Storage compression and dynamic synchronization method for multi-source heterogeneous driving data

Abstract

The invention belongs to the technical field of driving data processing, and particularly discloses a multi-source heterogeneous driving data storage compression and dynamic synchronization method, which comprises the following steps of: preprocessing original driving data, and converting the original driving data into a uniform format; according to the characteristics of the data, adaptively selecting a time-frequency transformation method to deeply mine redundant information of the data, and performing quantization processing on the mined redundant information; using a clock synchronization algorithm to accurately calibrate the clock of each data source; according to the method, the characteristics of data can be better captured through self-adaptive time-frequency transformation, redundant information in a frequency domain can be more accurately mined, the compression ratio is further improved on the premise of ensuring the data quality, and the space required by storage is reduced; by adopting the clock synchronization algorithm, the error of data synchronization can be controlled within a smaller range, the alignment precision of the data in time is improved, and the consistency of the data of different data sources in time is ensured to be higher.

Description

Technical Field

[0001] The present invention belongs to the technical field of vehicle data processing, and in particular relates to a storage compression and dynamic synchronization method for multi-source heterogeneous vehicle data. Background Art

[0002] Multi-source heterogeneous driving data is data related to vehicle driving that is collected from a variety of different types of data sources and has different data structures and formats. These data sources include but are not limited to vehicle-mounted sensors, driving recorders, navigation systems, roadside sensors in intelligent transportation systems, and data from other vehicles or infrastructure obtained by the Internet of Vehicles platform. In order to effectively store and transmit these large-scale data, these data need to be compressed.

[0003] At present, the storage and compression method for multi-source heterogeneous driving data is usually to convert the driving data from the time domain to the frequency domain, compress it using the redundant information in the frequency domain, and then synchronize and update the data of each data source according to the set time interval; however, since multi-source heterogeneous data may come from different sensors and devices, there are differences in their collection methods, sampling rates and accuracy, which may lead to inconsistency problems in the data compression process, resulting in low accuracy and reliability of data synchronization. Therefore, we need to propose a storage compression and dynamic synchronization method for multi-source heterogeneous driving data to solve the above problems, so that it can control the error of data synchronization within a smaller range, improve the accuracy of data alignment in time, ensure that the data from different data sources are more consistent in time, and improve the accuracy and reliability of data synchronization. Summary of the invention

[0004] The purpose of the present invention is to provide a storage compression and dynamic synchronization method for multi-source heterogeneous driving data, to control the error of data synchronization within a smaller range, to improve the temporal alignment accuracy of data, to ensure that data from different data sources are more consistent in time, to improve the accuracy and reliability of data synchronization, so as to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention adopts the following technical solutions: The storage compression and dynamic synchronization method of multi-source heterogeneous driving data includes the following steps: S1. Preprocessing the original driving data to convert the original driving data into a unified format; S2. According to the characteristics of the data, the time-frequency transformation method is adaptively selected to deeply mine the redundant information of the data, and the mined redundant information is quantified; S3. Use a clock synchronization algorithm to accurately calibrate the clocks of each data source; S4, extract representative feature data from data from different data sources; S5. Establishing a correspondence between data from different data sources by calculating the similarity between different features to perform feature matching; S6. Fine-tune the data based on the feature matching results to ensure accurate synchronization of the data in time; S7. Establish a multi-source heterogeneous driving database and store the compressed data on different nodes; S8. Establish metadata information for each data source to facilitate query, retrieval and management of the data; S9. Set a data update trigger mechanism and update the data in the database.

