A method, system, device and medium for detecting water leakage of a shield tunnel joint

Abstract

The application discloses a kind of shield tunnel joint leakage water detection method, system, equipment and medium, it is related to tunnel joint leakage water detection field, the first feature group containing detection object characteristics is determined by extracting the interest point feature of image data in the present application, and the second feature group possibly containing detection object characteristics, and the contour information of detection object and possibly containing detection object characteristics is obtained by the interest point feature in two feature groups, the corresponding first contour label group and second contour label group are generated, then the contour information with correlation degree in two contour label groups is determined, so the correlation degree distribution result of two contour label groups is obtained, with this result to calibrate the features of image data with correlation degree, form calibration feature parameter set, so as to avoid the training of neural network model is affected due to the quality problem of image, so that the detection accuracy of final neural network model is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of tunnel joint leakage water detection, more particularly, it relates to a shield tunnel joint leakage water detection method, system, device and medium. BACKGROUND

[0002] Leakage water disease is one of the common diseases of shield tunnel. Leakage water disease will cause steel and bolt corrosion, cause lining strength to decrease, and serious leakage water will cause soil loss, reduce the ground resistance, and aggravate the deformation of the tunnel, so leakage water disease is an important content of daily detection of the tunnel.

[0003] At present, the leakage water detection mainly relies on manual inspection, that is, the detection personnel visually finds the leakage water, records the leakage water position, type, severity and other information. This method needs to rely on the personal experience of the detection personnel, and different detection personnel have different judgments on the severity of the leakage water. In related technologies, a leakage water detection technology based on image recognition technology is also disclosed. However, in the actual shield tunnel, light, shielding, various stains and scratches will affect the quality of the images collected by the image collection device, reduce the recognition accuracy of the features in the images, and thus reduce the detection accuracy of whether the leakage occurs at the joint of the shield tunnel. Moreover, the leakage water detection technology based on image recognition technology cannot predict the formation of the leakage water channel between the contact surfaces of the sealing pads before the formation of the leakage water channel.

[0004] Therefore, how to improve the detection accuracy of the leakage at the joint of the shield tunnel and predict the formation of the leakage water channel between the contact surfaces of the sealing pads before the formation of the leakage water channel is a problem to be solved at present. SUMMARY

[0005] In order to solve the problems of the related art, the present application provides a shield tunnel joint leakage water detection method, system, device and medium. The present application determines a first feature group containing the characteristics of the detection object and a second feature group possibly containing the characteristics of the detection object by extracting the interest point features of the image data, obtains the contour information of the detection object and the characteristics possibly containing the detection object from the interest point features in the two feature groups, generates the corresponding first contour label group and second contour label group, determines the contour information with correlation degree in the two contour label groups, and obtains the correlation degree distribution result of the two contour label groups. The features with correlation degree in the image data are calibrated by the result, a calibrated feature parameter set is formed, the training of the neural network model is affected due to the quality problem of the image is avoided, and the detection accuracy of the final neural network model is improved.

[0006] The above technical purposes of the present application are achieved by the following technical solutions:

[0007] In a first aspect of the present application, a method for detecting leakage water of a shield tunnel joint is provided, and the method comprises the following steps:

[0008] acquiring image data of a surface of the shield tunnel;

[0009] extracting interest point features of the image data, determining a first feature group and a second feature group, wherein the first feature group contains a feature subset of object characteristics to be detected, and the second feature group represents a feature subset that may contain the object characteristics to be detected;

[0010] acquiring contour information of the interest point features in the first feature group to generate a first contour label group, and acquiring contour information of the interest point features in the second feature group to generate a second contour label group;

[0011] determining a correlation degree distribution result of the first contour label group and the second contour label group, calibrating the correlation degree distribution result, and obtaining a calibration feature parameter set from the image data based on the correlation degree distribution result;

[0012] using the calibration feature parameter set as an input training parameter of a neural network model to obtain a trained neural network model;

[0013] detecting joint leakage water based on the trained neural network model.

