Bridge autonomous evaluation method and system based on big data

Abstract

The invention relates to the technical field of intelligent assessment, and particularly discloses a bridge autonomous assessment method and system based on big data, and the method comprises the steps: collecting the visual information of a bridge at regular time, recognizing the visual information, and determining the type of a defect containing a position; reading monitoring data of all monitors; creating a sample set according to the defect types and the read monitoring data, clustering all samples according to the defect types to obtain a first clustering result, and determining data singleness of each defect type according to the first clustering result; clustering all the samples according to the monitoring data to obtain a second clustering result, and determining the defect type distribution condition of each type of monitoring data according to the second clustering result; acquiring monitoring data of the monitor in real time, and generating an inspection instruction according to the monitoring data, the data singleness and the defect type distribution condition; the invention provides an integrated data analysis architecture, which is higher in accuracy and higher in integrity, and better conforms to the actual situation.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent evaluation, and specifically to a bridge autonomous evaluation method and system based on big data. Background Art

[0002] Bridge autonomous evaluation refers to the use of sensing technology, artificial intelligence (AI), big data analysis and other means to enable the bridge to monitor, analyze and evaluate its structural health status by itself, reduce the dependence on manual inspection, and improve safety and maintenance efficiency.

[0003] Most of the existing bridge autonomous evaluation schemes are relatively single. For example, some thresholds are set, and when the data collected by a certain sensor reaches this threshold, an alarm message is generated. In fact, the evaluation process of the bridge is a whole, and a defect type corresponds to the sensing data of a certain area. Therefore, how to provide an integrated bridge autonomous evaluation scheme based on areas is the technical problem that the technical solution of the present invention wants to solve. Summary of the Invention

[0004] The purpose of the present invention is to provide a bridge autonomous evaluation method and system based on big data to solve the problems raised in the above background art.

[0005] To achieve the above purpose, the present invention provides the following technical solutions:

[0006] A bridge autonomous evaluation method based on big data, the method includes:

[0007] Regularly collect the visual information of the bridge, identify the visual information, and determine the defect type containing the position;

[0008] Take the previous recognition moment as the initial moment, take the current recognition moment as the end moment, create a time range, broadcast the time range to all monitors of the bridge, and read the monitoring data of all monitors;

[0009] Create a sample set according to the defect type and the read monitoring data, cluster all samples according to the defect type to obtain the first clustering result, and determine the data singularity of each defect type according to the first clustering result;

[0010] Cluster all samples according to the monitoring data to obtain the second clustering result, and determine the defect type distribution of each type of monitoring data according to the second clustering result;

[0011] Real-time obtain the monitoring data of the monitor, and generate an inspection instruction according to the monitoring data, data singularity and defect type distribution;

[0012] Among them, the data singularity is determined by the number of types of monitoring data corresponding to the same defect type, and the defect type distribution is determined by the number of samples of the defect type corresponding to the same type of monitoring data.

[0013] As a further solution of the present invention: the steps of regularly collecting visual information of the bridge, identifying the visual information, and determining the defect type containing the position include:

[0014] Generating a visual information acquisition instruction based on a preset period;

[0015] Each time a visual information acquisition instruction is generated, acquiring a remote sensing image of the bridge;

[0016] Identifying the remote sensing image and marking the defect area;

[0017] Generating an acquisition path according to the defect area, sending the acquisition path to the acquisition end, and acquiring a close-up image of the defect area;

[0018] Identifying the close-up image, determining the defect type and its image position, and determining the position of the defect type according to the relative position between the image position and the defect area.

[0019] As a further solution of the present invention: the steps of creating a sample set according to the defect type and the read monitoring data, clustering all samples according to the defect type to obtain a first clustering result, and determining the data singularity of each defect type according to the first clustering result include:

[0020] For any defect type, determining the acquisition range according to its position;

[0021] Obtaining the monitoring data of the monitor at each moment within the acquisition range, creating a data matrix at each moment according to the position relationship of the monitors, and arranging the data matrix in time sequence to obtain a three-dimensional matrix;

[0022] Taking a defect type and a three-dimensional matrix as a sample, counting the obtained defect types and three-dimensional matrices to obtain a sample set;

[0023] Clustering the three-dimensional matrix based on the defect type to obtain a first clustering result;

[0024] In the first clustering result of each defect type, clustering the three-dimensional matrix, determining the number of clusters, and determining the data singularity according to the number of clusters.

