Method and device for extracting information of structure plane of working face based on pixel difference network model

Abstract

The application provides a pixel difference network model-based information extraction method and device for a structure plane of a tunnel face, and belongs to the field of tunnel engineering. Image data of the tunnel face is collected according to a standardized collection method, and then, after screening and cropping, feature labeling is used to label the joint features of the tunnel face to form a database. Then, a pixel difference network edge detection model architecture is combined with the database to establish a joint recognition model for the tunnel face. The joint information of the tunnel face is extracted using the recognition model for the tunnel face photos. According to a completeness evaluation index for the tunnel face, the pixel calculation method for extracting the number of structure planes per unit area, the number of structure plane groups and the average distance between structure planes is proposed in combination with the model recognition result, so that the information of the structure plane of the tunnel face is extracted. The problem that the quality of the collected samples is extremely susceptible to dust and dim lighting environments, resulting in low recognition accuracy and difficulty in meeting the requirements of intelligent mechanization construction of tunnels is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of tunnel engineering, in particular, to a pixel difference network model-based information extraction method and device for a structure surface of a tunnel face. BACKGROUND

[0002] In the current theory of surrounding rock classification of underground tunnel structure, the number of rock mass volume joints, the number of structure surface groups, and the average spacing of structure surfaces in the structure surface information of the tunnel face are important indicators for dividing the completeness of the tunnel rock mass, and the basic classification of the surrounding rock of the tunnel is determined by the hardness of the rock and the completeness of the rock mass. It can be seen that the structure surface information of the tunnel face is not only needed for tunnel construction and tunnel geological advanced prediction, but also an important factor to be considered in the design and construction stages of the tunnel.

[0003] Currently, the extraction method of the structure surface information of the tunnel face mainly uses the manual measurement method, i.e., a geological personnel collects the structure surface information of the tunnel face by means of a geological compass and the like, and then performs geological recording to form a geological sketch to guide the construction. This method relying on manpower is low in efficiency, greatly influenced by subjectivity, and easy to cause delay in the construction period, and when the surrounding rock condition is poor, unsafe factors such as collapse and blockage are easy to threaten the safety of the geological survey personnel. Therefore, it is necessary to provide an extraction method for the structure surface information of the tunnel face. Although the method for identifying the surrounding rock information of the tunnel face by means of a tunnel face picture has been widely applied, due to the poor environment in the tunnel, there is a lack of a standardized collection method, and the quality of the collected sample is easily influenced by dust and dim light, resulting in low recognition accuracy and difficulty in meeting the requirements of the mechanized and intelligent construction of the tunnel. SUMMARY

[0004] This part of the disclosure is provided to briefly introduce the concepts, which will be described in detail in the following specific embodiments. This part of the disclosure is not intended to identify the key features or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0005] The embodiment of the present disclosure provides a pixel difference network model-based information extraction method and device for a structure surface of a tunnel face, aiming to improve the problem that the quality of the collected sample is easily influenced by dust and dim light, resulting in low recognition accuracy and difficulty in meeting the requirements of the mechanized and intelligent construction of the tunnel.

[0006] In a first aspect, the embodiment provides a pixel difference network model-based information extraction method for a structure surface of a tunnel face, comprising:

[0007] According to the adaptability test of the tunnel face image collection, a standardized collection method for the tunnel face image is proposed, and according to the standardized collection method, tunnel face image data is collected;

[0008] The collected data is screened and cropped, and then feature labeling is used to label the joint features of the working face to form a working face standard sample database;

[0009] In combination with the sample database, a joint crack recognition model of the working face is established based on the algorithm technical principle and a pixel difference network edge detection model architecture, and the joint information of the working face is extracted from the working face photo using the recognition model;

[0010] According to the working face integrity evaluation index and in combination with the model recognition result, a pixel calculation method for extracting the number of structure planes per unit area, the number of structure plane groups and the average distance of structure planes is proposed, and the structure plane information of the working face is extracted.

[0011] In combination with the embodiments of the first aspect, in some embodiments, a working face image standardized collection method is proposed according to a working face image collection adaptability test, and the tunnel working face image data is collected according to the standardized collection method, including:

[0012] The working face image collection range is the local working face and the whole working face;

[0013] According to the working face image collection adaptability test, after taking the first photo, the photo position is moved in a small range, and the second and third photos are taken.

[0014] In combination with the embodiments of the first aspect, in some embodiments, the collected data is screened and cropped, and then feature labeling is used to label the joint features of the working face to form a working face standard sample database, including:

[0015] In the taken photos, samples with obvious working face joint features and large joint coverage area are selected, and the floor area and the contour area outside the working face of part of the working face photos are cropped to form a sample data set;

[0016] The sample is read, the working face joint is labeled with a multi-segment line, and is named, and after the labeling is completed, it is saved as a json file;

[0017] The labeled working face image json file is converted into a png image file through Python, all working face original images and their corresponding labeled png images are uniformly cropped to 512*512 pixel size pictures, and finally the cropped images are divided into a training set and a verification set in a ratio of 9:1 to form a sample standard database.