[0014] In some possible implementation manners, the feature subset of the first feature group is gradient-divided, and typical sorting is performed between each gradient-divided group to obtain a first gradient group data set; the feature subset of the second feature group is gradient-divided, and typical sorting is performed between each gradient-divided group to obtain a second gradient group data set.

[0015] The objects in the first gradient group data set are respectively compared with the objects in the second gradient group data set in terms of similarity, and the objects in the first gradient group data set that satisfy a first preset requirement are used as the interest point features in the first feature group.

[0016] In some possible implementation manners, the similarity comparison includes at least one of contour comparison, inclination degree comparison, and distribution area comparison.

[0017] The first gradient group data set includes transverse joints and ring joints; and the second gradient group data set includes fire pipes, electric wires, grouting holes, and bolt holes.

[0018] In some possible implementation manners, the similarity comparison of the transverse joints includes comparison of chroma, inclination degree, and distribution area.

[0019] The similarity comparison of the ring joints includes comparison of chroma, inclination degree, and distribution area.

[0020] In some possible implementation manners, after the similarity comparison of the objects in the first echelon dataset, the method further includes the following steps:

[0021] obtaining a difference description information set, and generating a difference distribution matrix based on the difference description information set;

[0022] obtaining a difference correlation quantity of each object in the second echelon dataset according to the difference distribution matrix;

[0023] taking the object in the second echelon dataset that satisfies a second preset requirement on the difference correlation quantity as a point-of-interest feature in the second feature group; the difference correlation quantity includes at least one of a difference generation probability, a difference distance, or a difference proportion.

[0024] In some possible implementation manners, the determining of the correlation degree distribution result of the first contour label group and the second contour label group includes the following steps:

[0025] dividing node information of the first contour label group to obtain a plurality of first explicit node groups based on an object identity rule, and dividing node information of the second contour label group to obtain a plurality of second explicit node groups;

[0026] judging an Euclidean distance between each node in each first explicit node group and each node in each second explicit node group to generate a corresponding correlation reference coefficient;

[0027] performing screening processing on the correlation reference coefficient to obtain the correlation degree distribution result.

[0028] In some possible implementation manners, after the detection of the joint water leakage based on the trained neural network model, the method further includes:

[0029] manufacturing a segment joint specimen based on a segment actually used in a shield tunnel joint and a waterproof sealing gasket, recording a change process of humidity and stress of the segment joint specimen in a waterproof failure process, and obtaining a humidity-stress reference curve;

[0030] obtaining humidity data and stress data of a position of the joint;

[0031] fitting the change of the humidity data and the stress data along a thickness direction of the shield tunnel in the humidity-stress reference curve to obtain a prediction result of the joint water leakage.

[0032] A second aspect of the present application further provides a shield tunnel joint water leakage detection system, including:

[0033] an image data acquisition module, configured to acquire image data of a surface of the shield tunnel;

[0034] An object feature extraction module is used to extract interest point features from image data and determine a first feature group and a second feature group, wherein the first feature group contains a feature subset of the characteristics of the detected object and the second feature group represents a feature subset that may contain the characteristics of the detected object.

[0035] The object contour extraction module is used to obtain the contour information of the interest point features in the first feature group and generate a first contour label group, and to obtain the contour information of the interest point features in the second feature group and generate a second contour label group.

[0036] The parameter calibration module is used to determine the correlation distribution result between the first contour label group and the second contour label group, calibrate the correlation distribution result, and obtain a calibration feature parameter set from the image data based on the correlation distribution result;

[0037] The model training module is used to take the calibration feature parameter set as the input training parameters of the neural network model to obtain the trained neural network model.

[0038] The detection module is used to detect water leakage at joints based on a trained neural network model.

[0039] A third aspect of this application provides an electronic device comprising a processor, a memory, and a computer program stored in the memory and executable by the processor, wherein the computer program, when executed by the processor, implements the steps of a method for detecting water leakage at joints in a shield tunnel as described in any one of the first aspects of this application.