[0025] As a further solution of the present invention: the steps of clustering all samples according to the monitoring data to obtain a second clustering result, and determining the defect type distribution of each type of monitoring data according to the second clustering result include:

[0026] Read the sample set, cluster all samples according to the three-dimensional matrix of the monitoring data to obtain the second type of clustering result;

[0027] Query the defect types of each sample in the second type of clustering result, and calculate the proportion of each defect type;

[0028] Sort the defect types according to the data singularity of the defect types, and count the sorted defect types and their proportions to obtain the distribution of the defect types.

[0029] As a further solution of the present invention: the steps of obtaining the monitoring data of the monitor in real time and generating the inspection instruction according to the monitoring data, data singularity and defect type distribution include:

[0030] Obtain the monitoring data of the monitor in real time, take the current moment as the end moment, and determine the analysis range according to the end moment and the acquisition period of the visual information;

[0031] Obtain the monitoring data at each moment within the analysis range, perform three-dimensional matrix processing on the monitoring data to obtain a feature matrix;

[0032] Read the average three-dimensional matrix of each type of data in the second clustering result, traverse the feature matrix based on the average three-dimensional matrix, and when the matching degree reaches the preset threshold, read the defect type distribution corresponding to the average three-dimensional matrix;

[0033] Generate an inspection instruction based on the defect type distribution; the inspection instruction is used to control the acquisition end to obtain the close-range image at the matching position and upload the close-range image to the master control end.

[0034] As a further solution of the present invention: the comparison process of the three-dimensional matrix is as follows:

[0035] For two three-dimensional matrices to be compared, calculate the number of elements of the two three-dimensional matrices, divide the smaller value by the larger value, and use the obtained ratio as the intersection-over-union ratio;

[0036] Take the three-dimensional matrix with a larger number of elements as the basic matrix, traverse the larger three-dimensional matrix based on the three-dimensional matrix with a smaller number of elements, and calculate the similarity in real time;

[0037] Select the maximum value of the calculated similarities as the maximum similarity, and use the product of the intersection-over-union ratio and the maximum similarity as the comparison result;

[0038] Among them, the similarity calculation process is as follows:

[0039] In the formula, S is the similarity, N and M are the two dimensions representing the spatial position of the three-dimensional matrix with a smaller number of elements, T is the dimension representing the time of the three-dimensional matrix with a smaller number of elements, Δx ijkDenote the difference between the element at the row-column position (i, j, t) in a three-dimensional matrix with a smaller quantity and the corresponding element during the traversal process.

[0040] The technical solution of the present invention also provides a bridge autonomous evaluation system based on big data, and the system includes:

[0041] A visual recognition module, configured to periodically collect visual information of the bridge, recognize the visual information, and determine the defect types containing positions.

[0042] A monitoring data acquisition module, configured to use the previous recognition moment as the initial moment, use the current recognition moment as the end moment, create a time range, broadcast the time range to all monitors of the bridge, and read the monitoring data of all monitors.

[0043] A first sample clustering module, configured to create a sample set according to the defect types and the read monitoring data, cluster all samples according to the defect types to obtain a first clustering result, and determine the data singularity of each defect type according to the first clustering result.

[0044] A second sample clustering module, configured to cluster all samples according to the monitoring data to obtain a second clustering result, and determine the distribution of defect types of various monitoring data according to the second clustering result.

[0045] An inspection determination module, configured to obtain the monitoring data of the monitor in real time, and generate an inspection instruction according to the monitoring data, data singularity, and defect type distribution.

[0046] Wherein, the data singularity is determined by the number of types of monitoring data corresponding to the same defect type, and the defect type distribution is determined by the number of samples of defect types corresponding to the same type of monitoring data.

[0047] As a further solution of the present invention: the visual recognition module includes:

[0048] An instruction generation unit, configured to generate a visual information acquisition instruction based on a preset period.

[0049] An image acquisition unit, configured to acquire a remote sensing image of the bridge every time a visual information acquisition instruction is generated.

[0050] An identification and marking unit, configured to identify the remote sensing image and mark the defect area.

[0051] A path generation unit, configured to generate an acquisition path according to the defect area, send the acquisition path to the acquisition end, and obtain a close-up image of the defect area.

[0052] An identification execution unit, configured to identify the close-up image, determine the defect type and its image position, and determine the position of the defect type according to the relative position between the image position and the defect area.

