Small-radius curve structure deformation intelligent monitoring method and system based on intelligent camera and machine vision
By acquiring data on the curvature and water seepage distribution of tunnel tracks, extracting continuous transmission features, and performing signal processing and purification, the environmental interference problem in the identification of small-radius curves of tunnel tracks was solved, enabling accurate assessment of track safety status and identification of potential hazards, and improving the accuracy and continuity of monitoring.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-25
- Publication Date
- 2026-03-10
AI Technical Summary
Existing technologies suffer from information distortion due to environmental interference in the identification and analysis of small-radius curves in tunnel railway tracks. They are difficult to accurately identify curve types and their impact from water seepage, and lack the ability to transmit the continuity and integrity of curvature changes, thus affecting the accuracy and continuity of track safety assessments.
By acquiring curvature and water seepage distribution data through track monitoring equipment, extracting continuous transmission characteristics, using signal processing and data purification methods to separate environmental interference signals, calculating curvature gradient and water seepage influence coefficient, and generating a risk assessment report, a precise assessment of track safety status can be achieved.
This improves the accuracy and continuity of track safety monitoring, enables early identification of potential hazards, and ensures the stability and operational safety of tunnel tracks.
Smart Images
Figure CN121639766A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of rail transit information technology, and in particular to a small-radius curve structure deformation intelligent monitoring method and system based on intelligent cameras and machine vision. BACKGROUND
[0002] In the field of tunnel rail operation and maintenance, it is particularly important to ensure the stability and safety of the track structure.
[0003] In particular, in small-radius curve sections, such sections exhibit various morphological characteristics, including circular curves, transition curves, and compound curves. Due to their special geometric shapes, the track deformation and damage risk is higher, directly affecting the safety and efficiency of train operation.
[0004] Therefore, studying how to accurately identify and analyze the types of these curves and their potential risks has become an important issue that cannot be ignored in rail transit safety management.
[0005] However, current technical means for identifying and analyzing small-radius curves in tunnel rails still have obvious deficiencies.
[0006] Many methods are often limited to surface observation of track morphology, lacking dynamic capture capability for deep changes in curve geometric features, especially in complex environments.
[0007] This limitation leads to low accuracy in curve type judgment, making it more difficult to deal with interference from special factors such as water seepage in tunnels, thereby affecting comprehensive assessment of track safety status.
[0008] Focusing on technical difficulties, the identification and analysis of small-radius curves in tunnels face the core problem of complexity in the transmission of curvature changes.
[0009] As a key indicator reflecting curve morphology, the continuity and integrity of curvature changes in different sections are often disturbed by environmental factors, leading to information distortion in the identification process.
[0010] For example, water seepage in tunnels can cause local deformation of the track, resulting in deviations in the transmission of curvature changes from local to global, thereby affecting accurate judgment of curve types.
[0011] This deviation further exacerbates the difficulty of assessing the extent of water seepage, as the correlation between water seepage areas and curvature change anomalies has not been effectively revealed.
[0012] Therefore, how to capture the continuous transmission characteristics of curvature changes in the tunnel rail scenario, accurately identify the types of small-radius curves, and on this basis analyze the specific impact of water seepage on curve morphology, has become a key problem that needs to be solved.
[0013] The solution to this problem not only needs to overcome the technical challenges brought by environmental interference, but also needs to deeply explore the internal relationship between curvature change and external factors to provide reliable guarantee for track safety. SUMMARY
[0014] Therefore, the present application provides a small-radius curve structure deformation intelligent monitoring method based on intelligent cameras and machine vision, mainly comprising: The track monitoring device obtains the original curvature data and the water seepage distribution data from the surface of the tunnel track to obtain an initial curvature change sequence and a water seepage intensity map. The initial curvature change sequence represents a time sequence record of the track bending degree, and the water seepage intensity map reflects the density distribution of water seepage. According to the initial curvature change sequence, continuous transmission features are extracted, and a signal processing analysis method is used to investigate the distribution characteristics to determine the continuity and overall transmission mode of the curvature change between different road sections. The local deformation area is obtained from the water seepage intensity map. If the local deformation area and the curvature change anomaly overlap and meet the safety monitoring standard, it is marked as an abnormal interference area, and an interference area curvature deviation value is obtained. For the interference area curvature deviation value, a data purification method is used to separate different components, distinguish environmental interference signals and real curvature signals, and obtain a purified curvature change sequence. The curvature gradient is calculated from the purified curvature change sequence. If the curvature gradient meets the evaluation standard in a specific road section, it is classified as a high-risk curve type, and a curve type label is obtained. According to the curve type label and the water seepage intensity map, a correlation analysis method is used to calculate the influence coefficient of water seepage on curvature change, and an influence coefficient distribution map is obtained. From the influence coefficient distribution map, a high-influence area is extracted. If the continuity of the high-influence area meets the risk evaluation standard, a risk evaluation report is generated, which summarizes potential safety hazards and determines the overall track safety status.
