Tunnel health monitoring system and method based on computer vision

Abstract

The invention relates to the technical field of tunnel engineering monitoring, and discloses a tunnel health monitoring system and method based on computer vision, and the method comprises the steps: collecting the initial state data of a tunnel, and building a tunnel health reference model and a tunnel health state judgment rule base; original image data of the inner wall of the tunnel are collected and processed in real time, and tunnel actual state data and periodic actual state data are obtained; determining the tunnel health state based on the quantitative difference value of the actual state data and the health reference model and a health state determination rule base; and obtaining tunnel periodic state data, generating a health degree distribution diagram, and predicting the health state trend of the tunnel in combination with the health state judgment rule base. By constructing the dynamic reference model and combining real-time and periodic monitoring, quantitative difference judgment and trend prediction, accurate monitoring, graded early warning and preventive maintenance of the tunnel health state are realized, and the monitoring efficiency and the tunnel operation safety are improved.

Description

Technical Field

[0001] This invention relates to the field of tunnel engineering monitoring technology, and more specifically, to a tunnel health monitoring system and method based on computer vision. Background Technology

[0002] With the rapid advancement of transportation infrastructure construction in my country, the number of tunnel projects has surged and their structures have become increasingly complex. The geological conditions they traverse have shifted from conventional soil and rock to high-risk karst, high-stress, and water-rich strata. The service environment is characterized by high humidity, strong vibration, and high dust levels, posing stringent challenges to the real-time performance, accuracy, and durability of structural health monitoring. Computer vision technology, due to its non-contact, full-coverage, and high-resolution characteristics, has gradually become a core method for tunnel monitoring.

[0003] Traditional tunnel health monitoring methods suffer from several technical bottlenecks: First, traditional manual inspections are inefficient, limited by the experience of inspectors, and slow to identify early-stage defects such as minute cracks and hidden seepage, and cannot achieve full tunnel coverage. Second, existing sensor monitoring relies heavily on the collection of single physical quantities (such as strain and displacement), lacking comprehensive analysis of visual features such as lining surface texture and temperature distribution, making it prone to misjudgments due to incomplete data. Third, health status assessments are mostly based on fixed thresholds, failing to consider the differences in characteristics of different tunnel lining materials, and lacking a dynamic benchmark update mechanism, making it difficult to adapt to the complex process of natural structural aging and gradual development of defects. Fourth, monitoring data is disconnected from maintenance decisions, lacking quantitative prediction of defect development trends, and failing to achieve the shift from "post-event maintenance" to "pre-event prevention." These problems make it difficult to detect tunnel safety hazards in a timely manner, resulting in high operation and maintenance costs, and even potentially leading to major risks such as structural instability due to the deterioration of defects.

[0004] Therefore, it is necessary to design a tunnel health monitoring method based on computer vision to solve the problems of insufficient monitoring accuracy, vague health status judgment criteria, and lack of trend prediction ability in the existing technology. Summary of the Invention

[0005] In view of this, the present invention proposes a tunnel health monitoring system and method based on computer vision, aiming to solve the problems of insufficient monitoring accuracy, vague health status judgment criteria, and lack of trend prediction ability in the prior art.

[0006] In one aspect, the present invention proposes a tunnel health monitoring method based on computer vision, comprising:

[0007] Initial tunnel status data during the initial operation phase is collected by a track-mounted mobile detection unit at the top of the tunnel.

[0008] A tunnel health baseline model is established based on the initial state data of the tunnel, and a tunnel health status determination rule base is established based on the tunnel health baseline model.

[0009] The raw image data of the tunnel inner wall is collected in real time by distributed static monitoring nodes on the tunnel inner wall, and the raw image data is preliminarily processed to obtain the actual state data of the tunnel. The actual state data of the period within a preset period is recorded and saved.

[0010] The tunnel health status is determined based on the quantitative difference between the actual state data and the health benchmark model, as well as the health status determination rule base.

[0011] When the tunnel is determined to be healthy, the actual state data is stored in the data of the tunnel health benchmark model.

[0012] When tunnel defects are identified, maintenance suggestions will be sent and an alarm will be triggered;

[0013] The tunnel is scanned within a preset period based on the mobile detection unit to obtain tunnel periodic status data, and a health distribution map is generated based on the periodic status data, actual periodic status data and health benchmark model.

[0014] The health status trend of the tunnel is predicted based on the health distribution map and the health status determination rule base.

[0015] Furthermore, the initial state data of the tunnel includes lining surface image data and three-dimensional point cloud data; the process of establishing a tunnel health benchmark model based on the initial state data of the tunnel, and establishing a tunnel health status judgment rule base according to the tunnel health benchmark model, includes:

[0016] A tunnel lining unfolded diagram is generated based on the lining surface image;

[0017] A three-dimensional model of the tunnel structure is constructed based on the aforementioned three-dimensional point cloud data;

[0018] A tunnel health benchmark model is established based on the tunnel lining development diagram and the three-dimensional model of the tunnel structure.