[0040] A fourth aspect of this application provides a computer-readable storage medium storing a computer program, wherein when executed by a processor, the computer program implements the steps of a method for detecting water leakage at joints in a shield tunnel as described in any one of the first aspects of this application.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] 1、The present application determines a first feature group containing the detection object characteristics and a second feature group possibly containing the detection object characteristics by extracting the interest point features of the image data, and obtains the contour information of the detection object and the possible detection object characteristics from the interest point features in the two feature groups, generates the corresponding first contour label group and the second contour label group, then determines the contour information with correlation degree in the two contour label groups, thereby obtaining the correlation degree distribution result of the two contour label groups, and using the result to calibrate the features with correlation degree in the image data, and forms the calibration feature parameter set, thereby avoiding the influence of the quality problem of the image on the training of the neural network model, and improving the detection accuracy of the final neural network model.

[0043] 2、The present application also considers that the image recognition can only recognize the surface of the shield tunnel joint, and cannot make corresponding prediction for the middle part of the joint, therefore, after completing the detection based on the neural network model, the present application makes a segment joint test piece based on the segment actually used in the shield tunnel joint and the waterproof sealing pad, records the change process of the humidity and stress of the segment joint test piece in the waterproof failure process, thereby obtaining the humidity-stress reference curve, and obtaining the stress data and humidity data of the joint, fitting the change of the humidity data and stress data along the thickness direction of the shield tunnel in the humidity-stress reference curve, and judging whether the water seepage channel is formed between the joint sealing pad contact surfaces based on the fitting result of the stress and humidity in the humidity-stress reference curve, thereby realizing the prediction before the water seepage channel is formed between the joint sealing pad contact surfaces. BRIEF DESCRIPTION OF DRAWINGS

[0044] The accompanying drawings, which are included to provide a further understanding of the embodiments of the present application and constitute a part of the application, illustrate embodiments of the present application and are used to explain the principle of the present application, but do not limit the present application. In the drawings:

[0045] Figure 1 A flowchart of a shield tunnel joint water leakage detection method provided by the embodiments of the present application is shown in the figure.

[0046] Figure 2 A structure block diagram of a shield tunnel joint water leakage detection system provided by the embodiments of the present application is shown in the figure. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical scheme and advantages of the present application more clear, the present application is further described in detail below with reference to the embodiments and the drawings, the illustrative embodiments of the present application and the description thereof are used to explain the present application, and do not limit the present application.

[0048] It should be understood that the terms "first", "second", etc. are used for descriptive purposes only and cannot be construed as indicating or implying relative importance or an indicated number of technical features. 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 "a plurality of" is two or more, unless otherwise specifically limited.

[0049] The common waterproof technology for shield tunnel is single or double sealing pads. Double sealing pads are often used in high water pressure and large diameter working conditions to provide double lines of defense. The main implementation of joint waterproof is to generate a certain contact stress between sealing pads during segment assembly under pressure to resist external water pressure. For large diameter traffic tunnels, the economic and social impact of leakage during the service life is huge, so an effective leakage detection and monitoring scheme is needed to predict the change of the waterproof capacity of the segment joint.

[0050] Currently, leakage detection mainly relies on manual inspection, that is, detecting personnel visually find leakage, record the position, type, severity and other information of the leakage. This method relies on the personal experience of the detection personnel, and different detection personnel have different judgments on the severity of the leakage. In related technologies, there is also a leakage detection technology based on image recognition technology. However, in actual shield tunnels, light, shielding, and various stains and scratches can affect the quality of the images collected by the image collection device, reducing the accuracy of feature recognition in the image, thereby reducing the accuracy of the final detection of whether leakage occurs at the joint of the shield tunnel. Moreover, the leakage detection technology based on image recognition technology cannot predict the formation of a leakage channel between the contact surfaces of the sealing pads at the joint.

