Multi-modal data high-precision time synchronization acquisition system and method based on Syntaclos

Abstract

The invention discloses a multi-modal data high-precision time synchronization acquisition system and method based on Syntaclos, and relates to the technical field of time synchronization, the system comprises a target determination module, a time division module, an index establishment module, a state evaluation module and a time synchronization module; the method is technically characterized by comprising the following steps: sending a time synchronization message to slave clock equipment through master clock equipment, extracting time granularity, dividing time partitions, establishing a corresponding time partition index for each time partition, and outputting a prediction index table through a synchronous topology network; determining an initial time partition of the multi-modal data based on the prediction index table, introducing a machine learning model, building a rule engine, analyzing the degree of influence, predicting a next time partition with deviation, marking a time deviation event once, and performing synchronous processing on the detected time deviation event in a Syntacalos synchronization interface; the time synchronization efficiency is improved, and time synchronization errors are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of time synchronization, and in particular to a Syntalos-based multimodal data high-precision time synchronization acquisition system and method. Background Art

[0002] The default version of Syntalos already implements many commonly used input sources, including electrophysiological recording modules, in vivo calcium imaging (UCLA Miniscope), and various video cameras. That is, the timestamps of all data sources are aligned to a globally shared master clock. This facilitates analysis and data interoperability by combining the directory structure of data from different sources and their metadata, and storing similar types of data in the same format. Traditional time synchronization acquisition methods, such as LED pulse triggering, only provide initial synchronization and cannot dynamically correct time drift between devices, resulting in significant cumulative errors in long-term recordings. In Syntalos, on the one hand, time synchronization usually relies on a master clock. When the master clock device fails, the single point of failure will cause synchronization failure of the entire network. The fixed time window division is unable to adapt to the differences in multimodal data acquisition frequency (such as high-frequency acquisition of electrophysiological signals and low-frequency acquisition of video images), which can easily lead to timestamp misalignment. On the other hand, the existing Syntalos system has a high degree of functional coupling. Adding new devices or adjusting synchronization strategies requires extensive code modifications, which greatly reduces system expansion efficiency. In addition, data storage lacks structured indexes, which is inefficient when querying multimodal data for specific time intervals and specific devices, and greatly increases retrieval latency. Summary of the Invention

[0003] (1) Technical problems solved In response to the shortcomings of the existing technology, the present invention provides a high-precision time synchronization acquisition system and method for multimodal data based on Syntalos, which includes a target determination module, a time division module, an index establishment module, a status assessment module and a time synchronization module. The system sends a time synchronization message to the slave clock device through the master clock device, extracts the time granularity, divides the time partition, establishes a corresponding time partition index for each time partition, and constructs a synchronization topology network with different priorities, thereby solving the problems raised in the background technology.

[0004] (2) Technical solution To achieve the above objectives, the present invention is implemented through the following technical solutions: In a first aspect, the present application provides a high-precision time synchronization acquisition system for multimodal data based on Syntalos, the system comprising: A target determination module is configured to determine a first master clock device, a plurality of second master clock devices, and a plurality of slave clock devices. The first master clock device sends a time synchronization message to the plurality of second master clock devices and the plurality of slave clock devices. The time synchronization message includes at least a time granularity. The slave clock devices are configured to collect multimodal data. A time partitioning module is used to divide time partitions based on the time granularity of the time synchronization message; An index building module is used to build a corresponding time partition index based on each time partition and output a prediction index table through a synchronous topology network; A state assessment module is used to determine the initial time partition of the multimodal data based on the prediction index table, introduce a machine learning model, build a rule engine, analyze the impact, and predict the next time partition where the deviation will occur; The time synchronization module is used to mark a time deviation event based on the next time partition where the deviation occurs, and synchronously process the detected time deviation event in the Syntalos synchronization interface.

[0005] Furthermore, determining a first master clock device, a plurality of second master clock devices, and a plurality of slave clock devices includes: When receiving the power-on signal, immediately obtain the device's MAC address and clock code; A unique hash value is formed based on the MAC address and the clock code, and the device type is determined based on the hash value, including a first master clock device, a second master clock device, and a slave clock device.

[0006] Furthermore, the time partitions are divided into: Based on the multimodal data, obtaining time information including the full acquisition period of each slave clock device, including at least the acquisition timestamp of the slave clock device; Based on the slave clock device and the corresponding time information, combined with the time granularity of the first master clock device, the multimodal data is divided into multiple time partitions, and the division is performed according to the sampling frequency range determined by the slave clock device type and the time granularity of the first master clock device.