[0019] Furthermore, the process of establishing a tunnel health benchmark model based on the initial tunnel state data, and establishing a tunnel health state determination rule base based on the tunnel health benchmark model, also includes:

[0020] Feature parameters of different lining materials are extracted from the tunnel health benchmark model, and an initial tunnel health benchmark library is established based on the feature parameters.

[0021] Based on the tunnel initial health benchmark library, a multi-level health threshold is set, and a multi-level health status determination rule library is established according to the multi-level health threshold.

[0022] Furthermore, the characteristic parameters include: data parameters of surface texture, color, temperature, and structural dimensions of linings made of different materials;

[0023] The process of extracting feature parameters of different lining materials from the tunnel health benchmark model and establishing an initial tunnel health benchmark library based on these feature parameters includes:

[0024] Image data of different lining materials were selected from the tunnel health benchmark model, and the dataset was divided according to material type;

[0025] Extract the feature parameters of the dataset and calculate the baseline range of the feature parameters;

[0026] The data of the material type and its corresponding characteristic parameters and reference range are stored to form an initial health reference library.

[0027] Furthermore, the process of setting multi-level health thresholds based on the initial tunnel health benchmark library, and establishing a multi-level health status determination rule library based on the multi-level health thresholds, includes:

[0028] Based on the benchmark range, a multi-level health threshold is set for the characteristic parameters. The multi-level health threshold includes a first-level health threshold, a second-level health threshold, a third-level health threshold, and a fourth-level health threshold.

[0029] The tunnel health status determination rule base includes Level 1 health status, Level 2 health status, Level 3 health status, and Level 4 health status;

[0030] The four health status levels are established based on the corresponding health thresholds for the four levels.

[0031] Furthermore, the original image data of the tunnel inner wall includes: image data of the texture, color, temperature, and structural dimensions of the tunnel inner wall;

[0032] The process of acquiring raw image data of the tunnel interior wall in real time through distributed static monitoring nodes, and performing preliminary processing on the raw image data to obtain actual tunnel condition data includes:

[0033] Collect raw image data of the texture, color, temperature, and structural dimensions of the tunnel interior walls;

[0034] An image denoising algorithm was used to remove noise from the original image, and an image segmentation algorithm was used to divide the image data of lining regions with different materials.

[0035] Feature parameters are extracted from image data of lining areas with different materials to obtain actual tunnel condition data.

[0036] Furthermore, the process of determining the tunnel's health status based on the quantitative difference between the actual state data and the health benchmark model, and the health status determination rule base, includes:

[0037] The characteristic parameters of the actual status data of a certain monitoring node are compared and quantified with the characteristic parameters of the health benchmark model to obtain different quantification difference values;

[0038] When all quantitative difference values ​​are within the corresponding first-level health threshold, the monitoring node is determined to be healthy, and the actual status data of the monitoring node is stored in the health benchmark model.

[0039] If any quantitative difference value is greater than the corresponding first-level health threshold but less than the second-level health threshold, the monitoring node is determined to be abnormal.

[0040] When any quantitative difference value is greater than the corresponding level health threshold but less than the level 3 health threshold, the monitoring node is determined to be severely abnormal.

[0041] When any quantitative difference value is greater than the corresponding level health threshold but less than the level 3 health threshold, it is determined to be a disease at that monitoring node.

[0042] When a monitoring node is determined to be abnormal, severely abnormal, or diseased, its location information is recorded and prevention suggestions and alarms are sent.

[0043] Furthermore, the process of generating a health distribution map based on the aforementioned periodic state data, periodic actual state data, and health benchmark model includes:

[0044] The tunnel is divided into multiple monitoring units, and the structural difference data between the periodic state data and the health benchmark model and the characteristic difference data between the periodic actual state data and the health benchmark model are calculated in each monitoring unit.

[0045] A comprehensive difference value for each monitoring unit is generated based on the structural difference data and feature difference data.

[0046] Based on multi-level health thresholds, the comprehensive difference value is divided into four health levels, and a corresponding color is assigned to each of the four health levels to form a tunnel health distribution map.

[0047] Furthermore, the prediction of the tunnel's health status trend based on the health distribution map and the health status determination rule base includes:

[0048] Obtain a health status distribution map and historical comprehensive difference values ​​over a continuous preset time period;

[0049] The historical comprehensive difference value is linearly fitted, and based on the fitting result, it is determined whether the comprehensive difference value within a preset time period exceeds the health threshold of the health status judgment rule base.

[0050] If the health threshold is exceeded, tunnel defects will be predicted within a preset timeframe, and maintenance suggestions and alarms will be sent.

[0051] If the health threshold is not exceeded, a health status trend prediction will be made for the tunnel in the next cycle.

[0052] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0053] This invention establishes a health benchmark model and judgment rule base by collecting initial data through a track-mounted mobile detection unit. Combined with distributed static monitoring nodes to collect and process data in real time, the health status is determined based on quantitative difference values. When the tunnel is healthy, the benchmark data is updated; when defects are found, suggestions are pushed and alarms are issued. Furthermore, a health distribution map is generated through periodic scanning to predict trends. This invention achieves dynamic monitoring, accurate judgment, and trend prediction of the tunnel's health status, improving the comprehensiveness and timeliness of monitoring. It helps to promptly detect and handle defects, ensuring the safe operation of the tunnel.