[0015] The present application provides a small-radius curve structure deformation intelligent monitoring system based on intelligent cameras and machine vision, mainly comprising: The track monitoring data acquisition module is used to obtain the original curvature data and the water seepage distribution data from the surface of the tunnel track by the track monitoring device to obtain an initial curvature change sequence and a water seepage intensity map. The initial curvature change sequence represents a time sequence record of the track bending degree, and the water seepage intensity map reflects the density distribution of water seepage. a curvature transmission feature analysis module configured to extract continuous transmission features from the initial curvature change sequence, investigate distribution characteristics by using a signal processing analysis method, and determine the continuity of curvature change between different road segments and the overall transmission mode; an abnormal interference area marking module configured to obtain a local deformation area from the water seepage intensity map, mark the local deformation area as an abnormal interference area if the local deformation area overlaps with the abnormal curvature change and meets a safety monitoring standard, and obtain an interference area curvature deviation value; a curvature data purification module configured to separate different components by using a data purification method for the interference area curvature deviation value, distinguish between environmental interference signals and real curvature signals, and obtain a purified curvature change sequence; a high-risk curve classification module configured to calculate a curvature gradient from the purified curvature change sequence, classify a curve type as a high-risk curve type if the curvature gradient meets an evaluation standard at a specific road segment, and obtain a curve type label; a water seepage influence analysis module configured to calculate an influence coefficient of water seepage on curvature change by using a correlation analysis method according to the curve type label and the water seepage intensity map, and obtain an influence coefficient distribution map; a risk assessment report generation module configured to extract a high-influence area from the influence coefficient distribution map, generate a risk assessment report if the continuity of the high-influence area meets a risk assessment standard, and summarize potential safety hazards and determine an overall track safety state in the report.
[0016] The technical scheme provided by the embodiment of the present application can include the following beneficial effects: The present application discloses a tunnel rail safety monitoring method, which is aimed at the unique business scenario problem that the curvature change of rails in a tunnel environment is easily interfered by water seepage to cause deformation abnormalities, thereby causing overall track safety hazards. This problem integrates logical correlation challenges such as inaccurate data acquisition, mixed interference signals, and discontinuous risk assessment. The present application collects original curvature data and water seepage distribution data by a track monitoring device, forms an initial sequence and a map, extracts continuous transmission features and analyzes distribution characteristics, identifies a local deformation area overlapping with a curvature abnormality as an interference area, separates environmental interference by using a data purification method, obtains a purified curvature sequence, calculates a curvature gradient to classify a high-risk curve type, performs correlation analysis in combination with a water seepage intensity map, extracts a high-influence area to generate a risk assessment report, thereby accurately distinguishing between real signals and interference, and realizing quantitative evaluation of the influence of water seepage on curvature. The technical effect of the present application lies in improving the accuracy and continuity of track safety monitoring, identifying potential hazards as early as possible, and ensuring the overall stability and operation safety of tunnel rails. BRIEF DESCRIPTION OF DRAWINGS
[0017] The above and other objects, features and advantages of the present application will become more apparent from the following description of embodiments of the present application taken in conjunction with the accompanying drawings, in which: Figure 1 A flow chart of a small-radius curve structure deformation intelligent monitoring method based on intelligent cameras and machine vision according to the present application.
[0018] Figure 2 A schematic diagram of a small-radius curve structure deformation intelligent monitoring method based on intelligent cameras and machine vision according to the present application.
[0019] Figure 3 Another schematic diagram of a small-radius curve structure deformation intelligent monitoring method based on intelligent cameras and machine vision according to the present application.
[0020] Figure 4 A structural schematic diagram of a small-radius curve structure deformation intelligent monitoring system based on intelligent cameras and machine vision according to the present application.
[0021] Figure 5 A geometric shape comparison diagram of different types of small-radius curves.
[0022] Figure 6 A track deformation cross-sectional shape diagram under the influence of water seepage.
[0023] Figure 7 A spatial transmission pattern diagram of curvature variation.
[0024] Figure 8 A comparison diagram of signal processing analysis effects of curvature variation sequences.
[0025] Figure 9 A spatial distribution shape diagram of abnormal interference zones.
[0026] Figure 10 A comparison diagram of interference zone curvature data before and after purification.
[0027] Figure 11 A spatial variation pattern diagram of curvature gradient.
[0028] Figure 12 A comparison diagram of multi-dimensional risk assessment of different curve types.
[0029] Figure 13 A water seepage influence coefficient spatial distribution heat map.
[0030] Figure 14 A performance index comparison diagram of the method according to the present application and a traditional method. DETAILED DESCRIPTION
[0031] The present application is described in detail below based on examples, but the present application is not limited to these examples. In the following detailed description of the present application, some specific details are described in detail. The present application can also be fully understood without the description of these details. In order to avoid confusion of the essence of the present application, the well-known methods, processes, flows, elements and circuits are not described in detail.
[0032] In addition, those skilled in the art should understand that the drawings provided herein are for illustrative purposes only, and the drawings are not necessarily drawn to scale.
[0033] The technical solutions of the present application will be described below in detail in conjunction with examples. Obviously, the described examples are only a part of the examples of the present application, not all examples. Based on the examples in the present application, all other examples obtained by those skilled in the art without creative labor are within the scope of protection of the present application.
[0034] As shown in Figure 1 The present embodiment of a small radius curve structure deformation intelligent monitoring method and system based on intelligent camera and machine vision can specifically include: S101, obtaining original curvature data and water seepage distribution data from the surface of the tunnel track by the track monitoring device, obtaining an initial curvature change sequence and a water seepage intensity mapping, the initial curvature change sequence represents a time sequence record of the track bending degree, and the water seepage intensity mapping reflects the density distribution of water seepage.
