Traffic facility monitoring data denoising method and device, equipment and medium

Abstract

The application discloses a noise reduction method and device for traffic facility monitoring data, equipment and medium, relates to the technical field of digital data processing, and the method comprises the steps of obtaining a plurality of monitoring sequences and response constraint records of a traffic facility, generating a sample time domain interval, a measurement interval, a path time domain interval and a co-time unit according to a time stamp, an error thereof, a measurement error and a time delay constraint, determining a to-be-processed sample and a time sequence reference interval according to a target monitoring point sample and front and rear samples, converting a source monitoring point measurement interval into a path mapping interval, generating a response consistent interval in combination with a time domain coincidence relationship and a path checking weight, determining a noise sample and a replacement value according to the intersection relationship of the measurement interval, the time sequence reference interval and the response consistent interval, and generating a noise reduction monitoring sequence. The application can identify noise samples by using corresponding monitoring data of the same facility response when the sampling time and the sampling frequency of each monitoring sequence are inconsistent, and reduces the single-sequence judgment deviation.

Description

Technical Field

[0001] This application relates to the field of electronic digital data processing technology, and in particular to noise reduction methods, devices, equipment and media for traffic facility monitoring data. Background Technology

[0002] Transportation facilities such as bridges, tunnels, and roads typically have multiple monitoring points set up during operation to collect data on strain, displacement, vibration, and temperature. To reduce the impact of sensor errors and transmission interference on the monitoring results, it is necessary to identify and correct noise samples in the monitoring data.

[0003] Currently, filtering is usually performed based on the changes in adjacent samples in a single monitoring sequence, or the sampling times of different monitoring sequences are first unified through resampling, interpolation, etc., and then the correlation between multiple monitoring sequences is used to identify abnormal data.

[0004] However, the sampling times and frequencies of different monitoring sequences may differ, and direct resampling or interpolation can easily introduce new errors. Furthermore, processing each monitoring sequence individually makes it difficult to utilize the correlation between the responses of the same facility at different monitoring points, potentially misclassifying normal response changes as noise. Therefore, when the sampling times and frequencies of different monitoring sequences for traffic facilities are inconsistent, identifying noise samples based on the monitoring data corresponding to the responses of the same facility becomes an urgent problem to be solved. Summary of the Invention

[0005] The purpose of this application is to provide a noise reduction method, apparatus, equipment and medium for traffic facility monitoring data, which aims to solve the technical problem of how to identify noise samples based on the monitoring data corresponding to the response of the same facility when the sampling time and sampling frequency of different monitoring sequences of traffic facilities are inconsistent.

[0006] To achieve the above objectives, this application proposes a noise reduction method for traffic facility monitoring data, the method comprising: Acquire multiple monitoring sequences and response constraint records for traffic facilities, wherein the response constraint records include response paths, time delay constraints, conversion constraints, and path verification weights; Based on the timestamp, timestamp error, measurement error, and time delay constraint of the monitoring sequence, a sample time domain interval, a measurement interval, a path time domain interval, and a synchronous unit are generated. Based on the target monitoring point samples and previous and subsequent samples associated with the synchronous unit, determine the sample to be processed, the synchronous unit to be processed, and the time series reference interval. According to the conversion constraints, the measurement interval of the source monitoring point associated with the synchronous unit to be processed is converted into a path mapping interval, and then aggregated according to the temporal overlap relationship and the path verification weight to obtain the response consistency interval; Based on the intersection relationship between the measurement interval of the sample to be processed, the time series reference interval, and the response consistency interval, noise samples and replacement values ​​are determined, and a noise reduction monitoring sequence is generated.

[0007] Furthermore, to achieve the above objectives, this application also proposes a noise reduction device for traffic facility monitoring data, the device comprising: The data acquisition module is used to acquire multiple monitoring sequences and response constraint records of traffic facilities. The response constraint records include response paths, time delay constraints, conversion constraints, and path verification weights. The interval generation module is used to generate sample time domain intervals, measurement intervals, path time domain intervals, and synchronous units based on the timestamps, timestamp errors, measurement errors, and time delay constraints of the monitoring sequence. The sample screening module is used to determine the sample to be processed, the time unit to be processed, and the time series reference interval based on the target monitoring point sample and the samples before and after the synchronous unit. The interval aggregation module is used to convert the measurement interval of the source monitoring point associated with the synchronous unit to be processed into a path mapping interval according to the conversion constraint, and to aggregate it according to the temporal overlap relationship and the path verification weight to obtain the response consistency interval; The data replacement module is used to determine noise samples and replacement values ​​based on the intersection relationship between the measurement interval of the sample to be processed, the time series reference interval, and the response consistency interval, and to generate a noise reduction monitoring sequence.

[0008] In addition, to achieve the above objectives, this application also proposes a noise reduction device for traffic facility monitoring data, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the noise reduction method for traffic facility monitoring data as described above.

[0009] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the noise reduction method for traffic facility monitoring data as described above.

[0010] One or more technical solutions proposed in this application have at least the following technical effects: First, multiple monitoring sequences and response constraint records of traffic facilities are acquired. Response paths, time delay constraints, conversion constraints, and path verification weights are used to characterize the response transmission relationship between different monitoring points, providing a constraint basis for data comparison across monitoring sequences. Then, sample time domain intervals and measurement intervals are generated based on timestamps, timestamp errors, and measurement errors. Combined with time delay constraints, path time domain intervals and synchronous units are generated, thus establishing the time correspondence between different monitoring sequences while considering sampling time errors and measurement errors. Next, based on the target monitoring point samples and preceding and following samples associated with the synchronous units, the samples to be processed, the synchronous units to be processed, and the time series reference interval are determined. The process involves filtering samples that require further evaluation based on the changes in the monitoring points themselves. Further, the source monitoring point measurement intervals are converted into path mapping intervals according to conversion constraints, and then aggregated using temporal overlap relationships and path verification weights to obtain response-consistent intervals supported by multiple source monitoring points. Finally, noise samples and replacement values ​​are determined based on the intersection relationships between the measurement intervals of the samples to be processed and the temporal reference intervals and response-consistent intervals, generating a noise-reduced monitoring sequence. This allows for the identification of noise samples based on monitoring data corresponding to the response of the same facility when the sampling times and frequencies of different monitoring sequences for traffic facilities are inconsistent, reducing the bias caused by judging solely based on a single monitoring sequence. Attached Figure Description

[0011] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0012] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart illustrating the first embodiment of the noise reduction method for traffic facility monitoring data provided in this application; Figure 2 This is a schematic diagram illustrating the generation of response constraint records provided in the first embodiment of the noise reduction method for traffic facility monitoring data in this application; Figure 3 This is a schematic diagram illustrating the determination of the sample to be processed and the time series reference interval provided in the first embodiment of the noise reduction method for traffic facility monitoring data in this application. Figure 4 This is a schematic diagram of the response consistency interval and replacement value generation provided in the first embodiment of the noise reduction method for traffic facility monitoring data in this application; Figure 5This is a flowchart illustrating the second embodiment of the noise reduction method for traffic facility monitoring data provided in this application; Figure 6 This is a schematic diagram of the synchronous unit generation provided in the second embodiment of the noise reduction method for traffic facility monitoring data in this application; Figure 7 This is a schematic diagram of the module structure of a noise reduction device for traffic facility monitoring data according to an embodiment of this application; Figure 8 This is a schematic diagram of the equipment structure of the hardware operating environment involved in the noise reduction method for traffic facility monitoring data in the embodiments of this application.

[0014] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0015] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0016] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0017] It should be noted that the executing entity of this application embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device or traffic data processing system capable of realizing the above functions. The following uses a traffic data processing system as an example to describe this embodiment and the subsequent embodiments.

[0018] Based on this, the first embodiment of this application provides a noise reduction method for traffic facility monitoring data, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the noise reduction method for traffic facility monitoring data in this application.

[0019] In this embodiment, the noise reduction method for traffic facility monitoring data includes steps S10 to S50: Step S10: Obtain multiple monitoring sequences and response constraint records for traffic facilities. The response constraint records include response paths, time delay constraints, conversion constraints, and path verification weights. It should be noted that multi-class monitoring sequences refer to time-series data sets organized according to monitoring points and monitoring types, which may include strain, displacement, acceleration, tilt angle, cable force, and temperature monitoring sequences; response constraint records refer to records used to describe the data correspondence when different monitoring points monitor the response of the same facility; response paths refer to the component sequences from the source monitoring point to the target monitoring point along the component connection relationship; time delay constraints refer to the allowable range of time required for the response to be transmitted along the corresponding response path; conversion constraints refer to the conversion relationship and parameter range for converting the monitoring values ​​of the source monitoring point to the corresponding physical quantities of the target monitoring point; path verification weight refers to the reliability of the response path determined based on the calibration data, with a value range of [0, 1]. The larger the value, the higher the calibration matching degree of the corresponding response path.

[0020] Understandably, the traffic data processing system receives data packets uploaded by various monitoring devices, extracts monitoring samples according to the monitoring point identifier and monitoring type, and forms corresponding monitoring sequences. Each monitoring sample records at least the monitoring point identifier, monitoring type, timestamp, monitoring value, timestamp error, and measurement error. The timestamp error can originate from device clock synchronization deviation, sampling trigger deviation, or time stamp deviation. Subsequent calculations uniformly use a non-negative upper limit for the timestamp error. When the data packet records a time error interval, the larger of the absolute values ​​of the two endpoints of the interval is taken as the upper limit of the timestamp error. The measurement error can originate from sensor accuracy, calibration results, and the resolution of the acquisition device. Subsequent calculations uniformly use a non-negative absolute measurement error in the same unit as the monitoring value. When the data packet records a measurement error interval, the larger of the absolute values ​​of the two endpoints of the interval is taken as the absolute measurement error.

[0021] Facility response refers to changes in facility status caused by the same process such as traffic load, environmental effects, or component vibration, and can be reflected by multiple monitoring points. Target monitoring point refers to the monitoring point currently requiring noise identification, while source monitoring point refers to other monitoring points that record the same facility response as the target monitoring point in the calibration response segment and can provide response references. Component connection relationship refers to structured data recording the identification, connection type, and transferable response type of adjacent components. Transferable response type refers to the type of monitored physical quantity that can form a correlated response between adjacent components. Source and target monitoring points are determined based on the component where the monitoring point is located, the monitoring type, the monitoring direction, the calibration response segment, and the component connection relationship. Response constraint records are generated only for monitoring types with corresponding conversion relationships and that have passed calibration verification.

[0022] Response constraint records can be pre-generated based on historical calibration response segments, recording the time delay range, conversion relationship, conversion parameter range, and path verification weight corresponding to each response path. When the units of different types of monitoring data are inconsistent, they are converted into the physical quantities and units corresponding to the target monitoring point through conversion constraints. This step establishes the response transmission relationship between different monitoring points through response constraint records, providing a constraint basis for subsequent comparison of multiple types of monitoring data at different sampling times and sampling frequencies.

[0023] Step S20: Based on the timestamp, timestamp error, measurement error, and time delay constraint of the monitoring sequence, generate the sample time domain interval, measurement interval, path time domain interval, and synchronous unit; It should be noted that the sample time domain interval refers to the time range that can be attributed to the current sample, formed by shrinking inward according to the midpoint of the timestamps of the current sample and the adjacent samples, and according to the upper limit of the timestamp error of each sample; the measurement interval refers to the numerical range in which the actual state quantity of the monitored object may be located after considering the measurement error; the path time domain interval refers to the time range formed by moving the sample time domain interval of the source monitoring point sample according to the response transmission delay; and the synchronous unit refers to the time unit that records the index relationship between the target monitoring point sample and the source monitoring point sample with time correspondence.