[0054] On the other hand, this application also provides a computer vision-based tunnel health monitoring system, applied to the aforementioned computer vision-based tunnel health monitoring method, including: a data detection module and a control analysis module;

[0055] The data detection module is used to set up a track-mounted mobile detection unit at the top of the tunnel and deploy distributed static monitoring nodes on the inner wall of the tunnel. It is also used to collect initial state data of the tunnel and original image data of the inner wall of the tunnel during the initial stage of operation; to scan the tunnel within a preset period to obtain periodic state data of the tunnel; and to transmit the collected data to the control module.

[0056] The control and analysis module is used to establish a tunnel health benchmark model based on the initial state data of the tunnel, and to establish a tunnel health status judgment rule base according to the tunnel health benchmark model; it is also used to perform preliminary processing on the original image data to obtain actual tunnel state data, and to record and save the periodic actual state data within a preset period.

[0057] The control and analysis module also determines the tunnel's health status based on the quantitative difference between the actual state data and the health benchmark model, as well as the health status determination rule base; it also generates a health distribution map based on the periodic state data, periodic actual state data, and the health benchmark model; and it predicts the tunnel's health status trend based on the health distribution map and the health status determination rule base.

[0058] It is understandable that the aforementioned computer vision-based tunnel health monitoring system and method have the same beneficial effects, and will not be elaborated further here. Attached Figure Description

[0059] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0060] Figure 1 A flowchart of a tunnel health monitoring method based on computer vision provided in an embodiment of the present invention. Detailed Implementation

[0061] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art. It should be noted that, unless otherwise specified, embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0062] Reference Figure 1 In some embodiments of this application, a tunnel health monitoring method based on computer vision is proposed, including:

[0063] Step S100: Collect initial tunnel status data during the initial operation phase using a track-mounted mobile detection unit at the top of the tunnel.

[0064] Step S200: Establish a tunnel health benchmark model based on the initial state data of the tunnel, and establish a tunnel health status judgment rule base according to the tunnel health benchmark model.

[0065] Step S300: Collect raw image data of the tunnel inner wall in real time through distributed static monitoring nodes on the tunnel inner wall, perform preliminary processing on the raw image data to obtain actual tunnel status data, and record and save the periodic actual status data within a preset period.

[0066] Step S400: Determine the tunnel health status based on the quantitative difference value between the actual state data and the health benchmark model, and the health status determination rule base;

[0067] When the tunnel is determined to be healthy, the actual status data is stored in the data of the tunnel health benchmark model; when the tunnel is determined to be defective, maintenance suggestions are pushed and an alarm is triggered.

[0068] Step S500: Scan the tunnel within a preset period based on the mobile detection unit to obtain tunnel periodic status data, and generate a health distribution map based on the periodic status data, actual periodic status data and health benchmark model.

[0069] Step S600: Predict the health status trend of the tunnel based on the health distribution map and the health status determination rule base.

[0070] Specifically, during the initial operation of the tunnel, or within one month after maintenance, high-precision data collection is conducted on the entire tunnel using a track-mounted mobile detection unit; high-definition images of the lining surface are captured using a 360-degree panoramic camera to generate a panoramic unfolded image of the tunnel in a defect-free state; three-dimensional point cloud data is collected using lidar to establish a benchmark three-dimensional model of the tunnel structure, including parameters such as lining thickness and flatness. Visible / infrared images are continuously collected for 72 hours using dual-spectrum monitoring nodes to record temperature distribution and material texture characteristics under normal conditions; the above data are categorized and stored according to "section-material-environmental benchmark values" to form an initial health benchmark library.

[0071] Understandably, baseline data collected during the initial operation of a tunnel or after maintenance reflects structural characteristics in a disease-free or normal state and can be used as health baseline data to form an initial health baseline database.

[0072] Understandably, the tunnel health benchmark model is a digital model built based on initial state data, which includes features such as the surface texture of the tunnel lining, three-dimensional structural dimensions, and material distribution.

[0073] Specifically, a multi-level health threshold is set based on the health benchmark model, including different diseases such as surface texture width, temperature deviation, structural size and color deviation. A multi-level health status judgment rule base is established based on the multi-level health threshold.

[0074] Specifically, dual-spectrum monitoring nodes are deployed on the inner wall of the tunnel. Each node includes: a visible light lens to capture the texture and color changes of the lining surface in real time; an infrared lens to record the temperature field distribution; and a built-in sensor-linked heating and defogging module to ensure that the lens can still provide clear imaging when the humidity is >85%. Each node is distributed at 10-meter intervals and synchronizes data through an industrial Ethernet network to form a 24-hour continuous monitoring network.