[0006] Preferably, in step S1, the process of preprocessing the original driving data is as follows: S11, the original driving data is processed by moving average filtering to obtain filtered data. The formula of moving average filtering is as follows: , in, and , is the number of time series data, is the number of data points involved in the moving average calculation, After moving average filtering, the position The filtering result at is the jth data point in the original time series data, The position index of the current data point in the entire data sequence; S12. Sort the filtered data from small to large, and then calculate the positions of the quartiles respectively. The calculation formula is as follows: , ,in, is the upper quartile position, is the position of the lower quartile, is the number of filtered data; S13. Calculate the lower quartiles according to the position of the quartiles and upper quartile , the calculation formula is as follows: like and is an integer, then , ; like or If it is not an integer, set or The decimal part is a and the integer part is b, then the quartile calculation formula is used to calculate and , the calculation formula is: ,in, The lower quartile or upper quartile , and After sorting, they are and The data value of the location; S14, the upper quartile Subtract the lower quartile Get the interquartile range IQ and set the normal data range to , delete the data that exceeds the normal data range; S15, using FFmpeg library function to convert video files into mp4 format; S16. Perform linear normalization processing on the numerical data to convert the data into a unified format. The normalization formula is as follows: ,in, is the output value after normalization. is numeric data, is the minimum value of numeric data. The maximum value of numeric data.

[0007] Preferably, in step S2, the process of mining redundant information of data is as follows: A1. Obtain frequency domain information data by Fourier transform or wavelet transform; A2. Calculate the mean of the frequency domain coefficients corresponding to different frames in the frequency domain. The mean calculation formula is: , where M is the total number of data frames, For the The frequency domain coefficients of the frame, is the frequency domain coordinate, For coordinates The mean value at A3. According to the mean The calculated coordinates are The variance at , the variance calculation formula is: ,in, The coordinates are The variance at A4. Set the quantization interval to [c, d], the quantization level to L, and calculate the quantization step length Log. The calculation formula is: ,in, is the lower limit of the quantization interval, is the upper limit of the quantization interval; A5. Calculate the quantized output value based on the quantization step length Log. The calculation formula is: ,in, is the frequency domain coefficient, To map continuous frequency domain coefficients to a finite number of discrete values, is the lower limit of the quantization interval, is the quantized output value.

[0008] Preferably, in step S3, the process of calibrating the clocks of each data source is as follows: S31, initializing the clock of each data source according to whether the clock of the data source is a master clock or a slave clock; S32, the master clock periodically sends synchronization messages to the slave clocks in the network, and the synchronization messages include the local timestamp of the master clock sending the synchronization messages ; S33. When receiving a synchronization message from the clock, record the local time at which the message is received ; S34. After receiving the synchronization message, the slave clock sends a delay request message to the master clock and records the local time when the delay request message is sent. ; S35: After receiving the delay request message, the master clock sends a delay response message to the slave clock, and the delay response message includes the local time of the master clock when the delay request message is received. ; S36. The slave time calculates the clock offset Offset and the network transmission delay Delay between the slave clock and the master clock according to the received timestamp information. The calculation formula is as follows: , ; S37. The slave clock adjusts the local clock according to the calculated clock offset Offset, so that the slave clock is synchronized with the master clock.

[0009] Preferably, in step S4, the process of feature data extraction is as follows: S41, determining the specific time at which the feature is to be extracted according to the timestamp of the data, where the determination of the specific time satisfies ,in, is the time series of known data, To select data at a specific time, For a specific moment The index of nearby data points for traversal and summation; S42. According to a specific time Arrange the nearby data from small to large, calculate the median of the arranged data, and take the median as the feature data.

[0010] Preferably, in step S5, the process of performing feature matching is as follows: S51. At a specific time As the center, set the time window to ,in, is the time window size; S52. Calculate the similarity between the feature data of different data sources in the time window according to the feature data of different data sources. The similarity calculation formula is: , where r is the dimension of the feature vector, is the i-th element of the eigenvector A, is the i-th element of the eigenvector B, is the similarity between feature vector A and feature vector B; S53: According to the calculated similarity values, feature data pairs with high similarity are selected to establish corresponding relationships between the feature data pairs.

[0011] Preferably, in step S6, the process of fine-tuning the data is as follows: S61. Compare the timestamps of the data from different data sources according to the result of feature matching, and calculate the time deviation between the different data sources; S62. Adjust according to the time deviation to minimize the time deviation between different data sources, and update the data records of related data sources. When the time deviation is small, use the interpolation method to make adjustments; when the time deviation is large, it is necessary to further check the accuracy of the data or re-match the features.