[0018] In combination with the embodiments of the first aspect, in some embodiments, in combination with the sample database, a joint crack recognition model of the working face is established based on the algorithm technical principle and a pixel difference network edge detection model architecture, and the joint information of the working face is extracted from the working face photo using the recognition model, including:

[0019] A joint recognition model based on pixel difference network structure is established from three aspects of backbone structure, side structure and loss function;

[0020] The side structure includes an expansion convolution module for enriching multi-scale edge information, a compact spatial filter module for eliminating background noise, a 1x1 convolution layer for further reducing the feature volume to a single channel mapping, and then interpolation to the original size, and finally using the sigmoid function to create an edge map.

[0021] The loss function calculation method of the model is:

[0022]

[0023] In the formula, y i is the probability of the real edge; η is the defined threshold; β is the percentage of negative pixel samples, α = λ(1-β); and the final total loss of the model is

[0024] Based on the above established sample standard database, 80% is randomly extracted as the training set sample and 20% is randomly extracted as the test set sample to train the model.

[0025] Then the joint fracture recognition model of the working face is used to extract the joint information of the working face.

[0026] In combination with the embodiments of the first aspect, in some embodiments, according to the working face integrity evaluation index, combined with the model recognition result, a pixel calculation method for extracting the number of unit area structure planes, the number of structure plane groups and the average spacing of structure planes is proposed, and the extraction of the working face structure plane information is realized, including:

[0027] The recognition result is cleaned of the error nodes on the working face by image processing technology, and then the recognition image after cleaning the error nodes is skeletonized by the "zhang" method to obtain a single-pixel structure plane image.

[0028] Assuming that the longitudinal extension length of the working face structure plane is 1m, the number of unit area structure planes in the two-dimensional working face image is used to replace the number of rock mass volume joints to divide the working face rock integrity, and the number of unit area structure planes of the working face is obtained by calculating the actual area of the working face image and the total number of working face joints, and the calculation method is as follows: In the formula, A is the total pixel points of the working face image, a is the area of the unit pixel area in the actual working face, and N is the total number of structure planes in the working face image.

[0029] The structure planes with a difference of 10° in inclination angle are divided into a group, and the inclination angle of the structure plane is obtained through the pixel coordinates of the two endpoints of the structure plane, and the calculation method is as follows: In the formula, (x1, x2) and (y1, y2) are the starting point coordinates and the end point coordinates of a structural plane on the tunnel face, respectively, and a is the inclination angle of a structural plane on the tunnel face;

[0030] The average spacing of the structural plane is obtained by calculating the maximum and minimum intercept values of the structural plane in the same group, and the calculation method is as follows: In the formula, a is the length of a unit pixel value in meters in reality, b max is the maximum intercept value of the structural plane in the same group, b min is the minimum intercept value of the structural plane in the same group, N is the total number of the structural plane in the same group, is the average inclination angle of the structural plane in the same group.

[0031] In a second aspect, the embodiment of the present disclosure provides a tunnel face structural plane information extraction device based on a pixel difference network model, which comprises:

[0032] The acquisition unit proposes a tunnel face image standardized acquisition method according to the tunnel face image acquisition adaptability test, and acquires tunnel face image data according to the standardized acquisition method;

[0033] The sample database unit labels the tunnel face joint features by using feature labeling after screening and cropping processing of the acquired data, and forms a tunnel face standard sample database;

[0034] The recognition unit, in combination with the sample database, establishes a tunnel face joint crack recognition model based on the algorithm technical principle by using a pixel difference network edge detection model architecture, and extracts tunnel face joint information by using the recognition model on the tunnel face photo;

[0035] The extraction unit proposes a pixel calculation method for extracting the number of structural planes per unit area, the number of structural plane groups, and the average spacing of the structural plane according to the tunnel face completeness evaluation index and in combination with the model recognition result, and realizes the extraction of the tunnel face structural plane information.

[0036] In combination with the embodiment of the second aspect, in some embodiments, the recognition unit, in combination with the sample database, establishes a tunnel face joint crack recognition model based on the algorithm technical principle by using a pixel difference network edge detection model architecture, and extracts tunnel face joint information by using the recognition model on the tunnel face photo, which comprises:

[0037] A joint recognition model based on a pixel difference network structure is established from three aspects of backbone structure, side structure, and loss function;

[0038] The side structure includes an expanded convolution module for enriching multi-scale edge information, a compact spatial filter module for eliminating background noise, a 1x1 convolution layer for further reducing the feature volume into a single channel map, interpolation to the original size, and a sigmoid function for creating an edge map;

[0039] The loss function of the model is calculated as follows:

[0040]

[0041] where y i is the probability of the real edge; η is a defined threshold; β is the percentage of negative pixel samples, and α = λ(1-β); the final total loss of the model is

[0042] Based on the above established sample standard database, 80% is randomly selected as the training set sample and 20% is randomly selected as the test set sample for model training;

[0043] Then, the joint fracture recognition model is used to extract joint information from the photos of the working face.

[0044] In combination with the embodiments of the second aspect, in some embodiments, the extraction unit proposes a pixel calculation method for extracting the number of structure planes per unit area, the number of structure plane groups, and the average distance of structure planes according to the working face completeness evaluation index and the model recognition result, realizing the extraction of working face structure plane information, including:

[0045] The image processing technique is used to remove the error nodes on the working face, and then the "zhang" method is used to skeletonize the recognition image after removing the error nodes to obtain a single-pixel structure plane image.