[0035] The original curvature data and water seepage distribution data are obtained from the surface of the tunnel track by the track monitoring device, and the data is recorded in real time by using an automatic acquisition module to obtain the initial curvature change sequence and the water seepage intensity mapping. For the initial curvature change sequence, a time series analysis method is used to extract the time sequence change characteristics to determine the fluctuation of the track bending degree in different time periods. According to the time sequence change characteristics, if the fluctuation amplitude exceeds the preset threshold, an abnormal marking module is triggered to determine whether there is a potential track deformation risk area. For the water seepage intensity mapping, the density distribution is processed by layering through a spatial distribution analysis tool to obtain the distribution of high-density areas of water seepage. According to the high-density area distribution, if the water seepage intensity of some areas exceeds the predetermined standard, the specific water seepage risk points are determined by a region positioning module. By spatially superimposing and analyzing the track deformation risk area and the water seepage risk point, a logical matching method is used to determine whether there is an overlapping area between the two, and a comprehensive risk distribution result is obtained. According to the comprehensive risk distribution result, a dynamic distribution map of the risk area is generated by a data visualization tool to determine the priority of the key monitoring area.
[0036] As shown in Figure 5As shown in the attached figure, this illustration compares the geometric features of three main types of small-radius curves in a rail transit monitoring system, showcasing the unique geometric properties and technical parameters of each curve from a top-down planar view.
[0037] like Figure 6 As shown in the attached figure, this illustrates a comparison of track cross-sectional deformation under the influence of water seepage. The left side shows a standard I-shaped cross-section in its normal state, while the right side shows a deformed cross-section after water seepage erosion. It can be seen from the figure that water seepage causes the track bottom to sink by Δh and produces a tilt deformation of approximately 3°. Water from the seepage area (the bottom point-like area) penetrates upwards, weakening the track support structure and ultimately leading to irregular deformation of the cross-sectional shape.
[0038] S102. Extract continuous transmission features based on the initial curvature change sequence, use signal processing analysis methods to examine distribution characteristics, and determine the continuity and overall transmission mode of curvature change in different road segments.
[0039] For the initial curvature data, signal processing methods are used to segment and analyze the change sequence, obtaining details of curvature changes in different road segments and recording the distribution of changes between road segments. Based on the distribution of changes between road segments, a preset threshold range is used. If the curvature change of a road segment exceeds the threshold range, it is marked as an abnormal road segment, determining the distribution of potential abnormal areas. For the distribution of abnormal areas, the specific manifestations of continuous transmission characteristics are obtained, and the continuity between abnormal road segments and adjacent road segments is analyzed to determine whether a transmission trend exists. Based on the judgment of the transmission trend, spatial correlation tools are used to perform hierarchical mapping of the overall transmission pattern, obtaining the superimposed distribution of abnormal areas and transmission trends. For the superimposed distribution, a data filtering module extracts high-risk transmission paths, determining the priority order of key road segments. Based on the priority order of key road segments, an automated monitoring module tracks curvature changes in real time, obtaining dynamic change data and determining whether subsequent changes continue to deviate from the normal range. For the dynamic change data, the continuity and transmission trend are periodically updated through the data storage unit to obtain the latest road segment risk distribution status.
[0040] like Figure 7 As shown in the attached figure, this figure illustrates the continuous transmission characteristics of curvature changes between different track sections in the track monitoring system.
[0041] like Figure 8 As shown, the signal processing analysis method significantly improves the processing effect on curvature variation sequences. Subfigure (a) shows the unprocessed original curvature variation sequence. Within the road segment range of 0 to 100 meters, the curvature variation values fluctuate wildly, exhibiting a large amount of high-frequency noise and anomalous spikes, with curvature values ranging from -0.02 to 0.05 m. The irregular fluctuations between the two curves make it difficult to identify the true continuous transmission characteristics between different road segments. Subfigure (b) shows the curvature variation sequence after signal processing analysis. The noise is effectively suppressed, the curve is smooth and continuous, and the curvature transmission patterns of the three different road segments can be clearly identified: road segment 1 (0-30 m) presents a stable low curvature feature, road segment 2 (30-60 m) presents a gradual increase in curvature and a gentle curve feature, and road segment 3 (60-100 m) presents a circular curve feature with a stable curvature reaching a peak. Through the signal processing analysis method, the continuous transmission characteristics are successfully extracted, providing a reliable data basis for subsequent curve type identification and risk assessment.
[0042] S103, obtaining a local deformation area from the water seepage intensity map, if the local deformation area overlaps with the abnormal curvature change and meets the safety monitoring standard, marking it as an abnormal interference area to obtain the curvature deviation value of the interference area.
[0043] Through the water seepage intensity data acquisition system, the water seepage intensity distribution information of the target area is obtained. The data is preliminarily processed using an automatic scanning tool to obtain a water seepage intensity distribution map. The local deformation area is extracted from the water seepage intensity distribution map, and the distribution map is regionally divided using image segmentation technology to determine the boundary range of the local deformation area. For the local deformation area, the corresponding curvature change data is obtained. If the curvature change value exceeds the preset threshold range, it is marked as an abnormal area to obtain the distribution information of the abnormal area. From the abnormal area distribution information, the overlapping part with the local deformation area is extracted. If the area proportion of the overlapping part meets the preset safety monitoring standard, it is marked as an abnormal interference area to determine the specific location of the abnormal interference area. According to the location information of the abnormal interference area, the curvature deviation value in the interference area is calculated, and a pre-established curvature analysis model is used to obtain the specific value of the curvature deviation. For the curvature deviation value, the comparison result with the safety monitoring standard is obtained. If the deviation value exceeds the standard range, a priority ranking of the abnormal interference area is generated to determine the subsequent processing order. Through the priority ranking result, a processing scheme for the abnormal interference area is generated, and an automatic recording tool is used to store the processing scheme to obtain the final processing scheme list.