[0051] To solve the technical problems in the related art, the embodiment of the present application provides a shield tunnel joint leakage water detection method, system, device and medium, the first feature group containing the detection object characteristics is determined by extracting the interest point features of the image data, and the second feature group possibly containing the detection object characteristics is determined, and the detection object and the contour information possibly containing the detection object characteristics are obtained from the interest point features in the two feature groups, the corresponding first contour label group and the second contour label group are generated, then the contour information with correlation degree in the two contour label groups is determined, thereby obtaining the correlation degree distribution result of the two contour label groups, the features with correlation degree of the image data are calibrated according to the result, and the calibration feature parameter set is formed, thereby avoiding the influence of the quality problem of the image on the training of the neural network model, so that the detection accuracy of the final neural network model is improved. In the embodiment of the present application, the shield tunnel joint leakage water detection method is suitable for terminal equipment or server electronic equipment such as a computer, a computer, etc. The operating system of the electronic equipment can include but is not limited to an Android operating system, an IOS operating system, a Synbian (Symbian) operating system, a Black Berry (BlackBerry) operating system, a Windows Phone 8 operating system, etc. The embodiment of the present application is not limited. In the embodiment of the present application, the electronic equipment can be provided with a user interface (User Interface, UI), an interface module and a processor (central processing unit, CPU).

[0052] The detection method provided by the embodiment of the present application will be explained and described below. Please refer to Figure 1 , Figure 1 The flowchart of the shield tunnel joint leakage water detection method provided by the embodiment of the present application is shown in Figure 1 The method comprises the following steps:

[0053] S110, acquiring image data of the surface of the shield tunnel.

[0054] In the embodiment, an image acquisition device can be used to acquire the image data of the surface of the shield tunnel. The image acquisition device can be an industrial camera, a camera, or a CCD image array, or a CMOS image array, or a device, or a means, or a component, or an instrument for converting an optical image into digital data.

[0055] The image data is a data set generated by the image acquisition device, including images, graphics, photos, images, graphics, photos converted from video data streams, etc., which are simply referred to as "image data". Preferably, the key frames in the camera video stream can be used as the acquired images to obtain the image data of the surface of the shield tunnel.

[0056] S120, extracting the interest point features of the image data, determining a first feature group and a second feature group, wherein the first feature group contains a feature subset of the detected object characteristics, and the second feature group represents a feature subset that may contain the detected object characteristics.

[0057] In this embodiment, the extraction of the interest point features of the image data can be achieved based on a convolutional neural network or the like, which is a prior art and thus will not be described herein.

[0058] The first feature group and the second feature group are determined based on the interest point features. Since the shield tunnel is formed by splicing a plurality of concrete segments, each of which contains some small cracks, grouting holes, hand holes and the like, and especially the small cracks are extremely similar to the joints between the segments in the vertical direction or along the length of the tunnel, the image data taken in the special scene of the tunnel will contain unnecessary features. Therefore, the first feature group and the second feature group corresponding to the feature groups containing the detected object characteristics and the feature groups possibly containing the detected object characteristics are distinguished in this embodiment. For example, the first feature group may only contain joint features, and the second feature group may contain a part of the joints, a part of the small cracks of the segments, or grouting holes, fire pipes, wires, bolt holes and the like.

[0059] S130, obtaining contour information of the interest point features in the first feature group to generate a first contour label group, and obtaining contour information of the interest point features in the second feature group to generate a second contour label group.

[0060] In this embodiment, the contour information of the interest point features is further recognized based on the feature groups in S120. It can be understood that the contour information is information of the features, such as length, width and other size information, or circular, square and other shape information, so as to obtain the contour label groups.

[0061] S140, determining a correlation degree distribution result of the first contour label group and the second contour label group, calibrating the correlation degree distribution result, and obtaining a calibration feature parameter set from the image data based on the correlation degree distribution result.

[0062] In this embodiment, the correlation degree distribution result can also be referred to as a similarity degree distribution result, which can be calculated based on the feature information in the two contour label groups, such as the calculation of the cosine similarity. This is a prior art and thus will not be described herein. Based on this result, a corresponding threshold value can be set, such as a filtering threshold value of 75%. Based on the correlation degree distribution result, all the feature objects in the image data with a distribution result lower than 75% are filtered, and the remaining 25% of the pictures are the input parameter set suitable for the neural network training.

[0063] S150, taking the set of calibrated feature parameters as input training parameters of the neural network model to obtain a trained neural network model.

[0064] Here, taking the set of parameters as the training parameters of the neural network model is common knowledge, and no redundant description is made. As can be understood by those skilled in the art, the backbone network of the neural network model can be a neural network such as CNN, RNN, etc.