[0007] Furthermore, the construction of the synchronous topology network includes: Mapping a time partition to a topological node, where the topological node includes a first master clock node, a second master clock node, and a slave clock node; extracting the hop count information and communication link information of the time partition, and constructing a topological edge between the topological nodes of adjacent time partitions, where the topological edge carries at least a link delay parameter; At the same time, the clock is configured for the first master clock node, and the time synchronization message is broadcast to the second master clock node and the slave clock node through the topological edge, and a synchronization signal is generated. The second master clock node receives the synchronization signal of the first master clock node and synchronizes in real time.

[0008] Furthermore, the step of outputting the prediction index table includes: Traverse all nodes in the synchronous topology network to obtain the device type and connection relationship of each node, and the connection relationship is obtained through a preset graph model; Through each time partition, based on the node connection relationship, determine the time partitions where all nodes connected to the nodes in the partition are located, and form a set of associated time partitions; The device type set, device quantity, and associated time partition set are integrated into a prediction index table.

[0009] Furthermore, we introduced a machine learning model and built a rule engine to analyze the impact and predict the time partition when the next deviation will occur, including: Mark the association between nodes and links in a time topology network; a node is a device involved in time synchronization in a time topology network, and a link is a path connecting two nodes corresponding to an actual communication line, including but not limited to transmission channels composed of communication media such as wired Ethernet, wireless Wi-Fi, and optical fiber, and each link carries at least a synchronization signal delay parameter; Setting a set of equations for node synchronization status: a first set of equations generated by analyzing node time deviations and a second set of equations generated by analyzing link delays; obtaining solutions to the first and second sets of equations by obtaining the timestamps of all clocks on the nodes and the receiving and sending timestamps of time synchronization messages; The feature matrix is obtained by performing de-joining processing on the corresponding nodes, and the feature matrix corresponding to each node is compared and analyzed with the preset feature threshold matrix. When the feature matrix is less than the preset feature threshold matrix, the corresponding position is marked as 1, otherwise it is marked as 0, and the state matrix of the corresponding node combination is formed in sequence. Based on the first set of equations, the time deviation sequence is obtained; based on the time deviation sequence, the drift parameter is obtained, and combined with the state matrix, both are used as input features of the machine learning model, and the deviation classification probability is output. It is compared with the preset comparison interval in the model to obtain the corresponding deviation set and impact degree, and combined with the prediction index table, it is quickly mapped to the time partition where the deviation occurs; among them, the impact degree includes first-level impact, second-level impact and third-level impact.

[0010] Furthermore, the machine learning model adopts a decision tree algorithm.

[0011] Furthermore, in the step of configuring the clock for the first master clock node, a primary link and a redundant link are introduced through the time topology network. If there is a secondary or tertiary impact on the primary link, the slave clock node automatically switches to the redundant link to receive the synchronization signal.

[0012] In a second aspect, the present application provides a method for high-precision time synchronization acquisition of multimodal data based on Syntalos, the method comprising: Determine a first master clock device, several second master clock devices, and several slave clock devices, and the first master clock device sends a time synchronization message to the several second master clock devices and the several slave clock devices; wherein the time synchronization message includes at least a time granularity, and the slave clock device is used to collect multimodal data; Divide time partitions based on the time granularity of time synchronization messages; Establish a corresponding time partition index based on each time partition, and output the prediction index table through the synchronization topology network; Determine the initial time partition of multimodal data based on the prediction index table, introduce a machine learning model, build a rule engine, analyze the impact, and predict the next time partition where deviation will occur; A time deviation event is marked based on the next time partition where the deviation occurs, and the detected time deviation event is synchronously processed in the Syntalos synchronization interface.

[0013] In a third aspect, the present application provides a computer-readable storage medium having a program stored thereon. When the program is executed by a processor, the Syntalos-based multimodal data high-precision time synchronization acquisition method provided in the second aspect above is implemented.