[0044] Referring to Figure 9 The figure shows the spatial geometric distribution characteristics of the abnormal interference area on the track. Using a top view perspective, the position, shape, and intensity distribution of the interference area on the curved track are clearly presented.
[0045] S104, for the curvature deviation value of the interference area, a data purification method is used to separate different components, distinguish between environmental interference signals and true curvature signals, and obtain a purified curvature change sequence.
[0046] For the curvature value data of the interference area, a pre-established signal decomposition method is used to preliminarily split the original data, separate the signal components that may contain environmental interference and the preliminary curvature signal part, and obtain a preliminarily separated signal set. From the preliminarily separated signal set, the signal components related to environmental interference are obtained, and by comparing a preset interference characteristic template, if the matching degree of the signal components and the template is higher than a preset threshold, it is determined that it is an interference signal part, and a signal subset labeled as interference is obtained. For the signal subset labeled as interference, a filtering processing method is used to remove the influence of the interference signal, and the residual true signal part is extracted therefrom, and signal data filtered of interference is obtained. According to the signal data filtered of interference, in combination with the original curvature value data, through a signal reconstruction technology, the true signal part and the curvature signal are integrated to determine a reconstructed curvature signal sequence. For the reconstructed curvature signal sequence, a smoothing processing method is used to optimize signal fluctuation, obtain a continuous and stable signal change trend, and obtain a smoothed curvature change sequence. From the smoothed curvature change sequence, through a time window analysis method, it is judged whether the signal change conforms to a preset reasonable range, and if the change amplitude exceeds a preset threshold, local correction is performed to obtain a final purification sequence.
[0047] As shown in Figure 10 The processing process of the data purification method for the curvature deviation value of the interference area is divided into three stages. Subgraph (a) shows the original curvature deviation value signal, which mixes environmental interference and true curvature change components, and the signal amplitude fluctuates sharply between-2.0 and 3.0, with multiple abnormal peak values appearing near 150 seconds, 300 seconds and 450 seconds, reaching 2.5 or more, which cannot be directly used for curve type judgment. Subgraph (b) shows the environmental interference signal component extracted by the signal decomposition method, which is mainly composed of local deformation interference caused by water seepage, showing irregular pulse characteristics and high-frequency oscillation, with an amplitude range of-1.5 to 2.0. Subgraph (c) shows the purified true curvature signal. After filtering processing and signal reconstruction, the interference components are effectively removed, the curvature signal is smooth and stable, the amplitude range is narrowed to-0.5 to 1.0, the signal change trend is continuous, and the accuracy requirement for subsequent curvature gradient calculation is met. Through the data purification method, the environmental interference signal and the true curvature signal are effectively separated, and the accuracy of curve type identification is significantly improved.
[0048] S105, calculating a curvature gradient from the purified curvature change sequence, and if the curvature gradient meets the evaluation standard on a specific road section, classifying it as a high-risk curve type to obtain a curve type label.
[0049] For details, see Figure 2, step one: from the purified curvature change sequence, the gradient data is extracted through the data processing tool, and the standardization processing is carried out for each group of data to obtain the normalized gradient data set. Step two: according to the normalized gradient data set, the distribution of the specific road section is segmented, the data is mapped to the corresponding road section interval by using the preset road section division rule, and the gradient distribution range of each road section is determined. Step three: if the gradient distribution range of each road section exceeds the preset evaluation standard threshold, it is marked as a potential risk road section, and the preliminary risk road section set is obtained combined with the road section analysis result. Step four: for the preliminary risk road section set, the historical data record related to the high risk class is obtained, and the similarity between the current gradient data and the historical data is compared to determine whether it meets the judgment condition of the high risk class. Step five: if it is judged as a road section meeting the high risk class, the feature data of its curve type is further extracted, the support vector machine algorithm is used for classification processing of the feature data, and the specific curve type classification result is obtained. Step six: according to the curve type classification result, the corresponding classification label is generated, and the storage and index processing are carried out for each class of label data to determine the final classification label data set. Step seven: through the final classification label data set, the structured data output of road section risk evaluation is generated, which provides queryable classification basis for subsequent business processes, and completes the whole risk evaluation process.
[0050] As Figure 11 , the figure shows the spatial geometric variation form of track curvature gradient between different road sections, and visualizes the shape transition characteristics of gradient from low to high and the gradient shape characteristics of high risk curve type.