[0075] Specifically, when performing preliminary processing on the data collected in real time by the monitoring nodes, the inter-frame difference method is used to compare consecutive frame images to eliminate dynamic interference from vehicles and lights; adaptive threshold segmentation technology is used to automatically identify suspected defect areas, such as crack outlines and water stain ranges; and the location coordinates and size parameters of defects, such as crack length and width, are extracted by combining tunnel mileage and circumferential angle positioning.

[0076] Understandably, actual state data refers to characteristic parameters that, after processing, reflect the current state of the tunnel, such as crack length and temperature distribution deviation. The processed actual state data is summarized at a preset period, such as hourly, and then marked with timestamps and location information, stored as periodic actual state data.

[0077] Specifically, the periodic status data is the data obtained by the track-mounted mobile detection unit according to a preset cycle, such as scanning the entire tunnel monthly, reflecting the overall status.

[0078] Specifically, the quantitative difference value is the difference between the actual state data and the corresponding parameters of the benchmark model, such as the deviation between the actual temperature and the benchmark temperature, or the difference between the crack width and the benchmark value. By comparing the periodic state data, the periodic actual state data and the health benchmark model within a continuous preset period, the health status of the tunnel is determined. The difference value is obtained by quantifying the comparison results. By fitting the difference value, a health distribution map is generated and the health status of the tunnel within the future preset period is predicted.

[0079] Understandably, the health status trend is an inference about the future direction of the tunnel's health status based on historical data.

[0080] The above embodiments quantify the tunnel health status into comparable parameter indicators by establishing a benchmark model and a multi-level judgment mechanism, thereby achieving a shift from passive detection to proactive prevention. Initial data collection ensures the accuracy of the benchmark, while real-time monitoring and periodic verification form a three-dimensional monitoring system combining "points" and "surfaces."

[0081] As can be seen, this invention establishes a health benchmark model and judgment rule base by collecting initial data through a track-mounted mobile detection unit, and combines it with distributed static monitoring nodes to collect and process data in real time. The health status is determined based on the quantitative difference value. When the tunnel is healthy, the benchmark data is updated; when there are defects, suggestions are pushed and alarms are triggered. Then, a health distribution map is generated through periodic scanning to predict trends. This invention achieves dynamic monitoring, accurate judgment, and trend prediction of the tunnel's health status, improves the comprehensiveness and timeliness of monitoring, helps to detect and deal with defects in a timely manner, and ensures the safe operation of the tunnel.

[0082] Reference Figure 1 In some embodiments of this application, the initial state data of the tunnel includes lining surface image data and three-dimensional point cloud data; the process of establishing a tunnel health benchmark model based on the initial state data of the tunnel, and establishing a tunnel health status judgment rule base according to the tunnel health benchmark model, includes:

[0083] A tunnel lining unfolded diagram is generated based on the lining surface image;

[0084] A three-dimensional model of the tunnel structure is constructed based on the aforementioned three-dimensional point cloud data;

[0085] A tunnel health benchmark model is established based on the tunnel lining development diagram and the three-dimensional model of the tunnel structure.

[0086] Specifically, the image information of the tunnel inner wall lining surface can reflect the surface features of the lining, such as texture, color, cracks, and spalling; three-dimensional point cloud data can accurately present the spatial structure, geometric dimensions, and surface morphology of the tunnel; the image of the tunnel's annular lining surface can be converted into a planar unfolded image through projection and other methods, which facilitates the overall observation and analysis of the lining surface features; the three-dimensional model of the tunnel structure is a three-dimensional digital model built based on three-dimensional point cloud data that can intuitively display the overall structure, dimensions, and spatial relationships of the tunnel.

[0087] Understandably, the tunnel health benchmark model is a benchmark model that integrates the tunnel lining development diagram and the three-dimensional model of the tunnel structure, containing information such as the surface features and structural dimensions of the tunnel under normal conditions, serving as a reference standard for judging the health status of the tunnel; based on the tunnel health benchmark model, a set of judgment criteria and rules are set for different health states, which are used to evaluate the health status of the tunnel.

[0088] Understandably, during the initial operation phase or after maintenance, a tunnel's structure is in a relatively stable and normal state. At this time, the collected lining surface image data and 3D point cloud data can accurately reflect the tunnel's initial state. By processing this data to generate a tunnel lining unfolded diagram and a 3D model of the tunnel structure, and then fusing these two to establish a tunnel health baseline model, the initial normal state of the tunnel can be digitally fixed. The baseline model includes various characteristic parameters of the tunnel's normal state. By setting the normal range and deviation thresholds of these parameters, a standard for judging the tunnel's health can be established, thus providing a basis for subsequent tunnel health monitoring.

[0089] Reference Figure 1 In some embodiments of this application, the process of establishing a tunnel health benchmark model based on the tunnel initial state data and establishing a tunnel health status determination rule base according to the tunnel health benchmark model further includes:

[0090] Feature parameters of different lining materials are extracted from the tunnel health benchmark model, and an initial tunnel health benchmark library is established based on the feature parameters.

[0091] Based on the tunnel initial health benchmark library, a multi-level health threshold is set, and a multi-level health status determination rule library is established according to the multi-level health threshold.