[0046] Assuming that the longitudinal extension length of the working face structure plane is 1m, the number of structure planes per unit area in the two-dimensional working face image is used to replace the number of rock mass volume joints to divide the working face rock completeness, the number of structure planes per unit area of the working face is obtained by calculating the actual area of the working face image and the total number of working face joints, and the calculation method is as follows: where A is the total pixel points of the working face image, a is the area of the unit pixel area in the actual working face, and N is the total number of structure planes in the working face image.

[0047] The structure planes with a difference in inclination angle within 10° are divided into a group, and the inclination angle of the structure plane is obtained through the pixel coordinates of the two endpoints of the structure plane, and the calculation method is as follows:

[0048] In the formula, (x1, x2) and (y1, y2) are the starting point coordinates and the end point coordinates of a structural plane on the tunnel face, respectively; and a is the inclination angle of the structural plane on the tunnel face.

[0049] The average spacing of the structural plane is obtained by calculating the maximum and minimum intercept values of the structural plane in the same group, and the calculation method is as follows: In the formula, a is the length of a unit pixel value in meters in reality, bmax is the maximum intercept value of the structural plane in the same group, bmin is the minimum intercept value of the structural plane in the same group, and N is the total number of the structural planes in the same group. is the average inclination angle of the structural plane in the same group.

[0050] In a third aspect, the embodiments of the present disclosure provide an electronic device, comprising:

[0051] one or more processors;

[0052] a storage device configured to store one or more programs,

[0053] When the one or more programs are executed by the one or more processors, the one or more processors implement the structural plane information extraction method based on the pixel difference network model according to the first aspect.

[0054] In a fourth aspect, the embodiments of the present disclosure provide a computer readable medium having a computer program stored thereon, wherein the program is executed by a processor to implement the structural plane information extraction method based on the pixel difference network model according to the first aspect.

[0055] The present application has the following advantages: a structural plane information extraction method and device based on a pixel difference network model belong to the field of tunnel engineering. The method proposes a standardized image collection method for the tunnel face according to the adaptability test of the tunnel face image collection, collects tunnel face image data according to the standardized collection method, labels the structural plane joint features after the collected data are screened and cropped using the feature labeling open source program Labelme, forms a tunnel face image standard sample database, establishes a tunnel face joint identification model based on the algorithm technical principle and using the pixel difference network edge detection model architecture, extracts the tunnel face joint information using the identification model for the tunnel face photo, and proposes a pixel calculation method for extracting the number of structural planes per unit area, the number of structural plane groups, and the average spacing of the structural planes according to the completeness evaluation index of the tunnel face and the model identification result, and realizes the extraction of the structural plane information of the tunnel face. The method solves the problem that the sample quality is easily affected by dust and dim light environment during the collection, which leads to low recognition accuracy and difficulty in meeting the requirements of intelligent construction of tunnel mechanization. BRIEF DESCRIPTION OF DRAWINGS

[0056] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some of the embodiments of the present application, and therefore should not be regarded as a limitation on the scope, and for those of ordinary skill in the art, other related drawings can also be obtained without creative labor.

[0057] Figure 1 is a flowchart of a pixel difference network edge detection model-based structure surface information extraction method provided by an embodiment of the present application;

[0058] Figure 2 is a sample data set result schematic diagram provided by an embodiment of the present application, which screens out samples with obvious joint features and larger joint coverage area in the photographed photos, and crops the floor area and the contour area outside the structure surface of part of the structure surface photos;

[0059] Figure 3 is a result schematic diagram provided by an embodiment of the present application, which uses the open source program Labelme to read samples and labels the structure surface joint with multiple line segments;

[0060] Figure 4 is a result schematic diagram provided by an embodiment of the present application, which adopts a pixel difference network edge detection model architecture to establish a structure surface joint crack identification model;

[0061] Figure 5 is a result schematic diagram provided by an embodiment of the present application, which uses the structure surface joint crack identification model to extract structure surface joint information using the identification model for the structure surface photos;

[0062] Figure 6 is a result schematic diagram provided by an embodiment of the present application, which uses image processing technology to remove error nodes on the structure surface according to the identification result, and then uses the “zhang” method skeleton thinning to obtain a single-pixel structure surface image from the identification image after removing the error nodes;

[0063] Figure 7 is a structure schematic diagram of a pixel difference network model-based structure surface structure information extraction device provided by an embodiment of the present application;

[0064] Figure 8 is a schematic diagram of the basic structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0065] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of the present application.

[0066] Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of the application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of the present application.

[0067] It should be noted that: similar reference numerals and letters indicate similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0068] In the description of the present application, it should be understood that the terms "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application.

[0069] In addition, the terms "first", "second" are only for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the technical features indicated. Therefore, the features defined with "first", "second" can explicitly or implicitly include one or more of the features. In the description of the present application, the meaning of "multiple" is two or more, unless otherwise explicitly specified and limited.

[0070] In the present application, unless otherwise explicitly specified and limited, the terms "mounting", "connection", "connection", "fixing" and the like should be understood broadly, for example, it can be fixed connection, or detachable connection, or integral; it can be directly connected, or indirectly connected through an intermediate medium; it can be the internal communication of two elements or the interaction relationship between two elements. For those of ordinary skill in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0071] In the present application, unless otherwise explicitly specified and limited, the first feature is "on" or "under" the second feature can include that the first and second features are in direct contact, or that the first and second features are not in direct contact but are in contact through another feature between them. Moreover, the first feature "on", "above" and "over" the second feature includes that the first feature is directly above and obliquely above the second feature, or only indicates that the first feature is higher in horizontal height than the second feature. The first feature "under", "below" and "under" the second feature includes that the first feature is directly below and obliquely below the second feature, or only indicates that the first feature is lower in horizontal height than the second feature.