[0051] As Figure 12As shown, the performance of the four curve types in the five risk assessment dimensions is compared and analyzed using a radar chart. The assessment dimensions include curvature gradient, deformation risk, water seepage sensitivity, monitoring difficulty, and safety level, with a score range of 0-10 for each dimension. The compound curve (dotted line + diamond marker) has the highest risk among all curve types, with a monitoring difficulty of 9.0, a curvature gradient of 8.0, a deformation risk of 8.5, a water seepage sensitivity of 7.5, and a safety level of only 3.0, with the largest overall risk profile area. The easement curve (dashed line + triangle marker) has the highest curvature gradient of 8.5, a deformation risk of 7.5, a monitoring difficulty of 7.0, and a safety level of 4.0, belonging to the high-risk curve type. The circular arc curve (dotted line + square marker) has a medium risk, with a curvature gradient of 6.0, a deformation risk of 6.5, a monitoring difficulty of 5.5, and a safety level of 5.5. The straight line segment (solid line + circle marker) as the control benchmark has the lowest risk in each dimension, with a curvature gradient of only 2.0, a deformation risk of 3.0, and a safety level of 8.5. Through multi-dimensional risk assessment comparison, differentiated monitoring strategies and maintenance plans can be developed for different curve types, with increased monitoring frequency and maintenance intensity for high-risk types such as compound curves and easement curves.
[0052] S106, according to the curve type label and the water seepage intensity mapping, an influence coefficient distribution map is obtained by using a correlation analysis method to calculate the influence of water seepage on curvature change.
[0053] The curve type label and water seepage intensity data are obtained, and the original data is cleaned and formatted by a pre-established data processing module to obtain a standardized data set. For the standardized data set, the correlation between water seepage intensity and curvature change is calculated using a correlation analysis method to determine the influence coefficient of water seepage intensity on curvature change. According to the calculated influence coefficient, combined with the water seepage intensity distribution data, a distribution chart of the influence coefficient is generated through a visualization tool to present the regional characteristics of water seepage influence. If the influence coefficient in some areas of the generated distribution chart exceeds the preset threshold, the water seepage intensity data in that area is extracted in depth to obtain a detailed data set of the high-impact area. For the detailed data set of the high-impact area, combined with the curve type label, the curvature change trend is subdivided through a classification processing module to determine the performance differences of different types of curves in the high-impact area. According to the subdivided curvature change trend, combined with the distribution characteristics of water seepage influence, a targeted analysis result is generated through a data integration tool to determine the specific influence mode of water seepage on different curve types. Through the analysis result, the influence coefficient, distribution chart, and specific influence mode are saved in a structured manner using a storage module to obtain comprehensive data records for subsequent queries.
[0054] As Figure 13The water seepage influence coefficient calculated by the correlation analysis method shows a significant regional distribution characteristic in the two-dimensional space composed of the longitudinal position (0-100 m) and the transverse position (0-10 m) of the track. The heat map uses gray gradient to represent the size of the influence coefficient. The dark area represents a high influence coefficient, and the light area represents a low influence coefficient. The statistical results show that the influence coefficient ranges from 0.147 to 0.826, with an average influence coefficient of 0.319. The black solid line in the figure marks the high-risk threshold area with an influence coefficient exceeding 0.6. Three main high-influence areas are identified: the first area is located at 20-35 m longitudinally and 4-7 m transversely, with an influence coefficient exceeding 0.75; the second area is located at 50-70 m longitudinally and 2-6 m transversely, with a maximum influence coefficient of 0.826; the third area is located at 85-95 m longitudinally and 5-8 m transversely, with an influence coefficient of about 0.68. The high-influence area accounts for 16.40% of the total track area. These areas have a greater seepage intensity and a significant impact on the curvature change, and need to be the focus of monitoring and maintenance. Through the influence coefficient distribution map, the spatial influence pattern of seepage on track curvature change can be intuitively identified, providing quantitative basis for risk assessment.
[0055] S107, extracting a high-influence area from the influence coefficient distribution map, and if the continuity of the high-influence area meets the risk assessment standard, generating a risk assessment report summarizing potential safety hazards and determining the overall track safety status.
[0056] For details Figure 3 , by obtaining key data from the influence coefficient distribution map, performing regional division processing on the data, and obtaining the preliminary range of the high-influence area. According to the preliminary range of the high-influence area, using spatial analysis tools to detect the continuity in the area, and judging whether the continuity meets the preset evaluation standard. If the continuity meets the preset evaluation standard, further digging into the data in the high-influence area to obtain the distribution information of potential safety hazards. By classifying and arranging the distribution information of potential safety hazards, determining the priority order of the hazards, and obtaining the detailed list of hazards. According to the detailed list of hazards, using a logistic regression model to predict the overall status of track safety, and judging the level of overall safety status. According to the level of overall safety status, generating a corresponding risk assessment report, summarizing the classification and priority of potential safety hazards, and determining the final safety assessment result. If the level of overall safety status is lower than the preset threshold, generating special processing suggestions for high-priority items in the hazard list to obtain the data basis for subsequent tracking.
[0057] As Figure 14As shown, the method of the present application is compared with the traditional manual inspection and simple sensor monitoring method in four key performance indicators. In terms of recognition accuracy, the traditional manual inspection is only 72.5%, the simple sensor monitoring is 85.2%, and the method of the present application reaches 96.8%, which is 33.5 percentage points higher than the traditional manual inspection and 11.6 percentage points higher than the simple sensor monitoring. In terms of false alarm rate, the traditional manual inspection is 18.3%, the simple sensor monitoring is 12.5%, and the method of the present application is reduced to 2.1%, which is 88.5% lower than the false alarm rate, significantly reducing the number of false alarms. In terms of the missing rate, the traditional manual inspection is 15.2%, the simple sensor monitoring is 8.7%, and the method of the present application is reduced to 1.8%, which is 88.2% lower than the missing rate, effectively avoiding the omission of potential safety hazards. In terms of processing timeliness, the relative value of the traditional manual inspection is only 45, the simple sensor monitoring is 75, and the method of the present application reaches 95, which is 111.1% higher than the processing efficiency, realizing nearly real-time monitoring response. Through performance comparison, it can be seen that the method of the present application comprehensively uses intelligent cameras, machine vision, signal processing and data purification technologies, and is significantly superior to traditional methods in terms of recognition accuracy, reliability and real-time performance, and can provide more accurate and efficient technical support for the safety monitoring of small radius curves of tunnel tracks.
