An adaptive verification and cloud collaborative analysis method for road surface measurement data

Abstract

The application discloses a kind of self-adapting verification and cloud collaborative analysis method of road surface measurement data, it is related to road surface data verification analysis field, comprising: obtaining the road surface measurement data of target section;According to historical credible measurement record, construct road surface state evolution model, determine the reasonable interval corresponding to current measurement data and obtain evolution support degree;Determine consensus support degree by obtaining multiple observation records in time and space range;According to evolution support degree and consensus support degree, generate credibility object and determine verification result, upload cloud for secondary evaluation to conflict data, review result drives model update and issues edge end;Credibility object is stored in historical credibility sequence and updates road surface state evolution model.The application reduces false alarm rate and false negative rate by evolution support degree and consensus support degree double verification mechanism, combined with cloud collaborative review and model closed loop update, improve the adaptive ability of system to complex working conditions.

Description

Technical Field

[0001] This invention relates to the field of pavement data inspection and analysis, specifically to an adaptive verification and cloud-based collaborative analysis method for pavement measurement data. Background Technology

[0002] With the development of LiDAR, industrial cameras, inertial navigation equipment, and vehicle-mounted edge computing technology, road surface condition detection has been widely applied in road maintenance and infrastructure inspection. Existing technologies typically analyze road surface smoothness, rut depth, crack severity, and pothole conditions by collecting data such as road surface point clouds, images, vibration, and positioning.

[0003] Existing methods for verifying pavement measurement data typically rely on single measurement results, statistical characteristics, or consistency of multi-source data for anomaly detection, often resulting in direct conclusions of normal or abnormal. On one hand, current technologies lack mechanisms to validate the rationality of measurement data using long-term pavement condition evolution patterns, making it difficult to determine whether the data reflects the actual deterioration process of the target road segment. On the other hand, existing technologies lack methods for independently quantifying and fusing the support from different observation sources, making it difficult to trace the reliability of the measurement data. Furthermore, reliability information in existing technologies is usually used as a one-time judgment result, lacking mechanisms for continuous accumulation and historical utilization. This prevents optimization of subsequent data verification processes based on historical reliability changes and adaptive updates to pavement condition models using historical reliability information. Summary of the Invention

[0004] To address the aforementioned technical problems, this paper provides an adaptive verification and cloud-based collaborative analysis method for pavement measurement data. This solution overcomes the limitations of existing pavement measurement data verification methods described in the background, which typically rely on single measurement results, statistical characteristics, or multi-source data consistency for anomaly assessment. These methods often result in direct conclusions of normal or abnormal data. On one hand, existing technologies lack a mechanism to validate the rationality of measurement data using long-term pavement condition evolution patterns, making it difficult to determine whether the measurement data conforms to the actual deterioration process of the target road section. On the other hand, existing technologies lack a method for independently quantifying and fusing the support levels of different observation sources for measurement results, making it difficult to trace the reliability of the measurement data. Furthermore, the reliability information in existing technologies is usually used as a one-time judgment result, lacking a mechanism for continuous accumulation and historical utilization. This makes it impossible to optimize subsequent data verification processes based on historical reliability changes, or to adaptively update the pavement condition model using historical reliability information.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: An adaptive verification and cloud-based collaborative analysis method for road surface measurement data includes: S1. Obtain the pavement measurement data corresponding to the target road segment, preprocess the pavement measurement data, and establish the historical reliability sequence of the target road segment; S2. Filter the historical confidence sequence of the target road segment, and perform evolutionary deduction starting from the most recent historical confidence sequence to generate the current road surface condition prediction value, fluctuation range, and the degree of matching between the fluctuation range and the measurement data, and generate evolutionary support. S3. Obtain multiple observation records for the target road segment and compare them with the current road surface measurement data. Combine the historical credibility information of multiple observation records to generate consensus support. S4. Construct a multi-dimensional credibility vector based on evolutionary support and consensus support, generate a structured credibility object, and bind the credibility object to the road surface measurement data. S5. Determine the verification result based on the credibility score, evolution support and consensus support in the credibility object and write it into the historical credibility sequence. Upload it to the cloud for verification and correct the historical credibility sequence based on the verification result. S6. Assign weights to historical credibility sequences based on credibility scores and update road condition projection parameters.

[0006] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention establishes a state evolution model reflecting the long-term change patterns of a target road segment, and uses this model to predict the reasonable state range of the target road segment at the current moment. It then matches and analyzes the current measurement data with the predicted reasonable state range, quantifying this as evolutionary support to verify the rationality of data based on the historical change patterns of the road segment. Compared to existing technologies that primarily rely on single measurement results and fixed thresholds for anomaly detection, this invention dynamically analyzes measurement data by incorporating the long-term change characteristics of the target road segment, reducing misjudgments caused by equipment errors, environmental disturbances, and occasional anomalies, and improving the accuracy and adaptability of measurement data verification.

[0007] This invention acquires multiple observation records within a spatiotemporal range and analyzes the consistency between each observation record and the current measurement data to form a consensus support level, thereby achieving data credibility verification based on group observation results. Compared to existing technologies that rely solely on a single observation source for judgment, this invention can fully utilize observation data from multiple vehicles, multiple time points, and multiple sources to form group consistency constraints, improve the ability to identify abnormal data, and reduce the impact of a single observation result on the verification conclusion.