[0092] Specifically, the feature parameters include: data parameters of surface texture, color, temperature and structural dimensions of linings made of different materials;

[0093] Image data of different lining materials were selected from the tunnel health benchmark model, and the dataset was divided according to material type;

[0094] Extract the feature parameters of the dataset and calculate the baseline range of the feature parameters;

[0095] The data of the material type and its corresponding characteristic parameters and reference range are stored to form an initial health reference library.

[0096] Specifically, a multi-level health threshold is set based on the benchmark range, and the multi-level health threshold includes a first-level health threshold, a second-level health threshold, a third-level health threshold, and a fourth-level health threshold.

[0097] The multi-level health status determination rule base includes Level 1 health status, Level 2 health status, Level 3 health status, and Level 4 health status;

[0098] The four health status levels are established based on the corresponding health thresholds for the four levels.

[0099] It is understandable that the tunnel health status determination rule base is a multi-level health status determination rule base. Each different level of health threshold includes four levels of health thresholds, and each different level of health status determination rule base includes four different levels of health status.

[0100] Reference Figure 1 In some embodiments of this application, the original image data of the tunnel inner wall includes: image data of the texture, color, temperature, and structural dimensions of the tunnel inner wall;

[0101] The process of acquiring raw image data of the tunnel interior wall in real time through distributed static monitoring nodes, and performing preliminary processing on the raw image data to obtain actual tunnel condition data includes:

[0102] Collect raw image data of the texture, color, temperature, and structural dimensions of the tunnel interior walls;

[0103] An image denoising algorithm was used to remove noise from the original image, and an image segmentation algorithm was used to divide the image data of lining regions with different materials.

[0104] Feature parameters are extracted from image data of lining areas with different materials to obtain the actual state data of the tunnel.

[0105] Understandably, distributed monitoring ensures full coverage and continuous updates of raw image data, avoiding the limitations of manual inspections and enabling timely capture of sudden disease characteristics. Structured actual condition data provides a quantitative basis for health status assessment, realizing the transformation from qualitative observation to quantitative analysis and laying the foundation for the automation and intelligence of tunnel health monitoring.

[0106] Reference Figure 1 In some embodiments of this application, the process of determining the tunnel health status based on the quantitative difference between the actual state data and the health benchmark model, and the health status determination rule base, includes:

[0107] The characteristic parameters of the actual status data of a certain monitoring node are compared and quantified with the characteristic parameters of the health benchmark model to obtain different quantification difference values;

[0108] When all quantitative difference values ​​are within the corresponding first-level health threshold, the monitoring node is determined to be healthy, and the actual status data of the monitoring node is stored in the health benchmark model.

[0109] If any quantitative difference value is greater than the corresponding first-level health threshold but less than the second-level health threshold, the monitoring node is determined to be abnormal.

[0110] When any quantitative difference value is greater than the corresponding level health threshold but less than the level 3 health threshold, the monitoring node is determined to be severely abnormal.

[0111] When any quantitative difference value is greater than the corresponding level health threshold but less than the level 3 health threshold, it is determined to be a disease at that monitoring node.

[0112] When a monitoring node is determined to be abnormal, severely abnormal, or diseased, its location information is recorded and prevention suggestions and alarms are sent.

[0113] Specifically, the quantitative difference value is the specific numerical difference of each characteristic parameter obtained by comparing the characteristic parameters of the actual state data of the monitoring node with the corresponding parameters in the health benchmark model one by one, and then calculating the difference or deviation rate. Examples include the difference in texture width, the deviation in color, the temperature difference, and the deviation in structural size.

[0114] Specifically, when the quantitative difference values ​​of all feature parameters are within the corresponding first-level health threshold, the tunnel area corresponding to the monitoring node is determined to be healthy. At the same time, the actual status data of the monitoring node is stored in the health benchmark model to realize the dynamic update of the benchmark model to adapt to the normal aging of the tunnel.

[0115] Specifically, the first-level health threshold is the upper limit of the normal fluctuation of the feature parameters in the health benchmark model. When the quantified difference value is within this threshold, it indicates that the tunnel area corresponding to the monitoring node is in a healthy state.

[0116] The secondary health threshold is a boundary value that is higher than the primary health threshold. When the quantified difference value exceeds the primary health threshold but is lower than this threshold, it indicates that there is a slight anomaly in the tunnel area.

[0117] The Level 3 health threshold is the boundary value that is higher than the Level 2 health threshold. When the quantitative difference value exceeds the Level 2 health threshold but is lower than this threshold, it means that a severe anomaly has occurred in the tunnel area.

[0118] The Level 4 health threshold is the boundary value that is higher than the Level 3 health threshold. When the quantitative difference value exceeds this threshold, it indicates that there is damage in the tunnel area.

[0119] Understandably, when a monitoring node is determined to be abnormal, severely abnormal, or defective, the system automatically records the specific location information of that monitoring node, such as tunnel mileage and circumferential angle. Simultaneously, based on different status levels, corresponding preventative recommendations are pushed out. For example, in abnormal states, it is recommended to increase the monitoring frequency; in severely abnormal or defective states, it is recommended to arrange for maintenance personnel to conduct on-site verification and activate alarm devices, such as audible and visual alarms and system pop-up alarms, to prompt relevant personnel to handle the situation promptly.

