Three-dimensional real scene dynamic monitoring system for tunnel based on multi-source information fusion

Abstract

The application relates to the technical field of dynamic monitoring, in particular to a tunnel three-dimensional real scene dynamic monitoring system based on multi-source information fusion, which comprises a data acquisition module, an abnormality analysis module, a splicing analysis module and a dynamic adjustment module, calculates internal quality abnormality scores and environmental abnormality scores, determines abnormality acquisition characteristic values of scanning sections, judges the abnormality tendency of the scanning sections, performs local registration on the scanning sections with high abnormality tendency and adjacent scanning sections, calculates splicing deviation characteristic values, determines the splicing state of the scanning sections, judges whether splicing state adjustment is needed, calculates deviation adjustment coefficients in combination with the abnormality acquisition characteristic values and the splicing deviation characteristic values for the scanning sections needing splicing state adjustment, and adjusts splicing parameters based on the deviation adjustment coefficients. The application analyzes the scanning sections from the data splicing angle, adjusts the parameters of each deviation splicing section, and improves the splicing precision and the data acquisition precision.

Description

Technical Field

[0001] This invention relates to the field of dynamic monitoring technology, and in particular to a three-dimensional real-scene dynamic monitoring system for tunnels based on multi-source information fusion. Background Technology

[0002] Traditional tunnel monitoring primarily relies on discrete point measurement methods such as total stations and convergence meters. This approach suffers from problems like "representing the entire surface with a single point," low efficiency, and incomplete information. It struggles to comprehensively and intuitively reflect the continuous deformation and health status of the tunnel structure in three-dimensional space. While the application of real-scene modeling technologies such as 3D laser scanning can acquire complete 3D models, during long-distance, dynamic monitoring, especially in complex tunnel environments with vibration and dust, the trajectory calculation of moving scans is prone to cumulative errors, leading to point cloud stitching deviations and severely impacting the model's accuracy and reliability. Furthermore, the limited data sources and lack of quantitative consideration of environmental interference factors make it difficult for the system to promptly identify and compensate for anomalies at the data acquisition and processing front end, hindering truly intelligent, interference-resistant, and dynamic precision monitoring.

[0003] Chinese Patent Publication No. CN115239879A discloses a method and system for monitoring tunnel boring machines (TBMs). The method includes the following steps: acquiring three-dimensional point cloud information of the tunnel; establishing a three-dimensional model of the tunnel based on the point cloud information; obtaining the center of the tunnel's cross-section based on the three-dimensional model; obtaining the measured axis of the tunnel based on the center of the cross-section; obtaining deformation data of the tunnel's cross-section at all angles based on the measured axis; and sending an alarm if the deformation data exceeds a preset threshold range. This invention solves the technical problems of low intelligence, poor accuracy, and narrow applicability in monitoring TBMs.

[0004] Chinese Patent Publication No. CN118965814A discloses an online monitoring system for tunnel boring machines (TBMs). The system includes a real-time data acquisition module that collects real-time data on cutterhead wear and geological conditions based on cutterhead wear status and geological data. It analyzes the wear patterns between the cutterhead and the soil / rock, identifies the current construction environment, and obtains real-time geotechnical parameters and construction conditions. This invention, through real-time monitoring and refined geological condition analysis, enables accurate grasp of key parameters during tunnel excavation, optimizes construction strategies, enhances adaptability to complex geological changes, and establishes risk assessment and early warning mechanisms to promptly identify and address potential safety issues, thereby reducing the accident rate and ensuring the safety of construction personnel and equipment. By analyzing predicted geological conditions, the excavation path can be adjusted to adapt to potential future geological difficulties, thus achieving greater flexibility in project management and reducing potential economic losses.

[0005] However, the following problems still exist in the existing technology.

[0006] When using mobile lasers to reconstruct 3D real-world tunnel scenes, the acquired point cloud data may have regional accuracy deficiencies due to external or internal factors. In subsequent data processing, these local data errors will be transmitted and accumulated during continuous frame registration and trajectory calculation, forming stitching errors. Especially in long tunnel projects, this error will diverge as the scanning distance increases, eventually leading to global distortions such as overall distortion and drift of the point cloud model, affecting the absolute accuracy and reliability of the model. Summary of the Invention

[0007] To address this issue, the present invention provides a tunnel 3D real-scene dynamic monitoring system based on multi-source information fusion. This system solves the problem that when reconstructing tunnel 3D real-scenes based on mobile lasers, the acquired point cloud data suffers from regional accuracy loss due to external or internal factors. In subsequent data processing, these local data errors are transmitted and accumulated during continuous frame registration and trajectory calculation, forming splicing errors. Especially in long tunnel projects, these errors diverge with the increase of scanning distance, ultimately leading to global distortions such as overall distortion and drift of the point cloud model, affecting the absolute accuracy and reliability of the model.

[0008] To achieve the above objectives, the present invention provides a tunnel 3D real-scene dynamic monitoring system based on multi-source information fusion, comprising:

[0009] The data acquisition module collects environmental data and scanning point data of the tunnel in real time. The environmental data includes vibration data of the tunnel wall and concentration of suspended particulate matter in the air. The scanning point data includes scanning angle and scanning distance, and acquires tunnel sequence point cloud data and platform motion trajectory.