[0058] Referring to Figure 4 The present application also provides a small radius curve structure deformation intelligent monitoring system based on intelligent cameras and machine vision, mainly comprising: A track monitoring data acquisition module is used to acquire original curvature data and water seepage distribution data from the surface of the tunnel track by a track monitoring device to obtain an initial curvature change sequence and a water seepage intensity map, wherein the initial curvature change sequence represents a time sequence record of the track bending degree, and the water seepage intensity map reflects the density distribution of water seepage. A curvature transmission feature analysis module is used to extract continuous transmission features from the initial curvature change sequence, investigate the distribution characteristics by using a signal processing analysis method, and determine the continuity and overall transmission mode of the curvature change between different road sections. An abnormal interference area marking module is used to acquire a local deformation area from the water seepage intensity map, and if the local deformation area and the curvature change abnormality overlap and meet the safety monitoring standard, it is marked as an abnormal interference area to obtain an interference area curvature deviation value. A curvature data purification module is used to separate different components by using a data purification method for the interference area curvature deviation value, distinguish between environmental interference signals and real curvature signals, and obtain a purified curvature change sequence. A high-risk curve classification module is used to calculate a curvature gradient from the purified curvature change sequence, and if the curvature gradient meets the evaluation standard in a specific road section, it is classified as a high-risk curve type to obtain a curve type label. a water seepage influence analysis module configured to calculate an influence coefficient of water seepage on curvature change according to the curve type label and the water seepage intensity map by using a correlation analysis method, and obtain an influence coefficient distribution map; a risk assessment report generation module configured to extract a high-influence area from the influence coefficient distribution map, and generate a risk assessment report if the continuity of the high-influence area meets a risk assessment standard, wherein the report summarizes potential safety hazards and determines an overall track safety state.
[0059] In addition, it should be noted that various technical features described in the foregoing embodiments can be combined in any appropriate way, provided that there is no conflict. In order to avoid unnecessary repetition, the present application will not describe various possible combinations again. Furthermore, various embodiments of the present application can be combined in any appropriate way, provided that there is no conflict. Such combinations should also be considered as disclosed in the present application.
Claims
1. A small-radius curve structure deformation intelligent monitoring method based on intelligent cameras and machine vision, characterized in that, The method comprises: obtaining original curvature data and water seepage distribution data from the surface of the tunnel track by a track monitoring device, to obtain an initial curvature change sequence and a water seepage intensity map, wherein the initial curvature change sequence represents a time sequence record of the track bending degree, and the water seepage intensity map reflects the density distribution of water seepage; extracting continuous transfer features from the initial curvature change sequence, investigating the distribution characteristics by using a signal processing analysis method, and determining the continuity and overall transfer mode of the curvature change between different road sections; obtaining a local deformation area from the water seepage intensity map, and if the local deformation area overlaps with the abnormal curvature change and meets the safety monitoring standard, marking it as an abnormal interference area to obtain an interference area curvature deviation value; for the interference area curvature deviation value, using a data purification method to separate different components, distinguishing between environmental interference signals and real curvature signals, and obtaining a purified curvature change sequence; calculating a curvature gradient from the purified curvature change sequence, and if the curvature gradient meets the evaluation standard in a specific road section, classifying it as a high-risk curve type to obtain a curve type label; according to the curve type label and the water seepage intensity map, using a correlation analysis method to calculate the influence coefficient of water seepage on the curvature change, and obtaining an influence coefficient distribution map; extracting a high-influence area from the influence coefficient distribution map, and if the continuity of the high-influence area meets the risk evaluation standard, generating a risk evaluation report, which summarizes potential safety hazards and determines the overall track safety status.
2. The method according to claim 1, wherein, The method comprises: obtaining original curvature data and water seepage distribution data from the surface of the tunnel track by a track monitoring device, to obtain an initial curvature change sequence and a water seepage intensity map, wherein the initial curvature change sequence represents a time sequence record of the track bending degree, and the water seepage intensity map reflects the density distribution of water seepage; for the initial curvature change sequence, using a time series analysis method to extract time sequence change features, and determining the fluctuation of the track bending degree in different time periods; according to the time sequence change features, if the fluctuation amplitude exceeds a preset threshold, triggering an abnormality marking module to determine whether there is a potential track deformation risk area; for the water seepage intensity map, using a spatial distribution analysis tool to perform hierarchical processing on the density distribution to obtain the distribution of high-density water seepage areas; according to the high-density area distribution, if the water seepage intensity of some areas exceeds a predetermined standard, using a region positioning module to determine specific water seepage risk points; by spatially superimposing and analyzing the track deformation risk area and the water seepage risk points, using a logical matching method to determine whether there is an overlapping area between the two, to obtain a comprehensive risk distribution result; according to the comprehensive risk distribution result, using a data visualization tool to generate a dynamic distribution map of the risk area to determine the priority of the key monitoring area.