[0072] Embodiment

[0073] Figure 1 An embodiment flow chart of a pixel difference network model-based structure plane information extraction method for a working face is shown. Please refer to Figure 1 The pixel difference network model-based structure plane information extraction method is used in the field of tunnel engineering, and the structure plane information of the working face is obtained through the image information of the working face.

[0074] Please refer to Figure 1 The pixel difference network model-based structure plane information extraction method includes the following steps:

[0075] Step 101, according to the working face image collection adaptability test, a working face image standardized collection method is proposed, and according to the standardized collection method, tunnel working face image data is collected.

[0076] Here, step 101 specifically includes:

[0077] The working face image collection range is the local working face and the whole working face;

[0078] According to the working face image collection adaptability test, the best shooting opportunity for the working face image is after the on-site blasting is completed and the working face is scraped and the residue is removed;

[0079] According to the working face image collection adaptability test, the position of the light source of the shooting light is required to be placed at a distance of 11m from the working face, at a distance of 1.5m from the center line of the tunnel on each side, and the light source illumination intensity standard is 10000Lux;

[0080] According to the working face image collection adaptability test, the shooting instrument requirement is a camera or a mobile phone with a pixel greater than 12 million, the shooting distance is 12m from the center line of the working face, and the shooting quality requirement is that the photo needs to cover the tunnel contour and exceed 1m around to ensure that the working face structure plane information of the tunnel is clear and there is no obstruction.

[0081] According to the adaptive test of the working face image acquisition, after taking one photo according to the above shooting requirements, the photo position is moved slightly, and the second and third photos are taken.

[0082] In step 102, after the collected data is screened and cropped, the feature labeling is used to label the working face joint feature, and a working face standard sample database is formed.

[0083] Here, step 102 specifically includes:

[0084] In the taken photos, the samples with obvious working face joint features and large joint coverage area are screened out, and the floor area of part of the working face photos and the contour area outside the working face are cropped to form a sample data set. Please refer to Figure 2 , Figure 2 The result schematic diagram of screening out samples with obvious working face joint features and large joint coverage area in the taken photos, and cropping the floor area of part of the working face photos and the contour area outside the working face to form a sample data set is given.

[0085] Read the sample, label the working face joint with multiple lines, and name it. After labeling, save it as a json file. Please refer to Figure 3 , Figure 3 The result schematic diagram of reading the sample using the open source program Labelme and labeling the working face joint with multiple lines is given.

[0086] The labeled working face image json file is converted into a png image file through Python. All working face original images and their corresponding labeled png images are uniformly cropped to 512x512 pixel size pictures. Finally, the cropped images are divided into training set and validation set in the ratio of 9:1 to form a sample standard database.

[0087] In step 103, combined with the sample database, based on the algorithm technical principle, a working face joint crack identification model is established using the pixel difference network edge detection model architecture, and the working face joint information is extracted from the working face photo using the identification model.

[0088] Here, step 103 specifically includes:

[0089] From the backbone structure, the side structure and the loss function, a joint identification model based on the pixel difference network structure is established. The whole backbone network has 4 stages, and each stage has 4 residual blocks. Except that the first stage has an initial convolution layer and 3 residual blocks, please refer to Figure 4 , Figure 4 The result schematic diagram of establishing a working face joint crack identification model using a pixel difference network edge detection model architecture is given.

[0090] The side structure includes a dilated convolution module CDCM to enrich multi-scale edge information, followed by a compact spatial filter module CSAM to eliminate background noise. Then, a 1×1 convolutional layer is used to further reduce the feature volume to a single-channel map, which is then interpolated to the original size. Finally, the sigmoid function is used to create the edge map.

[0091] The loss function of the model is calculated as follows:

[0092]

[0093] In the formula, y i α is the probability of a true edge; η is the defined threshold; β is the percentage of negative pixel samples, α = λ(1-β); the final total loss of the model is...

[0094] Based on the established sample standard database, 80% of the samples were randomly selected as the training set and 20% as the test set to train the model.

[0095] Subsequently, a joint and fracture identification model was used to extract joint information from the tunnel face photographs. Please refer to [link to relevant documentation]. Figure 5 , Figure 5 A schematic diagram is shown showing the results of extracting joint information from a tunnel face photograph using a joint and fracture identification model.

[0096] Step 104: Based on the evaluation index of the integrity of the working face and combined with the model recognition results, a pixel calculation method is proposed to extract the number of structural surface noodles per unit area, the number of structural surface groups, and the average spacing of structural surfaces, thus realizing the extraction of structural surface information of the working face.

[0097] Based on the adaptability test of tunnel face image acquisition, a standardized acquisition method for tunnel face images is proposed. Following this standardized method, tunnel face image data is collected. After screening and cropping, the acquired data is labeled with feature annotation to form a standard sample database of tunnel face images. Combining this sample database, and based on algorithmic principles, a pixel difference network edge detection model is used to establish a tunnel face joint recognition model. The model is then used to extract joint information from tunnel face photographs. Based on the tunnel face integrity evaluation index and the model recognition results, a pixel calculation method is proposed to extract the number of structural surface segments per unit area, the number of structural surface groups, and the average spacing between structural surfaces, thus achieving the extraction of tunnel face structural surface information. This solves the problem that the quality of the acquired samples is easily affected by dust and dim lighting conditions, leading to low recognition accuracy and failing to meet the needs of mechanized and intelligent tunnel construction.