[0024] Understandably, the time range corresponding to the sample is determined based on the timestamps and timestamp errors of each monitoring sequence, and the numerical range corresponding to the sample is determined based on the monitored values ​​and measurement errors. Combined with the response propagation delay, a time correspondence is established between the source monitoring point samples and the target monitoring point samples. The synchronous unit only records the sample correspondence between different monitoring sequences, without changing the original timestamps, monitored values, or sampling frequencies, and does not generate interpolated samples. Samples that cannot establish a time correspondence do not participate in the current cross-monitoring point comparison and remain in the original monitoring sequence. This step characterizes the uncertainty of the monitoring data through the sample time domain interval and measurement interval, enabling monitoring sequences with different sampling times and sampling frequencies to establish a correspondence.

[0025] Step S30: Based on the target monitoring point samples and previous and subsequent samples associated with the synchronous unit, determine the sample to be processed, the synchronous unit to be processed, and the time series reference interval. It should be noted that the time series reference interval refers to the reasonable range of monitoring values ​​at the current time calculated based on the changes of the target monitoring point sample before and after; the sample to be processed refers to the target monitoring point sample that deviates from the time series reference interval and needs to be further judged in conjunction with other monitoring points; the synchronous unit to be processed refers to the synchronous unit associated with the sample to be processed.

[0026] Understandably, based on the target monitoring point samples associated with the synchronous unit and the preceding and following samples in the same monitoring sequence, the reference value range of the target sample at the current timestamp is determined. Then, based on the relationship between the measurement interval of the target sample and this reference value range, samples requiring further judgment in conjunction with other monitoring points are selected. If preceding and following samples or a reference value range that meets the processing conditions cannot be obtained, the corresponding target monitoring point sample is not identified as a sample to be processed. This step first uses the temporal changes of the target monitoring point itself to screen the objects to be processed, avoiding direct identification of noise based solely on a single measurement result.

[0027] Step S40: Convert the measurement interval of the source monitoring point associated with the synchronous unit to be processed into a path mapping interval according to the conversion constraint, and aggregate them according to the time domain overlap relationship and the path verification weight to obtain the response consistency interval; It should be noted that the path mapping interval refers to the numerical range obtained after converting the measurement interval of the source monitoring point to the physical quantity space of the target monitoring point according to the conversion constraint; the temporal overlap relationship refers to the degree of common coverage between the temporal interval of the source monitoring point path and the temporal interval of the sample to be processed; the response consistency interval refers to the reasonable response range of the target monitoring point supported by different source monitoring points reaching the preset number of source points.

[0028] Understandably, according to the corresponding conversion constraints, the source monitoring point data associated with the synchronous unit to be processed is converted to the physical quantity and unit corresponding to the target monitoring point. Then, combining the time correspondence between the source monitoring point samples and the sample to be processed, as well as the verification results of the response path, the response ranges provided by different source monitoring points are aggregated to obtain a response consistency interval reflecting the results jointly supported by multiple source monitoring points. If a valid aggregation result cannot be formed, a response consistency interval is not generated, and the original monitoring value of the current sample to be processed is retained.

[0029] Step S50: Based on the intersection relationship between the measurement interval of the sample to be processed, the time series reference interval, and the response consistency interval, determine the noise sample and replacement value, and generate a noise reduction monitoring sequence.

[0030] It should be noted that a noise sample refers to a monitoring sample that deviates from both the time-series variation range of the target monitoring point and the range of consistent responses of multiple monitoring points; a replacement value refers to a value used to replace the monitoring value of the noise sample, which is simultaneously constrained by the time-series reference interval and the response consistency interval; and a noise reduction monitoring sequence refers to a monitoring sequence that has completed the noise sample replacement while retaining the original time sequence.

[0031] Understandably, based on the intersection between the measurement interval of the sample to be processed and the time-series reference interval and the response consistency interval, it is determined whether the sample to be processed belongs to noise samples, and a replacement value is determined within the range allowed by both reference results. When a valid replacement value can be obtained, the monitoring value of the noise sample is replaced, and the original timestamp, monitoring point identifier, and monitoring type are retained; when a valid replacement value cannot be obtained, the original monitoring value of the corresponding sample is retained. After processing, the samples are merged according to the original timestamp order to generate a noise-reduced monitoring sequence.

[0032] This embodiment establishes a correspondence between asynchronous monitoring samples by using the sample time domain interval and response delay. Then, it identifies noise samples by combining the range of change before and after the target monitoring point and the response mapping results of different source monitoring points. This does not require changing the original sampling time and sampling frequency of each monitoring sequence, nor does it require generating interpolated samples. At the same time, the monitoring value is only replaced when the time series reference result and the response result across monitoring points are mutually verified, which improves the reliability of the noise reduction process of traffic facility monitoring data and provides a data foundation for subsequent facility status analysis.

[0033] As an example, the steps of obtaining multiple monitoring sequences and response constraint records for traffic facilities include: obtaining monitoring data packets uploaded by each monitoring device of the traffic facility; extracting monitoring point identifiers, monitoring types, timestamps, monitoring values, timestamp errors, and measurement errors from the monitoring data packets; grouping the extracted results according to the monitoring point identifiers and monitoring types, and arranging the monitoring values ​​in each group according to the timestamps to obtain multiple monitoring sequences; reading the correspondence between monitoring points and facility components, component connection relationships, monitoring directions, and calibration response segments; determining the response path, time delay constraints, conversion constraints, and path verification weights from the source monitoring point to the target monitoring point based on the calibration response segments, component connection relationships, and monitoring directions, and generating response constraint records; and associating the response constraint records with the corresponding monitoring sequences according to the monitoring point identifiers and monitoring types of the source and target monitoring points.

[0034] It should be noted that a monitoring data packet refers to a data record sent by a monitoring device, data acquisition gateway, or traffic facility monitoring platform that contains monitoring samples and their collected information; a calibration response segment refers to a monitoring data segment extracted from the source monitoring point and the target monitoring point when the traffic facility is in a normal response state, corresponding to the same facility response process, which can come from facility loading tests, equipment calibration processes, or historical monitoring periods where no sensor anomalies have been confirmed; and the monitoring direction refers to the spatial direction corresponding to when the monitoring device collects physical quantities such as displacement, strain, acceleration, or tilt angle.

[0035] Understandably, the traffic data processing system receives monitoring data packets uploaded by various monitoring devices in traffic facilities. These data packets can be sent directly by the monitoring devices or collected and sent by the data acquisition gateway. The monitoring data packets use a fixed field order or key-value field structure, and at least include the device identifier, data packet sequence number, monitoring point identifier, monitoring type, timestamp, monitoring value, timestamp error, measurement error, and data integrity verification fields. The system parses the data packets according to the communication protocol of the corresponding monitoring device, converts the timestamp to a unified time base, and converts the monitoring value and measurement error to a unified unit used for the corresponding monitoring type.

[0036] After analysis, the monitoring samples are grouped according to the monitoring point identifier and monitoring type. For example, when strain and temperature data are collected at monitoring point P01, the strain monitoring values ​​corresponding to P01 form one strain monitoring sequence, and the temperature monitoring values ​​corresponding to P01 form another temperature monitoring sequence. When the same monitoring type is collected by multiple monitoring points, each monitoring point forms its own monitoring sequence. Subsequently, the monitoring samples are arranged from earliest to latest according to the timestamp to obtain multiple types of monitoring sequences.

[0037] For data packets with duplicate timestamps, first determine whether they are duplicate transmissions based on the device identifier, data packet sequence number, and data content. If the data packet sequence number is the same and the data content is identical, only one copy is retained. When multiple different monitoring values ​​correspond to the same monitoring point, the same monitoring type, and the same timestamp, prioritize retaining samples that have passed integrity verification and have normal quality indicators. If a unique valid sample cannot be determined based on this, the relevant data packets are recorded as timestamp conflict data, and the relevant samples are not used to generate sample time domain intervals, but the original data packets are still saved. For data packets arriving out of order, they are rearranged according to the parsed timestamps. For data segments with non-contiguous data packet sequence numbers or adjacent timestamp intervals exceeding the normal sampling interval, the corresponding missing time period is recorded, and no supplementary samples are generated. The normal sampling interval is determined based on the sampling frequency configured in the monitoring equipment, and the data missing judgment threshold can be set to 1 to 3 times the normal sampling interval.

[0038] Next, the pre-saved correspondence between monitoring points and facility components, component connection relationships, and monitoring directions are read. The correspondence between monitoring points and facility components records at least the monitoring point identifier, the identifier of the component it belongs to, the monitoring type, and the monitoring direction; the component connection relationship records the connection order between components according to the identifiers of adjacent components. Based on the target monitoring point that needs to be processed, other monitoring points that can transmit the same facility response to the target monitoring point are selected from the component connection relationships as source monitoring points, and a response path is formed according to the connection order from the component where the source monitoring point is located to the component where the target monitoring point is located.

[0039] For each candidate response path, the timestamps, monitoring values, timestamp errors, and measurement errors of the source and target monitoring points are extracted from the calibration response segment. The delay constraints, conversion constraints, and path calibration weights are determined according to the candidate combination verification method described below. If there are no valid calibration samples or no candidate combination reaches the preset calibration threshold, no response constraint record is generated for the corresponding candidate response path.

[0040] The response constraint record adopts a structured record format, including at least a record identifier, source monitoring point identifier, target monitoring point identifier, source monitoring type, target monitoring type, component identifier sequence, time delay constraint, conversion relationship, conversion parameter range, and path verification weight. The system first locates the corresponding monitoring sequence based on the source and target monitoring point identifiers. Then, based on the source and target monitoring types in the response constraint record, it determines the specific sequence to be associated and writes the response constraint record identifier into the association information of the corresponding monitoring sequence. Monitoring sequences without corresponding response constraint records retain their original monitoring samples but do not participate in subsequent response consistency interval calculations.

[0041] This example reduces the impact of out-of-order, duplicate, and missing data packets on the sample arrangement by uniformly parsing monitoring data packets and forming monitoring sequences according to monitoring point identifiers and monitoring types. At the same time, it establishes the association between monitoring sequences and response constraint records based on the connection relationship of facility components and the calibration response fragments, so that subsequent processing can call the time delay constraints and conversion constraints corresponding to the source monitoring point, target monitoring point, and response path, avoiding incorrect matching between different monitoring points or different monitoring types.

[0042] As an example, the step of determining the response path, time delay constraint, conversion constraint, and path verification weight from the source monitoring point to the target monitoring point based on the calibration response segment, the component connection relationship, and the monitoring direction, and generating a response constraint record includes: generating a calibration sample time domain interval and a calibration measurement interval based on the timestamp error and measurement error of the calibration response segment; generating a response path from the source monitoring point to the target monitoring point based on the component connection relationship, and configuring candidate time delay parameters, candidate conversion relationships, and candidate conversion parameters for each response path; moving the calibration sample time domain interval of the source monitoring point according to each candidate combination and converting the calibration measurement interval, calculating the candidate matching rate based on the intersection of the conversion result and the corresponding interval of the target monitoring point; selecting the target combination with the largest candidate matching rate from the candidate combinations that reach the preset calibration threshold, generating time delay constraints and conversion constraints based on the target combination, using the corresponding candidate matching rate as the path verification weight, and generating a response constraint record.