[0010] An anomaly analysis module, connected to the data acquisition module, is used to calculate an intrinsic quality anomaly score based on the scanning angle and the scanning distance, and to calculate an environmental anomaly score based on the vibration data and the suspended particulate matter concentration, so as to determine the anomaly acquisition characterization value of the scanning segment and determine the anomaly tendency of the scanning segment.

[0011] The splicing analysis module, which is connected to the data acquisition module and the anomaly analysis module, responds to the high anomaly tendency of the scan segment, performs local registration of the scan segment with adjacent scan segments based on the tunnel sequence point cloud data and the platform motion trajectory, calculates the splicing deviation characterization value, and uses it to determine the splicing status of the scan segment and determine whether the splicing status adjustment is needed.

[0012] The dynamic adjustment module, which is connected to the splicing analysis module, responds to the need to adjust the splicing state, calculates the deviation adjustment coefficient by combining the abnormal acquisition characterization value and the splicing deviation characterization value, and adjusts the splicing parameters based on the deviation adjustment coefficient.

[0013] Furthermore, the anomaly analysis module calculates the intrinsic quality anomaly score, including:

[0014] The ratio of the scanning angle to the reference scanning angle is determined as the angle influence factor;

[0015] The ratio of the scanning distance to the reference scanning distance is determined as the distance influence factor;

[0016] The average of the sum of the angle influence factor and the distance influence factor is determined as the intrinsic quality anomaly score.

[0017] Furthermore, the anomaly analysis module calculates the environmental anomaly score, including:

[0018] Based on the vibration data, a vibration time-domain curve is constructed, and the vibration time-domain curve is transformed in the frequency domain to determine the vibration frequency.

[0019] The ratio of the vibration frequency to the reference vibration frequency is determined as the vibration influence factor;

[0020] The ratio of the suspended particulate matter concentration to the reference suspended particulate matter concentration is determined as the concentration influence factor;

[0021] The average of the sum of the vibration influence factor and the concentration influence factor is determined as the environmental anomaly score.

[0022] Furthermore, the anomaly analysis module determines the anomaly acquisition characterization values ​​of the scan segment, including:

[0023] The ratio of the intrinsic quality anomaly score to the benchmark intrinsic quality anomaly score is determined as the first anomaly factor;

[0024] The ratio of the environmental anomaly score to the benchmark environmental anomaly score is determined as the second anomaly factor;

[0025] The weighted sum of the first abnormal factor and the second abnormal factor is determined to be the abnormal acquisition characterization value.

[0026] Furthermore, the anomaly analysis module determines the anomaly tendency of the scan segment, wherein,

[0027] If the abnormal acquisition characterization value is greater than the abnormal acquisition characterization value threshold, then the abnormal tendency of the scan segment is determined to be high abnormal tendency.

[0028] If the abnormal acquisition characterization value is less than or equal to the abnormal acquisition characterization value threshold, then the abnormal tendency of the scan segment is determined to be low abnormal tendency.

[0029] Furthermore, the stitching analysis module performs local registration between the scan segment and adjacent scan segments, including:

[0030] The point clouds of the scan segment and the adjacent scan segment are registered using the iterative nearest point algorithm, and the root mean square error of the registration is determined as the point cloud matching coefficient.

[0031] Based on the platform's motion trajectory, the difference in trajectory pose between the scan segment and adjacent scan segments in the overlapping region is calculated, and the magnitude of the difference is determined as the trajectory matching coefficient.

[0032] Furthermore, the splicing analysis module calculates the splicing deviation characterization value, including,

[0033] The ratio of the point cloud matching coefficient to the reference point cloud matching coefficient is determined as the first stitching influence factor;

[0034] The ratio of the trajectory matching coefficient to the baseline trajectory matching coefficient is determined as the second stitching influence factor;

[0035] The weighted sum of the first splicing influence factor and the second splicing influence factor is determined as the splicing deviation characterization value.

[0036] Furthermore, the splicing analysis module determines the splicing status of the scanned segments and determines whether splicing status adjustment is needed, wherein...

[0037] If the splicing deviation characterization value is greater than the splicing deviation characterization value threshold, then the splicing state of the scan segment is determined to be an abnormal splicing state, and splicing state adjustment is required.

[0038] If the splicing deviation characterization value is less than or equal to the splicing deviation characterization value threshold, then the splicing state of the scan segment is determined to be a normal splicing state, and no splicing state adjustment is required.

[0039] Furthermore, the dynamic adjustment module calculates its deviation adjustment coefficient, including...

[0040] The ratio of the abnormal acquisition characterization value to the baseline abnormal acquisition characterization value is determined as the first adjustment factor;

[0041] The ratio of the splicing deviation characterization value to the baseline splicing deviation characterization value is determined as the second adjustment factor;

[0042] The weighted sum of the first adjustment factor and the second adjustment factor is determined to be the deviation adjustment coefficient.

[0043] Furthermore, the dynamic adjustment module adjusts the splicing parameters based on the deviation adjustment coefficient, wherein,

[0044] The splicing angle is inversely proportional to the aforementioned deviation adjustment coefficient.