3. The method of claim 1, wherein the method is a small-radius curve structure deformation intelligent monitoring method based on an intelligent camera and machine vision. The continuous transmission characteristics are extracted according to the initial curvature change sequence, signal processing analysis method is used to investigate the distribution characteristics, the continuity of the curvature change between different road sections and the overall transmission mode are determined, including: For the initial curvature data, the change sequence is segmented and analyzed by the signal processing method, the curvature change details of different road sections are obtained, and the change distribution record between road sections is obtained; According to the change distribution record between road sections, if the curvature change of a road section exceeds the threshold range, it is marked as an abnormal road section, and the distribution of the potential abnormal area is determined; For the distribution of the abnormal area, the specific form of the continuous transmission characteristics is obtained, the continuity degree between the abnormal road section and the adjacent road section is analyzed, and whether there is a transmission trend is judged; According to the judgment result of the transmission trend, the overall transmission mode is mapped by using the spatial correlation tool, and the superimposed distribution of the abnormal area and the transmission trend is obtained; For the superimposed distribution, the high-risk transmission path is extracted by the data screening module, and the priority order of the road section to be focused on is determined; According to the priority order of the road section to be focused on, the curvature change is tracked in real time by using the automatic monitoring module, the dynamic change data is obtained, and whether the subsequent change deviates from the normal range is judged; For the dynamic change data, the continuity degree and the transmission trend are periodically updated by the data storage unit, and the latest road section risk distribution state is obtained.
4. The method of claim 1, wherein the method is a small-radius curve structure deformation intelligent monitoring method based on an intelligent camera and machine vision. The local deformation area is obtained from the water seepage intensity mapping, and if the local deformation area and the curvature change abnormal overlap meet the safety monitoring standard, it is marked as an abnormal interference area, the curvature deviation value of the interference area is obtained, including: Through the water seepage intensity data acquisition system, the water seepage intensity distribution information of the target area is obtained, and the data is preliminarily processed by using the automatic scanning tool to obtain the water seepage intensity distribution map; The local deformation area is extracted from the water seepage intensity distribution map, the distribution map is regionally divided by using image segmentation technology, and the boundary range of the local deformation area is determined; For the local deformation area, the corresponding curvature change data is obtained, and if the curvature change value exceeds the preset threshold range, it is marked as an abnormal area, and the distribution information of the abnormal area is obtained; From the abnormal area distribution information, the overlapping part with the local deformation area is extracted, and if the area proportion of the overlapping part meets the preset safety monitoring standard, it is marked as an abnormal interference area, and the specific position of the abnormal interference area is determined; According to the position information of the abnormal interference area, the curvature deviation value in the interference area is calculated, and the specific numerical value of the curvature deviation is obtained by using the pre-established curvature analysis model; For the curvature deviation value, the comparison result with the safety monitoring standard is obtained, and if the deviation value exceeds the standard range, the priority order of the abnormal interference area is generated, and the subsequent processing order is determined; Through the priority order result, the processing scheme distribution of the abnormal interference area is generated, the processing scheme is stored by using the automatic recording tool, and the final processing scheme list is obtained.
5. The method of claim 1, wherein the method is a small-radius curve structure deformation intelligent monitoring method based on an intelligent camera and machine vision. For the curvature deviation value of the interference area, different components are separated by using the data purification method, the environmental interference signal and the real curvature signal are distinguished, and the purified curvature change sequence is obtained, including: For the curvature value data of the interference area, a pre-established signal decomposition method is used to preliminarily split the original data, separate the signal components that may contain environmental interference and the preliminary curvature signal part, and obtain a preliminarily separated signal set; From the preliminarily separated signal set, the signal components related to environmental interference are obtained, and by comparing the preset interference characteristic template, if the matching degree of the signal components and the template is higher than the preset threshold, it is determined as the interference signal part, and a signal subset labeled as interference is obtained; For the signal subset labeled as interference, a filtering processing method is used to remove the influence of the interference signal, and the residual true signal part is extracted therefrom, to obtain signal data filtered from the interference; According to the signal data filtered from the interference, the true signal part is integrated with the curvature signal to determine the reconstructed curvature signal sequence; For the reconstructed curvature signal sequence, a smoothing processing method is used to optimize the signal fluctuation, to obtain a continuous and stable signal change trend, and a smoothed curvature change sequence is obtained; From the smoothed curvature change sequence, a time window analysis method is used to determine whether the signal change conforms to the preset reasonable range, and if the change amplitude exceeds the preset threshold, local correction is performed to obtain a final purification sequence.