[0098] Here, step 104 specifically includes:

[0099] The identification result is cleared by image processing technology, and then the single-pixel structure surface image is obtained by skeleton thinning of the identification image after removing the error nodes by using the "zhang" method. Please refer to Figure 6 , Figure 6 The identification result is cleared by image processing technology, and then the single-pixel structure surface image is obtained by skeleton thinning of the identification image after removing the error nodes by using the "zhang" method. Please refer to

[0100] Assuming that the longitudinal extension length of the structure surface of the tunnel face is 1 m, the number of structure surface strips per unit area in the two-dimensional tunnel face image is used to replace the joint volume of the rock mass in the "Railway Tunnel Design Specification" TB10003-2016 to divide the rock integrity of the tunnel face. The number of structure surface strips per unit area of the tunnel face is obtained by calculating the actual area of the tunnel face image and the total number of joints of the tunnel face. The calculation method is as follows: In the formula, A is the total pixel points of the tunnel face image, a is the area of the unit pixel area in the actual tunnel face, and N is the total number of structure surface strips in the tunnel face image.

[0101] The structure surfaces with a difference in inclination angle within 10° are divided into a group, and the inclination angle of the structure surface is obtained through the pixel coordinates of the two endpoints of the structure surface. The calculation method is as follows: In the formula, (x1, x2) and (y1, y2) are the starting point coordinates and end point coordinates of a structure surface on the tunnel face, and a is the inclination angle of the structure surface on the tunnel face.

[0102] The average spacing of the structure surface is obtained by calculating the maximum and minimum intercept values of the structure surface in the same group. The calculation method is as follows: In the formula, a is the length of the unit pixel value in meters, b max is the maximum intercept value of the structure surface in the same group, b min is the minimum intercept value of the structure surface in the same group, N is the total number of structure surfaces in the same group, and a is the average inclination angle of the structure surface in the same group.

[0103] A method for extracting information of a tunnel face structure surface based on a deep learning pixel difference network edge detection model, according to the adaptability test of tunnel face image collection, the tunnel face is photographed. In the photographed image, samples with obvious joint characteristics and large joint coverage area of the tunnel face are selected. The floor area of part of the tunnel face photo and the contour area outside the tunnel face are cropped to form a sample data set. An open source program Labelme is used to read the sample, the tunnel face joint is labeled with a polyline, and is named. After labeling, it is saved as a json file. The labeled tunnel face image json file is converted into a png image file through Python. All tunnel face original images and their corresponding labeled png images are uniformly cropped to 512x512 pixel images. Finally, the cropped images are divided into a training set and a validation set in a 9:1 ratio to form a sample standard database. A joint recognition model based on a pixel difference network structure is established from three aspects of backbone structure, side structure and loss function. Based on the above established sample standard database, 80% is randomly selected as a training set sample and 20% is randomly selected as a test set sample to train the model. A pixel difference network edge detection model architecture is used to establish a tunnel face joint crack recognition model. The tunnel face joint crack recognition model is used to extract tunnel face joint information from the tunnel face photo using the recognition model. The recognition result is cleaned by image processing technology to remove the error nodes on the tunnel face. Then, the recognition image after removing the error nodes is skeletonized by the "zhang" method to obtain a single-pixel structure surface graph. It is assumed that the longitudinal extension length of the tunnel face structure surface is 1m. The number of structure surface strips per unit area in the two-dimensional tunnel face image is used to replace the rock mass volume joint number in the "Railway Tunnel Design Specification" TB10003-2016 to divide the tunnel face rock integrity. Combined with the model recognition result, the number of structure surface strips per unit area of the tunnel face, the number of structure surface groups and the average distance of the structure surface are calculated by calculating the actual area of the tunnel face image and the total number of tunnel face joints, so as to realize the extraction of the tunnel face structure surface information. The sample quality collected by the method is easily affected by dust and dim light, resulting in low recognition accuracy, which cannot meet the requirements of tunnel mechanization and intelligent construction.

[0104] Further referring to Figure 7 , as an implementation of the method shown in the above figures, the present disclosure provides a device for extracting information of a tunnel face structure surface based on a deep learning pixel difference network edge detection model. The device embodiment corresponds to the method embodiment shown in Figure 1 . The device can be applied to various electronic devices.

[0105] As shown in Figure 7 , the device for extracting information of a tunnel face structure surface based on a pixel difference network model, comprising:

[0106] The collection unit 701 proposes a standardization collection method of the tunnel face image according to the face image collection adaptability test, and collects the tunnel face image data according to the standardization collection method;

[0107] The sample database unit 702 labels the joint features of the face according to the feature labeling after the screening and cropping processing of the collected data, and forms a standard sample database of the face;

[0108] The recognition unit 703 combines the sample database, establishes a joint crack recognition model of the face based on the algorithm technical principle and using a pixel difference network edge detection model architecture, and extracts the joint information of the face using the recognition model on the face photo;

[0109] The extraction unit 704 combines the model recognition result according to the completeness evaluation index of the face, and proposes a pixel calculation method of extracting the number of structure planes per unit area, the number of structure plane groups and the average distance of the structure planes, so as to realize the extraction of the structure plane information of the face.

