility that the edge line in the cofferdam image is a seam edge is further judged accordingly.

[0075] Specifically, for any edge line in any cofferdam image, the Sobel operator is used to obtain the gradient vector of each edge pixel on the edge line; the sum vector of the gradient vectors of all edge pixels on the edge line is denoted as the overall gradient vector of the edge line; among them, obtaining the pixel point gradient vector by Sobel is a well-known technology, and the specific method will not be introduced here.

[0076] It should be noted that since the seams of the cofferdam are parallel, there will be many edges in the cofferdam image whose overall gradient vectors are similar to that of any seam edge.

[0077] Specifically, any edge line in any cofferdam image is denoted as the target edge line, and the modulus length of the difference vector between the overall gradient vector of the target edge line and the overall gradient vector of any edge line other than the target edge line in the cofferdam image is denoted as the difference degree between the target edge line and the edge line; the inverse proportional normalization result of the average value of the difference degrees between the target edge line and all edge lines other than the target edge line in the cofferdam image is denoted as the regularity degree of the target edge line;

[0078] Furthermore, the product of the seam edge feature of the target edge line and the regularity degree is denoted as the seam probability of the target edge line.

[0079] Step S003: Obtain the deformation degree of each edge line in each cofferdam image according to the extension and bending relationship of each edge line; according to the correlation between the deformation degree of the edge line in the cofferdam image and the water pressure index of the cofferdam image, as well as the joint probability of the edge line in the cofferdam image, obtain the temporary deformation degree of each edge line in each cofferdam image and the temporary deformation index of each cofferdam image.

[0080] It should be noted that the water flow impact will cause deformation in the joint area of the deep-water cofferdam of the bridge, and the deformation type is divided into temporary deformation and permanent deformation. Among them, the temporary deformation is the temporary deformation caused by the water flow impact. The deformation degree of the temporary deformation is small, and this deformation is reversible. After the water flow weakens, the joint will return to its original state. The temporary deformation belongs to normal deformation, while the permanent deformation is caused by the damage of the cofferdam structure by the water flow impact and belongs to abnormal deformation. Therefore, it is necessary to judge the type of deformation in the current cofferdam image to avoid misjudgment and ensure the accuracy of the detection results.

[0081] It should be further noted that when the water flow applies pressure to the cofferdam, the joint of the cofferdam shows a bending feature. Therefore, according to the bending degree of the edge line in the cofferdam image, the deformation degree of the edge line is obtained.

[0082] Specifically, the calculation method of the deformation degree of the b-th edge line in the a-th cofferdam image is as follows:

[0083]

[0084] In the formula, F a,b is the deformation degree of the b-th edge line in the a-th cofferdam image; G a,b is the maximum value of the difference angle between the directions of all edge pixel points in the b-th edge line in the a-th cofferdam image; E a,b is the number of edge pixel points on the b-th edge line in the a-th cofferdam image; α a,b,f is the minimum included angle between the direction of the gradient vector of the f-th edge pixel point on the b-th edge line in the a-th cofferdam image and the direction of the gradient vector of its characteristic adjacent point; norm() is a linear normalization function, and the normalization object is (cos(G a,b )) of all edge lines in the a-th cofferdam image; cos() is the cosine function.

[0085] It should be noted that the larger norm(cos(G a,b )) is, the greater the overall deformation degree of the b-th edge line in the a-th cofferdam image; The larger it is, the greater the local deformation degree of the b-th edge line in the a-th cofferdam image.

[0086] It should be noted that when the edge line deforms, it is necessary to determine whether the deformation of the edge line is a temporary deformation or a permanent deformation. If the deformation of the edge line is a temporary deformation, the degree of deformation of the edge line has a high correlation with the water pressure.

[0087] Specifically, any cofferdam image is denoted as the main image; any edge line in the main image is denoted as the main edge line; among all the edge lines of any cofferdam image other than the main image, the edge line with the number of pixel points overlapping the pixel points of the main edge line greater than the preset overlap threshold is denoted as the matching line of the main edge line; where the preset overlap threshold is taken as 30, and this embodiment is described by taking this as an example.

[0088] The sequence formed by the degrees of deformation of the main edge line and its matching line is denoted as the deformation sequence of the main edge line.

[0089] According to the order of the edge lines corresponding to the degrees of deformation in the deformation sequence of the main edge line, the water pressure indices of the cofferdam images where the main edge line and its matching line are located are arranged to form the water pressure sequence of the main edge line.

[0090] The Pearson correlation coefficient of the deformation sequence and the water pressure sequence of the main edge line is denoted as the water pressure influence index of the main edge line; among them, the method for obtaining the Pearson correlation coefficient is a well-known technology, and the specific method is not introduced here.

[0091] The product of the joint probability of the main edge line and the water pressure influence index is denoted as the temporary deformation degree of the main edge line.

[0092] The mean value of the temporary deformation degrees of all the edge lines in the main image is denoted as the temporary deformation index of the main image.

[0093] Step S004: Obtain the tightness of each edge line in each cofferdam image according to the temporary deformation index of the cofferdam image, the difference in the temporary deformation degree of the edge lines in the cofferdam image, and the degree of deformation of the edge lines.