[0043] It should be noted that the calibration sample time domain interval refers to the possible time range of the sampling time corresponding to the calibration sample after considering timestamp errors; the calibration measurement interval refers to the possible numerical range of the actual measured value corresponding to the calibration sample after considering measurement errors; the candidate delay parameter refers to the delay value to be verified set within the allowable delay range of the response path; the candidate conversion relationship can be a proportional conversion relationship or a single conversion relationship with bias; the candidate conversion parameter refers to the verification value of conversion relationship parameters such as proportional coefficient and bias; the candidate combination refers to the parameter combination formed by a candidate delay parameter, a candidate conversion relationship, and the corresponding candidate conversion parameter; the candidate matching rate refers to the proportion of the number of calibration samples that simultaneously meet the time domain intersection condition and the measurement interval intersection condition after processing with candidate combinations to the total number of source monitoring point calibration samples participating in the verification; the target combination refers to the candidate combination that reaches the preset calibration threshold and has the largest candidate matching rate; the preset calibration threshold refers to the minimum candidate matching rate for judging whether the candidate combination can represent the corresponding response path.

[0044] Please refer to Figure 2 , Figure 2 This diagram illustrates the generation of response constraint records provided in the first embodiment of the noise reduction method for traffic facility monitoring data in this application. The diagram shows the process of filtering response paths and corresponding constraint parameters using calibration response segments, component connection relationships, and monitoring directions. First, calibration sample time-domain intervals and calibration measurement intervals are generated based on the calibration response segments. Candidate response paths between the source monitoring point and the target monitoring point are determined based on component connection relationships. Candidate time delay parameters, candidate conversion relationships, and candidate conversion parameters are configured for each candidate response path in conjunction with the monitoring direction. The calibration sample time-domain intervals of the source monitoring point are moved according to each candidate combination, and the corresponding calibration measurement intervals are converted using the candidate conversion relationships and candidate conversion parameters. The candidate matching rate of each candidate combination is calculated based on the intersection of the conversion result with the interval corresponding to the target monitoring point. The candidate matching rate is compared with a preset calibration threshold. If no candidate combination reaches the preset calibration threshold, no corresponding response constraint record is generated. If a candidate combination reaches the preset calibration threshold, the target combination with the highest candidate matching rate is selected. Time delay constraints and conversion constraints are generated based on the target combination. The candidate matching rate corresponding to the target combination is used as the path verification weight to form a response constraint record.

[0045] Understandably, the process begins by reading calibration samples from both the source and target monitoring points within the same calibration response segment, retaining complete samples with timestamps, monitored values, timestamp errors, and measurement errors. Using the timestamps of each calibration sample as the center, the time domain interval of the calibration sample is determined based on the timestamp error. Then, a calibration measurement interval is generated with the monitored value as the center and the measurement error as the interval radius. If the source and target monitoring points use different time units or measurement units, they are first converted to a unified time unit, and the measured values ​​and measurement errors are converted to the unified unit used for the corresponding monitoring type.

[0046] In this example, the facility response type can be strain response, displacement response, acceleration response, tilt response, or cable force response caused by traffic loads, or temperature response caused by environmental changes. Based on the component connection relationships, a connection path is searched from the component containing the source monitoring point to the component containing the target monitoring point. Each path does not repeatedly pass through the same facility component; only paths where each component connection record supports the current facility response type and where there is a valid calibrated response segment between the source and target monitoring points are retained. For monitoring types with monitoring directions, such as displacement, acceleration, and tilt, the direction conversion relationship should also be determined according to the monitoring direction and conversion constraints; scalar monitoring types such as temperature do not undergo direction filtering.

[0047] The candidate delay parameter ranges from 0 to the smaller of the maximum allowable response propagation time of the response path and the duration of the calibration response segment. The maximum response propagation time is determined based on facility loading tests, the difference in response start times between the source and target monitoring points in historical calibration response segments, or structural analysis results, and is recorded along with the determination criteria. The candidate delay step size is not less than the time resolution of the monitoring equipment and not greater than the minimum sampling interval of the calibration samples. The candidate conversion relationship can be a proportional relationship or a linear relationship with bias, expressed as y=ax and y=ax+b respectively, where x represents the monitoring value at the source monitoring point, y represents the monitoring value converted to the physical quantity space of the target monitoring point, a represents the proportionality coefficient, and b represents the bias. The dimension of a is the target monitoring value unit divided by the source monitoring value unit, and b uses the same unit as the target monitoring value.

[0048] The range of values ​​for the proportional coefficient and bias is read from the facility loading test results or equipment calibration records. If no valid value range is recorded, no corresponding candidate combination is generated. Let the maximum absolute value of the endpoints of the source monitoring point calibration measurement interval be... The smallest absolute change in measurement that the target monitoring point can distinguish is , This represents the step size of the proportionality coefficient, and its dimensions are the same as those of the proportionality coefficient. same, This indicates the offset step size, and its unit is the same as the offset amount. Same, and satisfy When using a proportional relationship, both the offset and the offset step size are 0.

[0049] Each candidate delay parameter is combined with its respective candidate conversion relationship and corresponding candidate conversion parameter to obtain a candidate combination. For example, multiple candidate delay values ​​are configured for the same path, and each is combined with the proportional coefficient of the proportional conversion relationship, the proportional coefficient of the single conversion relationship with bias, and the bias amount. Each candidate combination is checked separately, and different conversion relationships are not mixed in the same check.

[0050] For any candidate combination, the calibration sample time domain interval of the source monitoring point is moved according to the candidate time delay parameter, and then the corresponding calibration measurement interval is converted according to the candidate conversion relationship and candidate conversion parameter. The source monitoring point calibration samples are processed sequentially according to the timestamp order, and samples whose time domain intervals intersect with the moved time domain intervals are searched from the target monitoring point calibration samples for which no corresponding relationship has yet been established. If multiple target monitoring point calibration samples meet the conditions, the sample with the longest time domain overlap is selected; if the overlap durations are the same, the sample with the smaller absolute value of the timestamp difference is selected, and the selected target monitoring point calibration sample is marked as matched, ensuring that during the verification process of the same candidate combination, one target monitoring point calibration sample corresponds to at most one source monitoring point calibration sample. If no time domain intersecting sample is found, the current source monitoring point calibration sample is recorded as an unmatched sample.

[0051] Further compare the converted calibration measurement interval with the corresponding target monitoring point calibration measurement interval. When the two intervals intersect, the corresponding source monitoring point calibration sample is recorded as a matched sample; when the two intervals do not intersect, they are recorded as unmatched samples. Divide the number of matched samples by the total number of source monitoring point calibration samples participating in the current candidate combination verification to obtain the candidate matching rate; the total number of calibration samples participating in the verification does not include samples with missing fields or invalid calibration measurement intervals.

[0052] The preset calibration threshold is greater than 0 and not greater than 1, and can be determined based on the lower limit of the candidate matching rate of the confirmed effective response path in multiple calibration response segments. When there are many calibration response segments, the 5% to 20% quantile of the candidate matching rate of the effective response path can be used. The specific quantile is determined based on the number of calibration response segments and the dispersion of the candidate matching rate, and remains unchanged during the verification process of the same batch of response paths. The candidate combination with the highest candidate matching rate is selected as the target combination from the candidate combinations that reach the preset calibration threshold. When the candidate matching rates are the same, the absolute value of the difference between the midpoint of the converted calibration measurement interval and the midpoint of the corresponding target monitoring point calibration measurement interval is calculated, and the candidate combination with the smaller average absolute value of the difference is selected first. When the average absolute value of the difference is still the same, the candidate conversion relationship with fewer parameters is selected first.

[0053] Candidate delay parameters in the target combination As the delay center, the delay is extended to both sides by half the candidate delay step size, and restricted to the range of candidate delay values ​​to obtain delay constraints. For the q-th candidate conversion parameter in the target combination, the delay is extended to both sides by half the corresponding candidate parameter step size, and restricted to the range of candidate values ​​for that parameter to obtain the conversion parameter interval. When a candidate parameter is located at the endpoint of its value range, the extension result is truncated using the endpoint of the value range. The candidate matching rate corresponding to the target combination is recorded as the path verification weight; when no candidate combination reaches the preset calibration threshold, no response constraint record is generated from the current source monitoring point to the target monitoring point.

[0054] This example represents calibration samples as time-domain intervals and measurement intervals using timestamp errors and measurement errors, ensuring that the verification of candidate combinations does not depend on the calibration sample timestamps or identical monitoring values. Simultaneously, it uses the candidate matching rate to filter delay parameters and conversion parameters, and uses the candidate matching rate as a path verification weight, enabling the generated response constraint records to reflect the degree of matching of different response paths in the calibration data, reducing the impact of paths lacking stable response correspondences on subsequent calculations of consistent response intervals.

[0055] As an example, the step of determining the sample to be processed, the time unit to be processed, and the time series reference interval based on the target monitoring point sample and the samples before and after the time unit includes: determining the target monitoring point sample whose timestamp is located within the time unit as the target sample, and determining the corresponding time unit as the target time unit; generating a first rate of change interval based on the two adjacent samples before the target sample, extending the measurement interval of the previous sample to the timestamp of the target sample to obtain a first reference interval; generating a second rate of change interval based on the two adjacent samples after the target sample, extending the measurement interval of the subsequent sample in the opposite direction to the timestamp of the target sample to obtain a second reference interval; taking the intersection of the first reference interval and the second reference interval as the time series reference interval; when the measurement interval of the target sample does not intersect with the time series reference interval, and the first rate of change interval and the second rate of change interval are located on both sides of zero, determining the target sample as the sample to be processed, and determining the target time unit as the time series reference interval to be processed.

[0056] It should be noted that the target sample refers to the target monitoring point sample whose timestamp falls within the time range of the synchronous unit and for which time series deviation judgment is currently required; the target synchronous unit refers to the synchronous unit containing the timestamp of the target sample; the first rate of change interval refers to the range of monitoring value change rate calculated based on the measurement interval of the two adjacent samples before the target sample; the second rate of change interval refers to the range of monitoring value change rate calculated based on the measurement interval of the two adjacent samples after the target sample.

[0057] The first reference interval refers to the range of monitored values ​​obtained by extending the measurement interval of the previous sample to the timestamp of the target sample according to the first rate of change interval; the second reference interval refers to the range of monitored values ​​obtained by extending the measurement interval of the subsequent sample in reverse according to the second rate of change interval to the timestamp of the target sample; the time-series reference interval refers to the range of monitored values ​​covered by the first reference interval and the second reference interval; being on both sides of zero means that the lower and upper endpoints of one rate of change interval are both greater than 0, and the lower and upper endpoints of the other rate of change interval are both less than 0.

[0058] Please refer to Figure 3 , Figure 3 This diagram illustrates the determination of the sample to be processed and the time-series reference interval provided in the first embodiment of the noise reduction method for traffic facility monitoring data of this application. The diagram shows the processing relationship of filtering the sample to be processed based on the adjacent samples before and after the target sample and the measurement interval of the target sample. A first rate of change interval is generated using the two adjacent samples before the target sample, and the measurement interval of the previous sample is extended to the timestamp of the target sample to obtain the first reference interval. A second rate of change interval is generated using the two adjacent samples after the target sample, and the measurement interval of the subsequent sample is extended in the opposite direction to the timestamp of the target sample to obtain the second reference interval. The intersection of the first reference interval and the second reference interval is taken as the time-series reference interval. The intersection of the measurement interval of the target sample and the time-series reference interval is judged, and it is determined whether the first rate of change interval and the second rate of change interval are on both sides of the zero value. If the measurement interval of the target sample and the time-series reference interval do not intersect, and the two rate of change intervals are on both sides of the zero value, then the sample to be processed and the synchronous unit to be processed are determined; otherwise, the target sample is retained.