[0045] Compared with existing technologies, this invention, by setting up a data acquisition module, an anomaly analysis module, a stitching analysis module, and a dynamic adjustment module, calculates intrinsic quality anomaly scores and environmental anomaly scores to determine the anomaly acquisition characterization values ​​of the scan segment, identifies the anomaly tendency of the scan segment, performs local registration of scan segments with high anomaly tendency with adjacent scan segments, calculates stitching deviation characterization values ​​to determine the stitching state of the scan segment, determines whether stitching state adjustment is needed, and for scan segments requiring stitching state adjustment, calculates a deviation adjustment coefficient based on the anomaly acquisition characterization values ​​and the stitching deviation characterization values, and adjusts the stitching parameters based on the deviation adjustment coefficient. This invention analyzes the scan segment from a data stitching perspective, improving stitching accuracy and data acquisition accuracy by adjusting the parameters of each deviation stitching segment.

[0046] In particular, by comprehensively analyzing environmental data and self-collected data, quality anomaly scores and environmental anomaly scores are determined to calculate anomaly collection characterization values, providing a data basis for judging the anomaly tendency of the scanned segment. It is understandable that to achieve high-precision 3D real-scene modeling of the tunnel, continuous, segmented data acquisition using a mobile laser scanning system is necessary. Subsequent point cloud registration and fusion algorithms then stitch the sequential scanned segments into a unified digital model. However, in reality, data acquisition is easily interfered with by complex working conditions within the tunnel. For example, continuous mechanical vibration induced by the operation of heavy equipment can be transmitted to the laser sensor through the scanning platform, introducing high-frequency, micro-amplitude random jitter. This disturbance, outside the calculation frequency of the inertial measurement unit, will directly cause coordinate drift and geometric distortion in the point cloud data. Simultaneously, the high concentration of dust and water vapor suspended matter generated during excavation will produce significant Mie scattering and energy attenuation effects on the laser beam, not only reducing its efficiency but also... Lowering the signal-to-noise ratio of the effective echo signal will generate a large number of unstructured noise points, which may even obscure the real tunnel wall features in severe cases. Furthermore, the repetition of tunnel wall features or the obstruction by temporary facilities will cause feature matching ambiguity. The combined effect of the above interference factors means that the original point cloud data contains inherent quality defects and external environmental disturbance errors from the source of acquisition. These subtle local errors existing in a single scan segment will be transmitted and nonlinearly accumulated through the algorithm chain in the subsequent pose map optimization or global registration process. This cumulative effect will lead to serious "error drift", causing the reconstructed 3D model to produce significant deformation, misalignment and other topological errors on a macroscopic scale. Based on this, this invention considers to quantify the abnormal acquisition characterization values ​​by pre-analyzing the environment and inherent data, providing data basis for determining the abnormal tendency of the scan segment, so as to carry out targeted analysis in subsequent matching calculations, improve the efficiency of abnormal analysis, and improve the data acquisition accuracy and stitching accuracy.

[0047] In particular, by analyzing tunnel sequence point cloud data and the local registration results of scan segments, the splicing deviation characterization value with high anomaly tendency is calculated, providing a data basis for subsequently determining the splicing state of scan segments. In practice, data splicing for model construction mostly adopts a strategy of local splicing and then integrating into the whole. However, sequence point cloud data has continuity in time and space. Therefore, any quality defect in a local scan segment will be transmitted and accumulated in subsequent splicing stages. In particular, forcibly registering low-quality abnormal data with high-quality data will not only distort the model in that local area, but also generate an incorrect transformation matrix. This incorrect transformation matrix will serve as the initial position for the next local splicing, causing the error to propagate and amplify continuously in the overall model, ultimately causing model misalignment, distortion, or even overall failure. Based on this, this invention considers pre-calculating the splicing deviation characterization value to quantify the specific deviation degree generated when an abnormal scan segment is locally registered with adjacent segments. This provides a data basis for subsequently judging whether splicing state adjustment is needed, achieving the goal of limiting splicing problems and errors to a local range and improving data acquisition accuracy and splicing accuracy.

[0048] In particular, by calculating the deviation adjustment coefficient, a data foundation is provided for subsequent stitching parameter adjustments. In practice, point cloud stitching mostly relies on preset, fixed parameter combinations. This "one-size-fits-all" strategy cannot distinguish the causes of anomalies. It is understandable that severe vibration and dense dust can both cause data anomalies, but their mechanisms of influence on point cloud data are completely different. Vibration mainly introduces high-frequency noise and geometric deformation, while dust mainly affects the data through missing points and noise points. In this case, fixed parameters cannot be adjusted specifically according to different causes of anomalies. At the same time, fixed parameters cannot quantify the severity of anomalies. Existing systems, upon detecting an anomaly, only know that "there is a problem," but not the severity of the problem. In this case, if the same conservative splicing parameters are used for both a slight vibration and a strong vibration, it will lead to over-adjustment and loss of accuracy for the slight vibration, and under-adjustment and inability to correct errors for the strong vibration, both of which will affect the accuracy of data splicing. Based on this, the present invention considers calculating the deviation adjustment coefficient. By fusing the anomaly acquisition characterization value and the splicing deviation characterization value, the complex adjustment problem is transformed into a quantifiable variable that is precisely matched with the current specific scenario, thereby further improving the accuracy of data acquisition and splicing. Attached Figure Description

[0049] Figure 1 This is a schematic diagram of the structure of a tunnel 3D real-scene dynamic monitoring system based on multi-source information fusion, as an embodiment of the invention.