[0014] (3) Beneficial effects The present invention provides a high-precision time synchronization acquisition system and method for multimodal data based on Syntalos, which has the following beneficial effects: The present invention designs a first master clock device, a second master clock device, and a slave clock device. The system divides the time into multiple time zones using the time granularity of the first master clock device and the slave clock devices and their corresponding time information. Within each time zone, the system can automatically adapt to new acquisition requirements to adjust the zone strategy without requiring any changes to the core synchronization algorithm, significantly improving the system's scalability. The present invention builds a synchronous topology network and introduces a primary link and a redundant link. When the primary link experiences a secondary or tertiary impact, the redundant link is automatically switched to receive the synchronization signal. At the same time, if the first master clock node experiences a tertiary impact, the second master clock node is promoted to the master clock node via the redundant link. To a certain extent, the time synchronization task can be seamlessly taken over, completely avoiding the risk of network-wide synchronization paralysis caused by master clock failure. The present invention marks the associations between nodes and links in a temporal topological network, sets a set of equations for node synchronization states, and structures and quantifies the complex relationships in the temporal topological network through mathematical modeling, facilitating subsequent prediction and analysis. The high-precision synchronization of multimodal data provides an interpretable and scalable analysis framework. The present invention integrates intuitive event marking and visual processing functions into the Syntalos synchronization interface, transforming complex operation processes into a simple and easy-to-understand graphical interface, greatly shortening processing time and providing a basic basis for subsequent synchronization optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a module diagram of a multimodal data high-precision time synchronization acquisition system according to an exemplary embodiment; Figure 2 The figure is a schematic diagram showing the steps of a method for high-precision time synchronization acquisition of multimodal data according to an exemplary embodiment. DETAILED DESCRIPTION

[0016] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0017] Example 1: The embodiment of the present invention provides a high-precision time synchronization acquisition system for multimodal data based on Syntalos; Figure 1 is a module diagram of a multimodal data high-precision time synchronization acquisition system according to an exemplary embodiment; see Figure 1 The system includes: a target determination module, a time division module, an index establishment module, a state evaluation module and a time synchronization module, and the target determination module, the time division module, the index establishment module, the state evaluation module and the time synchronization module are communicatively connected; The following is an explanation of each module: Target determination module: used to determine a first master clock device, several second master clock devices and several slave clock devices; The step of determining a first master clock device, a plurality of second master clock devices, and a plurality of slave clock devices includes: When receiving the power-on signal, immediately obtain the device's MAC address and clock code; A unique hash value is formed based on the MAC address and clock code, and the device type is determined based on the hash value, including a first master clock device, a second master clock device, and a slave clock device. The clock codes corresponding to the first master clock device, the second master clock device, and the slave clock device are all different, and the MAC address ranges corresponding to the first master clock device, the second master clock device, and the slave clock device are also different. Through hashing the MAC address and clock code, a unique identity can be formed, facilitating rapid identification of the device type. By pre-associating and storing the hash value (MAC and clock code combination) of known devices with the device type (e.g., primary master clock device, slave clock device), the device type can be determined by calculating the hash value of a new device and querying the mapping table. Time division module: a first master clock device sends a time synchronization message to a plurality of second master clock devices and slave clock devices; wherein the time synchronization message includes at least a time granularity; When several slave clock devices receive a time synchronization message, they extract the time granularity carried in the time synchronization message and divide the time partitions based on the time granularity; wherein, multimodal data of the several slave clock devices are collected by deploying multimodal sensors, and the multimodal data includes at least electrophysiological signals, video images, and acquisition timestamps; For example, multimodal sensors include electrode sensors (such as microelectrode arrays (MEA) and patch clamp electrodes) and cameras, and the above multimodal sensors are not drawn in the figure. They are adaptively installed to ensure that corresponding data from multiple slave devices are collected; The steps for dividing time partitions include: Based on the multimodal data, obtaining time information including the full acquisition period of each slave clock device, including at least the acquisition timestamp of the slave clock device; Based on the slave clock device and the corresponding time information, combined with the time granularity of the first master clock device, the multimodal data is divided into multiple time partitions through the sampling frequency range determined by the slave clock device type and the time granularity of the first master clock device. Specifically, it includes: based on the slave clock device and the corresponding time information, determining the sampling frequency range it supports, combining the time granularity of the first master clock device as the benchmark accuracy, time-aligning the multimodal data (such as: uniformly adapting the sampling frequencies of different modal data to the time granularity of the first master clock device), and finally dividing the multimodal data into multiple continuous time partitions according to the acquisition period corresponding to the sampling frequency or an integer multiple of the time granularity of the first master clock device.