[0053] As a further solution of the present invention: The first sample clustering module includes:

[0054] A range determination unit, configured to determine an acquisition range for any defect type according to its position;

[0055] A three-dimensional processing unit, configured to obtain the monitoring data of the monitors at each moment within the acquisition range, create a data matrix for each moment according to the positional relationship of the monitors, and arrange the data matrices in chronological order to obtain a three-dimensional matrix;

[0056] A sample set generation unit, configured to use a defect type and a three-dimensional matrix as a sample, count the obtained defect types and three-dimensional matrices, and obtain a sample set;

[0057] A first clustering execution unit, configured to cluster the three-dimensional matrix based on the defect type to obtain a first clustering result;

[0058] A class number application unit, configured to cluster the three-dimensional matrix in the first clustering result of each defect type, determine the number of clusters, and determine the data singularity according to the number of clusters.

[0059] As a further solution of the present invention: The second sample clustering module includes:

[0060] A second clustering execution unit, configured to read the sample set and cluster all samples according to the three-dimensional matrix of the monitoring data to obtain a second clustering result;

[0061] A proportion calculation unit, configured to query the defect types of each sample in the second clustering result and calculate the proportion of each defect type;

[0062] A sorting unit, configured to sort the defect types according to the data singularity of the defect types, count the sorted defect types and their proportions, and obtain the distribution of the defect types.

[0063] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention regularly collects visual information, detects defect conditions according to the visual information. At the same time, it reads the monitoring data of a corresponding area for each defect condition, uses the monitoring data as an independent variable and the defect condition as a dependent variable to establish samples. When the number of samples is large enough, it constructs the relationship between the monitoring data and the defect condition based on the samples, and autonomously evaluates the bridge, providing an integrated data analysis architecture with higher accuracy, higher integrity, and better fit with the actual situation. Description of the Drawings

[0064] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention.

[0065] Figure 1 It is a flowchart of a method for autonomous bridge assessment based on big data.

[0066] Figure 2 It is the first sub-flowchart of a method for autonomous bridge assessment based on big data.

[0067] Figure 3 It is the second sub-flowchart of a method for autonomous bridge assessment based on big data.

[0068] Figure 4 It is the third sub-flowchart of a method for autonomous bridge assessment based on big data.

[0069] Figure 5 It is the fourth sub-flowchart of a method for autonomous bridge assessment based on big data.

[0070] Figure 6 It is a block diagram of the composition structure of an autonomous bridge assessment system based on big data. Detailed implementation manners

[0071] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the following further details the present invention in conjunction with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0072] Figure 1 It is a flowchart of a method for autonomous bridge assessment based on big data. In an embodiment of the present invention, a method for autonomous bridge assessment based on big data, the method includes:

[0073] Step S100: Regularly collect the visual information of the bridge, identify the visual information, and determine the defect types containing positions;

[0074] Every preset period, collect the visual information of the bridge once. The visual information is obtained by an image acquisition terminal, and the image acquisition terminal includes a remote sensing image acquisition terminal and a drone acquisition terminal. By identifying the visual information, the defect types and their positions can be determined, that is, which positions of the bridge have which defect types. It should be noted that in the present invention, the visual information generally refers to the visual information on the surface of the bridge.

[0075] Step S200: Take the previous recognition moment as the initial moment, take the current recognition moment as the end moment, create a time range, broadcast the time range to all monitors of the bridge, and read the monitoring data of all monitors;

[0076] The previous recognition moment and the current recognition moment refer to the acquisition moments of visual information. Each time visual information is acquired, recognition is performed. At the same time, the monitoring data between the current recognition moment and the previous recognition moment is read out as the data corresponding to the defect type. Generally speaking, take the defect type as the dependent variable and the monitoring data within a previous period of time as the independent variable. Since the process of acquiring visual information of the bridge is timed, therefore, the time span of the monitoring data corresponding to each defect type is the same; in the technical solution of the present invention, the working frequencies of each monitor are the same. Thus, the number of monitoring data within the time span is the same; the monitor includes sensors installed on the bridge.

[0077] Step S300: Create a sample set according to the defect type and the read monitoring data, cluster all samples according to the defect type to obtain a first clustering result, and determine the data singularity of each defect type according to the first clustering result;

[0078] The defect type contains a location, and the monitoring data obtained in step S200 is the monitoring data of the entire bridge within a period of time. On this basis, locate a part of the data in the monitoring data of the entire bridge according to the location as the monitoring data corresponding to the defect type, and count all defect types and the corresponding monitoring data to obtain a sample set; cluster all samples in the sample set according to the defect type to obtain the monitoring data corresponding to each defect type, which is called the first clustering result. Determine the data singularity according to the type of the first clustering result. The data singularity is used to characterize how many types of monitoring data cause a certain defect type. The data singularity can be represented by a parameter of data singularity degree. The data singularity degree is inversely proportional to the number of types of monitoring data.