[0120] The above embodiments determine health status by quantifying differences, avoiding biases in subjective judgment and making the results more objective and accurate, thus improving the reliability of tunnel health monitoring. The tiered health status assessment and corresponding handling measures enable early detection and handling of tunnel anomalies. Different levels of recommendations and alarms also allow relevant personnel to take targeted measures, effectively reducing the risk of tunnel safety accidents.

[0121] Reference Figure 1 In some embodiments of this application, the process of generating a health distribution map based on the periodic state data, the periodic actual state data, and the health benchmark model includes:

[0122] The tunnel is divided into multiple monitoring units, and the structural difference data between the periodic state data and the health benchmark model and the characteristic difference data between the periodic actual state data and the health benchmark model are calculated in each monitoring unit.

[0123] A comprehensive difference value for each monitoring unit is generated based on the structural difference data and feature difference data.

[0124] Specifically, the panoramic images and 3D point cloud data in the periodic status data are mapped to the spatial coordinates of the health benchmark model; the location information of each monitoring node and the feature parameters of the corresponding area in the periodic actual status data are extracted, and the location information is mapped to the spatial coordinates of the benchmark model.

[0125] Specifically, for each monitoring unit, the corresponding periodic state data is retrieved and compared with the health benchmark model. Structural difference data, such as the deviation value of lining flatness and the average deviation of the three-dimensional point cloud, are calculated through three-dimensional point cloud registration and size measurement. At the same time, feature parameters in the actual periodic state data within the unit are extracted and compared with the feature parameters of the corresponding material in the health benchmark model to obtain feature difference data.

[0126] Specifically, different weights are set according to the degree of impact of structural difference data and feature difference data on tunnel health, such as a structural difference weight of 0.6 and a feature difference weight of 0.4. The comprehensive difference value of each monitoring unit is calculated by weighted summation, and the formula is: Comprehensive difference value = structural difference data × structural weight + feature difference data × feature weight.

[0127] Based on multi-level health thresholds, the comprehensive difference value is divided into four health levels, and a corresponding color is assigned to each of the four health levels to form a tunnel health distribution map.

[0128] Specifically, the overall difference value of each monitoring unit is compared with four health threshold levels. When the overall difference value is within the first-level health threshold, it is determined to be in a healthy state; between the first and second-level thresholds, it is in an abnormal state; between the second and third-level thresholds, it is in a severely abnormal state; and exceeding the third-level threshold, it is in a disease state. Different colors are assigned to the four health levels, such as green for healthy, yellow for abnormal, orange for severely abnormal, and red for disease. The corresponding colors are marked on the tunnel plan according to the location of the monitoring units, forming a tunnel health distribution map.

[0129] Reference Figure 1 In some embodiments of this application, predicting the health status trend of the tunnel based on the health distribution map and the health status determination rule base includes:

[0130] Obtain a health status distribution map and historical comprehensive difference values ​​over a continuous preset time period;

[0131] The historical comprehensive difference value is linearly fitted, and based on the fitting result, it is determined whether the comprehensive difference value within a preset time period exceeds the health threshold of the health status judgment rule base.

[0132] If the health threshold is exceeded, tunnel defects will be predicted within a preset timeframe, and maintenance suggestions and alarms will be sent.

[0133] If the health threshold is not exceeded, a health status trend prediction will be made for the tunnel in the next cycle.

[0134] Specifically, the continuous preset time health distribution map is a series of tunnel health distribution maps generated over a set continuous time period, such as the past 6 months, at certain cycles, such as monthly. These distribution maps record the changes in the health status of each monitoring unit of the tunnel over time. The historical comprehensive difference value is the stored comprehensive difference value data of each monitoring unit over multiple past periods, which is a quantitative indicator reflecting the historical changes in the tunnel's health status.

[0135] Specifically, for each monitoring unit, data points are marked on a coordinate system with time as the horizontal axis and historical comprehensive difference values ​​as the vertical axis. A linear regression algorithm is used for linear fitting to obtain the fitted line equation (e.g., y = kx + b, where y is the comprehensive difference value, x is time, k is the slope, and b is the intercept). When the slope k is positive and has a large absolute value, it indicates that the comprehensive difference value increases rapidly over time; when the slope is close to 0, it indicates that the difference value is basically stable.

[0136] Specifically, a future preset time, such as the next 6 months, is substituted into the fitted linear equation to calculate the predicted comprehensive difference value at that time point.

[0137] Understandably, the predicted value is compared with the three-level health threshold in the health status determination rule base: if the predicted comprehensive difference value exceeds the three-level health threshold, it is determined that the tunnel area corresponding to the monitoring unit will experience defects within a preset timeframe, and the system immediately generates targeted maintenance suggestions and triggers an alarm mechanism. If the predicted comprehensive difference value does not exceed the three-level health threshold, it indicates that the monitoring unit can remain in a healthy or slightly abnormal state within the preset timeframe, and the system does not push maintenance suggestions or alarms to it. Instead, it waits for the next monitoring cycle to obtain a new health distribution map and comprehensive difference value before re-predicting the health status trend.