[0012] Preferably, in step S7, the process of establishing a multi-source heterogeneous driving database is as follows: S71. Determine the overall architecture of the multi-source heterogeneous driving database, which includes storage methods and data organization forms; S72. Determine the number of nodes and node configuration according to the characteristics and scale of the data; S73. Store the compressed data on different nodes according to the planned architecture and storage method.

[0013] Preferably, in step S8, when establishing metadata information for the data of each data source, a detailed metadata structure is first defined for the data of each data source, and then the metadata information of each data source is collected during the data collection or storage process, and finally a relational database is used to store the collected metadata information in the metadata management module of the database to ensure that an effective association is established between the metadata and the actual data.

[0014] Preferably, in step S9, the process of setting the data update trigger mechanism is as follows: S91. Determine the frequency and timeliness requirements for data updates based on application requirements for real-time vehicle status monitoring, driving behavior analysis, or intelligent traffic management; S92, setting event triggering conditions, and monitoring the occurrence of related events in real time to determine whether the event triggering conditions are met. When the event triggering conditions are not met, continue to monitor the related events in real time. When the event triggering conditions are met, obtain the latest data from various data sources; S93, clean and convert the latest data obtained, and then synchronize the updated data in time sequence according to the timestamp information of the data to ensure that the data is inserted into the database in normal time sequence. If there is a deviation in the data timestamps of different data sources, calibration and adjustment are required to minimize the deviation; S94: Insert the synchronized updated data into the corresponding position of the database. If there is duplicate data or conflicting data when inserting the data, overwrite the duplicate or conflicting data.

[0015] Compared with the prior art, the storage compression and dynamic synchronization method of multi-source heterogeneous driving data proposed in the present invention has the following advantages: 1. The present invention can better capture the characteristics of data through adaptive time-frequency transformation, more accurately mine redundant information in the frequency domain, further improve the compression ratio while ensuring data quality, and reduce the space required for storage; by adopting a clock synchronization algorithm, the error of data synchronization can be controlled within a smaller range, the alignment accuracy of data in time can be improved, and the consistency of data from different data sources in time can be ensured to be higher.

[0016] 2. The present invention can overcome the data synchronization problems caused by sensor acquisition differences and network transmission delays to a certain extent through the mechanism of feature matching, enhance the stability and reliability of data synchronization, and ensure accurate data synchronization in complex driving environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A flowchart according to an embodiment of the present invention is shown; Figure 2A flowchart of preprocessing raw driving data according to an embodiment of the present invention is shown; Figure 3 A flowchart of redundant information mining of data according to an embodiment of the present invention is shown; Figure 4 A flowchart of calibrating the clocks of various data sources according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0018] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. The specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0019] The present invention provides Figure 1-4 The storage compression and dynamic synchronization method of multi-source heterogeneous driving data shown includes the following steps: S1. Preprocessing the original driving data to convert the original driving data into a unified format; like Figure 2 As shown in the figure, the process of preprocessing the original driving data is as follows: S11, the original driving data is processed by moving average filtering to obtain filtered data. The formula of moving average filtering is as follows: , in, and , is the number of time series data, is the number of data points involved in the moving average calculation, After moving average filtering, the position The filtering result at is the jth data point in the original time series data, The position index of the current data point in the entire data sequence; S12. Sort the filtered data from small to large, and then calculate the positions of the quartiles respectively. The calculation formula is as follows: , ,in, is the upper quartile position, is the position of the lower quartile, is the number of filtered data; S13. Calculate the lower quartiles according to the position of the quartiles and upper quartile , the calculation formula is as follows: like and is an integer, then , ; like or If it is not an integer, set or The decimal part is a and the integer part is b, then the quartile calculation formula is used to calculate and , the calculation formula is: ,in, The lower quartile or upper quartile , and After sorting, they are and The data value of the location; S14, the upper quartile Subtract the lower quartile Get the interquartile range IQ and set the normal data range to , delete the data that exceeds the normal data range; S15. Use FFmpeg library functions to uniformly convert video files into mp4 format, so as to improve the compatibility of video data on different devices and software platforms; S16. Perform linear normalization processing on the numerical data to convert the data into a unified format. The normalization formula is as follows: ,in, is the output value after normalization. is numeric data, is the minimum value of numeric data. The maximum value of the numeric data; By performing moving average filtering on the original driving data, the data is made smoother and more accurate, reducing the impact of random interference on data analysis. The quartile method is used to find and delete outliers, which can reduce the misleading effect of erroneous data on the overall analysis and ensure that the data can represent normal driving conditions. The video file format is converted so that the data can be better shared, transmitted and processed on different devices, software and platforms, improving the versatility of the data. Normalization processing can make data of different magnitudes and distributions comparable and consistent, which is conducive to subsequent data processing.