[0050] Figure 2 This is a logic block diagram for determining the abnormal tendency of the scan segment according to an embodiment of the invention;

[0051] Figure 3 This is a logic block diagram illustrating how to determine the splicing state of the scan segment and whether splicing state adjustment is needed, as per an embodiment of the invention. Detailed Implementation

[0052] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0053] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0054] It should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the term "connection" should be interpreted broadly. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be a connection within two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0055] Please see Figure 1 , Figure 1 This is a schematic diagram of the structure of a tunnel 3D real-scene dynamic monitoring system based on multi-source information fusion, according to an embodiment of the invention. The tunnel 3D real-scene dynamic monitoring system based on multi-source information fusion, according to an embodiment of the invention, includes:

[0056] The data acquisition module collects environmental data and scanning point data of the tunnel in real time. The environmental data includes vibration data of the tunnel wall and concentration of suspended particulate matter in the air. The scanning point data includes scanning angle and scanning distance, and acquires tunnel sequence point cloud data and platform motion trajectory.

[0057] An anomaly analysis module, connected to the data acquisition module, is used to calculate an intrinsic quality anomaly score based on the scanning angle and the scanning distance, and to calculate an environmental anomaly score based on the vibration data and the suspended particulate matter concentration, so as to determine the anomaly acquisition characterization value of the scanning segment and determine the anomaly tendency of the scanning segment.

[0058] The splicing analysis module, which is connected to the data acquisition module and the anomaly analysis module, responds to the high anomaly tendency of the scan segment, performs local registration of the scan segment with adjacent scan segments based on the tunnel sequence point cloud data and the platform motion trajectory, calculates the splicing deviation characterization value, and uses it to determine the splicing status of the scan segment and determine whether the splicing status adjustment is needed.

[0059] The dynamic adjustment module, which is connected to the splicing analysis module, responds to the need to adjust the splicing state, calculates the deviation adjustment coefficient by combining the abnormal acquisition characterization value and the splicing deviation characterization value, and adjusts the splicing parameters based on the deviation adjustment coefficient.

[0060] Specifically, the method for collecting vibration data of the tunnel wall is not limited. For example, it can be collected by a vibration sensing unit installed on the tunnel wall. The present invention does not limit the specific type, number and installation position of the vibration sensing unit. It can be any sensor or combination thereof known in the art that can realize vibration measurement, as long as it can effectively acquire the physical signal characterizing the vibration of the tunnel wall. Of course, those skilled in the art can also use other methods to collect the data, as long as they are reasonable, which will not be elaborated here.

[0061] Specifically, there are no restrictions on the method of collecting the concentration of suspended particulate matter. For example, it can be a sensor based on the principle of laser scattering, β-ray attenuation, or piezoelectric microbalance. That is, any known measurement method that can detect and output a signal characterizing the concentration of suspended particulate matter in the air in real time is acceptable, as long as it can detect the required data. This will not be elaborated further.

[0062] Specifically, there are no restrictions on the method of collecting scanning point data. For example, the collection points can be predetermined in the construction plan, and the scanning angle and scanning distance can be measured by laser. Of course, those skilled in the art can also use other methods to collect the data, as long as they are reasonable. This will not be elaborated further.

[0063] Specifically, there are no restrictions on the method of acquiring tunnel sequence point cloud data. For example, it can be acquired by a laser scanner. Of course, those skilled in the art can also use other methods to acquire it, as long as they are reasonable. This will not be elaborated further.

[0064] Specifically, the platform trajectory represents the position of the laser scanner during its movement. There are no restrictions on how the platform's motion trajectory is acquired. For example, it can be acquired through the positioning sensor built into the laser scanner. Of course, those skilled in the art can also acquire it in other ways, as long as it is reasonable. This will not be elaborated further.

[0065] Specifically, the anomaly analysis module calculates the intrinsic quality anomaly score, including:

[0066] The ratio of the scanning angle to the reference scanning angle is determined as the angle influence factor;

[0067] The ratio of the scanning distance to the reference scanning distance is determined as the distance influence factor;

[0068] The average of the sum of the angle influence factor and the distance influence factor is determined as the intrinsic quality anomaly score.

[0069] Specifically, the baseline scanning angle is pre-calculated by obtaining the historical scanning angles from several successful tunnel model constructions and determining the average of these historical scanning angles as the baseline scanning angle.

[0070] Specifically, the baseline scanning distance is pre-calculated by obtaining the historical scanning distances from several successful tunnel model constructions and determining the average of these historical scanning distances as the baseline scanning distance.