[0018] Index building module: creates a corresponding time partition index for each time partition and outputs a prediction index table through a synchronous topology network; The steps for building a synchronous topology network include: Mapping a time partition to a topological node, where the topological node includes a first master clock node, a second master clock node, and a slave clock node; extracting the hop count information and communication link information of the time partition, and constructing a topological edge between the topological nodes of adjacent time partitions, where the topological edge carries at least a link delay parameter; Extracting hop count information for time partitions involves measuring the hop count between nodes using ICMP Traceroute or specialized protocols (e.g., OSPF link-state protocol). If the hop count is less than a certain value (e.g., 3), the two time intervals are marked as adjacent. For example, in a multimodal data acquisition network, if the hop count between a sensor node in time partition A and a sink node in time partition B is 2, and a threshold of 3 is set, the two partitions meet the adjacent condition. A low hop count means that data packets pass through fewer intermediate nodes between the nodes in the two partitions, and the logical distance of data transmission is relatively close, indicating a strong direct or indirect correlation, providing the basis for defining them as adjacent time partitions. Communication link information, including: the timestamps of receiving and sending time synchronization messages, for example: the timestamp of the master sending the time synchronization message and the timestamp of the slave receiving the time synchronization message; At the same time, the clock is configured for the first master clock node, and a time synchronization message is broadcast to the second master clock node and the slave clock node through the topological edge, and a synchronization signal is generated. The second master clock node receives the synchronization signal from the first master clock node and synchronizes in real time. The steps for outputting the prediction index table include: Traverse all nodes in the synchronous topology network to obtain the device type and connection relationship of each node; The connection relationship is obtained through a preset graph model, including: Each device is treated as a target node in a graph model, and the time attribute of each device is used as the attribute of the node; the device types of all nodes are counted to obtain a device type set, including the slave clock device type, the first master clock device type, and the second master clock device type; By connecting nodes to build an edge, the edges of all nodes are counted to obtain an edge set. A graph model of the synchronous topology network is constructed, which is represented by: H = (C, E, R), where C is the device type set, E is the edge set, and R is the node attribute set. Through each time partition, based on the node connection relationship, determine the time partitions where all nodes connected to the nodes in the partition are located, and form a set of associated time partitions; Integrate the device type set, device quantity and associated time partition set into a prediction index table; In addition, in the step of configuring the clock for the first master clock node, a primary link and a redundant link are introduced through the time topology network to connect the first master clock node, the second master clock node and the slave clock node; If the primary link experiences a level 2 or level 3 impact, the slave clock node automatically switches to the redundant link to receive synchronization signals, including: A redundant link is established on the topological edge. The redundant link is used to monitor the synchronization status of the first master node in real time. When it is detected that it is in the third-level synchronization state, the second master node is promoted to the master node through the redundant link and the time synchronization message is rebroadcasted to the slave nodes through the redundant link. The slave clock node receives time synchronization messages through the topology edge, calculates the time deviation and link delay with the current master clock node, and adjusts the local clock to achieve synchronization; In this embodiment, the first master clock typically uses the Linux CLOCK_MONOTONIC clock type as the master clock. This clock is typically based on the system's CPU timestamp counter, has microsecond-level accuracy, and is defined as accurate. The second master clock is a candidate master clock and is typically a high-precision clock source, such as a rubidium atomic clock, a GPS receiver, or a BeiDou receiver. It should be noted that the clocks (time bases) of the primary, secondary, and slave clock devices are different. After determining the types of all devices, the time code corresponding to the device type is also obtained. The time code is essentially a specific field reserved in the synchronization message (such as the Syntalos time synchronization frame). The code value distinguishes the sending device type (such as primary, secondary, and slave) through the code value. The core is to add an identity bit or role field in addition to the timestamp data. When the first or second master clock device sends a synchronization message, it is mandatory to write the corresponding code (such as 00 / 01) in the identity bit and add its own priority level (for example, the first master clock device has the highest priority and is marked as 0). The second master clock device can also receive the synchronization message; When the second master clock is upgraded to the first master clock, the identification bit of the transmitted frame must be immediately changed from 01 to 00, and the priority field must be updated synchronously (for example, from 1 to 0) to ensure synchronous perception of the identity change; the slave clock device only receives synchronization messages and does not actively send frames with identity identification; if confirmation information needs to be sent back (such as delay measurement frame), 10 is written to the identification bit to indicate its own slave clock device identity.

[0019] Status Assessment Module: This module determines the initial time partition of multimodal data based on the prediction index table, introduces a machine learning model, builds a rule engine, analyzes the impact, and predicts the next time partition where a deviation will occur. Determine the initial time partition of the multimodal data based on the prediction index table, including: Based on multimodal data, extract the slave clock device type set M for each time partition; Traverse the prediction index table, for each time partition in the prediction index table, filter out the time partitions without master clocks, and for the remaining time partitions, check whether the device type of each slave clock in the slave clock node set belongs to M; Construct the modal matching function FT (M iNj , M): ; Where M iNj Indicates the type of the jth slave clock device in the i-th time partition, and the value of j is in the range [1, N], where N is a positive integer greater than 1; if F T (M iNj , M)>0, then the corresponding time partition i is used as the initial time partition, and finally an initial time partition set S={T i1 , T i2 ,……,T io}, and o is greater than 1; It should be noted that for each time partition and its corresponding prediction index table, they are first filtered by the master clock type and then modal matching is performed to ensure that the partition meets both the time base and data integrity requirements. That is, an initial time partition needs to have the time base of the first master clock device or the second master clock device. The initial time partition is determined by checking the relationship between the slave clock device type and the slave clock device type set within the time partition to achieve rapid screening.