[0065] S160, performing joint leakage detection based on the trained neural network model.

[0066] In this embodiment, based on the trained neural network model obtained in step S150, the leakage detection of the image data is common knowledge and no redundant description is made.

[0067] In one embodiment, the feature subsets of the first feature group are gradient-divided, and the typical sorting between each gradient team is obtained; the feature subsets of the second feature group are gradient-divided, and the typical sorting between each gradient team is obtained.

[0068] The objects in the first gradient team data set are compared with the objects in the second gradient team data set for similarity, and the objects in the first gradient team data set that meet the first preset requirement are taken as the interest point features in the first feature group.

[0069] In this embodiment, in order to provide a high calculation correlation distribution result, the feature subsets of the first feature group and the feature subsets of the second feature group are both gradient-divided and then sorted to obtain the corresponding gradient team data sets. In this way, the features in the two feature groups can be compared one by one, thereby ensuring the effective screening of features with large similarity differences.

[0070] If there are comparison results with large differences in similarity comparison, the comparison results with large differences are removed based on the first preset requirement. It should be understood here that the first preset requirement can be set to 30%, so that the features in the first feature group that do not meet the feature removal are removed, and the remaining features that meet the preset requirement are taken as the interest point features in the first feature group.

[0071] As a specific embodiment, the similarity comparison includes at least one of the comparison of the contour, the comparison of the inclination degree, and the comparison of the distribution area.

[0072] The first gradient team data set includes horizontal joints and ring joints; and the second gradient team data set includes fire pipes, electric wires, grouting holes, and bolt holes.

[0073] Specifically, since the distribution positions and specifications of the grouting holes and hand holes on the concrete segments are the same, and the distribution and specifications of the fire pipes, wires, bolt holes and the like in the tunnel are also the same, therefore, when performing the similarity comparison, only whether the contours are consistent, whether the inclination degrees are consistent, and whether the distribution regions are the same are compared.

[0074] The joints formed by the splicing of the multiple concrete segments together are consistent in length direction and consistent in width direction, i.e., the ring joint and the joint of the embodiment.

[0075] In one embodiment, the similarity comparison of the transverse joint includes comparison of chroma, comparison of inclination degree and comparison of distribution region.

[0076] The similarity comparison of the ring joint includes comparison of chroma, comparison of inclination degree and comparison of distribution region.

[0077] In the embodiment, the joint in the image data is a black joint with length, therefore, when performing the similarity comparison, the chroma, the inclination degree and the distribution region are completely different from those of the fire pipe, the wire, the grouting hole and the bolt hole, therefore, the above features are used to realize the preprocessing of the image data, which can greatly improve the screening rate of the image data and improve the fault tolerance rate during the image screening.

[0078] In one embodiment, after the similarity comparison of the objects in the first echelon data set, the following steps are further included:

[0079] Obtaining a difference description information set, and generating a difference distribution matrix based on the difference description information set;

[0080] Obtaining a difference related quantity of each object in the second echelon data set according to the difference distribution matrix;

[0081] Taking the object in the second echelon data set that satisfies a second preset requirement in the difference related quantity as a point of interest feature in the second feature group; wherein the difference related quantity includes at least one of a difference generation probability, a difference distance or a difference proportion.

[0082] In the embodiment, since the segment itself has some small cracks and the characteristics of the segment are extremely similar to the characteristics of the joint, and the characteristics of the electric wire in the tunnel are also similar to the characteristics of the joint, the embodiment defines a difference description information set, so as to further determine the difference between the detection object and the non-detection object, so that the difference distribution matrix is generated based on the difference description information set, and then the difference related quantity of each object in the second hierarchy data set is calculated based on the difference distribution matrix, so as to ensure the difference between the detection object and the non-detection object, and thus the image data corresponding to the feature subset in the second feature group is screened according to the second preset requirement, so as to improve the possibility that the features in the second feature group contain the detection object.