[0079] Step S400: Cluster all samples according to the monitoring data to obtain a second clustering result, and determine the defect type distribution of each type of monitoring data according to the second clustering result;

[0080] After obtaining the sample set, cluster the samples on another scale, that is, cluster all samples based on the monitoring data, which is called the second clustering result; determine the defect type distribution of each type of monitoring data according to the second clustering result. The defect type distribution is used to characterize which defect types the same type of monitoring data may correspond to.

[0081] Step S500: Real-time obtain the monitoring data of the monitor, and generate an inspection instruction according to the monitoring data, data singularity, and defect type distribution;

[0082] In the actual application stage, the monitoring data of the monitor is obtained in real time. The monitoring data is used as the cause. By analyzing the monitoring data, it can be determined whether certain defect types may occur in each area. The basis for the analysis of the monitoring data is the data uniformity and the distribution of defect types. If the judgment result is that a certain defect type may occur, at this time, an inspection instruction is generated, and the drone collection end performs close-up collection.

[0083] It should be noted that the period of the timed acquisition process in step S100 of the present application is generally long, mostly in hours, while the sensor data acquisition frequency is very high. The control instructions of the drone are generated based on the data acquired by the sensor, and then the close-up image is acquired according to the needs. An active detection process is introduced in the low-frequency timed acquisition process, thereby building a high-frequency bridge autonomous assessment architecture in disguise.

[0084] It can be understood that there are two types of clustering references in the above content, which respectively obtain data uniformity and defect type distribution. Data uniformity is determined by the number of types of monitoring data corresponding to the same defect type, and defect type distribution is determined by the number of samples of the defect type corresponding to the same type of monitoring data.

[0085] Figure 2 This is a first sub-flow chart of the bridge autonomous assessment method based on big data. The steps of regularly collecting visual information of the bridge, identifying the visual information, and determining the defect type containing the location include:

[0086] Step S101: generating a visual information acquisition instruction based on a preset cycle;

[0087] Step S102: each time a visual information acquisition instruction is generated, a remote sensing image of the bridge is acquired;

[0088] Step S103: identifying the remote sensing image and marking defective areas;

[0089] Step S104: generating a collection path according to the defect area, sending the collection path to the collection end, and acquiring a close-up image of the defect area;

[0090] Step S105: Identify the close-up image, determine the defect type and its image position, and determine the position of the defect type based on the image position and the relative position of the defect area.

[0091] In an example of the technical solution of the present invention, the timing visual acquisition process of the long period is described. Based on a preset period, a visual information acquisition instruction is generated to obtain a remote sensing image of the bridge. The accuracy of the remote sensing image is low. By identifying the remote sensing image, it is only possible to judge which areas may have defects. The areas that may have defects are marked as defect areas. The identification process of the remote sensing image is a forward identification process. Only when the identification result is completely problem-free will it not be marked as a defective area. According to the defect areas, an acquisition path is generated and sent to the acquisition end to obtain a close-up image of the defect areas. The acquisition end can be a drone. By identifying the close-up image, the defect type and its image position are determined. According to the image position and the relative position of the defect area, the position of the defect type is determined. Among them, the image position is the position of the defect type in the image, and the relative position is the position of the defect area on the bridge.

[0092] It is worth mentioning that when the acquisition end inspects the defect areas, it will also detect the non-defect areas. According to the detection results of the non-defect areas, the recognition accuracy of the remote sensing image can be judged, and then the accuracy of the remote sensing image can be adjusted according to the recognition accuracy of the remote sensing image. The higher the recognition accuracy, the smaller the accuracy can be.

[0093] Figure 3 For the second sub-flow block diagram of the bridge autonomous evaluation method based on big data, the steps of creating a sample set according to the defect type and the monitored data read, clustering all samples according to the defect type, and obtaining the first clustering result, and determining the data singularity of each defect type according to the first clustering result include:

[0094] Step S301: For any defect type, determine the acquisition range according to its position;

[0095] Step S302: Obtain the monitoring data of the monitors at each moment within the acquisition range, create a data matrix for each moment according to the position relationship of the monitors, and arrange the data matrix in chronological order to obtain a three-dimensional matrix;

[0096] Step S303: Take a defect type and a three-dimensional matrix as a sample, count the obtained defect types and three-dimensional matrices to obtain a sample set;

[0097] Step S304: Cluster the three-dimensional matrix based on the defect type to obtain the first clustering result;

[0098] Step S305: In the first clustering result of each defect type, cluster the three-dimensional matrix, determine the number of clusters, and determine the data singularity according to the number of clusters.