[0094] It should be noted that for any edge line, the greater the influence of the water flow on the edge line, that is, the greater the temporary deformation degree of the edge line and the smaller the degree of deformation, the better the tightness of the edge line.

[0095] Specifically, for any edge line in any cofferdam image, the linear normalization result of the difference between the temporary deformation degree of the edge line and the temporary deformation index of the cofferdam image is denoted as the water flow action index of the edge line; the object of normalization is the difference between the temporary deformation degrees of all the edge lines in the cofferdam image and the temporary deformation index of the cofferdam image.

[0096] The product of the inverse normalization result of the deformation degree of the edge line and the water flow action index is denoted as the sealing performance of the edge line.

[0097] Step S005: Obtain the construction quality inspection result of the bridge deep-water cofferdam according to the sealing performance of all edge lines in all cofferdam images.

[0098] It should be noted that after obtaining the sealing performance of the edge line, construction quality inspection needs to be carried out according to the sealing performance of the edge line.

[0099] Specifically, if there are edge lines with a sealing performance less than or equal to the sealing threshold in all cofferdam images, it indicates that the construction quality of the cofferdam is unqualified and further construction of the cofferdam is required; if the sealing performance of all edge lines in all cofferdam images is greater than the sealing threshold, it indicates that the construction quality of the cofferdam is qualified; among them, the value of the sealing threshold is 0.5, and this embodiment is described by taking this as an example.

[0100] This embodiment uses the exp(-MX) model to present the inverse proportional relationship and normalization process. MX is the input of the model, and the implementer can set the inverse proportional function and normalization function according to the actual situation.

[0101] Another embodiment of the present invention provides a construction quality inspection system for a bridge deep-water cofferdam based on machine vision. The system includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the above method steps S001 to S005 are implemented.

[0102] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for detecting the construction quality of deep-water cofferdams of bridges based on machine vision, characterized in that: The method comprises the following steps: Obtain several cofferdam images and their water pressure index; Obtain several cofferdam scale transformation images of each cofferdam image, and obtain several edge lines in each cofferdam image and its cofferdam scale transformation image; obtain the scale feature of each edge line in each cofferdam image according to the positional relationship of the edge lines between each cofferdam image and its cofferdam scale transformation image; obtain the seam edge feature of each edge line in each cofferdam image according to the extension offset relationship and the scale feature of each edge line in each cofferdam image; obtain the seam probability of each edge line in each cofferdam image according to the overall gradient relationship between the edge lines in each cofferdam image and the seam edge feature of each edge line; According to the extension and bending relationship of each edge line in each cofferdam image, the deformation degree of each edge line in each cofferdam image is obtained; according to the correlation between the deformation degree of the edge line in the cofferdam image and the water pressure index of the cofferdam image, and the seam probability of the edge line in the cofferdam image, the temporary deformation degree of each edge line in each cofferdam image and the temporary deformation index of each cofferdam image are obtained; According to the temporary deformation index of the cofferdam image and the difference in the temporary deformation degree of the edge lines in the cofferdam image, as well as the deformation degree of the edge lines, the sealing property of each edge line in each cofferdam image is obtained; According to the sealing of all edge lines in all cofferdam images, the construction quality inspection results of the bridge deep-water cofferdam are obtained.

2. The method for detecting the construction quality of deep water cofferdam of a bridge based on machine vision according to claim 1 is characterized in that: The method of obtaining the scale feature of each edge line in each cofferdam image according to the positional relationship of the edge lines between each cofferdam image and its cofferdam scale transformation image includes the following specific methods: For any edge line in any cofferdam image, edge lines whose positions in all cofferdam scale transformation images of the cofferdam image intersect with the position of the edge line are recorded as reference lines of the edge line; The calculation method of the scale feature of the bth edge line in the ath cofferdam image is: In the formula, A a,b is the scale feature of the bth edge line in the ath cofferdam image; B a,b is the number of reference lines of the bth edge line in the ath cofferdam image; C a,b,c is the DTW distance between the bth edge line in the ath cofferdam image and its cth reference line; exp() is an exponential function with a natural constant as the base.

3. The method for detecting the construction quality of deep water cofferdam of a bridge based on machine vision according to claim 1 is characterized in that: The method of obtaining the seam edge feature of each edge line in each cofferdam image according to the extension offset relationship and scale feature of each edge line in each cofferdam image includes the following specific methods: For any edge pixel point in any edge line in any cofferdam image, any endpoint of the edge line is recorded as the characteristic endpoint of the edge line, and the next adjacent pixel point on the shortest path from the edge pixel point along the edge line to the characteristic endpoint is recorded as the characteristic adjacent point of the edge pixel point; the direction from the edge pixel point to its characteristic adjacent point is recorded as the characteristic direction of the edge pixel point; The minimum angle between the characteristic direction of the edge pixel point and the characteristic direction of its adjacent characteristic point is recorded as the offset angle of the edge pixel point; The calculation method of the seam edge feature of the bth edge line in the ath cofferdam image is: Where D a,b A is the seam edge feature of the bth edge line in the ath cofferdam image; a,b is the scale feature of the bth edge line in the ath cofferdam image; E a,b is the number of edge pixels on the bth edge line in the ath cofferdam image; θ a,b,d is the offset angle of the dth edge pixel on the bth edge line in the ath cofferdam image; || is the absolute value function; exp() is the exponential function with a natural constant as the base; cos() is the cosine function.