[0071] Understandably, in actual measurement environments, the larger the scanning angle, the weaker the reflected signal, and the worse the positional accuracy and intensity information reliability of the point cloud. At the same time, the farther the distance, the larger the area covered by the same size laser spot, resulting in sparser points per unit area, worse measurement accuracy, and more abnormal measurement quality.

[0072] Specifically, the anomaly analysis module calculates environmental anomaly scores, including:

[0073] Based on the vibration data, a vibration time-domain curve is constructed, and the vibration time-domain curve is transformed in the frequency domain to determine the vibration frequency.

[0074] The ratio of the vibration frequency to the reference vibration frequency is determined as the vibration influence factor;

[0075] The ratio of the suspended particulate matter concentration to the reference suspended particulate matter concentration is determined as the concentration influence factor;

[0076] The average of the sum of the vibration influence factor and the concentration influence factor is determined as the environmental anomaly score.

[0077] Specifically, the reference vibration frequency is calculated in advance. The historical vibration frequencies of several successful tunnel model constructions are obtained in advance, and the average of each historical vibration frequency is determined as the reference vibration frequency.

[0078] Specifically, the baseline suspended particulate matter concentration is calculated in advance. The historical suspended particulate matter concentrations from several successful tunnel model constructions are obtained in advance, and the average of each historical suspended particulate matter concentration is determined as the baseline suspended particulate matter concentration.

[0079] Specifically, the anomaly analysis module determines the anomaly acquisition characteristics of the scan segment, including:

[0080] The ratio of the intrinsic quality anomaly score to the benchmark intrinsic quality anomaly score is determined as the first anomaly factor;

[0081] The ratio of the environmental anomaly score to the benchmark environmental anomaly score is determined as the second anomaly factor;

[0082] The weighted sum of the first abnormal factor and the second abnormal factor is determined to be the abnormal acquisition characterization value.

[0083] Specifically, the baseline intrinsic quality anomaly score is calculated in advance. The historical intrinsic quality anomaly scores of several successful tunnel model constructions are obtained in advance, and the average of each historical intrinsic quality anomaly score is determined as the baseline intrinsic quality anomaly score.

[0084] Specifically, the baseline environmental anomaly score is calculated in advance. Several historical environmental anomaly scores from successful tunnel model constructions are obtained in advance, and the average of each historical environmental anomaly score is determined as the baseline environmental anomaly score.

[0085] Specifically, the sum of the weight coefficients of the first abnormal factor and the second abnormal factor is 1. When adjusting the weights, considering that the data collection status is affected by both internal collection factors and environmental factors, the weight coefficients of the first abnormal factor and the second abnormal factor are both set to 0.5.

[0086] Specifically, by comprehensively analyzing environmental data and self-collected data, quality anomaly scores and environmental anomaly scores are determined to calculate anomaly collection characterization values, providing a data foundation for judging the anomaly tendency of the scanned segment. It is understandable that to achieve high-precision 3D real-scene modeling of the tunnel, continuous, segmented data acquisition using a mobile laser scanning system is necessary. Subsequent point cloud registration and fusion algorithms then stitch the sequential scanned segments into a unified digital model. However, in reality, data acquisition is easily interfered with by complex working conditions within the tunnel. For example, continuous mechanical vibration induced by the operation of heavy equipment can be transmitted to the laser sensor through the scanning platform, introducing high-frequency, micro-amplitude random jitter. This disturbance, outside the calculation frequency of the inertial measurement unit, will directly cause coordinate drift and geometric distortion in the point cloud data. Simultaneously, the high concentration of dust and water vapor suspended matter generated during excavation will produce significant Mie scattering and energy attenuation effects on the laser beam, not only... This reduces the signal-to-noise ratio of the effective echo signal and generates a large number of unstructured noise points, which in severe cases can even obscure the true features of the tunnel wall. Furthermore, the repetition of tunnel wall features or the ambiguity caused by temporary facilities obstructing the feature matching, along with the combined effect of these interference factors, means that the original point cloud data contains inherent quality defects and external environmental disturbance errors from the source of acquisition. These subtle local errors existing in a single scan segment will be transmitted and nonlinearly accumulated through the algorithm chain during subsequent pose map optimization or global registration. This cumulative effect will lead to severe "error drift," causing the reconstructed 3D model to produce significant deformation, misalignment, and other topological errors on a macroscopic scale. Based on this, this invention considers quantifying the abnormal acquisition characterization values ​​by pre-analyzing the environment and inherent data, providing data basis for determining the abnormal tendencies of the scan segment, so as to conduct targeted analysis during subsequent matching calculations, improve the efficiency of abnormal analysis, and improve the data acquisition accuracy and stitching accuracy.

[0087] Please see Figure 2 , Figure 2 This is a logic block diagram illustrating the determination of anomaly tendencies in a scan segment according to an embodiment of the invention. Specifically, the anomaly analysis module determines the anomaly tendencies in the scan segment, wherein...

[0088] If the abnormal acquisition characterization value is greater than the abnormal acquisition characterization value threshold, then the abnormal tendency of the scan segment is determined to be high abnormal tendency.

[0089] If the abnormal acquisition characterization value is less than or equal to the abnormal acquisition characterization value threshold, then the abnormal tendency of the scan segment is determined to be low abnormal tendency.