[0110] In some optional embodiments, the recognition unit 703 combines the sample database, establishes a joint crack recognition model of the face based on the algorithm technical principle and using a pixel difference network edge detection model architecture, and extracts the joint information of the face using the recognition model on the face photo, including:

[0111] A joint recognition model based on the pixel difference network structure is established from three aspects of the backbone structure, the side structure and the loss function;

[0112] The side structure includes an inflation convolution module CDCM for enriching multi-scale edge information, a compact spatial filter module CSAM for eliminating background noise, a 1x1 convolution layer for further reducing the feature volume to a single channel mapping, and then interpolation to the original size, and finally using a sigmoid function to create an edge map;

[0113] The loss function calculation method of the model is:

[0114]

[0115] In the formula, y i is the probability of the real edge; η is a defined threshold; β is the percentage of negative pixel samples, α = λ (1-β); and the final total loss of the model is

[0116] Based on the above established sample standard database, 80% is randomly extracted as a training set sample and 20% is randomly extracted as a test set sample to train the model;

[0117] Then, the joint crack recognition model of the face is used to extract the joint information of the face using the recognition model on the face photo.

[0118] In some optional embodiments, the extraction unit 704 proposes a pixel calculation method of the number of structural planes per unit area, the number of structural plane groups and the average spacing of structural planes according to the evaluation index of the degree of completeness of the tunnel face and in combination with the model recognition result, and realizes the extraction of the structural plane information of the tunnel face, including:

[0119] The image processing technology is used to remove the error nodes on the tunnel face according to the recognition result, and then the single-pixel structural plane image is obtained by using the “zhang” method skeleton thinning on the recognition image after removing the error nodes;

[0120] It is assumed that the longitudinal extension length of the structural plane of the tunnel face is 1 m, and the number of structural planes per unit area in the two-dimensional tunnel face image is used to replace the joint volume number of rock mass in the “Railway Tunnel Design Specification” TB10003-2016 to divide the rock mass completeness of the tunnel face. The number of structural planes per unit area of the tunnel face is obtained by calculating the actual area of the tunnel face image and the total number of structural planes, and the calculation method is as follows: In the formula, A is the total pixel points of the tunnel face image, a is the area of the unit pixel area in the actual tunnel face, and N is the total number of structural planes in the tunnel face image;

[0121] The structural planes with a difference in inclination angle within 10° are divided into a group, and the inclination angle of the structural plane is obtained through the pixel coordinates of the two endpoints of the structural plane, and the calculation method is as follows:

[0122] In the formula, (x1, x2) and (y1, y2) are the starting point coordinates and the end point coordinates of a structural plane on the tunnel face, respectively, and a is the inclination angle of the structural plane on the tunnel face.

[0123] The average spacing of the structural plane is obtained by calculating the maximum and minimum intercept values in the same group of structural planes, and the calculation method is as follows: In the formula, a is the length of the unit pixel value in meters in the actual tunnel face, bmax is the maximum intercept value in the same group of structural planes, bmin is the minimum intercept value in the same group of structural planes, and N is the total number of structural planes in the same group, is the average inclination angle of the same group of structural planes.

[0124] Reference will now be made to Figure 8 which shows a structural schematic diagram of an electronic device suitable for implementing the embodiments of the present disclosure. The electronic device in the embodiments of the present disclosure can include but is not limited to mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablets), PMPs (portable multimedia players) and the like, and fixed terminals such as digital TVs, desktop computers and the like. Figure 8The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein.

[0125] like Figure 8 As shown, the electronic device may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 901, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 902 or a program loaded from a storage device 908 into a random access memory (RAM) 903. The RAM 903 also stores various programs and data required for the operation of the electronic device 900. The processing unit 901, ROM 902, and RAM 903 are interconnected via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0126] Typically, the following devices can be connected to I / O interface 905: input devices 906 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 907 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 908 including, for example, magnetic tapes, hard disks, etc.; and communication devices 909. Communication device 909 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 8 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown. More or fewer devices may be implemented or have alternatively.

[0127] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication device 909, or installed from storage device 908, or installed from ROM 6902. When the computer program is executed by processing device 901, it performs the functions defined in the methods of embodiments of this disclosure.

[0128] It should be noted that the computer-readable medium of the present disclosure can be a computer-readable signal medium or a computer-readable storage medium or any combination thereof. The computer-readable storage medium may, for example, be, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or any suitable combination thereof. More specific examples of the computer-readable storage medium can include, but are not limited to, an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, the computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus or device. In the present disclosure, the computer-readable signal medium can include a data signal that propagates in a baseband or as part of a carrier wave, carrying computer-readable program code. Such a propagated data signal can take many forms, including but not limited to electro-magnetic, optical, or any suitable combination thereof. The computer-readable signal medium can also be any computer-readable medium that can send, propagate, or transfer a program for use by or in connection with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wire, cable, optical fiber, RF (radio frequency), or any suitable combination thereof.

[0129] In some embodiments, the client, server, or both can communicate using any current known or future developed network protocol, such as HTTP (HyperText Transfer Protocol), and can be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include local area networks ("LAN"), wide area networks ("WAN"), the Internet, and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any current known or future developed networks.