[0059] Understandably, the process begins by sequentially checking the target monitoring point samples associated with each time unit to determine if their timestamps fall within the time range of that time unit. When the time unit uses a left-closed, right-open time interval, target monitoring point samples with timestamps equal to the left endpoint are assigned to the current time unit, while those with timestamps equal to the right endpoint are not. If the right endpoint is also the left endpoint of the next time unit, the target monitoring point sample is assigned to that next time unit. Target monitoring point samples that meet the boundary assignment rules are identified as target samples, and their corresponding time units are identified as target time units. Target monitoring point samples whose timestamps do not belong to any time unit are not included in the current processing.

[0060] For the nth target sample, two valid samples are selected from the same target monitoring sequence whose timestamps are strictly earlier than the target sample and are adjacent to the target sample. These are denoted as the (n-2)th sample and the (n-1)th sample, respectively. The first rate of change interval is calculated based on the difference between the measurement intervals of the two samples and the timestamp interval. The calculation formula is as follows: The measurement interval of the (n-1)th sample is extended to the timestamp of the target sample according to the first rate of change interval to obtain the first reference interval. The calculation formula is as follows: Then, two valid samples with timestamps strictly later than the target sample and adjacent to the target sample are selected from the same target monitoring sequence, and are denoted as the (n+1)th sample and the (n+2)th sample, respectively. The second rate of change interval is calculated based on the difference between the measurement intervals of the two samples and the timestamp interval, using the following formula: Extend the measurement interval of the (n+1)th sample backward to the timestamp of the target sample to obtain the second reference interval. The calculation formula is as follows: In the formula, This represents the first rate of change interval corresponding to the nth target sample. This indicates the corresponding second rate of change interval. Indicates the first reference interval. Indicates the second reference interval. and These represent the measurement intervals of corresponding adjacent samples in the target monitoring sequence. and These represent the timestamps of the corresponding samples. Since all samples come from the same target monitoring sequence, the monitoring sequence index is omitted in this set of formulas. The dimension of the rate of change interval is the monitoring value unit divided by the time unit, and the dimension of the rate of change interval multiplied by the time interval is consistent with that of the measurement interval.

[0061] For the difference between two measurement intervals, the lower endpoint of the first interval minus the upper endpoint of the second interval is used as the lower endpoint of the difference interval, and the upper endpoint of the first interval minus the lower endpoint of the second interval is used as the upper endpoint of the difference interval. When multiplying an interval by a positive number, the order of the endpoints is maintained; when multiplying an interval by a negative number, the order of the endpoints after multiplication is swapped. The time difference in the first reference interval is positive, and the time difference in the second reference interval is negative, thus achieving forward extension and backward extension respectively.

[0062] Next, the intersection of the first reference interval and the second reference interval is calculated, and the non-empty intersection is taken as the time series reference interval corresponding to the nth target sample; when the first reference interval and the second reference interval do not intersect, the time series reference interval is recorded as empty, and the current target sample is not used to generate the sample to be processed.

[0063] Further compare the measurement interval of the target sample with the time series reference interval, and check the position of the first and second rate of change intervals relative to zero. When the measurement interval of the target sample does not intersect with the time series reference interval, and both endpoints of the first rate of change interval are greater than 0 and both endpoints of the second rate of change interval are less than 0, or both endpoints of the first rate of change interval are less than 0 and both endpoints of the second rate of change interval are greater than 0, the target sample is identified as the sample to be processed, and the target synchronous unit is identified as the synchronous unit to be processed. At the same time, the correlation between the sample to be processed and the synchronous unit to be processed is recorded.

[0064] When either the first or second rate of change interval contains 0, it indicates that the trend of change in the corresponding direction is not yet determined, and the target sample is not identified as a sample to be processed. Similarly, if the two rate of change intervals are on the same side of zero, the measurement interval of the target sample intersects with the time-series reference interval, the time-series reference interval is empty, or there are fewer than two valid samples on either side of the target sample, no corresponding sample to be processed will be generated. Let the normal sampling interval be... The maximum allowed interval between adjacent samples is Where c is a dimensionless coefficient, which can be set from 1 to 3 according to the number of consecutive missing values ​​allowed by the monitoring equipment; the time interval between any adjacent sample and the preceding and following samples is greater than If the timestamps are duplicated, samples with identical timestamps are not used to calculate the rate of change interval. If it is impossible to obtain two strictly increasing valid samples before and after the target sample, the same principle applies: insufficient sample quantity before and after the target sample.

[0065] This example generates a first reference interval and a second reference interval by taking adjacent samples on both sides of the target sample, and uses the intersection of the two reference intervals to limit the reasonable range of values ​​for the target sample at the current timestamp. At the same time, it filters samples whose time series change direction has reversed and whose measured values ​​deviate from the extended results by combining the conditions that the two rate of change intervals are located on both sides of the zero value. This provides objects to be processed for subsequent confirmation of noise samples by combining response data from other monitoring points, and reduces the bias caused by judging based on only one side of the change trend.

[0066] As an example, the step of converting the measurement intervals of the source monitoring points associated with the synchronous unit to be processed into path mapping intervals according to the conversion constraints, and aggregating them according to the temporal overlap relationship and the path verification weights to obtain response consistency intervals includes: determining the effective samples of source monitoring points according to the coverage relationship between the synchronous unit to be processed and each path temporal interval, and generating path mapping intervals according to the corresponding conversion constraints; determining the target path mapping interval and path support weight of each source monitoring point according to the degree of overlap between each path temporal interval and the sample temporal interval of the sample to be processed and the corresponding path verification weights; dividing candidate value segments according to the endpoints of each target path mapping interval, calculating the support rate of each candidate value segment and the corresponding number of different source monitoring points; filtering candidate value segments according to a preset support rate threshold and a preset number of source points, and merging adjacent candidate value segments to obtain candidate response intervals; when candidate response intervals exist, determining response consistency intervals from the candidate response intervals according to the length and support rate of the candidate value segments contained in each candidate response interval.

[0067] It should be noted that: a valid source monitoring point sample refers to a source monitoring point sample that has been identified as a valid sample, and whose corresponding path time domain interval covers the synchronous unit to be processed, the measurement interval is not empty, and the lower endpoint of the measurement interval is not greater than the upper endpoint; the target path mapping interval refers to the path mapping interval with the largest path support weight among the path mapping intervals corresponding to the same source monitoring point; the path support weight refers to the support level determined by combining the degree of overlap between the path time domain interval and the sample time domain interval of the sample to be processed, as well as the path verification weight; the candidate value segment refers to the continuous numerical range divided by the endpoints of each target path mapping interval. When there is only one endpoint after deduplication, the candidate value segment is the point interval formed by that endpoint; the support rate refers to the proportion of the sum of the path support weights covering the candidate value segment to the sum of all path support weights; the candidate response interval refers to the numerical range obtained by merging adjacent candidate value segments that meet the screening conditions; the interval support is the weighted average of the support rates of each candidate value segment within the candidate response interval, calculated according to the length of the candidate value segment.

[0068] Understandably, the process begins by reading the samples from each source monitoring point associated with the synchronous unit to be processed, along with their corresponding path time domain intervals. If the start time of the path time domain interval is no later than the start time of the synchronous unit to be processed, and the end time is no earlier than the end time of the synchronous unit, then the path time domain interval is determined to cover the synchronous unit to be processed. From the samples that meet the coverage condition, samples with non-empty measurement intervals, interval endpoints using the same physical quantity unit, and lower endpoints not greater than upper endpoints are retained to obtain valid source monitoring point samples.

[0069] The conversion constraints include conversion relationships and conversion parameter ranges. For each valid sample from a source monitoring point, the endpoints of the measurement interval and the conversion parameter range are substituted into the corresponding conversion relationship, and the minimum and maximum values ​​among the conversion results are taken as the lower and upper endpoints of the path mapping interval. When the conversion relationship is a proportional conversion relationship, the conversion result is determined by combining the endpoints of the measurement interval and the endpoints of the proportional coefficient range; when the conversion relationship is a linear conversion relationship with bias, the bias range is superimposed on the proportional conversion result. The generated path mapping interval uses the physical quantity and unit corresponding to the target monitoring point.

[0070] Then, based on the degree of overlap between the time domain intervals of each path and the time domain intervals of the sample to be processed, as well as the corresponding path verification weights, the path support weights of each path mapping interval are determined. When the same source monitoring point corresponds to multiple path mapping intervals, a target path mapping interval is determined from them, and the corresponding path support weight is used as the path support weight of the source monitoring point. The specific calculation method is explained below.

[0071] Next, the lower and upper endpoints of each target path mapping interval are collected. After removing duplicate endpoints, they are arranged in ascending order of value. The range of values ​​between two adjacent different endpoints is determined as the candidate value segment. For the j-th candidate value segment, the target path mapping intervals that completely cover this candidate value segment are counted, and the corresponding support rate is calculated. The calculation formula is as follows: In the formula, This represents the support rate for the j-th candidate value segment. This represents the j-th candidate value segment. This represents the set of indexes for the target path mapping interval corresponding to the nth sample to be processed. and These represent the mapping interval of the u-th target path in the set and its path support weight, respectively. This indicates that the target path mapping interval completely covers the candidate value segment. If only one endpoint exists after deduplication, the point interval formed by that endpoint is taken as the candidate value segment, and the support rate is calculated in the same way. When the support rate of the point interval reaches the preset support rate threshold and the number of different source monitoring points reaches the preset number of source points, the point interval is directly determined as the response consistency interval, and candidate response intervals and interval support are no longer calculated. If the screening conditions are not met, no response consistency interval is generated. The denominator of the support rate is the sum of the support weights of all paths after the same-source screening is completed; when the sum of the support weights of all paths is 0, no further candidate value segments and candidate response intervals are generated.

[0072] The number of monitoring points from different sources is counted after deduplication based on the monitoring point identifiers corresponding to the target path mapping interval. Candidate value segments with a support rate reaching a preset support rate threshold and corresponding different source monitoring point counts reaching a preset source point count are retained. The preset support rate threshold is greater than 0 and not greater than 1, and can be determined based on the lower limit of the support rate of each candidate value segment in the normal calibration response segment. The preset source point count is an integer not less than 2 and not greater than the total number of currently available source monitoring points, and can be set according to the number of traffic facility monitoring points and the redundancy of the response path. When the number of currently available source monitoring points is less than the preset source point count, no candidate response interval is generated.

[0073] Candidate value segments with adjacent numerical ranges and no unreserved value segments at their common endpoints are merged sequentially to obtain candidate response intervals. If there are candidate value segments between two retained candidate value segments that do not meet the screening criteria, no cross-segment merging is performed. The interval support is calculated based on the length and support rate of the candidate value segments contained in each candidate response interval, using the following formula: In the formula, This represents the interval support of the h-th candidate response interval. This represents the set of candidate value segments contained in the h-th candidate response interval. This represents the length of the j-th candidate value segment. The length of the candidate value segment uses the unit of the physical quantity corresponding to the target monitoring point, and the interval support is a dimensionless parameter.