[0099] In an example of the technical solution of the present invention, for any defect type, the acquisition range is determined according to its position. Generally, the acquisition range is a spherical region, and the radius of the spherical region is a preset value. The monitoring data of the monitor at each moment within the acquisition range is obtained. According to the positional relationship of the monitors, a data matrix at each moment is created. The shape of the data matrix is rectangular, but there may be some positions without data. At this time, default values can be used to replace the positions without data. The data matrices are arranged according to the time sequence to obtain a three-dimensional matrix.

[0100] Regarding a defect type and a three-dimensional matrix as a sample, the obtained defect types and three-dimensional matrices are statistically analyzed to obtain a sample set. Based on the defect type, the three-dimensional matrices are clustered to obtain a first clustering result. Since the number of types of defect types is limited, there are only a limited number of them. Therefore, the process of clustering the three-dimensional matrices based on the defect type is very simple, which is a string comparison process and can be achieved by using a logical operation process.

[0101] In the first clustering result of each defect type, the three-dimensional matrices are clustered to determine the number of clusters, and the data singularity is determined according to the number of clusters. The meaning of this process is that for a set of three-dimensional matrices corresponding to each defect type, the three-dimensional matrices are clustered, and a further classification result can be obtained. The number of clusters is determined, and the data singularity degree is calculated according to the inverse ratio of the number of clusters to characterize the data singularity.

[0102] Figure 4 It is the third sub-process block diagram of the autonomous bridge evaluation method based on big data. The steps of clustering all samples according to the monitoring data to obtain a second clustering result and determining the distribution of defect types of various monitoring data according to the second clustering result include:

[0103] Step S401: Read the sample set, and cluster all samples according to the three-dimensional matrix of the monitoring data to obtain a second type of clustering result;

[0104] Step S402: Query the defect type of each sample in the second type of clustering result, and calculate the proportion of each defect type;

[0105] Step S403: Sort the defect types according to the data singularity of the defect types, and statistically analyze the sorted defect types and their proportions to obtain the distribution of defect types.

[0106] In an example of the technical solution of the present invention, a sample set is read, all samples are clustered according to the three-dimensional matrix of the monitoring data to obtain a second type of clustering result. The defect types of each sample are queried in the second type of clustering result, the number of each type of defect is calculated, and the total number is divided by the total number to obtain the proportion of various defect types corresponding to the same type of three-dimensional matrix. Its practical significance is what is the probability of different defect types occurring for a certain type of monitoring data; finally, the defect types are sorted according to the data singularity of the defect types, and the sorting rule is to arrange them in descending order based on the data singularity, and the sorted defect types and their proportions are counted to obtain the distribution of defect types.

[0107] Among them, the advantage of descending order is that the defect types with simpler causes are ranked more forward, and in subsequent analysis, they are also analyzed earlier.

[0108] Figure 5 It is the fourth sub-process block diagram of the bridge independent evaluation method based on big data. The steps of obtaining the monitoring data of the monitor in real time and generating an inspection instruction according to the monitoring data, data singularity, and defect type distribution are as follows:

[0109] Step S501: Obtain the monitoring data of the monitor in real time. Taking the current moment as the end moment, determine the analysis range according to the end moment and the acquisition period of the visual information.

[0110] Step S502: Obtain the monitoring data at each moment within the analysis range, perform three-dimensional matrix processing on the monitoring data to obtain a feature matrix.

[0111] Step S503: Read the average three-dimensional matrix of each type of data in the second clustering result, traverse the feature matrix based on the average three-dimensional matrix, and when the matching degree reaches the preset threshold, read the defect type distribution corresponding to the average three-dimensional matrix.

[0112] Step S504: Generate an inspection instruction based on the defect type distribution; the inspection instruction is used to control the acquisition end to obtain the close-range image at the matching position and upload the close-range image to the total control end.

[0113] In an example of the technical solution of the present invention, the monitoring data of the monitor is obtained in real time. Taking the current moment as the end moment, a regular acquisition period (the parameter in step S100) is inversely deduced forward to obtain the analysis range, and the monitoring data at each moment within the analysis range is obtained. At this time, the monitoring data is the monitoring data of the entire bridge. The monitoring data is three-dimensionally processed in the manner of step S302 to obtain the three-dimensional matrix corresponding to the monitoring data of the entire bridge, which is called the feature matrix; its time dimension is the same as that of the three-dimensional matrix of each sample in the sample set, but the other two dimensions representing positions are different. The dimensions representing positions of the feature matrix are much larger than those of the three-dimensional matrix.