[0020] S2, adaptively select the time-frequency transformation method according to the characteristics of the data to deeply mine the redundant information of the data, and quantify the mined redundant information; The time-frequency transformation method is set to Fourier transform or wavelet transform, wherein Fourier transform is applicable to vehicle speed data with obvious periodicity, and the vehicle speed data in the time domain is converted to the frequency domain through Fourier transform to analyze the frequency components in the frequency domain and find the features corresponding to the periodic features; wavelet transform is applicable to brake data with mutation characteristics, and by continuously performing the next layer of discrete wavelet transform on the low-frequency approximate coefficients, the features of the signal at different scales can be analyzed to capture the mutation information in the brake data; like Figure 3 As shown in the figure, the process of redundant information mining of data is as follows: A1. Obtain frequency domain information data by Fourier transform or wavelet transform; Among them, the Fourier transform formula is: ,in, is the frequency domain data after Fourier transform, is the vehicle speed value collected at discrete time point p, p = 0, 1, ..., N-1, that is, the vehicle speed data sequence from the 0th time point to the N-1th time point, N is the total number of points for collecting vehicle speed data, is an imaginary unit, satisfying , is the complex exponential term in the DFT transform, which is used to convert the time domain signal a into the frequency domain; The formula for wavelet transform is as follows: , ,in, =0, 1, …, , N is the total number of brake signal data points, After DWT decomposition, The low-frequency component coefficients at the positions; After DWT decomposition, The high frequency component coefficients at the positions are Used to identify the low-frequency approximation coefficient and high frequency detail factor The index variable, is the even number of sample data in the original signal, is the odd number of sample data in the original discrete signal; A2. Calculate the mean of the frequency domain coefficients corresponding to different frames in the frequency domain. The mean calculation formula is: , where M is the total number of data frames, For the The frequency domain coefficients of the frame, is the frequency domain coordinate, For coordinates The mean value at A3. According to the mean The calculated coordinates are The variance at , the variance calculation formula is: ,in, The coordinates are The variance at A4. Set the quantization interval to [c, d], the quantization level to L, and calculate the quantization step length Log. The calculation formula is: ,in, is the lower limit of the quantization interval, is the upper limit of the quantization interval; A5. Calculate the quantized output value based on the quantization step length Log. The calculation formula is: ,in, is the frequency domain coefficient, To map continuous frequency domain coefficients to a finite number of discrete values, is the lower limit of the quantization interval, is the quantized output value; S3. Use a clock synchronization algorithm to accurately calibrate the clocks of each data source; like Figure 3 As shown, the process of calibrating the clocks of each data source is as follows: S31, initializing the clock of each data source according to whether the clock of the data source is a master clock or a slave clock; S32, the master clock periodically sends synchronization messages to the slave clocks in the network, and the synchronization messages include the local timestamp of the master clock sending the synchronization messages ; S33. When receiving a synchronization message from the clock, record the local time at which the message is received ; S34. After receiving the synchronization message, the slave clock sends a delay request message to the master clock and records the local time when the delay request message is sent. ; S35: After receiving the delay request message, the master clock sends a delay response message to the slave clock, and the delay response message includes the local time of the master clock when the delay request message is received. ; S36. The slave time calculates the clock offset Offset and the network transmission delay Delay between the slave clock and the master clock according to the received timestamp information. The calculation formula is as follows: , ; S37, the slave clock adjusts the local clock according to the calculated clock offset Offset, so that the slave clock is synchronized with the master clock; Through the above calibration method, the error of data synchronization can be controlled within a smaller range, the time alignment accuracy of data can be improved, and the time consistency of data from different data sources can be ensured to be higher.