[0074] When one or more candidate response intervals exist, the candidate response interval with the highest interval support is determined as the consistent response interval. When the interval support is the same, the minimum number of different source monitoring points corresponding to each candidate value segment within the candidate response interval is taken as the source point support number of that candidate response interval. Candidate response intervals with a larger source point support number are selected first. If the source point support numbers are still the same, the candidate response interval with the smaller interval width is selected. When no candidate response interval exists, a consistent response interval is not generated, and the current sample to be processed does not enter the noise confirmation process based on the consistent response interval.

[0075] This example adjusts the support level of each response path by the degree of overlap between the path's time domain interval and the sample's time domain interval. It also reduces the impact of paths with low time domain correspondence or low calibration matching on the aggregation results by combining path verification weights. At the same time, it retains only one target path mapping interval for the same source monitoring point, and then forms a response consistency interval based on the candidate value segments commonly covered by different source monitoring points. This reduces the bias caused by a single monitoring point repeatedly participating in aggregation through multiple response paths, and enables the response consistency interval to reflect the common support range of multiple source monitoring points for the current facility response.

[0076] As an example, the step of determining the target path mapping interval and path support weight of each source monitoring point based on the degree of overlap between each path time domain interval and the sample time domain interval of the sample to be processed and the corresponding path verification weight includes: obtaining each path mapping interval corresponding to the same source monitoring point, the path time domain interval and path verification weight corresponding to each path mapping interval, and the sample time domain interval of the sample to be processed; calculating the path support weight based on each path time domain interval, the sample time domain interval of the sample to be processed, and the corresponding path verification weight; comparing the path support weights of each path mapping interval corresponding to the same source monitoring point, determining the path mapping interval with the largest path support weight as the target path mapping interval, and determining the corresponding path support weight as the path support weight of the source monitoring point.

[0077] It should be noted that the time domain overlap ratio refers to the proportion of the intersection time of the path time domain interval and the sample time domain interval of the sample to be processed to the union time, and the value ranges from 0 to 1.

[0078] Understandably, firstly, the path mapping intervals associated with the synchronous units to be processed are grouped according to the source monitoring point identifiers, and the response path identifier, path time domain interval, and path verification weight corresponding to each path mapping interval are read. The path time domain interval and the sample time domain interval of the sample to be processed use the same time unit; if different time units are used, they are first uniformly converted to seconds, milliseconds, or other pre-selected time units.

[0079] The path verification weight is the candidate matching rate obtained when generating response constraint records, with a value range of 0 to 1, and is not re-normalized within the processing time unit; different calibration batches use the same candidate matching rate calculation method and preset calibration threshold, so that the path verification weight of each response path has the same meaning, and the response path with a path verification weight of 0 does not participate in the determination of the target path mapping interval.

[0080] For the k-th valid sample of the i-th source monitoring sequence in the p-th response path, calculate the intersection and union durations of the path's time domain interval and the sample time domain interval of the n-th sample to be processed. Then, multiply the time domain overlap ratio by the corresponding path verification weight to obtain the path support weight. The calculation formula is as follows: In the formula, This represents the path support weight corresponding to the k-th valid sample of the i-th source monitoring sequence in the p-th response path. This represents the path verification weight of the p-th response path. Indicates the corresponding path time domain interval. This represents the time-domain interval of the nth sample to be processed, with the vertical line outside the interval indicating the interval duration. Both the path verification weights and the time-domain overlap ratio are dimensionless parameters; therefore, the path support weights are also dimensionless parameters.

[0081] Both the path time domain interval and the sample time domain interval use valid time intervals with a start time no later than the end time, and the union duration should be greater than 0. When two intervals do not have a common time range, the intersection duration is 0, and the path support weight is 0; when two intervals only touch at their endpoints, the intersection is a point interval, and its duration is also recorded as 0; when two intervals completely overlap, the time domain overlap ratio is 1, and the path support weight is equal to the corresponding path verification weight.

[0082] After calculating the path support weights, among all path mapping intervals corresponding to the same source monitoring point, the path mapping interval with the largest path support weight is selected as the target path mapping interval, and the corresponding path support weight is used as the path support weight for that source monitoring point. When the same source monitoring point corresponds to only one path mapping interval, that path mapping interval is directly determined as the target path mapping interval.

[0083] When multiple path mapping intervals have the same maximum path support weight, the path mapping interval with the larger path verification weight is selected first; if the path verification weights are still the same, the path mapping interval with the smaller interval width is selected. After each source monitoring point completes the same-source comparison, the monitoring point identifier, target path mapping interval, and path support weight are output respectively for subsequent candidate value segment division and support rate calculation.

[0084] This example reflects the temporal correspondence between valid samples and samples to be processed at source monitoring points by the temporal overlap ratio, and uses path verification weights to reflect the matching degree of response paths in the calibration process, so that path support weights are constrained by both the current temporal relationship and the reliability of historical paths. On this basis, each source monitoring point retains only the path mapping interval with the largest path support weight, reducing the bias caused by the same source monitoring point repeatedly participating in subsequent aggregation through multiple response paths.

[0085] As an example, the step of determining noise samples and replacement values ​​based on the intersection relationship between the measurement interval of the sample to be processed, the time-series reference interval, and the response consistency interval, and generating a noise reduction monitoring sequence includes: when there is a response consistency interval, the measurement interval of the sample to be processed does not intersect with the response consistency interval, and the time-series reference interval intersects with the response consistency interval, the intersection of the time-series reference interval and the response consistency interval is determined as a feasible replacement interval; the intersection of each target path mapping interval and the feasible replacement interval is calculated respectively, and the non-empty intersection is determined as a replacement contribution interval; when there is a replacement contribution interval and the sum of the corresponding path support weights is not zero, the sample to be processed is determined as a noise sample, and the midpoint of each replacement contribution interval is weighted according to the corresponding path support weight to obtain a replacement value; the sample to be processed that is not determined as a noise sample is determined as a retained sample, the noise sample is replaced with the corresponding replacement value, and the replaced noise sample, the retained sample, and the original sample that is not determined as a sample to be processed are merged according to the original timestamp order to generate a noise reduction monitoring sequence.

[0086] It should be noted that the feasible replacement interval refers to the numerical range jointly covered by the time series reference interval and the response consistency interval; the replacement contribution interval refers to the non-empty intersection of the target path mapping interval and the feasible replacement interval, which is used to provide candidate values ​​for replacement; and the retained sample refers to the unprocessed sample that does not meet the conditions for determining noise samples and continues to retain the original monitoring value.

[0087] Please refer to Figure 4 , Figure 4 This diagram illustrates the generation of response consistency intervals and replacement values ​​provided in the first embodiment of the noise reduction method for traffic facility monitoring data in this application. The diagram shows the process of forming a response judgment criterion based on the measurement intervals of valid samples from source monitoring points, and using this criterion to process the samples to be processed. First, the measurement intervals of valid samples from source monitoring points are converted into path mapping intervals according to conversion constraints. Based on the temporal overlap between the path time domain interval and the sample time domain interval of the samples to be processed, as well as the path verification weight, the target path mapping interval and path support weight for each source monitoring point are determined. Candidate value segments are divided according to the endpoints of each target path mapping interval. The support rate of each candidate value segment and the corresponding number of different source monitoring points are calculated. Candidate value segments are filtered according to a preset support rate threshold and a preset number of source points, and adjacent retained candidate value segments are merged to obtain candidate response intervals. The interval support of each candidate response interval is calculated based on the length and support rate of the candidate value segment. The candidate response interval with the highest interval support is determined as the response consistency interval.

[0088] The intersection of the measurement interval, time-series reference interval, and response consistency interval of the sample to be processed is determined. If the measurement interval of the sample to be processed does not intersect with the response consistency interval, but the time-series reference interval intersects with the response consistency interval, then the feasible replacement interval is obtained based on the intersection of the time-series reference interval and the response consistency interval. The non-empty intersection of each target path mapping interval with the feasible replacement interval is taken as the replacement contribution interval. The midpoint of each replacement contribution interval is weighted according to the corresponding path support weight to obtain the replacement value, and the replacement value is used to generate a noise reduction monitoring sequence. If the intersection judgment condition is not met, the sample to be processed is retained, and the retained sample is incorporated into the noise reduction monitoring sequence.

[0089] Understandably, the process begins by sequentially checking whether each sample to be processed has a corresponding consistent response interval, and then comparing the measurement interval, time series reference interval, and consistent response interval of the sample to be processed. When the measurement interval and the consistent response interval do not intersect, but the time series reference interval and the consistent response interval do intersect, the intersection of the time series reference interval and the consistent response interval is calculated, and this intersection is determined as the feasible replacement interval.

[0090] Each interval is treated as a closed interval including its endpoints. When the time-series reference interval and the response-consistent interval intersect at their endpoints, the feasible interval is replaced with the point interval formed by those endpoints. When the measurement interval and the response-consistent interval intersect at their endpoints, the two intervals are determined to intersect, and the corresponding sample to be processed is not identified as a noise sample. If there is no response-consistent interval, the measurement interval intersects with the response-consistent interval, or the time-series reference interval and the response-consistent interval do not intersect, the corresponding sample to be processed is identified as a retained sample.

[0091] Then, the intersection of each target path mapping interval and the feasible replacement interval is calculated one by one, and the non-empty intersection is determined as the replacement contribution interval. The target path mapping interval and the feasible replacement interval use the same physical quantity and unit corresponding to the target monitoring point; if the units are inconsistent, the interval intersection is not judged, and the corresponding target path mapping interval is excluded. The target path mapping interval and the feasible replacement interval only intersect at their endpoints, and the interval formed by those endpoints is taken as the replacement contribution interval.

[0092] Each replacement contribution interval inherits the source monitoring point identifier and path support weight of the corresponding target path mapping interval, thereby establishing a one-to-one correspondence between replacement contribution intervals and path support weights. Replacement contribution intervals with a path support weight of 0 do not participate in the replacement value calculation; when there is no replacement contribution interval, or the sum of the path support weights corresponding to all replacement contribution intervals is 0, the current sample to be processed is determined as a retained sample.

[0093] When there is a replacement contribution interval and the sum of the corresponding path support weights is greater than 0, the current sample to be processed is determined as a noise sample, and the average of the lower and upper endpoints of each replacement contribution interval is taken as the midpoint of the corresponding interval; when the replacement contribution interval is a point interval, the midpoint of the interval is the endpoint.

[0094] The replacement value of the nth noise sample is obtained by weighting the midpoint of each replacement contribution interval according to the path support weights corresponding to those intervals, using the following formula: In the formula, This represents the replacement value when the nth sample to be processed is determined to be a noise sample. This indicates the number of replacement contribution intervals corresponding to this sample. This represents the midpoint of the v-th replacement contribution interval. This represents the path support weight corresponding to the v-th replacement contribution interval. The path support weight is a dimensionless parameter, and the midpoint and replacement value of the replacement contribution interval use the same physical quantity and unit as the monitoring value of the target monitoring point. Since each replacement contribution interval is located within the replacement feasible interval, and each path support weight is non-negative, the calculated replacement value is located within the replacement feasible interval, and no further interval endpoint correction is required.

[0095] The monitoring values ​​of noise samples are replaced with corresponding replacement values, while retaining the original timestamps, monitoring point identifiers, and monitoring types. Retained samples continue to use their original monitoring values, and original samples not identified as pending processing are left unmodified. Finally, the replaced noise samples, retained samples, and other original samples are compiled according to their monitoring point identifiers and monitoring types, and arranged from earliest to latest according to their original timestamps to generate corresponding noise reduction monitoring sequences. When identical timestamps exist, the order within the original monitoring sequence is maintained.