[0090] Specifically, the anomaly acquisition characterization value threshold represents the boundary of the range of differences in the acquired data that can be used to construct a 3D model. It is calculated in advance by obtaining several historical anomaly acquisition characterization values ​​corresponding to the successful construction of the tunnel 3D model. The product of the mean of each historical anomaly acquisition characterization value and the anomaly coefficient is determined as the anomaly acquisition characterization value threshold. The anomaly coefficient is selected in the interval [0.8, 1.0]. In practice, in order to improve the accuracy of the data, the anomaly coefficient is determined to be 0.9.

[0091] Specifically, the stitching analysis module performs local registration between the scan segment and adjacent scan segments, including:

[0092] The point clouds of the scan segment and the adjacent scan segment are registered using the iterative nearest point algorithm, and the root mean square error of the registration is determined as the point cloud matching coefficient.

[0093] Based on the platform's motion trajectory, the difference in trajectory pose between the scan segment and adjacent scan segments in the overlapping region is calculated, and the magnitude of the difference is determined as the trajectory matching coefficient.

[0094] Specifically, the root mean square error is calculated as follows:

[0095] Identify the nearest neighboring scan points in adjacent scan segments to each scan point within the scan segment, forming several point pairs;

[0096] Determine the transformation matrix that minimizes the average distance between each point;

[0097] The transformation matrix is ​​placed in the point cloud of the scan segment, and then moved and rotated to bring it closer to the point cloud data of the adjacent scan segment in order to determine the final matching point pair.

[0098] The root mean square error is determined by the average of the squared distances between the final matched point pairs.

[0099] Specifically, the modulus of the difference is calculated as follows:

[0100] The Euclidean distance between a scan trajectory segment and its adjacent trajectory segments is determined as the distance difference.

[0101] The difference between the three Euler angles of the scan trajectory segment and the adjacent trajectory segment is defined as the angle difference value;

[0102] The square root of the weighted sum of the distance difference and the angle difference is determined as the modulus of the difference.

[0103] Specifically, the splicing analysis module calculates the splicing deviation characterization value, including,

[0104] The ratio of the point cloud matching coefficient to the reference point cloud matching coefficient is determined as the first stitching influence factor;

[0105] The ratio of the trajectory matching coefficient to the baseline trajectory matching coefficient is determined as the second stitching influence factor;

[0106] The weighted sum of the first splicing influence factor and the second splicing influence factor is determined as the splicing deviation characterization value.

[0107] Specifically, the baseline point cloud matching coefficient is calculated in advance. Several historical point cloud matching coefficients corresponding to successfully constructed 3D tunnel models are obtained in advance, and the average of each historical point cloud matching coefficient is determined as the baseline point cloud matching coefficient.

[0108] Specifically, the baseline trajectory matching coefficient is calculated in advance. Several historical trajectory matching coefficients corresponding to successfully constructed 3D tunnel models are obtained in advance, and the average of each historical trajectory matching coefficient is determined as the baseline trajectory matching coefficient.

[0109] Specifically, the sum of the weight coefficients of the first and second splicing influence factors is 1. When adjusting the weights, considering that the point cloud values ​​can directly affect the model's construction accuracy, the weight coefficient of the first splicing influence factor is set to 0.6 and the weight coefficient of the second splicing influence factor is set to 0.4.

[0110] Specifically, by analyzing tunnel sequence point cloud data and the local registration results of scan segments, the splicing deviation characterization value with high anomaly tendency is calculated, providing a data basis for subsequently determining the splicing state of scan segments. In practice, data splicing for model construction mostly adopts a strategy of local splicing and then integrating into the whole. However, sequence point cloud data has continuity in time and space. Therefore, any quality defect in a local scan segment will be transmitted and accumulated in subsequent splicing stages. In particular, forcibly registering low-quality abnormal data with high-quality data will not only distort the model in that local area, but also generate an incorrect transformation matrix. This incorrect transformation matrix will serve as the initial position for the next local splicing, causing the error to continuously propagate and amplify in the overall model, ultimately causing model misalignment, distortion, or even overall failure. Based on this, this invention considers pre-calculating the splicing deviation characterization value to quantify the specific deviation degree generated when an abnormal scan segment is locally registered with adjacent segments. This provides a data basis for subsequently judging whether splicing state adjustment is necessary, achieving the goal of limiting splicing problems and errors to a local range and improving data acquisition accuracy and splicing accuracy.

[0111] Please see Figure 3 , Figure 3 This is a logic block diagram illustrating how an embodiment of the invention determines the splicing state of a scan segment and whether splicing state adjustment is needed. Specifically, the splicing analysis module determines the splicing state of the scan segment and determines whether splicing state adjustment is needed, wherein...

[0112] If the splicing deviation characterization value is greater than the splicing deviation characterization value threshold, then the splicing state of the scan segment is determined to be an abnormal splicing state, and splicing state adjustment is required.

[0113] If the splicing deviation characterization value is less than or equal to the splicing deviation characterization value threshold, then the splicing state of the scan segment is determined to be a normal splicing state, and no splicing state adjustment is required.