6. The method of claim 1, wherein the method is a small-radius curve structure deformation intelligent monitoring method based on an intelligent camera and machine vision. The curvature gradient is calculated from the purified curvature change sequence, and if the curvature gradient meets the evaluation standard at a specific road section, it is classified as a high-risk curve type to obtain a curve type label, including: Step one: From the purified curvature change sequence, gradient data is extracted by a data processing tool, and each group of data is standardized to obtain a normalized gradient data set; Step two: According to the normalized gradient data set, the distribution of the specific road section is processed in sections, and the data is mapped to the corresponding road section interval using a preset road section division rule to determine the gradient distribution range of each road section; Step three: If the gradient distribution range of each road section exceeds the preset evaluation standard threshold, it is marked as a potential risk road section, and the preliminary risk road section set is obtained in combination with the road section analysis result; Step four: For the preliminary risk road section set, the historical data records related to the high-risk class are obtained, and the similarity between the current gradient data and the historical data is compared to determine whether the judgment condition of the high-risk class is met; Step five: If it is determined that the road section meets the high-risk class, the feature data of the curve type is further extracted, and a support vector machine algorithm is used to classify the feature data to obtain a specific curve type classification result; Step six: According to the curve type classification result, the corresponding classification label is generated, and each class of label data is stored and indexed to determine the final classification label data set; Step seven: Through the final classification label data set, a structured data output of road section risk assessment is generated to provide a queryable classification basis for subsequent business processes, and the entire risk assessment process is completed.
7. The method of claim 1, wherein the method is a small-radius curve structure deformation intelligent monitoring method based on an intelligent camera and machine vision. According to the curve type label and the water seepage intensity mapping, a correlation analysis method is used to calculate the influence coefficient of water seepage on curvature change to obtain an influence coefficient distribution map, including: The curve type label and the water seepage intensity data are acquired, original data is processed by pre-established data processing modules to clean and format, and a standardized data set is obtained; For the standardized data set, the correlation between water seepage intensity and curvature change is calculated by using a correlation analysis method, and an influence coefficient of water seepage intensity on curvature change is determined; According to the calculated influence coefficient, combined with the water seepage intensity distribution data, a distribution chart of the influence coefficient is generated by a visualization tool to present the regional characteristics of water seepage influence; If the influence coefficient of some areas in the generated distribution chart exceeds a preset threshold value, the water seepage intensity data of the area is extracted in depth to obtain a detailed data set of a high-influence area; For the detailed data set of the high-influence area, combined with the curve type label, the curvature change trend is subdivided by a classification processing module to judge the performance difference of different types of curves in the high-influence area; According to the subdivided curvature change trend, combined with the distribution characteristics of water seepage influence, a targeted analysis result is generated by a data integration tool to determine the specific influence mode of water seepage on different curve types; Through the analysis result, the influence coefficient and the distribution chart and the specific influence mode are stored by a storage module to obtain comprehensive data records for subsequent query.
8. The method of claim 1, wherein the method is a small-radius curve structure deformation intelligent monitoring method based on an intelligent camera and machine vision. The high-influence area is extracted from the influence coefficient distribution chart, and if the continuity of the high-influence area meets the risk assessment standard, a risk assessment report is generated, which summarizes potential safety hazards and determines the overall track safety state, including: By obtaining key data from the influence coefficient distribution chart, the data is processed by area division to obtain the preliminary range of the high-influence area; According to the preliminary range of the high-influence area, a spatial analysis tool is used to detect the continuity in the area to determine whether the continuity meets the preset evaluation standard; If the continuity meets the preset evaluation standard, the data in the high-influence area is deeply mined to obtain the distribution information of potential safety hazards; By classifying and arranging the distribution information of potential safety hazards, the priority order of the hazards is determined to obtain a detailed list of hazards; For the detailed list of hazards, a logistic regression model is used to predict the overall state of track safety to determine the level of overall safety state; According to the level of overall safety state, a corresponding risk assessment report is generated to summarize the classification and priority of potential safety hazards and determine the final safety assessment result; If the level of overall safety state is lower than a preset threshold value, a special processing suggestion is generated for high-priority items in the hazard list to obtain data basis for subsequent tracking.
9. A small-radius curve structure deformation intelligent monitoring system based on intelligent cameras and machine vision, characterized in that, The system comprises: A track monitoring data acquisition module is configured to acquire original curvature data and water seepage distribution data from a tunnel rail surface by a track monitoring device to obtain an initial curvature change sequence and a water seepage intensity mapping, the initial curvature change sequence represents a time sequence record of the degree of track bending, and the water seepage intensity mapping reflects the density distribution of water seepage; The curvature transmission feature analysis module is configured to extract continuous transmission features according to the initial curvature change sequence, investigate distribution characteristics by using a signal processing analysis method, and determine the continuity of curvature change between different road sections and the overall transmission mode. The abnormal interference area marking module is configured to obtain a local deformation area from the water seepage intensity map, mark an abnormal interference area if the local deformation area and the abnormal curvature change overlap and meet a safety monitoring standard, and obtain a curvature deviation value of the interference area. The curvature data purification module is configured to separate different components by using a data purification method, distinguish environmental interference signals from real curvature signals, and obtain a purified curvature change sequence for the curvature deviation value of the interference area. The high-risk curve classification module is configured to calculate a curvature gradient from the purified curvature change sequence, classify a high-risk curve type if the curvature gradient meets an evaluation standard in a specific road section, and obtain a curve type label. The water seepage influence analysis module is configured to calculate an influence coefficient of water seepage on curvature change by using a correlation analysis method according to the curve type label and the water seepage intensity map, and obtain an influence coefficient distribution map. The risk assessment report generation module is configured to extract a high-influence area from the influence coefficient distribution map, generate a risk assessment report if the continuity of the high-influence area meets a risk assessment standard, and summarize potential safety hazards and determine an overall track safety state.
10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the method of any one of claims 1-8.