[0096] This example defines the feasible range of replacement values ​​by using both the time-series reference interval and the response consistency interval. Only target path mapping intervals falling within this range are used in the replacement value calculation, ensuring that the replacement values ​​simultaneously conform to the changing trends of the target monitoring points and the response relationships of multiple source monitoring points. At the same time, the midpoint of the replacement contribution interval is weighted according to the path support weight, reducing the impact of monitoring data with low time-domain correspondence or path verification on the replacement results. Furthermore, the original time sequence and monitoring point information are preserved while replacing the monitoring values ​​of noisy samples.

[0097] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to the first embodiment described above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 5 , Figure 5This is a flowchart illustrating the second embodiment of the noise reduction method for traffic facility monitoring data according to this application. Step S20 of the noise reduction method for traffic facility monitoring data includes steps S21 to S25: Step S21: Sort the samples in each monitoring sequence according to the timestamp to obtain an ordered monitoring sequence; Step S22: Generate sample time domain intervals based on the midpoint of the timestamps and the timestamp error of adjacent samples in each ordered monitoring sequence, and retain sample time domain intervals and corresponding samples whose left endpoint is not greater than the right endpoint to obtain valid samples; Step S23: Generate a measurement interval with the monitoring value of each valid sample as the interval center and the corresponding measurement error as the interval radius; Step S24: Obtain the path time domain interval according to the sample time domain interval corresponding to the effective sample of the time-delay constrained mobile source monitoring point; Step S25: Generate a synchronous unit based on the boundary of the time domain interval of the path and the sample time domain interval corresponding to the effective samples of the target monitoring point.

[0098] It should be noted that an ordered monitoring sequence refers to a monitoring sequence formed by arranging samples from earliest to latest according to their timestamps; a valid sample refers to a sample whose left endpoint of the corresponding time domain interval is not greater than its right endpoint and can participate in subsequent interval processing; a point interval refers to an interval whose left endpoint and right endpoint are equal and correspond to only one time point or one measurement value.

[0099] Understandably, firstly, each monitoring sequence grouped according to monitoring point identification and monitoring type is read separately. After completing the duplicate data processing and missing data marking, the samples in each group are arranged from earliest to latest according to the timestamp to obtain an ordered monitoring sequence with strictly increasing timestamps. Samples recorded as timestamp conflict data do not participate in the calculation of the sample time domain interval and rate of change interval, but continue to be stored in the original monitoring data.

[0100] For samples that are not at the beginning or end of each ordered monitoring sequence, a sample time domain interval is generated based on the timestamps of the current sample and its adjacent samples, as well as the upper limit of timestamp error. Sample time domain intervals with a left endpoint no greater than the right endpoint, along with their corresponding samples, are retained. The specific calculation method is explained below. If the number of samples in an ordered monitoring sequence is less than 3, no valid sample is generated, and the original samples are still saved in timestamp order.

[0101] The time domain intervals of samples whose left endpoint is not greater than the right endpoint, and the corresponding samples, are retained. The specific endpoint judgment and point interval processing methods are explained below.

[0102] For each valid sample, a measurement interval is generated with the monitored value as the center and the corresponding absolute measurement error as the interval radius. The calculation formula is as follows: In the formula, This represents the measurement interval of the k-th valid sample in the i-th ordered monitoring sequence. Indicates the corresponding monitoring value. This indicates the corresponding absolute measurement error. The monitored value, absolute measurement error, and measurement interval all use the same physical quantity and unit.

[0103] When the monitoring data packet records the relative measurement error, the relative measurement error is converted into a decimal form, and then combined with the absolute value of the monitored value to calculate the absolute measurement error. The calculation formula is as follows: In the formula, This represents the relative measurement error of the k-th valid sample in the i-th ordered monitoring sequence. When the relative measurement error is recorded as a percentage, it is first divided by 100 to convert it to a decimal before being used in the calculation of the absolute measurement error. When the monitoring equipment provides both relative measurement error and fixed absolute error, the absolute measurement error is determined according to the error expression method in the accuracy specification of the monitoring equipment. If the accuracy specification stipulates that the two errors are added together, the converted result of the relative measurement error is added to the fixed absolute error. If the accuracy specification treats the two errors as independent upper limits, the larger value of the two errors is taken as the radius of the measurement interval.

[0104] Next, the sample time domain interval corresponding to the valid samples of the source monitoring point is read, as well as the time delay constraint corresponding to the current response path. The sample time domain interval is shifted using interval addition to obtain the path time domain interval. The calculation formula is as follows: In the formula, This represents the time-domain interval corresponding to the k-th valid sample of the i-th ordered monitoring sequence in the p-th response path. Indicates the time domain interval of the corresponding sample. ⊕ represents the time delay constraint for the p-th response path, and ⊕ represents interval addition. The sample time domain interval and time delay constraint use the same time unit. The lower endpoint of the time delay constraint is not greater than the upper endpoint, and the direction of interval movement is consistent with the time direction of the response from the source monitoring point to the target monitoring point.

[0105] Then, the sample time domain interval boundaries of valid samples from the target monitoring points and the path time domain interval boundaries of valid samples from each source monitoring point are collected. Duplicate boundaries are deleted and arranged in chronological order, and the time periods between adjacent different boundaries are used as candidate time periods. For each candidate time period, the target monitoring point sample time domain interval and the source monitoring point path time domain interval that completely cover the time period are determined. When there is valid sample coverage of the target monitoring points and the number of different source monitoring points that are covered together reaches the preset number of source points, the corresponding time period and the identifier of each valid sample are recorded.

[0106] If the sets of valid sample identifiers for target monitoring points and valid sample identifiers for source monitoring points are identical for adjacent candidate time periods, and each source monitoring point has at least one path time interval covering the complete time period after merging, then adjacent candidate time periods are merged into the same synchronous unit, and an index relationship is established between the synchronous unit and the corresponding valid sample. If the coverage condition is not met, no corresponding synchronous unit is generated, and the relevant valid samples do not participate in the current cross-monitoring point response comparison, but remain in the original monitoring sequence.

[0107] This embodiment generates the sample time domain interval by using the midpoint of the timestamp and the timestamp error, eliminating invalid intervals caused by time errors, and retaining the uncertain range of the monitoring values ​​by using the measurement interval. On this basis, the sample time domain interval of the source monitoring point is moved according to the time delay constraint of the response path, and then a synchronous unit is established according to the common coverage relationship of the intervals of the source monitoring point and the target monitoring point, so that monitoring sequences with different sampling times and sampling frequencies can establish a sample correspondence without changing the original timestamp or generating interpolated samples.

[0108] As an example, the step of generating sample time domain intervals based on the midpoint of the timestamps and timestamp errors of adjacent samples in each ordered monitoring sequence, and retaining sample time domain intervals and corresponding samples whose left endpoint is not greater than the right endpoint, to obtain valid samples includes: calculating the sample time domain interval of the k-th sample based on the timestamps and upper limits of timestamp errors of the k-th sample and adjacent samples in the i-th ordered monitoring sequence; comparing the left and right endpoints of each sample time domain interval, retaining sample time domain intervals and corresponding samples whose left endpoint is not greater than the right endpoint, and determining the retained corresponding samples as valid samples, thereby obtaining the sample time domain interval and valid samples of the i-th ordered monitoring sequence.

[0109] It should be noted that the front boundary refers to the time boundary between the previous sample and the current sample after timestamp error correction. It is located to the left of the current sample's time domain interval and is used to limit the time range covered by the current sample. The back boundary refers to the time boundary between the current sample and the next sample after timestamp error correction. It is located to the right of the current sample's time domain interval and is used to limit the time range covered by the current sample.

[0110] Understandably, from the i-th ordered monitoring sequence, intermediate samples with the preceding and following samples are selected sequentially. The timestamps and upper limits of timestamp errors for the three samples are read, and all time parameters are converted to the same time unit before calculating the sample time domain interval for the current sample. The upper limit of timestamp error is a non-negative value. If the timestamp or the upper limit of timestamp error is missing, the sample time domain interval for the corresponding sample is not calculated, and the corresponding sample is not determined as a valid sample.

[0111] For the k-th sample in the i-th ordered monitoring sequence, the sample time domain interval of the k-th sample is calculated based on the timestamps of the previous sample, the current sample, and the next sample, as well as the upper limit of the timestamp error corresponding to the three samples. The calculation formula is as follows: In the formula, This represents the time-domain interval of the k-th sample in the i-th ordered monitoring sequence. , and These represent the timestamps of the previous sample, the current sample, and the next sample, respectively. , and These represent the corresponding timestamp error limits.

[0112] Without considering timestamp errors, the front boundary of the current sample is located at the midpoint between the timestamps of the previous sample and the current sample, and the back boundary is located at the midpoint between the timestamps of the current sample and the next sample. After considering timestamp errors, the front boundary is moved towards the current sample by half the sum of the upper limits of the timestamp errors of the previous sample and the current sample, and the back boundary is moved towards the current sample by half the sum of the upper limits of the timestamp errors of the current sample and the next sample, thereby reducing the time range to which the current sample can be assigned.

[0113] After calculation, the left and right endpoints of the sample time domain interval are read respectively. If the left endpoint is less than the right endpoint, the sample time domain interval and the corresponding sample are retained. If the left endpoint is equal to the right endpoint, the sample time domain interval is retained as a point interval, and the corresponding sample is determined as a valid sample. If the left endpoint is greater than the right endpoint, it indicates that the interval between adjacent samples is insufficient to accommodate the corresponding timestamp error. Therefore, the corresponding sample time domain interval is not retained, the corresponding sample is not determined as a valid sample, and the corresponding sample is not used to generate measurement intervals, path time domain intervals, and synchronous units.

[0114] Since the first and last samples of each ordered monitoring sequence only have one-sided adjacent samples, it is impossible to determine the front and back boundaries simultaneously using the same formula. Therefore, this example does not utilize the first and last samples to generate the sample time domain interval, nor does it determine the first and last samples as valid samples. The first and last samples, along with other samples that are not determined as valid samples, are retained in the original monitoring sequence and do not participate in the current cross-monitoring point response comparison. When generating subsequent noise-reduced monitoring sequences, they are still retained according to the original timestamp order.

[0115] This example determines the initial time boundary between samples by using the midpoint of the timestamps of adjacent samples, and then uses the upper limit of the timestamp error of adjacent samples to narrow the boundary range inward, so that the time domain interval of the sample can reflect the time range that can still be attributed to the current sample when there is an error in the sampling time. At the same time, by eliminating the intervals where the left endpoint is greater than the right endpoint, it avoids using samples that cannot form an effective time range for subsequent establishment of synchronous relationships when the timestamp error is too large relative to the sampling interval.

[0116] As an example, the step of generating a synchronous unit based on the boundaries of the path time domain interval and the sample time domain interval corresponding to the valid samples of the target monitoring point includes: collecting the sample time domain interval boundaries corresponding to the valid samples of the target monitoring point and the path time domain interval boundaries corresponding to the valid samples of different source monitoring points, and arranging them in chronological order; determining the time periods between adjacent boundaries as candidate time periods, and recording the valid sample identifiers of the target monitoring point and the set of valid sample identifiers of the source monitoring point covering each candidate time period; retaining the candidate time periods covered by the sample time domain intervals corresponding to the valid samples of the target monitoring point and jointly covered by the path time domain intervals corresponding to different source monitoring points reaching a preset number of source points; merging adjacent candidate time periods with the same set of valid sample identifiers of the target monitoring point and the same set of valid sample identifiers of the source monitoring point, and establishing an index relationship between the merged time period and the corresponding valid sample when each source monitoring point's valid sample has at least one path time domain interval covering the merged complete time period, thereby obtaining the synchronous unit.