[0021] S4, extract representative feature data from data from different data sources; like Figure 4 As shown in the figure, the process of feature data extraction is as follows: S41, determining the specific time at which the feature is to be extracted according to the timestamp of the data, where the determination of the specific time satisfies ,in, is the time series of known data, To select data at a specific time, For a specific moment The index of nearby data points for traversal and summation; S42. According to a specific time Arrange the nearby data from small to large, calculate the median of the arranged data, and take the median as the feature data; S5. Establishing a correspondence between data from different data sources by calculating the similarity between different features to perform feature matching; The process of feature matching is as follows: S51. At a specific time As the center, set the time window to ,in, is the time window size; S52. Calculate the similarity between the feature data of different data sources in the time window according to the feature data of different data sources. The similarity calculation formula is: , where r is the dimension of the feature vector, is the i-th element of the eigenvector A, is the i-th element of the eigenvector B, is the similarity between feature vector A and feature vector B; S53: According to the calculated similarity values, feature data pairs with high similarity are selected to establish corresponding relationships between the feature data pairs.

[0022] S6. Fine-tune the data based on the feature matching results to ensure accurate synchronization of the data in time; The process of fine-tuning the data is as follows: S61. Compare the timestamps of the data from different data sources according to the result of feature matching, and calculate the time deviation between the different data sources; S62, adjusting according to the time deviation to minimize the time deviation between different data sources, and updating the data records of the relevant data sources. When the time deviation is small, the interpolation method is used for adjustment; when the time deviation is large, the accuracy of the data needs to be further checked or feature matching needs to be re-performed; Through the feature matching mechanism, it is possible to overcome data synchronization problems caused by differences in sensor collection and network transmission delays to a certain extent, enhance the stability and reliability of data synchronization, and ensure accurate data synchronization even in complex driving environments.

[0023] S7. Establish a multi-source heterogeneous driving database and store the compressed data on different nodes to improve the reliability and scalability of storage; The process of establishing a multi-source heterogeneous driving database is as follows: S71. Determine the overall architecture of the multi-source heterogeneous driving database, where the overall architecture includes storage methods and data organization forms, wherein the storage methods include file storage and relational database storage; the data organization forms include classification by data source and classification by time sequence; S72. Determine the number of nodes and node configuration according to the characteristics and scale of the data; S73, storing the compressed data on different nodes according to the planned architecture and storage method; S8. Establish metadata information for each data source to facilitate query, retrieval and management of the data. The metadata information includes the source of the data, collection time, sampling rate and sampling accuracy; When establishing metadata information for the data of each data source, first define a detailed metadata structure for the data of each data source (the metadata structure contains fields for data source, collection time, sampling rate and accuracy), then collect the metadata information of each data source during the data collection or storage process, and finally use a relational database to store the collected metadata information in the metadata management module of the database to ensure that an effective association is established between the metadata and the actual data to facilitate query and retrieval.

[0024] S9. Set a data update trigger mechanism and update the data in the database.

[0025] The process of setting the data update trigger mechanism is as follows: S91. Determine the frequency and timeliness requirements for data updates based on application requirements for real-time vehicle status monitoring, driving behavior analysis, or intelligent traffic management; S92, setting event triggering conditions, and monitoring the occurrence of related events in real time to determine whether the event triggering conditions are met. When the event triggering conditions are not met, continue to monitor the related events in real time. When the event triggering conditions are met, obtain the latest data from various data sources; S93, clean and convert the latest data obtained, and then synchronize the updated data in time sequence according to the timestamp information of the data to ensure that the data is inserted into the database in normal time sequence for subsequent analysis and query. If there is a deviation in the data timestamps of different data sources, calibration and adjustment are required to minimize the deviation; S94: Insert the synchronized updated data into the corresponding position of the database. If there is duplicate data or conflicting data when inserting the data, overwrite the duplicate or conflicting data.