[0008] This invention integrates evolutionary support and consensus support to generate a credibility object containing multiple credibility components and source information. This allows the credibility of measurement data to not only be quantitatively represented but also traced back to its formation basis. Compared to existing technologies that only output a single verification result, this invention can simultaneously retain contribution information from different verification dimensions, improving the interpretability and traceability of data verification results. Attached Figure Description

[0009] Figure 1 This is an overall flowchart of the collaborative analysis method in this invention; Figure 2 This is a flowchart illustrating the evolutionary support generation process of the collaborative analysis method in this invention. Figure 3 This is a flowchart illustrating the consensus support generation process of the collaborative analysis method in this invention. Figure 4 A flowchart for generating a credibility object verification process for the collaborative analysis method in this invention is provided. Detailed Implementation

[0010] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0011] Reference Figures 1-4 As shown, an adaptive verification and cloud-based collaborative analysis method for road surface measurement data specifically includes: S1. Obtain the pavement measurement data corresponding to the target road segment, preprocess the pavement measurement data, and establish the historical reliability sequence of the target road segment; S2. Filter the historical confidence sequence of the target road segment, and perform evolutionary deduction starting from the most recent historical confidence sequence to generate the current road surface condition prediction value, fluctuation range, and the degree of matching between the fluctuation range and the measurement data, and generate evolutionary support. S3. Obtain multiple observation records for the target road segment and compare them with the current road surface measurement data. Combine the historical credibility information of multiple observation records to generate consensus support. S4. Construct a multi-dimensional credibility vector based on evolutionary support and consensus support, generate a structured credibility object, and bind the credibility object to the road surface measurement data. S5. Determine the verification result based on the credibility score, evolution support and consensus support in the credibility object and write it into the historical credibility sequence. Upload it to the cloud for verification and correct the historical credibility sequence based on the verification result. S6. Assign weights to historical credibility sequences based on credibility scores and update road condition projection parameters.

[0012] Furthermore, step S1 specifically includes: Acquire raw road surface measurement data synchronously collected by multiple source sensors while the data collection vehicle is traveling on the target road section; The original road surface measurement data is timestamped and spatially calibrated to obtain a standard measurement data frame; The attitude information of the acquisition vehicle is used to perform motion distortion correction on the standard measurement data frame to obtain distortion-free road surface measurement data; The distortion-free road surface measurement data is preprocessed, and the preprocessed road surface measurement data is used as the road surface measurement data corresponding to the target road segment. In this embodiment, the vehicle uses multi-source sensors to continuously measure the road surface during its journey on the target road segment, thereby acquiring raw measurement data. The original measurement data includes at least one or more of the following: road surface image information, spatial location information, vehicle motion information, and vibration information; The original measurement data is synchronized to align different types of data in time and unify them in space to the same coordinate system, thereby forming a standard measurement data frame. Furthermore, due to the influence of road surface undulations, acceleration, deceleration, and steering during actual driving, the vehicle body undergoes attitude changes such as pitch, roll, and yaw. These attitude changes result in differences in the sensor poses at different measurement points within the same scanning cycle at the moment of acquisition. The real-time acquired vehicle attitude information, including pitch angle, roll angle, and acceleration data, is used to perform point-by-point motion compensation for each measurement point in the standard measurement data frame: based on the precise acquisition time of each measurement point, the corresponding vehicle attitude is interpolated and calculated, and each measurement point is inversely transformed from the sensor coordinate system to a unified reference coordinate system to eliminate measurement deviations caused by vehicle motion and obtain distorted road measurement data. Statistical filtering is used to remove outlier noise points in road measurement data. The height difference between adjacent measurement points is calculated point by point along the driving direction. Abnormal points that exceed the local reasonable variation range in both the height difference between a measurement point and the previous point and the height difference between the point and the next point are removed to obtain target measurement data for subsequent analysis. Optionally, the data acquisition vehicle is equipped with a camera, LiDAR, and inertial measurement unit (IMU) simultaneously during operation to continuously measure the target road segment. Different sensors output image information, point cloud information, and attitude acceleration information at different frequencies. Within a specific travel timeframe, such as 0.5 seconds, the camera acquires an image of the road surface crack area, the LiDAR scans the spatial point cloud of the road surface at the same location, and the IMU records slight pitch and roll changes of the vehicle. Due to slight differences in the sampling times of the various sensors, the different data are first interpolated to unify them to the same time point, and then analyzed using external parameters. The calibration relationship transforms the image coordinates and point cloud coordinates into the vehicle coordinate system. Then, for the attitude changes of the vehicle between 0.5 seconds and 1 second, the pitch angle, roll angle and acceleration data provided by the inertial measurement unit are used to perform point-by-point attitude compensation for each road measurement point. The measurement points that were originally offset in space due to the change of vehicle attitude are inversely transformed to the position under the unified reference attitude. Finally, on the basis of unified data, statistical filtering is used to remove discrete points that deviate significantly from the local mean, and abrupt changes in the height of adjacent measurement points are detected to remove abnormal points that exceed the reasonable range of change, thereby obtaining stable and consistent road measurement results.