[0130] The above computer-readable medium can be included in the above electronic device; or can exist separately, without being assembled into the electronic device.

[0131] The computer readable medium described above carries one or more programs, when the one or more programs are executed by the electronic device, cause the electronic device to: propose a tunnel face image standardized collection method according to a tunnel face image collection adaptive test, collect tunnel face image data according to the standardized collection method; after the collected data is screened and cropped, mark the tunnel face joint features using feature marking to form a tunnel face standard sample database; in combination with the sample database, based on the algorithm technical principle, use a pixel difference network edge detection model architecture to establish a tunnel face joint crack identification model, use the identification model to extract tunnel face joint information from the tunnel face photo; according to a tunnel face completeness evaluation index, in combination with the model identification result, propose a pixel calculation method for extracting the number of unit area structure planes, the number of structure plane groups and the average distance of structure planes, and realize the extraction of the tunnel face structure plane information.

[0132] Computer program code for carrying out operations of the present disclosure can be written in any one or more of a variety of programming languages or combinations of languages including an object oriented programming language such as Java, Smalltalk, C++, Python, as well as conventional procedural programming languages such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0133] The flow and block diagrams in the drawings show architectural, functional, and operational representations of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flow and block diagrams can represent a module, a segment, or a portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that in some alternative implementations, the functions noted in the blocks can occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently or the blocks may

[0134] The units described in the embodiments of the present disclosure can be implemented in the form of software, or can be implemented in the form of hardware. Among them, the name of the unit does not constitute a limitation on the unit itself in some cases. For example, the recognition unit can also be described as "a unit that, in combination with a sample database, establishes a rock face joint crack recognition model based on an algorithmic principle, using a pixel difference network edge detection model architecture, and extracts rock face joint information from a rock face photo using the recognition model".

[0135] In the context of the present disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the above. More specific examples of the machine-readable storage medium will include one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the above.

[0136] The above description is merely the preferred embodiments of the present disclosure and the explanation of the principles of the technology applied. It should be understood by those skilled in the art that the disclosure range involved in the present disclosure is not limited to the technical solutions formed by the specific combinations of the above technical features, and also covers other technical solutions formed by any combination of the above technical features or equivalent features without departing from the above disclosure concept. For example, the technical solutions formed by replacing the above features with the technical features disclosed in the present disclosure but not limited to the technical features with similar functions.

[0137] In addition, although each operation is described in a specific order, this should not be understood as requiring the operations to be performed in the specific order shown or in a sequential order. In certain circumstances, multitasking and parallel processing can be advantageous. Similarly, although several implementation details are included in the above discussion, these should not be interpreted as limiting the scope of the present disclosure. Certain features described in the context of separate embodiments can also be combined in a single embodiment. Conversely, various features described in the context of a single embodiment can also be separated and implemented in multiple embodiments.

[0138] Although the subject matter has been described in language specific to structural features and / or methodological acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are disclosed as example forms of implementing the claims.

[0139] The preferred embodiments of the application are described above in detail for the purposes of clarity and understanding. It should be appreciated that the application can be practiced in a variety of ways, some of which have not been described. It should be further appreciated that the application is not limited to any particular embodiment, but is applicable to any apparatus, system, or method that falls within the scope of the appended claims.

Claims

1. A method for extracting information of a structure plane of a working face based on a pixel difference network model, characterized in that, The method comprises: According to the adaptability test of the working face image acquisition, a working face image standardized acquisition method is proposed, and working face image data of the tunnel is acquired according to the standardized acquisition method; After the collected data is screened and cropped, the working face joint features are labeled using feature labeling to form a working face standard sample database; Combined with the sample database, a working face joint crack identification model is established based on the algorithm technical principle using a pixel difference network edge detection model architecture to extract working face joint information using the identification model on the working face photo; including: A joint identification model based on the pixel difference network structure is established from three aspects of backbone structure, side structure and loss function; The side structure includes an inflation convolution module for enriching multi-scale edge information, a compact spatial filter module for eliminating background noise, a 1x1 convolution layer for further reducing the feature volume to a single channel mapping, and then interpolation to the original size, and finally using the sigmoid function to create an edge map; The loss function calculation method of the model is: where y i is the probability of being a true edge; η is a defined threshold; β is the percentage of negative pixel samples, and α = λ(1 - β); the final total loss of the model is Based on the above established sample standard database, 80% is randomly extracted as the training set sample, and 20% is randomly extracted as the test set sample to train the model; Then the working face joint crack identification model is used to extract the working face joint information using the identification model on the working face photo; According to the working face completeness evaluation index, combined with the model recognition result, a pixel calculation method for extracting the number of structure planes per unit area, the number of structure plane groups and the average distance of structure planes is proposed to realize the extraction of the working face structure plane information; including: The recognition result is cleaned of the error nodes on the working face by image processing technology, and then the single-pixel structure plane image is obtained by skeleton thinning of the recognition image after the error nodes are removed; Assuming that the longitudinal extension length of the structure plane of the working face is 1m, the number of structure planes per unit area in the two-dimensional working face image is used to replace the volume joint number of the rock mass to divide the rock mass integrity of the working face, and the number of structure planes per unit area of the working face is obtained by calculating the actual area of the working face image and the total number of joints of the working face, and the calculation method is as follows: In the formula, A is the total pixel points of the working face image, a is the area of the unit pixel area in the actual working face, and N is the total number of structure planes in the working face image. Structural planes with a difference in inclination angle within 10° are divided into a group, and the inclination angle of the structural plane is obtained through the pixel coordinates of the two end points of the same structural plane, and the calculation method is as follows: In the formula, (x1, x2) and (y1, y2) are the starting point coordinates and the end point coordinates of a structural plane on the tunnel face, respectively, and a is the inclination angle of the structural plane on the tunnel face. The average interval of the structural plane is obtained by calculating the maximum and minimum intercept values in the same group of structural planes, and the calculation method is shown as follows: In the formula, a is the length of a unit pixel value in meters in reality, b max is the maximum intercept value in the same group of structural planes, b min is the minimum intercept value in the same group of structural planes, N is the total number of the same group of structural planes, is the average inclination of the same group of structural planes.