4. The method for detecting the construction quality of deep water cofferdam of a bridge based on machine vision according to claim 1 is characterized in that: The method of obtaining the seam probability of each edge line in each cofferdam image according to the overall gradient relationship between the edge lines in each cofferdam image and the seam edge feature of each edge line includes the following specific methods: Record any edge line in any cofferdam image as a target edge line; obtain the gradient vector of each edge pixel point on the target edge line; record the sum vector of the gradient vectors of all edge pixels on the target edge line as the overall gradient vector of the target edge line; The modulus length of the difference vector between the overall gradient vector of the target edge line and the overall gradient vector of any edge line in the cofferdam image except the target edge line is recorded as the degree of difference between the target edge line and the edge line; the inverse proportional normalization result of the mean of the degree of difference between the target edge line and all edge lines in the cofferdam image except the target edge line is recorded as the regularity of the target edge line; The product of the seam edge feature and the regularity of the target edge line is recorded as the seam probability of the target edge line.

5. The method for detecting the construction quality of deep water cofferdam of bridge based on machine vision according to claim 3 is characterized in that: The method of obtaining the deformation degree of each edge line in each cofferdam image according to the extension and bending relationship of each edge line in each cofferdam image includes the following specific methods: Obtain the gradient vector of each edge pixel point on each edge line in each cofferdam image; In the formula, F a,b is the deformation degree of the bth edge line in the ath cofferdam image; G a,b is the maximum value of the angle difference between the directions of the gradient vectors of all edge pixels in the bth edge line in the ath cofferdam image; E a,b is the number of edge pixels on the bth edge line in the ath cofferdam image; α a,b,f It is the minimum angle between the direction of the gradient vector of the fth edge pixel point on the bth edge line in the ath cofferdam image and the direction of the gradient vector of its feature adjacent point; norm() is the linear normalization function; cos() is the cosine function.

6. The method for detecting the construction quality of deep water cofferdams of bridges based on machine vision according to claim 1 is characterized in that: The specific method for obtaining the temporary deformation degree of each edge line in each cofferdam image is as follows: Record any one of the cofferdam images as the main image; record any one of the edge lines in the main image as the main edge line; obtain the edge lines whose number of pixel points overlapping with the pixel point position of the main edge line is greater than the preset overlap threshold among all the edge lines of any cofferdam image except the main image, and record them as the matching lines of the main edge line; The sequence consisting of the deformation degree of the main edge line and its matching line is recorded as the deformation sequence of the main edge line; According to the order of edge lines corresponding to the deformation degree in the deformation sequence of the main edge lines, the water pressure indexes of the cofferdam images where the main edge lines and their matching lines are located are arranged to form a water pressure sequence of the main edge lines; The Pearson correlation coefficient between the deformation sequence and the water pressure sequence of the main edge line is recorded as the water pressure influence index of the main edge line; The product of the joint probability of the main edge line and the water pressure influence index is recorded as the temporary deformation degree of the main edge line.

7. The method for detecting the construction quality of deep water cofferdams of bridges based on machine vision according to claim 6 is characterized in that: The specific method for obtaining the temporary deformation index of each cofferdam image is as follows: The average of the temporary deformation degrees of all edge lines in the main image is recorded as the temporary deformation index of the main image.

8. The method for detecting the construction quality of deep water cofferdam of bridge based on machine vision according to claim 1 is characterized in that: The method of obtaining the sealing property of each edge line in each cofferdam image according to the temporary deformation index of the cofferdam image and the difference in temporary deformation degree of the edge lines in the cofferdam image and the deformation degree of the edge lines includes: For any edge line in any cofferdam image, the linear normalization result of the difference between the temporary deformation degree of the edge line and the temporary deformation index of the cofferdam image is recorded as the water flow index of the edge line; The product of the inversely proportional normalized result of the deformation degree of the edge line and the water flow index is recorded as the sealing performance of the edge line.

9. The method for detecting the construction quality of deep water cofferdams of bridges based on machine vision according to claim 1 is characterized in that: The construction quality inspection result of the deep water cofferdam of the bridge is obtained according to the sealing of all edge lines in all cofferdam images, and the specific method includes: If there are edge lines with sealing properties less than or equal to the sealing threshold in all cofferdam images, the cofferdam construction quality is unqualified; if there are edge lines with sealing properties greater than the sealing threshold in all cofferdam images, the cofferdam construction quality is qualified.

10. A machine vision-based bridge deepwater cofferdam construction quality inspection system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the method for detecting the construction quality of deep-water cofferdams of bridges based on machine vision as described in any one of claims 1 to 9 are implemented.

Images