[0013] Furthermore, step S2 specifically includes: The historical reliability sequence of the target road segment is parsed to extract historical measurement records, and the historical measurement records are filtered for reliability based on the historical reliability sequence to obtain a set of historical reliable measurement records; The historical reliable measurement record set is reconstructed by time series, sorted according to the collection time order, and trajectory alignment is performed by combining spatial location information to form structured road surface state sequence data; Based on the road surface state sequence data, multidimensional road surface state feature vectors are extracted, and the state correspondence of the feature vectors in the time dimension is constructed. A state evolution model of road surface state changing over time is established based on the state correspondence in the time dimension. In this embodiment, to ensure that the subsequent model is built on a reliable data foundation, the historical records corresponding to the target road segment are first analyzed. The historical records contain the confidence scores, measurement data and verification conclusions corresponding to different collection times. Extract corresponding records according to the target road segment and filter them according to the credibility score: sort the historical credibility scores from high to low, take the adjacent scores that show a significant decrease after sorting as the boundary, retain the records with scores not lower than the boundary and without conflict markers, and remove low scores and controversial records to obtain a set of historical credibility measurement records, so as to reduce the impact of abnormal data on subsequent model construction. After obtaining a set of historical reliable measurement records, they are sorted according to the collection time order, and combined with spatial location information for location matching and trajectory alignment, so that the measurement results of the same road segment acquired at different times can establish a correspondence and form a structured road surface state sequence data that can reflect the historical state change process of the target road segment. Furthermore, multidimensional pavement condition feature vectors are extracted from the pavement condition sequence data, including crack length, crack width and pothole area reflecting the degree of pavement damage, as well as rut ​​depth and smoothness indicators reflecting driving quality. The feature vectors corresponding to each time point form a change sequence according to the collection time order. Subsequently, by calculating the absolute changes, relative rates of change, and correlation changes between different features between adjacent time points, multi-dimensional change information is obtained. The calculation method is as follows: take the change sequence of two features in multiple adjacent time intervals, calculate the ratio of the changes of the two features in the same time interval, and take the average of the ratios of multiple intervals as the correlation change. The more stable the ratio, the stronger the correlation. Based on this, a state correspondence relationship between different time nodes is established, and a state evolution model that can describe the long-term change trend of the target road segment is formed according to the state correspondence relationship in the continuous historical record. Optionally, the construction of the state evolution model does not depend on a specific type of neural network. Its implementation method can be flexibly selected according to the accumulation of historical data of the target road segment: in the data sparse stage, linear regression or multinomial fitting can be used to model the deterioration trend of each feature separately; after sufficient data accumulation, a time-series recursive method can be used, taking the road surface state feature vector at the current moment as input and the state feature vector at the next moment as output, and fitting the state transition relationship between adjacent time nodes through weighted regression; The training and updating of the model are entirely based on historical reliable measurement records selected from the historical reliability sequence: during the initial construction, after accumulating a sufficient number of reliable records from the first collection, feature vectors at each time point are extracted and formed into a change sequence in chronological order. State transition parameters are established through least squares fitting or weighted regression. After each subsequent verification, newly added reliable objects and measurement data are written into the historical reliability sequence. When updating the model, differentiated weights are assigned according to the proportion of the reliability score of each record to the total score of all records participating in the update. Through weighted fitting, the update direction of the model parameters is dominated by high reliability data, and the contribution of low reliability data is suppressed, thereby realizing the online incremental update of the model. For example, the historical records of a target road segment contain six detection data points, with corresponding confidence scores ranked as 0.96, 0.94, 0.93, 0.91, 0.62, and 0.58. There is a significant drop between 0.91 and 0.62. Taking 0.91 as the dividing line, the records with scores of 0.62 and 0.58 are removed. After filtering, the four retained records were sorted by acquisition time, and the crack width feature values ​​of each detection were extracted as 2.1mm, 2.3mm, 2.4mm, and 2.6mm, respectively. By calculating the changes in adjacent time nodes (+0.2mm, +0.1mm, +0.2mm) and the average rate of change (approximately 0.017mm / day), a state correspondence between crack width and time was established. The resulting state evolution model can infer the reasonable crack width range at the current moment based on the most recent reliable record, providing a basis for the generation of subsequent evolution support. Since the state evolution model is based on reliable records, it can reduce the impact of abnormal data and noise on model construction, making the state evolution model more consistent with the actual deterioration process of the target road segment, thereby improving the accuracy and stability of subsequent verification.

[0014] Furthermore, step S2 also includes: Based on the state evolution model, the most recent reliable measurement record is used to perform evolutionary deduction, generate the road surface state prediction value corresponding to the current time, and simultaneously output the fluctuation range corresponding to the prediction value to obtain the initial reasonable range. Obtain the set of environmental impact factors corresponding to the current moment, and adjust the reasonable range to generate a corrected reasonable range; The degree of interval matching is obtained based on the road surface measurement data and the corrected reasonable interval. Parameters are calibrated based on the matching relationship between historical pavement condition data and corresponding reasonable intervals to obtain evolutionary support. In this embodiment, after the state evolution model is constructed, the state evolution model is used to perform evolution consistency analysis on the current data to be verified. Starting from the most recent reliable record, and combining it with the road surface state change pattern represented by the state evolution model, the road surface state of the target road segment at the current moment is predicted to obtain the predicted value. Since road conditions are affected by factors such as long-term service, traffic load and natural environment, a single predicted value is difficult to accurately reflect the actual range of change. Therefore, while generating the predicted value, the deviation between the model's predicted value and the actual measured value in the historical reliable records is statistically analyzed. Based on the deviation, the allowable fluctuation range of the predicted value is determined, and the corresponding fluctuation range is formed with the predicted value as the center, as the initial reasonable interval. The fluctuation range is used to reflect the possible deviation between historical evolution patterns and actual road surface conditions; Obtain the environmental influencing factors corresponding to the current moment, including one or more of temperature, rainfall, humidity, season, traffic load and regional climate. Compare the measurement deviations of the current environmental conditions with the same environmental conditions in the historical reliable records, calculate the correction amount under the current environmental conditions, and use the correction amount to adjust the upper and lower boundaries of the initial reasonable interval to obtain the corrected reasonable interval. After obtaining the corrected reasonable interval, the currently collected measurement data is matched and analyzed with the interval. The matching is measured from two aspects: first, whether the measurement data falls within the interval; second, when the measurement data is outside the interval, the degree of deviation from the nearest interval boundary is calculated. When the measured data is within the interval and the deviation is low, it indicates that the data has a high degree of consistency with the historical evolution pattern of the target road segment; otherwise, it indicates that the consistency is reduced. Subsequently, based on the matching relationship between historical data and corresponding reasonable intervals, parameter calibration is performed to obtain the mapping relationship. The matching degree corresponding to each historical record and the final quality conclusion confirmed after subsequent verification are extracted. Taking the matching degree as input and the support label determined based on the quality conclusion as output, a continuous and monotonically non-decreasing mapping function is established through regression fitting, which converts the matching situation into evolutionary support. The evolutionary support increases monotonically with the increase of the matching degree. The higher the evolutionary support, the more the current measurement data conforms to the normal state change trend of the target road segment; the lower the support, the greater the deviation. Evolutionary support is calculated using the following formula: ; In the formula, For evolutionary support, For current measurement data, This represents the theoretical state value at the current moment, predicted based on the road surface state evolution model. The reasonable range half-width after environmental factor correction; optional, the most recent reliable record of a certain target road segment shows that the crack width is 2.6mm, the state evolution model deduces the current crack width prediction value as 2.8mm, and the fluctuation range obtained by statistical historical deviation is ±0.3mm, then the initial reasonable range is [2.5mm, 3.1mm]; The ambient temperature detected at the current acquisition time is 8℃, accompanied by continuous rainfall. The average measurement deviation under the same environmental conditions in the historical records is +0.2mm. Therefore, the upper and lower boundaries of the initial reasonable range are both increased by 0.2mm to obtain the corrected reasonable range [2.7mm, 3.3mm]. The crack width in the current measurement data is 3.0 mm, which falls within the reasonable range after correction, and the deviation from the predicted value of 2.8 mm is 0.2 mm, which is less than the fluctuation range of 0.3 mm. This indicates that the measurement data has a high consistency with the historical evolution pattern, and a high evolution support is obtained after matching relationship mapping. Through the above processing, the judgment of whether the data is abnormal is no longer based solely on a single measurement result, but rather on the rationality of the measurement data by combining the long-term state change patterns of the target road section itself.