2. The method according to claim 1, wherein, According to the adaptability test of the working face image acquisition, a working face image standardized acquisition method is proposed, and working face image data of the tunnel is acquired according to the standardized acquisition method, including: The working face image acquisition range is the local working face and the whole working face; According to the adaptability test of the working face image acquisition, one photo is taken according to the shooting requirements, the shooting position is moved slightly, and the second and third photos are taken.

3. The method according to claim 1, wherein, After the collected data is screened and cropped, the working face joint features are labeled using feature labeling to form a working face standard sample database, including: In the taken photos, the samples with obvious working face joint features and large joint coverage area are screened, and the floor area of part of the working face photos and the contour area outside the working face are cropped to form a sample data set; The samples are read, the working face joints are labeled with multiple line segments, and are named, and after the labeling is completed, the json file is saved; The labeled tunnel face image json file is converted into a png image file through Python. All the tunnel face images and their corresponding labeled png images are uniformly cropped to 512*512 pixel size pictures. Finally, the cropped images are divided into training set and validation set in the ratio of 9:1 to form a sample standard database.

4. A tunnel face structure information extraction device based on a pixel difference network model, comprising: A collection unit proposes a tunnel face image standardization collection method according to the tunnel face image collection adaptability test, and collects tunnel face image data according to the standardization collection method; A sample database unit labels the tunnel face joint features using feature labeling after screening and cropping the collected data, and forms a tunnel face standard sample database; An identification unit combines the sample database, establishes a tunnel joint crack identification model based on the algorithm technical principle using a pixel difference network edge detection model architecture, and extracts tunnel joint information using the identification model on the tunnel face photo; comprising: A joint identification model based on a pixel difference network structure is established from three aspects of backbone structure, side structure and loss function; The side structure includes an inflation convolution module for enriching multi-scale edge information, a compact spatial filter module for eliminating background noise, a 1*1 convolution layer for further reducing the feature volume to a single channel mapping, and a sigmoid function for creating an edge mapping; The loss function calculation method of the model is: where y i is the probability of being a true edge; η is a defined threshold; β is the percentage of negative pixel samples, and α = λ(1 - β); the final total loss of the model is Based on the above established sample standard database, 80% is randomly extracted as a training set sample and 20% is randomly extracted as a test set sample to train the model; Then the tunnel joint crack identification model is used to extract the tunnel joint information using the identification model on the tunnel face photo; An extraction unit proposes a pixel calculation method for extracting the number of structure planes per unit area, the number of structure plane groups and the average distance of structure planes according to the tunnel face completeness evaluation index combined with the model recognition result, and realizes the extraction of the tunnel face structure information; comprising: The recognition result is cleaned of error nodes on the tunnel face through image processing technology, and then a single-pixel structure plane image is obtained by using the "zhang" method skeleton thinning on the recognition image after removing the error nodes; Assuming that the longitudinal extension length of the tunnel face structure plane is 1m, the number of structure planes per unit area in the two-dimensional tunnel face image is used to replace the number of rock mass volume joints to divide the rock completeness of the tunnel face, the number of structure planes per unit area of the tunnel face is obtained by calculating the actual area of the tunnel face image and the total number of tunnel face joints, and the calculation method is as follows: In the formula, A is the total pixel points of the palm face image, a is the area of the unit pixel area in the actual palm face, and N is the total number of structural surfaces in the palm face image. Structural planes with a difference in inclination angle within 10° are divided into a group, and the inclination angle of the structural plane is obtained through the pixel coordinates of the two endpoints of the same structural plane, and the calculation method is as follows: In the formula, (x1, x2) and (y1, y2) are respectively the starting point coordinate and the end point coordinate of a structural plane on the tunnel face, and a is the inclination angle of the structural plane on the tunnel face. The average interval of the structural plane is obtained by calculating the maximum and minimum intercept values in the same group of structural planes, and the calculation method is shown as follows: In the formula, a is the length of a unit pixel value in meters in reality, b max is the maximum intercept value in the same group of structural planes, b min is the minimum intercept value in the same group of structural planes, N is the total number of the same group of structural planes, is the average inclination of the same group of structural planes.

5. An electronic device, comprising: comprising: one or more processors; a storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the method of any one of claims 1-3.

6. A computer readable medium having stored thereon a computer program, characterized in that, The program, when executed by the processor, implements the method of any one of claims 1-3.

Images