[0138] The above embodiments, by predicting in advance whether defects will occur within a preset timeframe, can buy time for tunnel maintenance, allowing managers to formulate maintenance plans and allocate resources in advance, thus shifting from "post-event maintenance" to "pre-event prevention" and reducing losses caused by defects. By combining health distribution maps and health status judgment rule bases, trend predictions are kept consistent with health status judgment standards, ensuring the reliability of prediction results and providing strong support for the safe and stable operation of tunnels.

[0139] In another preferred embodiment based on the above embodiments, this embodiment provides a tunnel health monitoring system based on computer vision, which is applied to the above-mentioned tunnel health monitoring method based on computer vision, and includes: a detection module and a control prediction module;

[0140] The detection module is used to set up a track-mounted mobile detection unit at the top of the tunnel and deploy distributed static monitoring nodes on the inner wall of the tunnel. It is also used to collect initial state data of the tunnel and raw image data of the inner wall of the tunnel during the initial stage of operation; to scan the tunnel within a preset period to obtain periodic state data of the tunnel; and to transmit the collected data to the control module.

[0141] The control prediction module is used to establish a tunnel health benchmark model based on the tunnel's initial state data, and to establish a tunnel health status judgment rule base based on the tunnel health benchmark model.

[0142] The control module is also used to establish a tunnel health benchmark model based on the initial state data of the tunnel, and to establish a tunnel health status judgment rule base according to the tunnel health benchmark model; it is also used to perform preliminary processing on the original image data to obtain the actual state data of the tunnel, and to record and save the periodic actual state data within a preset period.

[0143] The control prediction module is also used to determine the tunnel's health status based on the quantitative difference between the actual state data and the health benchmark model, as well as the health status determination rule base; it also generates a health distribution map based on periodic state data, periodic actual state data, and the health benchmark model; and it predicts the tunnel's health status trend based on the health distribution map and the health status determination rule base.

[0144] It is understandable that the aforementioned computer vision-based tunnel health monitoring system and method have the same beneficial effects, and will not be elaborated further here.

[0145] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program goods. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program goods embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0146] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program goods according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0147] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0148] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0149] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A tunnel health monitoring method based on computer vision, characterized in that, include: Initial tunnel status data during the initial operation phase is collected by a track-mounted mobile detection unit at the top of the tunnel. A tunnel health baseline model is established based on the initial state data of the tunnel, and a tunnel health status determination rule base is established based on the tunnel health baseline model. The raw image data of the tunnel inner wall is collected in real time by distributed static monitoring nodes on the tunnel inner wall, and the raw image data is preliminarily processed to obtain the actual state data of the tunnel. The actual state data of the period within a preset period is recorded and saved. The tunnel health status is determined based on the quantitative difference between the actual state data and the health benchmark model, as well as the health status determination rule base. When the tunnel is determined to be healthy, the actual state data is stored in the data of the tunnel health benchmark model. When tunnel defects are identified, maintenance suggestions will be sent and an alarm will be triggered; The tunnel is scanned within a preset period based on the mobile detection unit to obtain tunnel periodic status data, and a health distribution map is generated based on the periodic status data, actual periodic status data and health benchmark model. The health status trend of the tunnel is predicted based on the health distribution map and the health status determination rule base.

2. The tunnel health monitoring method based on computer vision according to claim 1, characterized in that, The initial state data of the tunnel includes lining surface image data and three-dimensional point cloud data; The process of establishing a tunnel health baseline model based on the initial tunnel state data, and establishing a tunnel health status determination rule base based on the tunnel health baseline model, includes: A tunnel lining unfolded diagram is generated based on the lining surface image; A three-dimensional model of the tunnel structure is constructed based on the aforementioned three-dimensional point cloud data; A tunnel health benchmark model is established based on the tunnel lining development diagram and the three-dimensional model of the tunnel structure.

3. The tunnel health monitoring method based on computer vision according to claim 2, characterized in that, The process of establishing a tunnel health baseline model based on the initial tunnel state data and establishing a tunnel health status determination rule base based on the tunnel health baseline model also includes: Feature parameters of different lining materials are extracted from the tunnel health benchmark model, and an initial tunnel health benchmark library is established based on the feature parameters. Based on the tunnel initial health benchmark library, a multi-level health threshold is set, and a multi-level health status determination rule library is established according to the multi-level health threshold.

4. The tunnel health monitoring method based on computer vision according to claim 3, characterized in that, The characteristic parameters include: data parameters of surface texture, color, temperature and structural dimensions of linings made of different materials; The process of extracting feature parameters of different lining materials from the tunnel health benchmark model and establishing an initial tunnel health benchmark library based on these feature parameters includes: Image data of different lining materials were selected from the tunnel health benchmark model, and the dataset was divided according to material type; Extract the feature parameters of the dataset and calculate the baseline range of the feature parameters; The data of the material type and its corresponding characteristic parameters and reference range are stored to form an initial health reference library.