[0015] Furthermore, step S3 specifically includes: Acquire multiple road surface observation records for the target road segment within a spatiotemporal range, and process the multiple road surface observation records; Based on road surface observation records, corresponding road surface condition observation features are extracted, and the consistency relationship between the road surface condition observation features and the current road surface measurement data is determined. Based on the aforementioned consistency relationship and combined with the historical reliability information corresponding to the road surface observation records, the multiple observation records are processed to generate a comprehensive observation result. Based on the degree of matching between the comprehensive observation results and the road surface measurement data, the overall consistency level of the road surface measurement data is determined, and a consensus support level is generated based on the overall consistency level. In this embodiment, after completing the evolutionary support calculation, multiple observation records for the target road segment are further acquired within the spatiotemporal range, and the current measurement data are compared and analyzed. First, the spatial range is determined based on the target road segment location corresponding to the current measurement data, and the time range is determined with the current measurement time as the center. The spatial range is the actual range of the target road segment, and the time range is from the last major change or repair of the target road segment to the present. Multiple observation records are extracted from the historical records located within the spatial range and whose collection time falls within the time range, and these records are processed for time synchronization and spatial location alignment. Since observation records from different sources differ in terms of acquisition equipment, acquisition time, and driving trajectory, the road surface condition features corresponding to each record are extracted after alignment processing, and the extraction method is consistent with the current measurement data. Based on the consistency relationship between the characteristics of each record and the current measurement data, the degree of consistency of each record is calculated: the smaller the difference, the higher the degree of consistency. After obtaining the consistency level of each record, the historical credibility information accumulated by each record in the previous verification process is combined for weighted fusion. After normalizing the consistency level and historical credibility information respectively, the smaller value of the two is taken as the fusion weight of the record. Based on the fusion weight of each record, the weighted fusion is performed to generate a comprehensive observation result. After calculating the deviation between the current measurement data and the comprehensive observation results, the deviation is compared with the deviation corresponding to each historical record in the road segment's historical reliability sequence. After sorting the deviations from smallest to largest, the proportion of historical records following the current deviation to the total number of records is used as a quantitative value of the overall consistency level, which is the consensus support. The formula for calculating the comprehensive observation results is as follows: ; In the formula, To synthesize the observation results, For the first The data corresponding to each observation record For the first The fusion weights corresponding to each observation record, and , This indicates the degree of consistency between the current observation record and the current measurement data. This provides historical reliability information corresponding to the observation record; Finally, the consensus support is calculated as follows: ; In the formula, For consensus support, For current measurement data, To synthesize the observation results, This represents the maximum deviation between all observation records and the current measurement data; Optionally, if the current measurement data for a target road section shows a rut depth of 3.0 mm, three observation records are obtained from the spatiotemporal range: Record A is from another detection vehicle, with a rut depth of 2.9 mm and a historical reliability of 0.92; Record B is from a roadside sensing device, with a rut depth of 3.2 mm and a historical reliability of 0.85; Record C is from a historical database, with a rut depth of 2.8 mm and a historical reliability of 0.90. Calculate the degree of consistency between each record and the current measurement data: Record A has a difference of 0.1 mm, and the degree of consistency after normalization is 0.95; Record B has a difference of 0.2 mm, and the degree of consistency after normalization is 0.90; Record C has a difference of 0.2 mm, and the degree of consistency after normalization is 0.90. The smaller value between the consistency of each record and the historical reliability is used as the fusion weight: 0.92 for record A, 0.85 for record B, and 0.90 for record C; The weighted fusion yielded a comprehensive observation result of (2.9×0.92+3.2×0.85+2.8×0.90) / (0.92+0.85+0.90)=2.96mm, which is highly consistent with the current measurement data of 3.0mm. The overall consistency level is high, generating a high degree of consensus support, indicating that the current measurement data has strong support in the group observation.