[0114] Read the second clustering result. The distribution of defect types corresponding to each type of monitored data (3D matrix) represented by the second clustering result is read. The average 3D matrix of each type of monitored data is read (calculate the average matrix of the 3D matrices of each sample). Based on the average 3D matrix, traverse the feature matrix. When the matching degree reaches the preset threshold, read the distribution of defect types corresponding to the average 3D matrix. The distribution of defect types is the arrangement process of defect types and their proportions according to data singularity. Read the defect types in sequence, determine the detection target according to the defect types, use the matched position as the end point, generate a patrol inspection instruction, and send it to the acquisition end. The acquisition end obtains the close-up image at the matched position and uploads the close-up image to the master control end. The master control end is used to identify the close-up image, which can be an artificial end or an AI end. For the present invention, it only needs to upload the close-up image to the master control end, and the working process of the master control end is not limited.

[0115] As a preferred embodiment of the technical solution of the present invention, the comparison process of the 3D matrix is as follows:

[0116] For two 3D matrices to be compared, calculate the number of elements of the two 3D matrices, divide the smaller value by the larger value, and use the obtained ratio as the intersection-over-union ratio;

[0117] Use the 3D matrix with a larger number of elements as the basic matrix, traverse the larger 3D matrix based on the 3D matrix with a smaller number of elements, and calculate the similarity in real time;

[0118] Select the maximum value of the calculated similarities as the maximum similarity, and use the product of the intersection-over-union ratio and the maximum similarity as the comparison result;

[0119] Among them, the similarity calculation process is as follows:

[0120] In the formula, S is the similarity, N and M are the two dimensions representing the spatial position of the 3D matrix with a smaller number of elements, T is the dimension representing the time of the 3D matrix with a smaller number of elements, and Δx ijk represents the difference between the element at the row-column position (i, j, t) in the 3D matrix with a smaller number of elements and the corresponding element during the traversal process.

[0121] In an example of the technical solution of the present invention, the comparison process of the 3D matrix is described. The comparison process of the 3D matrix is actually the comparison process of monitored data, which is used when clustering samples based on monitored data and is also used when traversing the feature matrix based on the average 3D matrix. Therefore, a specific description is required.

[0122] Since the period for collecting monitoring data in this application is fixed, the T values of various three-dimensional matrices are the same. However, N and M represent the distribution positions of the monitors corresponding to the monitoring data at a certain moment, and the values of N and M for different three-dimensional matrices may be different, which corresponds to the acquisition range. Of course, this application can be completely set to the same N and M. In this way, during the matrix comparison process in the clustering process, N, M, and T are all the same, and the comparison process can be directly compared without the need for traversal and selection processes. However, the process of traversing the feature matrix based on the average three-dimensional matrix must involve a traversal and matching process, and its process can adopt the above scheme.

[0123] Figure 6 It is a block diagram of the composition structure of a bridge autonomous evaluation system based on big data. In an embodiment of the present invention, a bridge autonomous evaluation system based on big data, the system 10 includes:

[0124] A visual recognition module 11, configured to periodically collect visual information of the bridge, recognize the visual information, and determine the defect types containing positions;

[0125] A monitoring data acquisition module 12, configured to use the previous recognition moment as the initial moment, the current recognition moment as the end moment, create a time range, broadcast the time range to all monitors of the bridge, and read the monitoring data of all monitors;

[0126] A first sample clustering module 13, configured to create a sample set according to the defect types and the read monitoring data, cluster all samples according to the defect types to obtain a first clustering result, and determine the data singularity of each defect type according to the first clustering result;

[0127] A second sample clustering module 14, configured to cluster all samples according to the monitoring data to obtain a second clustering result, and determine the distribution of defect types of various monitoring data according to the second clustering result;

[0128] An inspection judgment module 15, configured to obtain the monitoring data of the monitors in real time, and generate an inspection instruction according to the monitoring data, data singularity, and defect type distribution;

[0129] Among them, the data singularity is determined by the number of types of monitoring data corresponding to the same defect type, and the defect type distribution is determined by the number of samples of the defect types corresponding to the same type of monitoring data.

[0130] Further, the visual recognition module 11 includes:

[0131] An instruction generation unit, configured to generate a visual information acquisition instruction based on a preset period;

[0132] An image acquisition unit, configured to acquire a remote sensing image of the bridge every time a visual information acquisition instruction is generated;

[0133] An identification marking unit for identifying the remote sensing image and marking the defective area;

[0134] A path generation unit for generating a collection path based on the defective area, sending the collection path to the collection end, and obtaining a close-up image of the defective area;

[0135] An identification execution unit for identifying the close-up image, determining the defect type and its image position, and determining the position of the defect type according to the relative position between the image position and the defective area.