[0026] Adaptive time-frequency transformation can better capture the characteristics of data, more accurately mine redundant information in the frequency domain, further improve the compression ratio while ensuring data quality, and reduce the space required for storage; by adopting a clock synchronization algorithm, the error of data synchronization can be controlled within a smaller range, improving the accuracy of data alignment in time and ensuring that data from different data sources are more consistent in time.

[0027] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for storage compression and dynamic synchronization of multi-source heterogeneous driving data, characterized in that: The steps include: S1. Preprocessing the original driving data to convert the original driving data into a unified format; S2. According to the characteristics of the data, the time-frequency transformation method is adaptively selected to deeply mine the redundant information of the data, and the mined redundant information is quantified; S3. Use a clock synchronization algorithm to accurately calibrate the clocks of each data source; S4, extract representative feature data from data from different data sources; S5. Establishing a correspondence between data from different data sources by calculating the similarity between different features to perform feature matching; S6. Fine-tune the data based on the feature matching results to ensure accurate synchronization of the data in time; S7. Establish a multi-source heterogeneous driving database and store the compressed data on different nodes; S8. Establish metadata information for each data source to facilitate query, retrieval and management of the data; S9. Set a data update trigger mechanism and update the data in the database.

2. The method for storage compression and dynamic synchronization of multi-source heterogeneous driving data according to claim 1, characterized in that: In step S1, the process of preprocessing the original driving data is as follows: S11, the original driving data is processed by moving average filtering to obtain filtered data. The formula of moving average filtering is as follows: ， in, and , is the number of time series data, is the number of data points involved in the moving average calculation, After moving average filtering, the position The filtering result at is the jth data point in the original time series data, The position index of the current data point in the entire data sequence; S12. Sort the filtered data from small to large, and then calculate the positions of the quartiles respectively. The calculation formula is as follows: ， ,in, is the upper quartile position, is the position of the lower quartile, is the number of filtered data; S13. Calculate the lower quartiles according to the position of the quartiles and upper quartile , the calculation formula is as follows: like and is an integer, then , ; like or If it is not an integer, set or The decimal part is a and the integer part is b, then the quartile calculation formula is used to calculate and , the calculation formula is: ,in, The lower quartile or upper quartile , and After sorting, they are and The data value of the location; S14, the upper quartile Subtract the lower quartile Get the interquartile range IQ and set the normal data range to , delete the data that exceeds the normal data range; S15, using FFmpeg library function to convert video files into mp4 format; S16. Perform linear normalization processing on the numerical data to convert the data into a unified format. The normalization formula is as follows: ,in, is the output value after normalization. is numeric data, is the minimum value of numeric data. The maximum value of numeric data.

3. The method for storage compression and dynamic synchronization of multi-source heterogeneous driving data according to claim 1, characterized in that: In step S2, the process of mining redundant information of data is as follows: A1. Obtain frequency domain information data by Fourier transform or wavelet transform; A2. Calculate the mean of the frequency domain coefficients corresponding to different frames in the frequency domain. The mean calculation formula is: , where M is the total number of data frames, For the The frequency domain coefficients of the frame, is the frequency domain coordinate, For coordinates The mean value at A3. According to the mean The calculated coordinates are The variance at , the variance calculation formula is: ,in, The coordinates are The variance at A4. Set the quantization interval to [c, d], the quantization level to L, and calculate the quantization step length Log. The calculation formula is: ,in, is the lower limit of the quantization interval, is the upper limit of the quantization interval; A5. Calculate the quantized output value based on the quantization step length Log. The calculation formula is: ,in, is the frequency domain coefficient, To map continuous frequency domain coefficients to a finite number of discrete values, is the lower limit of the quantization interval, is the quantized output value.