[0016] Furthermore, step S4 specifically includes: Obtain the evolutionary support and the consensus support, and normalize the evolutionary support and the consensus support. A multi-dimensional credibility vector is constructed based on the normalized evolutionary support and the normalized consensus support. A structured credibility object is generated based on the multidimensional credibility vector, and the source markers corresponding to each contribution component are recorded in the credibility object; The credibility object is bound to the road surface measurement data to form a complete data unit containing source information and contribution weight; In this embodiment, after calculating the evolutionary support and consensus support, the scores of the two dimensions are fused. First, normalize the two to put them on the same numerical benchmark, thus eliminating the differences in numerical range and distribution caused by different calculation paths. The normalized evolutionary support is used as the evolutionary dimension component, and the normalized consensus support is used as the consensus dimension component. They are combined according to a unified data field to construct a multi-dimensional credibility vector. Unlike existing technologies that directly fuse multi-source information into a single value, this embodiment retains independent components in two dimensions, enabling subsequent processing to distinguish the degree of contribution of each component. In one implementation method, the credibility score is calculated as follows: ; In the formula, This indicates the credibility score. Indicates evolutionary support. Indicates the degree of consensus support; A structured credibility object is generated based on the multidimensional credibility vector, and the source markers corresponding to each component are recorded in it. The evolution dimension component is derived from the calculation results of the state evolution model, and the consensus dimension component is derived from the calculation results of multi-source observation comparison. When an abnormal credibility score is subsequently discovered for a certain credibility object, the corresponding calculation steps and data sources can be traced back through source marking. The credibility object is then bound to the measurement data, forming a complete data unit containing source information and contribution weight. Optionally, in a certain verification, the evolution support is 0.85 and the consensus support is 0.72. After normalization, the evolution dimension component is 0.85 and the consensus dimension component is 0.72, and a multidimensional credibility vector [0.85, 0.72] is constructed. The generated credibility object contains a credibility score of 0.79 and records that the source marker evolution component is calculated from the state evolution model and the consensus component is derived from multi-source observation comparison. Once the credibility object is bound to the original measurement data, it forms a complete data unit containing the measurement value, credibility score, evolution dimension contribution, consensus dimension contribution, and source tag. Subsequent verification decisions and model updates can be based on this complete data unit.

[0017] Furthermore, step S5 specifically includes: Based on the credibility score, evolutionary support, and consensus support contained in the credibility object, the verification result of the road surface measurement data is determined; Set a confidence interval, compare the current confidence score with the confidence interval, and process the road surface measurement data based on the comparison results; The credibility object is associated with the verification results of the road surface measurement data and stored, and written into the historical credibility sequence according to the target road segment and the collection time; In this embodiment, the credibility score, evolutionary support, and consensus support are extracted from the credibility object, and the measurement data is graded based on these three aspects of information. The confidence score is compared with the preset confidence interval, which includes an upper limit and a lower limit. The confidence interval is dynamically set according to the historical score distribution characteristics of the target road segment: the historical confidence scores are sorted from high to low, and the values ​​in the higher distribution position after sorting are taken as the upper limit and the values ​​in the lower distribution position are taken as the lower limit, so that the interval division adapts to the actual data quality level of different road segments. When the confidence score is higher than the upper limit, it indicates that the measurement data has high support at both the evolutionary law level and the population observation level, and it is judged as valid data and directly enters the subsequent processing flow. When the confidence score is below the lower limit, it indicates that the measurement data has not received sufficient support in both dimensions. It is judged as invalid data and directly removed from the subsequent process, and will not participate in the road condition assessment and model update. When the confidence score is between the upper and lower limits, it indicates that the measurement data has a certain degree of deficiency in a certain dimension but has not yet reached the level of direct elimination. At this time, joint correction processing is triggered: the relative magnitude of evolution support and consensus support is used as the weight coefficient, and the predicted value output by the state evolution model is weighted and combined with the comprehensive results of multi-source observations to generate corrected data. When evolutionary support is higher than consensus support, the revised result is closer to the model prediction; conversely, it is closer to the group observation result. The corrected data is marked as pending review and uploaded to the cloud for inspection. The cloud inspection must meet two conditions simultaneously: First, the deviation between the current measurement data and the cross-vehicle historical measurement data does not exceed the deviation level corresponding to most historical records in the historical reliability sequence of this road segment; Second, the deviation between the current measurement data and the predicted value of the state evolution model does not exceed the fluctuation range corresponding to this prediction. Changes that pass the inspection are considered valid data and incorporated into the normal process, while those that fail are removed from subsequent processes. After completing the classification judgment, the credibility object is associated with the verification result and stored. A tag field of the verification result is added to the credibility object to record the judgment result, judgment time and correction parameters of this verification. The credibility is written into the historical credibility sequence according to the target road segment and collection time dimension to form a credibility record arranged in chronological order, which provides a data foundation for subsequent model updates and credibility trend analysis. Optionally, if the historical reliability scores of a target road segment are sorted and the values ​​at a higher distribution position are 0.88 and the values ​​at a lower distribution position are 0.45, then the upper limit is set to 0.88 and the lower limit is set to 0.45. The current measurement data has a confidence score of 0.62, which falls between 0.45 and 0.88, triggering joint correction processing. At this point, the evolutionary support is 0.72 and the consensus support is 0.50. The evolutionary support is higher than the consensus support, and the corrected measurement data is closer to the predicted value of the state evolution model. The corrected data was marked as pending review and uploaded to the cloud. After consistency comparison and secondary evaluation in the cloud, it was confirmed to pass the inspection and changed to valid data. The credibility object of this verification was marked with the following fields: Judgment result: effective after correction, Judgment time: 2025-06-15-14:30, Correction weight: Evolution 0.72 / Consensus 0.50, and written into the historical credibility sequence according to the target road segment K12+300 and the collection time.