[0083] In one embodiment, determining the correlation degree distribution result of the first profile label group and the second profile label group comprises the following steps:

[0084] Based on the object identity rule, the node information of the first profile label group is divided to obtain a plurality of first explicit node groups, and the node information of the second profile label group is divided to obtain a plurality of second explicit node groups;

[0085] The Euclidean distances between each node in each first explicit node group and each node in each second explicit node group are determined to generate corresponding correlation reference coefficients;

[0086] The correlation reference coefficients are screened to obtain the correlation degree distribution result.

[0087] Specifically, whether it is a detection object or a non-detection object, in the shield tunnel, it has the same identity rule, for example, the fire pipe has the same pipe diameter and color, for example, the joint, whether it is a joint along the direction of the shield tunnel, the size and color are the same, for example, the joint perpendicular to the direction of the tunnel, the size and color are also the same.

[0088] Therefore, the node information of the first profile label group is divided, and the node information of the same node group is set as an explicit node group, so as to obtain a plurality of first explicit node groups; similarly, a plurality of second explicit node groups are obtained.

[0089] The Euclidean distances between each node in each first explicit node group and each node in each second explicit node group are determined to generate corresponding correlation reference coefficients. It should be understood that the calculation method of the Euclidean distance is the existing calculation method, which can be realized based on the clustering method, and the embodiment will not be described in detail. The correlation reference coefficient of the corresponding node is obtained based on the size of the Euclidean distance, which can be calculated according to the calculation method of the simple correlation coefficient, which is the existing calculation method, so the embodiment will not be described in detail.

[0090] The image recognition can only recognize the surface of the shield tunnel joint, and cannot make corresponding prediction for the middle part of the joint. Therefore, after the joint water leakage detection based on the trained neural network model, the method further includes:

[0091] A segment joint specimen is made based on the segment and waterproof sealing gasket actually used in the shield tunnel joint, and the change process of humidity and stress of the segment joint specimen in the waterproof failure process is recorded to obtain a humidity-stress reference curve;

[0092] Obtain the humidity data and stress data of the position of the joint;

[0093] Fit the change of the humidity data and stress data along the thickness direction of the shield tunnel in the humidity-stress reference curve to obtain the prediction result of the joint leakage.

[0094] Specifically, (1) a segment joint specimen is made based on the segment and waterproof sealing gasket actually used in the construction of the shield tunnel joint.

[0095] (2) The segment contact surface (segment-segment contact surface, sealing gasket-sealing gasket contact surface) in the thickness direction of the segment is densely covered with humidity and stress sensors. For the segment joint, the monitoring points are arranged at intervals of 5 cm between the contact surfaces in the thickness direction of the segment, and the sealing gasket contact surface needs to be avoided to prevent the change of the contact surface material properties from causing the waterproof performance to decrease. Due to the limited width of the segment, only one monitoring point needs to be arranged at the midpoint in the width direction for one segment joint. For the inter-ring joint, the monitoring points are arranged at intervals of 60° in the ring direction (the thickness direction of the point arrangement is consistent with that of the segment joint), and there are six radial monitoring point bands.

[0096] (3) According to the design water pressure and the open fault working condition, indoor waterproof test is carried out, the humidity stress reference relationship curve is obtained by increasing the water pressure until the waterproof failure.

[0097] (4) Record the change process of humidity and stress during the failure process of the waterproof joint to obtain the humidity-stress reference curve of the indoor test.

[0098] (5) The change of the humidity data and stress data of the obtained monitoring points along the thickness direction of the tunnel is fitted by means of the humidity-stress reference curve, the historical data set is established based on the deep learning algorithm, and the hydrogeological conditions of the observed joint are predicted by artificial intelligence.

[0099] (6) The corrected data is plotted, the humidity and stress alarm thresholds are set according to the actual requirements of the project, and the mutual prediction between stress and humidity is carried out through the reference curve and the deep learning algorithm, which improves the accuracy of the early warning.