[0117] It should be noted that the candidate time period refers to the continuous time range divided according to the boundaries of the time domain intervals of each sample and the time domain interval of the path; covering the candidate time period means that the start time of the interval is no later than the start time of the candidate time period, and the end time of the interval is no earlier than the end time of the candidate time period; the preset number of source points refers to the minimum number of different source monitoring points required to form a synchronous unit; the index relationship refers to the association record between the synchronous unit and the effective samples of the target monitoring point, the effective samples of the source monitoring point, and the corresponding original data.

[0118] Please refer to Figure 6 , Figure 6This diagram illustrates the generation of synchronous units in the second embodiment of the noise reduction method for traffic facility monitoring data in this application. The diagram shows the processing relationship for generating synchronous units based on the sample time-domain interval boundaries of the target monitoring point and the path time-domain interval boundaries of different source monitoring points. First, the two types of interval boundaries are collected and arranged chronologically. The time periods between adjacent boundaries are divided into candidate time periods. The effective sample identifiers of the target monitoring point and the set of effective sample identifiers of the source monitoring point covering each candidate time period are recorded. The sample time-domain interval coverage corresponding to the effective samples of the target monitoring point, the number of different source monitoring points counted after deduplication according to the monitoring point identifier, and the preset number of source points are compared. If the target monitoring point… If the time domain interval corresponding to a valid sample does not cover a candidate time period, or the number of different source monitoring points does not reach the preset number of source points, the corresponding candidate time period is not retained. If the time domain interval corresponding to a valid sample of a target monitoring point covers a candidate time period, and the number of different source monitoring points reaches the preset number of source points, the corresponding candidate time period is retained. For adjacent candidate time periods with the same set of valid sample identifiers of target monitoring points and valid sample identifiers of source monitoring points, it is checked whether each source monitoring point's valid sample has at least one path time domain interval covering the merged complete time period. If the coverage condition is met, adjacent candidate time periods are merged, and an index relationship between the merged time period and the corresponding valid sample is established to obtain a synchronous unit.

[0119] Understandably, firstly, for a target monitoring point, the sample time domain interval and path time domain interval corresponding to each valid sample are read, and the interval boundaries are uniformly converted to the same time unit. Then, the left and right endpoints of the target monitoring point's sample time domain interval, as well as the left and right endpoints of each source monitoring point's path time domain interval, are collected. The time unit can be uniformly set to seconds or milliseconds, and the time precision is retained according to the timestamp precision in the monitoring data packet. Different time units are not mixed in the same synchronous unit generation process.

[0120] After data aggregation, duplicate boundaries with identical values ​​are removed, retaining only one corresponding time value. The remaining boundaries are then arranged from earliest to latest time. The time interval between two adjacent boundaries with different values ​​is identified as a candidate time interval, and candidate time intervals uniformly adopt a left-closed, right-open boundary form. No candidate time intervals are generated when adjacent boundaries are equal. Time intervals with zero duration are not included in the generation of synchronous units, and the corresponding point intervals and original samples are continued to be stored in the valid sample record.

[0121] For each candidate time period, the sample time domain interval of the target monitoring point and the path time domain interval of each source monitoring point are checked separately. The valid sample identifiers of the target monitoring point that fully covers the candidate time period, as well as the valid sample identifiers of the source monitoring points that fully cover the candidate time period, are recorded. If no sample time domain interval of the target monitoring point covers the candidate time period, the current candidate time period is not retained; if there are valid sample intervals of the target monitoring point that cover the candidate time period, the number of different source monitoring points that cover the current candidate time period is continued to be counted.

[0122] The number of different source monitoring points is counted after deduplication based on the monitoring point identifier. If there are multiple response paths for the same source monitoring point, or if the same valid sample corresponds to multiple path time domain intervals, the source monitoring point is counted only once. The set of valid sample identifiers for source monitoring points records the valid sample identifiers covering the candidate time period. When multiple path time domain intervals correspond to the same valid sample, the valid sample identifier is not written repeatedly, while the corresponding response path identifier is retained for subsequent path mapping interval calculation.

[0123] The preset number of source points is an integer not less than 2 and not greater than the total number of source monitoring points that can be associated with the current target monitoring point. This number can be determined based on the number of monitoring points deployed, response path coverage, and minimum data redundancy requirements. For example, if the target monitoring point can be associated with 3 source monitoring points, the preset number of source points can be set to 2, indicating that at least 2 different source monitoring points need to establish a time correspondence with the target monitoring point. If the total number of currently associatable source monitoring points is less than the preset number of source points, no corresponding synchronous unit is generated, and cross-monitoring point response comparisons are not performed using the current time range.

[0124] The candidate time period is retained by the sample time domain interval covered by the valid samples of the target monitoring point and by the path time domain interval of different source monitoring points that reach the preset number of source points. The valid sample identifier of the target monitoring point and the set of valid sample identifiers of the source monitoring point are recorded.

[0125] Then, adjacent candidate time periods are checked in chronological order. If the right endpoint of the previous candidate time period equals the left endpoint of the next candidate time period, and the sets of valid sample identifiers of the target monitoring points and the source monitoring points associated with the two candidate time periods are the same, the two candidate time periods are merged. Before merging, it is further checked whether each source monitoring point has at least one path time interval that covers the complete time period after merging. If the coverage condition is met, merging is performed; otherwise, they are retained as two independent candidate time periods to avoid the merged time period lacking a response path that can be completely covered.

[0126] For the merged time period, a time unit identifier is generated, and an index relationship is established between the time unit and each valid sample. The index relationship records at least the start and end times of the time unit, the monitoring point identifier, the valid sample identifier, the original timestamp, and the original monitoring value. It may also record the monitoring type, response path identifier, and corresponding interval identifier. When establishing the index relationship, the original data content corresponding to the valid samples is not modified. The original timestamp, original monitoring value, and sampling order remain unchanged. The time unit is only used to record the time correspondence between different monitoring sequences.

[0127] This example divides the time range corresponding to different sampling times and frequencies into candidate time periods by uniformly arranging the boundaries of the sample time domain intervals and path time domain intervals. Then, it filters the candidate time periods that meet the conditions based on the coverage of the target monitoring point and the number of different source monitoring points. At the same time, it deduplicates the coverage relationship formed by the same source monitoring point through multiple response paths and retains the correspondence between the synchronous unit and the original sample through the index relationship. This reduces the impact of repeated paths on the statistics of the number of source monitoring points and avoids the synchronous processing from changing the original monitoring data.

[0128] As a calculation example, the timestamps of five consecutive valid samples from the target monitoring point are 8s, 9s, 10s, 11s, and 12s, corresponding to measurement intervals of [8.9, 9.1], [9.9, 10.1], [14.8, 15.2], [10.0, 10.2], and [9.0, 9.2], respectively. The sample with timestamp 10s is the target sample. Based on the two preceding samples, the first rate of change interval [0.8, 1.2] is obtained. Extending the sample with timestamp 9s to 10s yields the first reference interval [10.7, 11.3]. Based on the two following samples, the second rate of change interval [-1.2, -0.8] is obtained. Extending the sample with timestamp 11s in reverse direction to 10s yields the second reference interval [10.8, 11.4]. Thus, the time series reference interval [10.8, 11.3] is obtained. The measurement interval of the target sample [14.8, 15.2] does not intersect with the time series reference interval, and the two rate of change intervals are located on both sides of the zero value. Therefore, the target sample is determined as the sample to be processed. The target path mapping intervals obtained by the conversion of the three source monitoring points are [10.7, 11.2], [10.9, 11.4] and [10.8, 11.3], respectively, with corresponding path support weights of 0.8, 0.7 and 0.6, respectively. The preset support rate threshold is 0.6, and the preset number of source points is 2. After dividing the candidate value segments according to the endpoints of each interval, [10.8, 10.9], [10.9, 11.2] and [11.2, 11.3] all meet the screening criteria and are merged to obtain the response consistency interval [10.8, 11.3]. The intersection of the time-series reference interval and the response consistency interval is [10.8, 11.3]. The three replacement contribution intervals are [10.8, 11.2], [10.9, 11.3], and [10.8, 11.3], with corresponding midpoints of 11.0, 11.1, and 11.05, respectively. The replacement value, weighted according to path support, is approximately 11.05. This 11.05 is used to replace the original monitoring value of the target sample, while retaining the timestamp, monitoring point identifier, and monitoring type. A denoised monitoring sequence is then generated according to the original time sequence.

[0129] This application also provides a noise reduction device for traffic facility monitoring data; please refer to... Figure 7The noise reduction device for traffic facility monitoring data includes: Data acquisition module 10 is used to acquire multiple monitoring sequences and response constraint records of traffic facilities. The response constraint records include response paths, time delay constraints, conversion constraints, and path verification weights. The interval generation module 20 is used to generate sample time domain intervals, measurement intervals, path time domain intervals, and synchronous units based on the timestamps, timestamp errors, measurement errors, and time delay constraints of the monitoring sequence. The sample screening module 30 is used to determine the sample to be processed, the time unit to be processed, and the time series reference interval based on the target monitoring point sample and the samples before and after the synchronous unit. The interval aggregation module 40 is used to convert the measurement interval of the source monitoring point associated with the synchronous unit to be processed into a path mapping interval according to the conversion constraint, and to aggregate it according to the time domain coincidence relationship and the path verification weight to obtain the response consistency interval; The data replacement module 50 is used to determine the noise sample and replacement value based on the intersection relationship between the measurement interval of the sample to be processed, the time series reference interval, and the response consistency interval, and to generate a noise reduction monitoring sequence.

[0130] The noise reduction device for traffic facility monitoring data provided in this application, employing the noise reduction method for traffic facility monitoring data in the above embodiments, can solve the technical problem of how to identify noise samples based on the monitoring data corresponding to the response of the same facility when the sampling times and sampling frequencies of different monitoring sequences of traffic facilities are inconsistent. Compared with the prior art, the beneficial effects of the noise reduction device for traffic facility monitoring data provided in this application are the same as those of the noise reduction method for traffic facility monitoring data provided in the above embodiments, and other technical features in the noise reduction device for traffic facility monitoring data are the same as those disclosed in the above method embodiments, and will not be repeated here.

[0131] This application provides a noise reduction device for traffic facility monitoring data. The noise reduction device for traffic facility monitoring data includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the noise reduction method for traffic facility monitoring data in the first embodiment described above.

[0132] The following is for reference. Figure 8The diagram illustrates a structural schematic of a noise reduction device suitable for implementing the embodiments of this application for traffic facility monitoring data. The noise reduction device for traffic facility monitoring data in the embodiments of this application can be a server, edge computing gateway, industrial computer, or other electronic device with data storage and processing capabilities deployed on a traffic facility monitoring platform. The specific hardware configuration can be determined based on the number of monitoring points, data sampling frequency, and processing cycle.