[0018] Furthermore, step S6 specifically includes: Historical credibility objects are extracted from the historical credibility sequence, and the credibility score, evolution support and consensus support corresponding to the historical credibility objects are analyzed to construct a model update sample set; Based on the confidence scores in the updated sample set, the historical measurement records are assigned dynamic update weights. The historical measurement records are fitted with the dynamically updated weights to update the parameters or state transition relationships of the pavement state evolution model and calculate the latest pavement deterioration evolution trend. The updated road surface state evolution model is fed back to the edge or acquisition end; In this embodiment, the state evolution model is updated using historical credibility sequences to form a closed-loop mechanism; Extract the stored credibility objects from the historical credibility sequence of the target road segment, analyze their credibility scores, evolution support and consensus support one by one, and build a model to update the sample set based on each credibility object and its corresponding measurement record. To reflect the differentiated contributions of data with different levels of credibility to model updates, each record is assigned a different update weight based on its credibility score: records with higher scores are given higher weights, enabling them to play a greater guiding role in the process of updating model parameters; records with lower scores are given lower weights, suppressing their contribution to model updates. Specifically, the total confidence score of all records involved in the update is calculated, and the ratio of the confidence score of each historical measurement record to the total is used as its update weight. The updated weights are used as weighting coefficients for each sample to perform weighted fitting on the state evolution model, making the fitting curve closer to the high-confidence sample points and allowing for greater tolerance of deviation for the low-confidence sample points. Through weighted fitting, the parameters or state transition relationships of the model are updated, and the latest pavement deterioration evolution trend is calculated. Finally, the updated model is fed back to the edge or collection end. The cloud sends the updated model parameters or model files to the edge computing devices deployed on the collection vehicles or roadside nodes through the wireless communication network. The edge replaces or overwrites the local model, so that the updated model can be directly used for reasonable interval prediction and evolution support calculation in the subsequent data collection process, forming a continuous collaborative closed loop from the cloud to the edge. Optionally, if the confidence scores of four historical records in the historical confidence sequence of a certain target road segment are 0.95, 0.88, 0.72, and 0.45 respectively, with a total of 3.00, then the update weights corresponding to the four records are 0.317, 0.293, 0.240, and 0.150 respectively; Taking the rut depth feature as an example, the rut depth measurements corresponding to the four records are 8.2mm, 8.5mm, 9.1mm, and 11.3mm, respectively; When the fitting is unweighted, all four records participate equally, and the degradation trend obtained by fitting is affected by the low confidence record of 11.3 mm, resulting in a larger degradation rate. After weighted fitting, the combined weights of the high-confidence records 8.2 mm and 8.5 mm reached 0.61, dominating the fitting direction, while the low-confidence record 11.3 mm accounted for only 0.15, and its influence on the fitting results was greatly suppressed. The updated model parameters are closer to the actual deterioration pattern of the target road section, and the deterioration trend calculation result is about 0.08 mm / day, which is consistent with the long-term change pattern of the target road section's historical records. Once the edge receives the updated model, it completes the local replacement. Subsequent measurement data can be directly used to perform reasonable interval prediction and evolutionary support calculation using the updated model.

[0019] Furthermore, an adaptive verification and cloud-based collaborative analysis method for road surface measurement data also includes: After generating the credibility object, the degree of difference between the evolutionary support and the consensus support is obtained; When the degree of difference exceeds the judgment threshold, a conflict marker is added to the credibility object; Upload the credibility object carrying the conflict marker and the corresponding original road surface measurement data to the cloud; The cloud platform performs consistency comparison analysis on the original road surface measurement data based on the cross-vehicle historical measurement data set, and evaluates it through the road surface state evolution model to generate a review judgment result; Based on the verification and judgment results, the original road surface measurement data are processed, and the corresponding historical records in the historical confidence sequence are marked, corrected, and their weights are updated. Based on the corrected historical reliability sequence, the road surface state evolution model is incrementally updated and corrected; The updated road surface state evolution model is then distributed to the edge. In this embodiment, after generating the credibility object, the degree of difference between the evolutionary support and consensus support is calculated based on the latter. The degree of difference is compared with a judgment threshold obtained by statistical analysis of historical records through a sliding time window. The judgment threshold is determined according to the statistical distribution of the degree of difference corresponding to each historical record in the historical credibility sequence and is dynamically adjusted as the sequence is updated: the degree of difference corresponding to each historical record in the historical credibility sequence of the target road segment is statistically analyzed, sorted from smallest to largest, and searched from the smallest to the largest end. The position where the increase between adjacent degree of difference first exceeds the average increase between adjacent degree of difference in the sequence is taken as the dividing point, and the degree of difference corresponding to the dividing point is taken as the judgment threshold. When the degree of difference exceeds the judgment threshold, a conflict mark is added to the credibility object, and the credibility object with the conflict mark and the corresponding original measurement data are uploaded to the cloud. The cloud platform performs consistency comparison analysis based on historical measurement data sets across vehicles and conducts evaluation in conjunction with road condition evolution models; Based on the evaluation results, a review judgment result is generated, and the original data is retained, removed, or marked as difficult. At the same time, the corresponding records in the historical credibility sequence are marked, corrected, and their weights are updated. The correction value required for the correction is calculated as follows: ; In the formula, Indicates the correction value. This represents the predicted value from the evolutionary model. This represents the comprehensive observation results. Indicates evolutionary support. Indicates the degree of consensus support. The theoretical state value at the current moment is predicted based on the road surface state evolution model; The state evolution model is incrementally updated and corrected based on the corrected historical credibility sequence, and the updated model is then distributed to the edge. Among them, the conflict markers and the conflict records in the historical credibility sequence together constitute the conflict change record; Optionally, in a certain verification, the evolutionary support is 0.45 and the consensus support is 0.82, with a difference of 0.37 between the two. The average increase in the degree of difference between adjacent records in the historical reliability sequence of this road segment is 0.06. Sorted by the degree of difference from smallest to largest, the values ​​are [0.03, 0.05, 0.07, 0.10, 0.12, 0.14, 0.21, 0.38]. Searching from the smallest end, the increase between 0.21 and 0.38 is 0.17, which exceeds the average increase of 0.06 for the first time. The dividing point of 0.21 is taken as the current judgment threshold. The current difference level of 0.37 exceeds the threshold of 0.21. Add a conflict flag to the trustworthy object and upload it to the cloud. The cloud retrieved measurement data from three other testing vehicles on the same road segment. Consistency comparison showed that the average measurement value of the three vehicles deviated from the current data by 0.41, which is far beyond the normal fluctuation range. At the same time, the cloud-based state evolution model predicted a value of 0.68 for this road segment, which also deviated significantly from the current measurement value. After comprehensive evaluation, a review judgment result is generated, the original data is marked as to be removed, and the corresponding record in the historical credibility sequence is marked as invalid and its weight is reduced to zero; The state evolution model is incrementally updated based on the corrected sequence, the parameter estimates of the road segment deterioration rate are corrected, and the updated model is sent to the edge.