[0100] Please refer toFigure 2 , Figure 2 A structural block diagram of a detection system for shield tunnel joint leakage water provided by an embodiment of the present application is shown in FIG. 1, which comprises: Figure 2

[0101] An image data acquisition module, configured to acquire image data of a shield tunnel surface;

[0102] An object feature extraction module, configured to extract interest point features of the image data, determine a first feature group and a second feature group, wherein the first feature group contains a feature subset of a detection object characteristic, and the second feature group represents a feature subset that may contain the detection object characteristic;

[0103] An object contour extraction module, configured to acquire contour information of the interest point features in the first feature group, generate a first contour label group, and acquire contour information of the interest point features in the second feature group, generate a second contour label group;

[0104] A parameter calibration module, configured to determine a correlation degree distribution result of the first contour label group and the second contour label group, calibrate the correlation degree distribution result, and obtain a calibrated feature parameter set from the image data based on the correlation degree distribution result;

[0105] A model training module, configured to take the calibrated feature parameter set as an input training parameter of a neural network model, and obtain a trained neural network model;

[0106] A detection module, configured to perform joint leakage water detection based on the trained neural network model.

[0107] As can be seen, the detection system of the embodiment has the following effects: the first feature group containing the detection object characteristic and the second feature group possibly containing the detection object characteristic are determined by extracting the interest point features of the image data, and the contour information of the detection object and the feature subset possibly containing the detection object characteristic is acquired from the interest point features in the two feature groups, the corresponding first contour label group and second contour label group are generated, then the contour information with correlation degree in the two contour label groups is determined, thereby obtaining the correlation degree distribution result of the two contour label groups, the features with correlation degree in the image data are calibrated based on the result, forming the calibrated feature parameter set, thereby avoiding the influence of the quality problem of the image on the training of the neural network model, and improving the detection accuracy of the final neural network model.

[0108] ​In another embodiment of the present application, an electronic device is provided, which includes one or more processors; a memory coupled with the processors, for storing one or more programs; and when the one or more programs are executed by the one or more processors, the one or more processors implement the steps of the method for detecting shield tunnel joint leakage water described in the above embodiments. The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, and are suitable for implementing one or more instructions, and are particularly suitable for loading and executing one or more instructions in the computer storage medium to implement a corresponding method flow or a corresponding function; the processor described in the embodiments of the present application can be used to perform the operations of the method for detecting shield tunnel joint leakage water.

[0109] In another embodiment of the present application, the present application further provides a readable storage medium, specifically a computer readable storage medium (Memory), which is a memory device in a computer device, and is used to store programs and data. It can be understood that the computer readable storage medium herein can include an internal storage medium in the computer device, and of course can also include an extended storage medium supported by the computer device. The computer readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer readable storage medium herein can be a high-speed RAM memory, or a non-volatile memory such as at least one disk memory. One or more instructions stored in the computer readable storage medium can be loaded and executed by the processor to implement the corresponding steps of the method for detecting shield tunnel joint leakage water described in the above embodiments. Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can be in the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can be in the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program codes.

[0110] The above detailed description has further explained the purpose, technical scheme and beneficial effects of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method for detecting water leakage at joints in shield tunnels, characterized in that, The methods include: Acquire image data of the shield tunnel surface; Extract interest point features from image data to determine a first feature group and a second feature group, wherein the first feature group contains a feature subset of the characteristics of the detected object, and the second feature group represents a feature subset that may contain the characteristics of the detected object. Obtain the contour information of the interest point features in the first feature group and generate a first contour label group; and obtain the contour information of the interest point features in the second feature group and generate a second contour label group. The process involves determining the correlation distribution between the first and second contour label groups, calibrating the correlation distribution, and obtaining a calibration feature parameter set from the image data based on the correlation distribution. Specifically, determining the correlation distribution between the first and second contour label groups includes the following steps: based on object identity rules, dividing the node information of the first contour label group to obtain multiple first explicit node groups; dividing the node information of the second contour label group to obtain multiple second explicit node groups; determining the Euclidean distance between each node in each of the first explicit node groups and each node in each of the second explicit node groups, generating corresponding correlation reference coefficients; and filtering the correlation reference coefficients to obtain the correlation distribution. The calibrated feature parameter set is used as the input training parameters of the neural network model to obtain the trained neural network model; Detection of water leakage at joints based on a trained neural network model.