4. The method for storage compression and dynamic synchronization of multi-source heterogeneous driving data according to claim 1, characterized in that: In step S3, the process of calibrating the clocks of each data source is as follows: S31, initializing the clock of each data source according to whether the clock of the data source is a master clock or a slave clock; S32, the master clock periodically sends synchronization messages to the slave clocks in the network, and the synchronization messages include the local timestamp of the master clock sending the synchronization messages ; S33. When receiving a synchronization message from the clock, record the local time at which the message is received ; S34. After receiving the synchronization message, the slave clock sends a delay request message to the master clock and records the local time when the delay request message is sent. ; S35: After receiving the delay request message, the master clock sends a delay response message to the slave clock, and the delay response message includes the local time of the master clock when the delay request message is received. ; S36. The slave time calculates the clock offset Offset and the network transmission delay Delay between the slave clock and the master clock according to the received timestamp information. The calculation formula is as follows: ， ； S37. The slave clock adjusts the local clock according to the calculated clock offset Offset, so that the slave clock is synchronized with the master clock.

5. The method for storage compression and dynamic synchronization of multi-source heterogeneous driving data according to claim 1, characterized in that: In step S4, the process of feature data extraction is as follows: S41, determining the specific time at which the feature is to be extracted according to the timestamp of the data, where the determination of the specific time satisfies ,in, is the time series of known data, To select data at a specific time, For a specific moment The index of nearby data points for traversal and summation; S42. According to a specific time Arrange the nearby data from small to large, calculate the median of the arranged data, and take the median as the feature data.

6. The method for storage compression and dynamic synchronization of multi-source heterogeneous driving data according to claim 1, characterized in that: In step S5, the process of feature matching is as follows: S51. At a specific time As the center, set the time window to ,in, is the time window size; S52. Calculate the similarity between the feature data of different data sources in the time window according to the feature data of different data sources. The similarity calculation formula is: , where r is the dimension of the feature vector, is the i-th element of the eigenvector A, is the i-th element of the eigenvector B, is the similarity between feature vector A and feature vector B; S53: According to the calculated similarity values, feature data pairs with high similarity are selected to establish corresponding relationships between the feature data pairs.

7. The method for storage compression and dynamic synchronization of multi-source heterogeneous driving data according to claim 1, characterized in that: In step S6, the process of fine-tuning the data is as follows: S61. Compare the timestamps of the data from different data sources according to the result of feature matching, and calculate the time deviation between the different data sources; S62. Adjust according to the time deviation to minimize the time deviation between different data sources, and update the data records of related data sources. When the time deviation is small, use the interpolation method to make adjustments; when the time deviation is large, it is necessary to further check the accuracy of the data or re-match the features.

8. The method for storage compression and dynamic synchronization of multi-source heterogeneous driving data according to claim 1, characterized in that: In step S7, the process of establishing a multi-source heterogeneous driving database is as follows: S71. Determine the overall architecture of the multi-source heterogeneous driving database, which includes storage methods and data organization forms; S72. Determine the number of nodes and node configuration according to the characteristics and scale of the data; S73. Store the compressed data on different nodes according to the planned architecture and storage method.

9. The method for storage compression and dynamic synchronization of multi-source heterogeneous driving data according to claim 1, characterized in that: In step S8, when establishing metadata information for the data of each data source, first define a detailed metadata structure for the data of each data source, then collect the metadata information of each data source during the data collection or storage process, and finally use a relational database to store the collected metadata information in the metadata management module of the database to ensure that an effective association is established between the metadata and the actual data.

10. The method for storage compression and dynamic synchronization of multi-source heterogeneous driving data according to claim 1, characterized in that: In step S9, the process of setting the data update trigger mechanism is as follows: S91. Determine the frequency and timeliness requirements for data updates based on application requirements for real-time vehicle status monitoring, driving behavior analysis, or intelligent traffic management; S92, setting event triggering conditions, and monitoring the occurrence of related events in real time to determine whether the event triggering conditions are met. When the event triggering conditions are not met, continue to monitor the related events in real time. When the event triggering conditions are met, obtain the latest data from various data sources; S93, clean and convert the latest data obtained, and then synchronize the updated data in time sequence according to the timestamp information of the data to ensure that the data is inserted into the database in normal time sequence. If there is a deviation in the data timestamps of different data sources, calibration and adjustment are required to minimize the deviation; S94: Insert the synchronized updated data into the corresponding position of the database. If there is duplicate data or conflicting data when inserting the data, overwrite the duplicate or conflicting data.

Images