[0020] Furthermore, an adaptive verification and cloud-based collaborative analysis method for road surface measurement data also includes: Obtain the identifier of the lifecycle state in the credibility object, wherein the lifecycle state includes at least the stable state, the wave dynamic state, and the disputed state; When the target road segment is in the stable state, subsequent data collection and verification are performed on the target road segment according to a preset regular cycle. When in the wave dynamic, shorten the next collection and verification time window for the target road segment, and increase the monitoring weight of the credibility object in the historical credibility sequence. The monitoring weight is used to adjust the influence factor of the credibility object in subsequent credibility calculations. When in the disputed state, the credibility object corresponding to the target road segment is temporarily suspended from participating in the update of the road surface state evolution model, and multi-source cross-verification of adjacent road segments is triggered. In this embodiment, after generating the credibility object, the data state corresponding to the credibility object is updated, and the updated data state is written into the credibility object as the identifier of the life cycle state. The life cycle state includes at least the stable state, the fluctuating state, and the disputed state, which are jointly determined based on the credibility score range, the credibility change trend, and the historical conflict marking situation. When the confidence score is in the high confidence range and the number of conflict events occurring within the preset time window is less than the historical average number of conflict events, it is determined to be in a stable state. At this time, the data collection and verification process is executed according to the regular sampling cycle, and the default monitoring weight remains unchanged. When the credibility score is in the middle range and the historical conflict markers appear intermittently, it is determined to be a wave dynamic. At this time, the next collection and verification time window is shortened, the monitoring weight of the corresponding credibility object in the historical credibility sequence is increased, and its monitoring weight in subsequent credibility calculations is also increased to improve the sensitivity to short-term fluctuations. The adjustment range of the monitoring weight is determined by the magnitude of the credibility change. The greater the fluctuation of the credibility score, the greater the increase in the corresponding monitoring weight. When the fluctuation decreases, the monitoring weight is gradually reduced and restored to the default level. When the credibility score continues to decrease and the conflict markers continue to increase, it is determined to be in a disputed state. At this time, the credibility object's participation in the parameter update of the state evolution model is temporarily suspended, and multi-source cross-checking is performed on the adjacent road segments to determine whether there is a regional systematic error. For example, the confidence scores of a target road segment in five consecutive verifications were 0.93, 0.91, 0.85, 0.71, and 0.64, respectively. The confidence scores of the last three verifications changed by -0.06, -0.14, and -0.07, respectively, showing a continuous downward trend. Historical conflict markers appeared 0, 1, 2, and 3 times in the last four verifications, respectively, showing a continuous upward trend. Based on the judgment criteria, the state of this road segment changed from the initial stable state to a dynamic state of controversy. When entering the wave dynamic, the next collection and verification time window is shortened from the original 30 days to 15 days, and the monitoring weight of the corresponding credibility object is increased from the default value of 0.25 to 0.42; the monitoring weight is continuously adjusted upward during the period of increased credibility score fluctuation, and gradually reduced to the default level after the fluctuation slows down; Once the disputed state is entered, the participation of the credibility object of that road segment in the state evolution model update is temporarily suspended. At the same time, multi-source cross-checking of the adjacent road segments within 500 meters before and after the road segment is triggered to determine whether the data anomaly of that road segment is an isolated phenomenon or a regional systematic error. By jointly driving the lifecycle state and conflict change records, the trustworthy object forms a continuous state change sequence in the time dimension, enabling adaptive control of the state evolution model update strategy.

[0021] The advantages of this invention are as follows: By establishing a state evolution model that reflects the long-term evolution law of the target road segment, the measurement data is matched and analyzed with the reasonable interval predicted by the model and quantified into evolution support. At the same time, the consistency of multiple observation records within the spatiotemporal range is independently quantified into consensus support, forming a dual verification mechanism that combines evolution law verification and group observation verification. This solves the problem that existing technologies rely solely on single measurement results and fixed thresholds for anomaly judgment, and are easily affected by equipment errors, environmental changes, and local anomaly disturbances, leading to misjudgments. On this basis, the evolution support and consensus support are integrated into a structured credibility object and continuously stored in the historical credibility sequence. Based on the credibility score, corresponding update weights are assigned to the historical records. Through cloud-edge collaboration, automatic discovery, verification processing, and continuous model correction of conflict data are achieved, solving the problem that credibility information is used only once and cannot be continuously accumulated and utilized for feedback in existing technologies. By constructing a closed-loop collaborative mechanism that generates credibility through data verification, drives model updates through credibility, and optimizes data verification through the updated model, the system can continuously learn the state change law of the target road segment, improving the ability to identify abnormal data, the reliability of data verification, and the adaptability to complex working conditions.

[0022] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for adaptive verification and cloud-based collaborative analysis of road surface measurement data, characterized in that, include: S1. Obtain the pavement measurement data corresponding to the target road segment, preprocess the pavement measurement data, and establish the historical reliability sequence of the target road segment; S2. Filter the historical confidence sequence of the target road segment, and perform evolutionary deduction starting from the most recent historical confidence sequence to generate the current road surface condition prediction value, fluctuation range, and the degree of matching between the fluctuation range and the measurement data, and generate evolutionary support. S3. Obtain multiple observation records for the target road segment and compare them with the current road surface measurement data. Combine the historical credibility information of multiple observation records to generate consensus support. S4. Construct a multi-dimensional credibility vector based on evolutionary support and consensus support, generate a structured credibility object, and bind the credibility object to the road surface measurement data. S5. Determine the verification result based on the credibility score, evolution support and consensus support in the credibility object and write it into the historical credibility sequence. Upload it to the cloud for verification and correct the historical credibility sequence based on the verification result. S6. Assign weights to historical credibility sequences based on credibility scores and update road condition projection parameters.