2. The method for detecting water leakage at the joints of a shield tunnel according to claim 1, characterized in that, The feature subset of the first feature group is divided by gradient partitioning, and the resulting tiers are sorted by typicality to obtain the first tier dataset; the feature subset of the second feature group is divided by gradient partitioning, and the resulting tiers are sorted by typicality to obtain the second tier dataset. Each object in the first tier dataset is compared with each object in the second tier dataset for similarity. The objects in the first tier dataset whose comparison results meet the first preset requirements are taken as interest point features in the first feature group.

3. The method for detecting water leakage at the joints of a shield tunnel according to claim 2, characterized in that, The similarity comparison includes at least one of the following: contour comparison, tilt comparison, and distribution area comparison. The first echelon dataset includes transverse seams and circumferential seams; the second echelon dataset includes fire pipes, electrical wires, grouting holes, and bolt holes.

4. The method for detecting water leakage at the joints of a shield tunnel according to claim 3, characterized in that, The similarity comparison of the transverse seams includes the comparison of color intensity, the comparison of tilt degree, and the comparison of distribution area; The similarity comparison of the circumferential seam includes the comparison of color intensity, the comparison of tilt degree, and the comparison of distribution area.

5. The method for detecting water leakage at the joints of a shield tunnel according to claim 4, characterized in that, After comparing the similarity of objects in the first tier dataset, the following steps are also included: Obtain a set of difference description information, and generate a difference distribution matrix based on the set of difference description information; The differential correlation quantity of each object in the second echelon dataset is obtained based on the differential distribution matrix. The objects in the second echelon dataset whose differential correlation values ​​meet the second preset requirements are used as interest point features in the second feature group; The difference-related quantities mentioned therein include at least one of the following: the probability of difference occurrence, the difference distance, or the difference proportion.

6. The method for detecting water leakage at the joints of a shield tunnel according to claim 1, characterized in that, After detecting joint leakage based on the trained neural network model, the method also includes: Based on the actual use of tunnel segments and waterproof gaskets in shield tunnel joints, segment joint specimens were made, and the changes in humidity and stress of the segment joint specimens during the waterproof failure process were recorded to obtain the humidity-stress baseline curve. Obtain humidity and stress data at the joint location; By fitting the changes in humidity and stress data along the thickness direction of the shield tunnel onto the humidity-stress baseline curve, the prediction results of joint leakage are obtained.

7. A detection system for water leakage at joints in shield tunnels, characterized in that, include: Image data acquisition module, used to acquire image data of the shield tunnel surface; An object feature extraction module is used to extract interest point features from image data and determine a first feature group and a second feature group, wherein the first feature group contains a feature subset of the characteristics of the detected object and the second feature group represents a feature subset that may contain the characteristics of the detected object. The object contour extraction module is used to obtain the contour information of the interest point features in the first feature group and generate a first contour label group, and to obtain the contour information of the interest point features in the second feature group and generate a second contour label group. A parameter calibration module is used to determine the correlation distribution result between the first contour label group and the second contour label group, calibrate the correlation distribution result, and obtain a calibration feature parameter set from the image data based on the correlation distribution result. The determination of the correlation distribution result between the first contour label group and the second contour label group includes the following steps: based on object identity rules, dividing the node information of the first contour label group to obtain multiple first explicit node groups; and dividing the node information of the second contour label group to obtain multiple second explicit node groups; determining the Euclidean distance between each node in each first explicit node group and each node in each second explicit node group, generating corresponding correlation reference coefficients; and filtering the correlation reference coefficients to obtain the correlation distribution result. The model training module is used to take the calibration feature parameter set as the input training parameters of the neural network model to obtain the trained neural network model. The detection module is used to detect water leakage at joints based on a trained neural network model.

8. An electronic device, characterized in that, The electronic device includes a processor, a memory, and a computer program stored in the memory and executable by the processor, wherein when the computer program is executed by the processor, it implements the steps of a method for detecting water leakage at the joints of a shield tunnel as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, it implements the steps of a method for detecting water leakage at the joints of a shield tunnel as described in any one of claims 1 to 6.

Images