[0114] Specifically, the splicing deviation characterization value threshold represents a boundary at which a three-dimensional model can be successfully constructed after data splicing. It is pre-calculated by obtaining several historical splicing deviation characterization values ​​corresponding to successfully constructed tunnel three-dimensional models in advance. The product of the mean of each historical splicing deviation characterization value and the splicing coefficient is determined as the splicing deviation characterization value threshold. The splicing coefficient is obtained in the interval [0.8, 1.0]. In practice, in order to improve the modeling accuracy, the splicing coefficient is determined to be 0.9.

[0115] Specifically, the dynamic adjustment module calculates its deviation adjustment coefficient, including

[0116] The ratio of the abnormal acquisition characterization value to the baseline abnormal acquisition characterization value is determined as the first adjustment factor;

[0117] The ratio of the splicing deviation characterization value to the baseline splicing deviation characterization value is determined as the second adjustment factor;

[0118] The weighted sum of the first adjustment factor and the second adjustment factor is determined to be the deviation adjustment coefficient.

[0119] Specifically, the baseline anomaly acquisition characterization value is the anomaly acquisition characterization value corresponding to the baseline intrinsic quality anomaly score and the baseline environmental anomaly score.

[0120] Specifically, the baseline stitching deviation characterization value is the stitching deviation characterization value corresponding to the baseline point cloud matching coefficient and the baseline trajectory matching coefficient.

[0121] Specifically, the sum of the weight coefficients of the first adjustment factor and the second adjustment factor is 1. When performing weighting, abnormal data will directly cause splicing errors. Therefore, the weight coefficient of the first adjustment factor is determined to be 0.6 and the weight coefficient of the second adjustment factor is 0.4.

[0122] Specifically, by calculating the deviation adjustment coefficient, a data basis is provided for subsequent stitching parameter adjustments. In practice, point cloud stitching mostly relies on preset, fixed parameter combinations. This "one-size-fits-all" strategy cannot distinguish the causes of anomalies. It is understandable that while severe vibration and dense dust can both cause data anomalies, their mechanisms of impact on point cloud data are completely different. Vibration mainly introduces high-frequency noise and geometric deformation, while dust mainly affects the data through missing points and noise points. In this case, fixed parameters cannot be adjusted specifically according to different causes of anomalies. Furthermore, fixed parameters cannot quantify the severity of the anomalies. Currently, existing systems, upon detecting anomalies, only know that "there is a problem," but not the severity of the problem. In this case, if the same conservative splicing parameters are used for both a slight vibration and a strong vibration, it will lead to over-adjustment and loss of accuracy for the slight vibration, and under-adjustment and inability to correct errors for the strong vibration, both affecting the accuracy of data splicing. Based on this, this invention considers calculating a deviation adjustment coefficient. By fusing the anomaly acquisition characteristic value and the splicing deviation characteristic value, the complex adjustment problem is transformed into a quantifiable variable that precisely matches the current specific scenario, further improving the accuracy of data acquisition and splicing.

[0123] Specifically, the dynamic adjustment module adjusts the splicing parameters based on the deviation adjustment coefficient, wherein...

[0124] The splicing angle is inversely proportional to the aforementioned deviation adjustment coefficient.

[0125] Specifically, when the deviation adjustment coefficient is large, it indicates poor data quality or large splicing error, resulting in poor splicing status. In this case, if the algorithm is allowed to perform large-angle searches (i.e., large splicing angles), it is very easy to be misled by noise, leading to complete splicing failure. Therefore, this application dynamically reduces the "splicing angle" through an inverse proportional relationship, making fine and conservative adjustments only near the current position, thereby sacrificing some search speed in exchange for splicing reliability and success rate, preventing error accumulation, and improving splicing accuracy.

[0126] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0127] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A tunnel 3D real-scene dynamic monitoring system based on multi-source information fusion, characterized in that, include: The data acquisition module collects environmental data and scanning point data of the tunnel in real time. The environmental data includes vibration data of the tunnel wall and concentration of suspended particulate matter in the air. The scanning point data includes scanning angle and scanning distance, and acquires tunnel sequence point cloud data and platform motion trajectory. An anomaly analysis module, connected to the data acquisition module, is used to calculate an intrinsic quality anomaly score based on the scanning angle and the scanning distance, and to calculate an environmental anomaly score based on the vibration data and the suspended particulate matter concentration, so as to determine the anomaly acquisition characterization value of the scanning segment and determine the anomaly tendency of the scanning segment. The splicing analysis module, which is connected to the data acquisition module and the anomaly analysis module, responds to the high anomaly tendency of the scan segment, performs local registration of the scan segment with adjacent scan segments based on the tunnel sequence point cloud data and the platform motion trajectory, calculates the splicing deviation characterization value, and uses it to determine the splicing status of the scan segment and determine whether the splicing status adjustment is needed. The dynamic adjustment module, which is connected to the splicing analysis module, responds to the need to adjust the splicing state by calculating the deviation adjustment coefficient based on the abnormal acquisition characterization value and the splicing deviation characterization value, and adjusts the splicing parameters based on the deviation adjustment coefficient. The anomaly analysis module determines the anomaly acquisition characterization values ​​of the scan segment, including: The ratio of the intrinsic quality anomaly score to the benchmark intrinsic quality anomaly score is determined as the first anomaly factor; The ratio of the environmental anomaly score to the benchmark environmental anomaly score is determined as the second anomaly factor; The weighted sum of the first abnormal factor and the second abnormal factor is determined to be the abnormal acquisition characterization value.