[0136] Specifically, the first sample clustering module 13 includes:

[0137] A range determination unit for determining the collection range for any defect type according to its position;

[0138] A three-dimensional processing unit for obtaining the monitoring data of the monitor at each moment within the collection range, creating a data matrix for each moment according to the position relationship of the monitors, and arranging the data matrices in chronological order to obtain a three-dimensional matrix;

[0139] A sample set generation unit for taking a defect type and a three-dimensional matrix as a sample, counting the obtained defect types and three-dimensional matrices, and obtaining a sample set;

[0140] A first clustering execution unit for clustering the three-dimensional matrix based on the defect type to obtain a first clustering result;

[0141] A class number application unit for clustering the three-dimensional matrix in the first clustering result of each defect type, determining the number of clusters, and determining the data singularity according to the number of clusters.

[0142] Furthermore, the second sample clustering module 14 includes:

[0143] A second clustering execution unit for reading the sample set and clustering all samples according to the three-dimensional matrix of the monitoring data to obtain a second clustering result;

[0144] A proportion calculation unit for querying the defect types of each sample in the second clustering result and calculating the proportion of each defect type;

[0145] A sorting unit for sorting the defect types according to the data singularity of the defect types, counting the sorted defect types and their proportions, and obtaining the distribution of the defect types.

[0146] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A bridge autonomous assessment method based on big data, characterized in that, The method includes: Regularly collecting visual information of the bridge, identifying the visual information, and determining the defect types containing positions; Taking the previous recognition moment as the initial moment, taking the current recognition moment as the end moment, creating a time range, broadcasting the time range to all monitors of the bridge, and reading the monitoring data of all monitors; Creating a sample set according to the defect types and the read monitoring data, clustering all samples according to the defect types to obtain a first clustering result, and determining the data singularity of each defect type according to the first clustering result; Clustering all samples according to the monitoring data to obtain a second clustering result, and determining the distribution of defect types of various monitoring data according to the second clustering result; Obtaining the monitoring data of the monitor in real time, and generating an inspection instruction according to the monitoring data, data singularity, and defect type distribution; Among them, the data singularity is determined by the number of types of monitoring data corresponding to the same defect type, and the defect type distribution is determined by the number of samples of the defect type corresponding to the same type of monitoring data.

2. The bridge autonomous assessment method based on big data according to claim 1, wherein The steps of regularly collecting visual information of the bridge, identifying the visual information, and determining the defect types containing positions include: Generating a visual information collection instruction based on a preset period; Each time a visual information collection instruction is generated, obtaining a remote sensing image of the bridge; Identifying the remote sensing image and marking the defect area; Generating a collection path according to the defect area, sending the collection path to the collection end, and obtaining a close-up image of the defect area; Identifying the close-up image, determining the defect type and its image position, and determining the position of the defect type according to the relative position between the image position and the defect area.

3. The bridge autonomous evaluation method based on big data according to claim 2, wherein The steps of creating a sample set according to the defect types and the read monitoring data, clustering all samples according to the defect types to obtain a first clustering result, and determining the data singularity of each defect type according to the first clustering result include: For any defect type, determining the collection range according to its position; Obtaining the monitoring data of the monitor at each moment within the collection range, creating a data matrix at each moment according to the positional relationship of the monitors, and arranging the data matrix in chronological order to obtain a three-dimensional matrix; Taking a defect type and a three-dimensional matrix as a sample, counting the obtained defect types and three-dimensional matrices to obtain a sample set; Clustering the three-dimensional matrix based on the defect type to obtain a first clustering result; In the first clustering result of each defect type, clustering the three-dimensional matrix, determining the number of clusters, and determining the data singularity according to the number of clusters.

4. The bridge autonomous assessment method based on big data according to claim 1, wherein The steps of clustering all samples according to the monitoring data to obtain a second clustering result, and determining the distribution of defect types of various monitoring data according to the second clustering result include: Reading the sample set, clustering all samples according to the three-dimensional matrix of the monitoring data to obtain a second clustering result; Querying the defect types of each sample in the second clustering result and calculating the proportion of each defect type; Sorting the defect types according to the data singularity of the defect types, counting the sorted defect types and their proportions to obtain the distribution of the defect types.