2. The adaptive verification and cloud-based collaborative analysis method for road surface measurement data according to claim 1, characterized in that, Step S1 specifically includes: Acquire raw road surface measurement data synchronously collected by multiple source sensors while the data collection vehicle is traveling on the target road section; The original road surface measurement data is timestamped and spatially calibrated to obtain a standard measurement data frame; The attitude information of the acquisition vehicle is used to perform motion distortion correction on the standard measurement data frame to obtain distortion-free road surface measurement data; The distortion-free road surface measurement data is preprocessed, and the preprocessed road surface measurement data is used as the road surface measurement data corresponding to the target road segment.

3. The adaptive verification and cloud-based collaborative analysis method for road surface measurement data according to claim 2, characterized in that, Step S2 specifically includes: The historical reliability sequence of the target road segment is parsed to extract historical measurement records, and the historical measurement records are filtered for reliability based on the historical reliability sequence to obtain a set of historical reliable measurement records; The historical reliable measurement record set is reconstructed by time series, sorted according to the collection time order, and trajectory alignment is performed by combining spatial location information to form structured road surface state sequence data; Based on the road surface state sequence data, multidimensional road surface state feature vectors are extracted, and the state correspondence of the feature vectors in the time dimension is constructed. A state evolution model of road surface state over time is established based on the state correspondence in the time dimension.

4. The adaptive verification and cloud-based collaborative analysis method for road surface measurement data according to claim 3, characterized in that, Step S2 further includes: Based on the state evolution model, the most recent reliable measurement record is used to perform evolutionary deduction, generate the road surface state prediction value corresponding to the current time, and simultaneously output the fluctuation range corresponding to the prediction value to obtain the initial reasonable range. Obtain the set of environmental impact factors corresponding to the current moment, and adjust the reasonable range to generate a corrected reasonable range; The degree of interval matching is obtained based on the road surface measurement data and the corrected reasonable interval. The parameters are calibrated based on the matching relationship between historical road surface condition data and corresponding reasonable intervals to obtain the evolutionary support.

5. The adaptive verification and cloud-based collaborative analysis method for road surface measurement data according to claim 4, characterized in that, Step S3 specifically includes: Acquire multiple road surface observation records for the target road segment within a spatiotemporal range, and process the multiple road surface observation records; Based on road surface observation records, corresponding road surface condition observation features are extracted, and the consistency relationship between the road surface condition observation features and the current road surface measurement data is determined. Based on the aforementioned consistency relationship and combined with the historical reliability information corresponding to the road surface observation records, the multiple observation records are processed to generate a comprehensive observation result. Based on the degree of matching between the comprehensive observation results and the road surface measurement data, the overall consistency level of the road surface measurement data is determined, and a consensus support level is generated based on the overall consistency level.

6. The adaptive verification and cloud-based collaborative analysis method for road surface measurement data according to claim 5, characterized in that, Step S4 specifically includes: Obtain the evolutionary support and the consensus support, and normalize the evolutionary support and the consensus support. A multi-dimensional credibility vector is constructed based on the normalized evolutionary support and the normalized consensus support. A structured credibility object is generated based on the multidimensional credibility vector, and the source markers corresponding to the contribution components are recorded in the credibility object; The credibility object is bound to the road surface measurement data to form a complete data unit containing source information and contribution weight.

7. The adaptive verification and cloud-based collaborative analysis method for road surface measurement data according to claim 6, characterized in that, Step S5 specifically includes: Based on the credibility score, evolutionary support, and consensus support contained in the credibility object, the verification result of the road surface measurement data is determined; Set a confidence interval, compare the current confidence score with the confidence interval, and process the road surface measurement data based on the comparison results; The credibility object is associated with the verification results of the road surface measurement data and stored, and written into the historical credibility sequence according to the target road segment and collection time.

8. The adaptive verification and cloud-based collaborative analysis method for road surface measurement data according to claim 7, characterized in that, Step S6 specifically includes: Historical credibility objects are extracted from the historical credibility sequence, and the credibility score, evolution support and consensus support corresponding to the historical credibility objects are analyzed to construct a model update sample set; Based on the confidence scores in the updated sample set, the historical measurement records are assigned dynamic update weights. The historical measurement records are fitted with the dynamically updated weights to update the parameters or state transition relationships of the pavement state evolution model and calculate the latest pavement deterioration evolution trend. The updated road surface state evolution model is fed back to the edge or acquisition end.

9. The adaptive verification and cloud-based collaborative analysis method for road surface measurement data according to claim 1, characterized in that, Also includes: After generating the credibility object, the degree of difference between the evolutionary support and the consensus support is obtained; When the degree of difference exceeds the judgment threshold, a conflict marker is added to the credibility object; Upload the credibility object carrying the conflict marker and the corresponding original road surface measurement data to the cloud; The cloud platform performs consistency comparison analysis on the original road surface measurement data based on the cross-vehicle historical measurement data set, and evaluates it through the road surface state evolution model to generate a review judgment result; Based on the verification and judgment results, the original road surface measurement data are processed, and the corresponding historical records in the historical confidence sequence are marked, corrected, and their weights are updated. Based on the corrected historical reliability sequence, the road surface state evolution model is incrementally updated and corrected; The updated road surface state evolution model is then distributed to the edge.

10. The adaptive verification and cloud-based collaborative analysis method for road surface measurement data according to claim 9, characterized in that, Also includes: Obtain the identifier of the lifecycle state in the credibility object, wherein the lifecycle state includes at least the stable state, the wave dynamic state, and the disputed state; When the target road segment is in the stable state, subsequent data collection and verification are performed on the target road segment according to a preset regular cycle. When in the wave dynamic, shorten the next collection and verification time window for the target road segment, and increase the monitoring weight of the credibility object in the historical credibility sequence. The monitoring weight is used to adjust the influence factor of the credibility object in subsequent credibility calculations. When in the disputed state, the credibility object corresponding to the target road segment is temporarily suspended from participating in the update of the road surface state evolution model, and multi-source cross-verification of adjacent road segments is triggered.

Images