2. The tunnel three-dimensional real-scene dynamic monitoring system based on multi-source information fusion according to claim 1, characterized in that, The anomaly analysis module calculates the intrinsic quality anomaly score, including: The ratio of the scanning angle to the reference scanning angle is determined as the angle influence factor; The ratio of the scanning distance to the reference scanning distance is determined as the distance influence factor; The average of the sum of the angle influence factor and the distance influence factor is determined as the intrinsic quality anomaly score.

3. The tunnel three-dimensional real-scene dynamic monitoring system based on multi-source information fusion according to claim 1, characterized in that, The anomaly analysis module calculates the environmental anomaly score, including: Based on the vibration data, a vibration time-domain curve is constructed, and the vibration time-domain curve is transformed in the frequency domain to determine the vibration frequency. The ratio of the vibration frequency to the reference vibration frequency is determined as the vibration influence factor; The ratio of the suspended particulate matter concentration to the reference suspended particulate matter concentration is determined as the concentration influence factor; The average of the sum of the vibration influence factor and the concentration influence factor is determined as the environmental anomaly score.

4. The tunnel three-dimensional real-scene dynamic monitoring system based on multi-source information fusion according to claim 1, characterized in that, The anomaly analysis module determines the abnormal tendency of the scan segment, wherein, If the abnormal acquisition characterization value is greater than the abnormal acquisition characterization value threshold, then the abnormal tendency of the scan segment is determined to be high abnormal tendency. If the abnormal acquisition characterization value is less than or equal to the abnormal acquisition characterization value threshold, then the abnormal tendency of the scan segment is determined to be low abnormal tendency.

5. The tunnel three-dimensional real-scene dynamic monitoring system based on multi-source information fusion according to claim 1, characterized in that, The stitching analysis module performs local registration between the scan segment and adjacent scan segments, including: The iterative nearest point algorithm is used to register the point clouds of the scan segment with those of adjacent scan segments, and the root mean square error of the registration is determined as the point cloud matching coefficient. Based on the platform's motion trajectory, the difference in trajectory pose between the scanned segment and adjacent scanned segments in the overlapping region is calculated, and the magnitude of the difference is determined as the trajectory matching coefficient. Among them, the nearest neighboring scan point to each scan point in the adjacent scan segment is determined, forming several point pairs; Determine the transformation matrix that minimizes the average distance between each point; The transformation matrix is ​​placed in the point cloud of the scan segment, and then moved and rotated to bring it closer to the point cloud data of the adjacent scan segment in order to determine the final matching point pair. The root mean square error is determined by the average of the squared distances between the final matched point pairs. The Euclidean distance between a scan trajectory segment and its adjacent trajectory segments is determined as the distance difference. The difference between the three Euler angles of the scan trajectory segment and the adjacent trajectory segment is defined as the angle difference value; The square root of the weighted sum of the distance difference and the angle difference is determined as the modulus of the difference.

6. The tunnel three-dimensional real-scene dynamic monitoring system based on multi-source information fusion according to claim 5, characterized in that, The splicing analysis module calculates the splicing deviation characterization value, including, The ratio of the point cloud matching coefficient to the reference point cloud matching coefficient is determined as the first stitching influence factor; The ratio of the trajectory matching coefficient to the baseline trajectory matching coefficient is determined as the second stitching influence factor; The weighted sum of the first splicing influence factor and the second splicing influence factor is determined as the splicing deviation characterization value.

7. The tunnel three-dimensional real-scene dynamic monitoring system based on multi-source information fusion according to claim 1, characterized in that, The stitching analysis module determines the stitching status of the scanned segments and determines whether stitching status adjustment is needed. If the splicing deviation characterization value is greater than the splicing deviation characterization value threshold, then the splicing state of the scan segment is determined to be an abnormal splicing state, and splicing state adjustment is required. If the splicing deviation characterization value is less than or equal to the splicing deviation characterization value threshold, then the splicing state of the scan segment is determined to be a normal splicing state, and no splicing state adjustment is required.

8. The tunnel three-dimensional real-scene dynamic monitoring system based on multi-source information fusion according to claim 1, characterized in that, The dynamic adjustment module calculates its deviation adjustment coefficient, including The ratio of the abnormal acquisition characterization value to the baseline abnormal acquisition characterization value is determined as the first adjustment factor; The ratio of the splicing deviation characterization value to the baseline splicing deviation characterization value is determined as the second adjustment factor; The weighted sum of the first adjustment factor and the second adjustment factor is determined to be the deviation adjustment coefficient.

9. The tunnel three-dimensional real-scene dynamic monitoring system based on multi-source information fusion according to claim 1, characterized in that, The dynamic adjustment module adjusts the splicing parameters based on the deviation adjustment coefficient, wherein... The splicing angle is inversely proportional to the aforementioned deviation adjustment coefficient.

Images