5. The bridge autonomous assessment method based on big data according to claim 2, wherein The steps of obtaining the monitoring data of the monitor in real time and generating an inspection instruction according to the monitoring data, data singularity, and defect type distribution are as follows: Obtain the monitoring data of the monitor in real time. Taking the current moment as the end moment, determine the analysis range according to the end moment and the acquisition period of visual information; Obtain the monitoring data at each moment within the analysis range, perform three-dimensional matrix processing on the monitoring data to obtain a feature matrix; Read the average three-dimensional matrix of each type of data in the second clustering result, traverse the feature matrix based on the average three-dimensional matrix. When the matching degree reaches the preset threshold, read the defect type distribution corresponding to the average three-dimensional matrix; Generate an inspection instruction based on the defect type distribution; the inspection instruction is used to control the acquisition end to obtain the close-range image at the matching position and upload the close-range image to the total control end.

6. The bridge autonomous assessment method based on big data according to any one of claims 3 to 5, characterized in that The comparison process of the three-dimensional matrix is as follows: For two three-dimensional matrices to be compared, calculate the number of elements in the two three-dimensional matrices, divide the smaller value by the larger value, and use the obtained ratio as the intersection over union; Take the three-dimensional matrix with a larger number of elements as the base matrix, traverse the larger three-dimensional matrix based on the three-dimensional matrix with a smaller number of elements, and calculate the similarity in real time; Select the maximum value of the calculated similarities as the maximum similarity, and use the product of the intersection over union and the maximum similarity as the comparison result; Among them, the similarity calculation process is as follows: Where S is the similarity, N and M are two dimensions representing the spatial position of the three-dimensional matrix with a smaller number of elements, T is the dimension representing the time of the three-dimensional matrix with a smaller number of elements, and Δx ijk represents the difference between the element at the row-column position (i, j, t) in the three-dimensional matrix with a smaller number of elements and the corresponding element during the traversal process.

7. A bridge autonomous evaluation system based on big data, characterized in that, The system includes: A visual recognition module, which is used to regularly collect the visual information of the bridge, identify the visual information, and determine the defect type containing the position; A monitoring data acquisition module, which is used to take the previous recognition moment as the start moment, take the current recognition moment as the end moment, create a time range, broadcast the time range to all monitors of the bridge, and read the monitoring data of all monitors; A first sample clustering module, which is used to create a sample set according to the defect type and the read monitoring data, cluster all samples according to the defect type to obtain a first clustering result, and determine the data singularity of each defect type according to the first clustering result; A second sample clustering module, which is used to cluster all samples according to the monitoring data to obtain a second clustering result, and determine the defect type distribution of each type of monitoring data according to the second clustering result; An inspection determination module, which is used to obtain the monitoring data of the monitor in real time and generate an inspection instruction according to the monitoring data, data singularity, and defect type distribution; Among them, the data singularity is determined by the number of types of monitoring data corresponding to the same defect type, and the defect type distribution is determined by the number of samples of the defect type corresponding to the same type of monitoring data.

8. The bridge autonomous evaluation system based on big data according to claim 7, characterized in that The visual recognition module includes: An instruction generation unit, which is used to generate a visual information acquisition instruction based on a preset period; An image acquisition unit, which is used to obtain the remote sensing image of the bridge every time a visual information acquisition instruction is generated; An identification marking unit, which is used to identify the remote sensing image and mark the defect area; A path generation unit, which is used to generate an acquisition path according to the defect area, send the acquisition path to the acquisition end, and obtain the close-range image of the defect area; An identification execution unit is used to identify the close-range image, determine the defect type and its image position, and determine the position of the defect type according to the relative positions of the image position and the defect area.

9. The bridge autonomous evaluation system based on big data according to claim 8, characterized in that, The first sample clustering module includes: A range determination unit is used to determine the acquisition range for any defect type according to its position; A three-dimensional processing unit is used to obtain the monitoring data of the monitor at each moment within the acquisition range, create a data matrix for each moment according to the position relationship of the monitors, and arrange the data matrices in chronological order to obtain a three-dimensional matrix; A sample set generation unit is used to take a defect type and a three-dimensional matrix as a sample, count the obtained defect types and three-dimensional matrices, and obtain a sample set; A first clustering execution unit is used to cluster the three-dimensional matrix based on the defect type to obtain a first clustering result; A class number application unit is used to cluster the three-dimensional matrix in the first clustering result of each defect type, determine the number of clusters, and determine the data singularity according to the number of clusters.

10. The bridge autonomous evaluation system based on big data according to claim 7, characterized in that, The second sample clustering module includes: A second clustering execution unit is used to read the sample set and cluster all samples according to the three-dimensional matrix of the monitoring data to obtain a second clustering result; A proportion calculation unit is used to query the defect types of each sample in the second clustering result and calculate the proportion of each defect type; A sorting unit is used to sort the defect types according to the data singularity of the defect types, count the sorted defect types and their proportions, and obtain the distribution of the defect types.

Images