5. The tunnel health monitoring method based on computer vision according to claim 4, characterized in that, The process of setting multi-level health thresholds based on the tunnel initial health benchmark library and establishing a multi-level health status determination rule library based on the multi-level health thresholds includes: Based on the benchmark range, a multi-level health threshold is set for the characteristic parameters. The multi-level health threshold includes a first-level health threshold, a second-level health threshold, a third-level health threshold, and a fourth-level health threshold. The tunnel health status determination rule base includes Level 1 health status, Level 2 health status, Level 3 health status, and Level 4 health status; The four health status levels are established based on the corresponding health thresholds for the four levels.

6. The tunnel health monitoring method based on computer vision according to claim 5, characterized in that, The original image data of the tunnel inner wall includes: image data of the texture, color, temperature, and structural dimensions of the tunnel inner wall; The process of acquiring raw image data of the tunnel interior wall in real time through distributed static monitoring nodes, and performing preliminary processing on the raw image data to obtain actual tunnel condition data includes: Collect raw image data of the texture, color, temperature, and structural dimensions of the tunnel interior walls; An image denoising algorithm was used to remove noise from the original image, and an image segmentation algorithm was used to divide the image data of lining regions with different materials. Feature parameters are extracted from image data of lining areas with different materials to obtain actual tunnel condition data.

7. The tunnel health monitoring method based on computer vision according to claim 6, characterized in that, The process of determining the tunnel's health status based on the quantitative difference between the actual state data and the health benchmark model, and using the health status determination rule base, includes: The characteristic parameters of the actual status data of a certain monitoring node are compared and quantified with the characteristic parameters of the health benchmark model to obtain different quantification difference values; When all quantitative difference values ​​are within the corresponding first-level health threshold, the monitoring node is determined to be healthy, and the actual status data of the monitoring node is stored in the health benchmark model. If any quantitative difference value is greater than the corresponding first-level health threshold but less than the second-level health threshold, the monitoring node is determined to be abnormal. When any quantitative difference value is greater than the corresponding level health threshold but less than the level 3 health threshold, the monitoring node is determined to be severely abnormal. When any quantitative difference value is greater than the corresponding level health threshold but less than the level 3 health threshold, it is determined to be a disease at that monitoring node. When a monitoring node is determined to be abnormal, severely abnormal, or diseased, its location information is recorded and prevention suggestions and alarms are sent.

8. The tunnel health monitoring method based on computer vision according to claim 7, characterized in that, The process of generating a health distribution map based on the aforementioned periodic state data, periodic actual state data, and health benchmark model includes: The tunnel is divided into multiple monitoring units, and the structural difference data between the periodic state data and the health benchmark model and the characteristic difference data between the periodic actual state data and the health benchmark model are calculated in each monitoring unit. A comprehensive difference value for each monitoring unit is generated based on the structural difference data and feature difference data. Based on multi-level health thresholds, the comprehensive difference value is divided into four health levels, and a corresponding color is assigned to each of the four health levels to form a tunnel health distribution map.

9. The tunnel health monitoring method based on computer vision according to claim 8, characterized in that, The prediction of tunnel health status trends based on health distribution maps and health status determination rule bases includes: Obtain a health status distribution map and historical comprehensive difference values ​​over a continuous preset time period; The historical comprehensive difference value is linearly fitted, and based on the fitting result, it is determined whether the comprehensive difference value within a preset time period exceeds the health threshold of the health status judgment rule base. If the health threshold is exceeded, tunnel defects will be predicted within a preset timeframe, and maintenance suggestions and alarms will be sent. If the health threshold is not exceeded, a health status trend prediction will be made for the tunnel in the next cycle.

10. A tunnel health monitoring system based on computer vision, characterized in that, The tunnel health monitoring method based on computer vision according to any one of claims 1 to 9 includes: a detection module and a control and analysis module; The detection module is used to set up a track-mounted mobile detection unit at the top of the tunnel and deploy distributed static monitoring nodes on the inner wall of the tunnel. It is also used to collect initial state data of the tunnel and original image data of the inner wall of the tunnel during the initial stage of operation; to scan the tunnel within a preset period to obtain periodic state data of the tunnel; and to transmit the collected data to the control module. The control and analysis module is used to establish a tunnel health benchmark model based on the tunnel initial state data, and to establish a tunnel health status judgment rule base according to the tunnel health benchmark model. The control module is also used to establish a tunnel health benchmark model based on the initial state data of the tunnel, and to establish a tunnel health status judgment rule base according to the tunnel health benchmark model; it is also used to perform preliminary processing on the original image data to obtain actual tunnel state data, and to record and save the periodic actual state data within a preset period. The control and analysis module is also used to determine the tunnel's health status based on the quantitative difference between the actual state data and the health benchmark model, as well as the health status determination rule base; it also generates a health distribution map based on the periodic state data, periodic actual state data, and the health benchmark model; and it predicts the tunnel's health status trend based on the health distribution map and the health status determination rule base.

Images