[0133] like Figure 8 As shown, a noise reduction device for traffic facility monitoring data may include a processing unit 1001, a ROM 1002, a storage unit 1003, a RAM 1004, a bus 1005, an I / O interface 1006, an input device 1007, an output device 1008, and a communication device 1009. The processing unit 1001 reads the computer program and monitoring data from the ROM 1002, RAM 1004, or storage unit 1003 via the bus 1005, and executes a noise reduction method for the traffic facility monitoring data. The communication device 1009 receives monitoring data packets and outputs noise-reduced monitoring sequences. The input device 1007 and output device 1008 receive device configuration and output processing status. The device can be configured with servers, edge computing gateways, or industrial computers depending on the deployment environment, and the corresponding processing and storage resources can be configured according to the number of monitoring points, data sampling frequency, and processing cycle.

[0134] The noise reduction device for traffic facility monitoring data provided in this application, employing the noise reduction method for traffic facility monitoring data in the above embodiments, can solve the technical problem of how to identify noise samples based on the monitoring data corresponding to the response of the same facility when the sampling times and sampling frequencies of different monitoring sequences of traffic facilities are inconsistent. Compared with the prior art, the beneficial effects of the noise reduction device for traffic facility monitoring data provided in this application are the same as those of the noise reduction method for traffic facility monitoring data provided in the above embodiments, and other technical features in the noise reduction device for traffic facility monitoring data are the same as those disclosed in the above method embodiments, and will not be repeated here.

[0135] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the noise reduction method for traffic facility monitoring data in the above embodiments.

[0136] The computer-readable storage medium provided in this application can be a tangible medium capable of storing computer programs, such as a hard disk, random access memory, read-only memory, flash memory, optical storage device, or magnetic storage device. When the computer program is executed by a processor, it implements the above-described noise reduction method for traffic facility monitoring data.

[0137] The aforementioned computer-readable storage medium may be included in a noise reduction device for traffic facility monitoring data; or it may exist independently and not be assembled into a noise reduction device for traffic facility monitoring data.

[0138] The aforementioned computer-readable storage medium carries one or more programs. When these programs are executed by a noise reduction device for traffic facility monitoring data, the noise reduction device for traffic facility monitoring data causes the following: it acquires multiple monitoring sequences and response constraint records for traffic facilities, the response constraint records including response paths, time delay constraints, conversion constraints, and path verification weights; based on the timestamps, timestamp errors, measurement errors, and time delay constraints of the monitoring sequences, it generates sample time domain intervals, measurement intervals, path time domain intervals, and synchronous units; based on the target monitoring point samples and preceding and following samples associated with the synchronous units, it determines the samples to be processed, the synchronous units to be processed, and the time series reference interval; it converts the source monitoring point measurement intervals associated with the synchronous units to be processed into path mapping intervals according to the conversion constraints, and aggregates them according to the time domain overlap relationship and the path verification weights to obtain response consistency intervals; based on the intersection relationship between the measurement intervals of the samples to be processed and the time series reference intervals and the response consistency intervals, it determines noise samples and replacement values, and generates a noise reduction monitoring sequence.

[0139] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described noise reduction method for traffic facility monitoring data. This solves the technical problem of identifying noise samples based on monitoring data corresponding to the response of the same facility when the sampling times and frequencies of different monitoring sequences of traffic facilities are inconsistent. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the noise reduction method for traffic facility monitoring data provided in the above embodiments, and will not be repeated here.

[0140] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A noise reduction method for traffic facility monitoring data, characterized in that, The method includes: Acquire multiple monitoring sequences and response constraint records for traffic facilities, wherein the response constraint records include response paths, time delay constraints, conversion constraints, and path verification weights; Based on the timestamp, timestamp error, measurement error, and time delay constraint of the monitoring sequence, a sample time domain interval, a measurement interval, a path time domain interval, and a synchronous unit are generated. Based on the target monitoring point samples and previous and subsequent samples associated with the synchronous unit, determine the sample to be processed, the synchronous unit to be processed, and the time series reference interval. According to the conversion constraints, the measurement interval of the source monitoring point associated with the synchronous unit to be processed is converted into a path mapping interval, and then aggregated according to the temporal overlap relationship and the path verification weight to obtain the response consistency interval; Based on the intersection relationship between the measurement interval of the sample to be processed, the time series reference interval, and the response consistency interval, noise samples and replacement values ​​are determined, and a noise reduction monitoring sequence is generated.

2. The method as described in claim 1, characterized in that, The step of generating the sample time domain interval, measurement interval, path time domain interval, and synchronous unit based on the timestamp, timestamp error, measurement error, and time delay constraint of the monitoring sequence includes: The samples in each monitoring sequence are sorted according to their timestamps to obtain an ordered monitoring sequence; The sample time domain interval is generated based on the midpoint of the timestamps and the timestamp error of adjacent samples in each ordered monitoring sequence, and the sample time domain interval and corresponding sample with the left endpoint not greater than the right endpoint are retained to obtain the effective sample; A measurement interval is generated with the monitoring value of each valid sample as the interval center and the corresponding measurement error as the interval radius; The path time domain interval is obtained according to the sample time domain interval corresponding to the effective sample of the time-delay constrained mobile source monitoring point; Based on the boundaries of the time domain intervals of the path and the effective samples of the target monitoring point, a synchronous unit is generated.

3. The method as described in claim 2, characterized in that, The step of generating a synchronous unit based on the boundary of the time domain interval of the path and the sample time domain interval corresponding to the effective samples of the target monitoring point includes: The sample time domain interval boundaries corresponding to the valid samples of the target monitoring point and the path time domain interval boundaries corresponding to the valid samples of different source monitoring points are collected and arranged in chronological order. The time period between adjacent boundaries is determined as the candidate time period, and the effective sample identifiers of the target monitoring points and the set of effective sample identifiers of the source monitoring points covering each candidate time period are recorded. Retain candidate time periods that are covered by the sample time domain intervals corresponding to the valid samples of the target monitoring point and are jointly covered by the path time domain intervals corresponding to different source monitoring points that reach the preset number of source points; Adjacent candidate time periods with the same set of valid sample identifiers for target monitoring points and valid sample identifiers for source monitoring points are merged. When each source monitoring point has at least one path time domain interval covering the complete time period after merging, an index relationship between the merged time period and the corresponding valid sample is established to obtain the synchronous unit.

4. The method as described in claim 1, characterized in that, The step of determining the sample to be processed, the time unit to be processed, and the time series reference interval based on the target monitoring point sample and the samples before and after the synchronous unit includes: The target monitoring point sample whose timestamp is located within the synchronous unit is determined as the target sample, and the corresponding synchronous unit is determined as the target synchronous unit; A first rate of change interval is generated based on the two adjacent samples preceding the target sample. The measurement interval of the previous sample is extended to the timestamp of the target sample to obtain a first reference interval. A second rate of change interval is generated based on the two adjacent samples following the target sample. The measurement interval of the subsequent sample is then extended in reverse to the timestamp of the target sample to obtain a second reference interval. The intersection of the first reference interval and the second reference interval is taken as the time series reference interval; When the measurement interval of the target sample does not intersect with the time series reference interval, and the first rate of change interval and the second rate of change interval are located on both sides of zero, the target sample is determined as the sample to be processed, and the target synchronous unit is determined as the synchronous unit to be processed.

5. The method as described in claim 1, characterized in that, The step of converting the measurement interval of the source monitoring point associated with the synchronous unit to be processed into a path mapping interval according to the conversion constraint, and aggregating them according to the temporal overlap relationship and the path verification weight to obtain the response consistency interval includes: The effective samples of the source monitoring points are determined based on the coverage relationship between the synchronous unit to be processed and the time domain intervals of each path, and the path mapping interval is generated according to the corresponding conversion constraints. Based on the degree of overlap between the time domain interval of each path and the time domain interval of the sample to be processed, and the corresponding path verification weight, the target path mapping interval and path support weight of each source monitoring point are determined. Candidate value segments are divided according to the endpoints of each target path mapping interval, and the support rate of each candidate value segment and the number of different source monitoring points are calculated. Candidate value segments are filtered based on a preset support rate threshold and a preset number of source points, and adjacent candidate value segments are merged to obtain a candidate response interval; When candidate response intervals exist, response consistency intervals are determined from the candidate response intervals based on the length of the candidate value segments contained in each candidate response interval and the support rate.

6. The method as described in claim 5, characterized in that, The step of determining noise samples and replacement values ​​and generating a noise reduction monitoring sequence based on the intersection relationship between the measurement interval of the sample to be processed, the time series reference interval, and the response consistency interval includes: When there is a consistent response interval, the measurement interval of the sample to be processed does not intersect with the consistent response interval, and the time reference interval intersects with the consistent response interval, the intersection of the time reference interval and the consistent response interval is determined as the feasible replacement interval. Calculate the intersection of each target path mapping interval with the alternative feasible interval, and determine the non-empty intersection as the alternative contribution interval; When there is a replacement contribution interval and the sum of the corresponding path support weights is not zero, the sample to be processed is determined as a noise sample, and the midpoint of each replacement contribution interval is weighted according to the corresponding path support weight to obtain the replacement value. The samples that were not identified as noise samples were identified as retained samples. The noise samples were replaced with the corresponding replacement values. The replaced noise samples, the retained samples, and the original samples that were not identified as noise samples were merged in the order of the original timestamps to generate a noise reduction monitoring sequence.

7. The method according to any one of claims 1 to 6, characterized in that, The steps for obtaining multiple monitoring sequences and response constraint records of traffic facilities include: Acquire monitoring data packets uploaded by various monitoring devices of traffic facilities, and extract monitoring point identifiers, monitoring types, timestamps, monitoring values, timestamp errors, and measurement errors from the monitoring data packets; The extracted results are grouped according to the monitoring point identifier and the monitoring type, and the monitoring values ​​in each group are arranged according to the timestamp to obtain multiple monitoring sequences; Read the correspondence between monitoring points and facility components, the connection relationship between components, the monitoring direction, and the calibration response segments; Based on the calibration response segment, the component connection relationship, and the monitoring direction, determine the response path, time delay constraint, conversion constraint, and path verification weight from the source monitoring point to the target monitoring point, and generate a response constraint record. Based on the monitoring point identifiers and monitoring types of the source monitoring point and the target monitoring point, the response constraint record is associated with the corresponding monitoring sequence.

8. A noise reduction device for traffic facility monitoring data, characterized in that, The device employs the noise reduction method for traffic facility monitoring data as described in any one of claims 1 to 7, and the device comprises: The data acquisition module is used to acquire multiple monitoring sequences and response constraint records of traffic facilities. The response constraint records include response paths, time delay constraints, conversion constraints, and path verification weights. The interval generation module is used to generate sample time domain intervals, measurement intervals, path time domain intervals, and synchronous units based on the timestamps, timestamp errors, measurement errors, and time delay constraints of the monitoring sequence. The sample screening module is used to determine the sample to be processed, the time unit to be processed, and the time series reference interval based on the target monitoring point sample and the samples before and after the synchronous unit. The interval aggregation module is used to convert the measurement interval of the source monitoring point associated with the synchronous unit to be processed into a path mapping interval according to the conversion constraint, and to aggregate it according to the temporal overlap relationship and the path verification weight to obtain the response consistency interval; The data replacement module is used to determine noise samples and replacement values ​​based on the intersection relationship between the measurement interval of the sample to be processed, the time series reference interval, and the response consistency interval, and to generate a noise reduction monitoring sequence.

9. A noise reduction device for traffic facility monitoring data, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the noise reduction method for traffic facility monitoring data as claimed in any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the noise reduction method for traffic facility monitoring data as described in any one of claims 1 to 7.

Images