[0020] Introducing a machine learning model and building a rule engine to predict the next time zone where a deviation will occur, including: Mark the association between nodes and links in a time topology network; a node is a device involved in time synchronization in the time topology network (e.g., a primary master clock, a secondary master clock, and a slave clock), and a link is a path connecting two nodes corresponding to an actual communication line, including but not limited to transmission channels composed of communication media such as wired Ethernet, wireless Wi-Fi, and optical fiber. Each link carries at least the link delay corresponding to the synchronization signal. Setting a set of equations for node synchronization status: a first set of equations generated by analyzing node time deviations and a second set of equations generated by analyzing link delays; obtaining solutions to the first and second sets of equations by obtaining the timestamps of all clocks on the nodes and the receiving and sending timestamps of time synchronization messages; The steps for generating the first set of equations include: Based on the time topology network, the clock of the master clock node (the first master clock or the second master clock) is determined as the reference clock and its clock is marked as T master ; At the same time, determine the set of all nodes (such as slave clocks) participating in synchronization; For each node participating in synchronization, define its time deviation t offest , represents the deviation between the slave clock and the master clock, forming a time deviation sequence. Among them, because the first master clock device has a higher priority than the second master clock device, if the first master clock and the second master clock exist in the master clock node, the first master clock is preferentially selected as the reference; Collect the time information of the slave clock device, including the acquisition timestamp, that is, the time is the current moment recorded by the slave clock node itself; using the time base (such as the time base of the first master clock as the global clock), establish the equation based on the relationship between a slave clock node and the base clock: T = T master +t offest ; By combining all nodes, the above equations are combined to form the first set of equations; The steps for generating the second set of equations include: Based on the time topology network, the communication link set between all nodes is determined, and each link has the characteristics of transmission delay; for each link, the link delay parameter t is defined delay , which represents the time required for the synchronization message to be transmitted from the master clock node to the slave clock node; In a certain link, it is assumed that the master clock node sends a synchronization signal to the slave clock node, and the time T received by the slave clock node is recv Send time T to the master clock node send The relationship is: T recv =T send +t delay ; By combining all links, the above equations are combined to form the second set of equations; By analyzing the time deviation and link delay parameters of each node, the network synchronization performance can be intuitively evaluated. If the time deviation of most nodes is large, it indicates that the network synchronization effect is poor. Subsequent analysis requires adjusting the synchronization strategy or troubleshooting the problem nodes. The feature matrix is obtained by processing the de-splicing of the corresponding nodes, and the feature matrix corresponding to each node is compared and analyzed with the preset feature threshold matrix. When the feature matrix is less than the preset feature threshold matrix, the corresponding position is marked as 1, indicating normal synchronization; otherwise, it is marked as 0, indicating abnormal synchronization, and the corresponding nodes are marked in sequence to form a state matrix; Based on the first set of equations, a time deviation series is obtained; For example: In a continuous time window, the time deviation value between the slave clock node and the master clock node (such as the first master clock) is periodically recorded to form a time deviation sequence Eliminate sudden interference (such as short-term deviation noise caused by transmission delay fluctuations) through sliding average and Kalman filtering methods; Obtain drift parameters based on a time deviation series: Build a drift model: ; Where, is the initial time deviation, is the drift parameter, t is the time, t0 is the reference time, is random noise and is 0 after preprocessing. Because the long-term change of node time deviation usually conforms to the linear model, the drift model is linearly fitted by the least squares method, with time t as the independent variable, and the variation series after preprocessing is As the dependent variable, construct the objective function: , by solving the objective function, the optimal slope can be obtained , which is the drift parameter of the node; Based on the drift parameter and the state matrix, the state matrix and the drift parameter are used as input features of the machine learning model, the deviation classification probability is output, and compared with the preset comparison interval in the model to obtain the corresponding deviation set, specifically including: comparing and analyzing the deviation classification probability with the preset first comparison interval, second comparison interval, and third comparison interval: If the deviation classification probability is in the first comparison interval, it is determined that the temporal topology network circle order is low and the impact of the deviation is relatively minor, and the corresponding node is marked as the first deviation set; at the same time, the deviation classification probability at this time is marked as 1, assigned a first-level character, and 1 is combined with the first-level character to generate a first-level synchronization state; If the deviation classification probability is in the second comparison interval, the temporal topology network is determined to be of medium order, and the impact of the deviation is moderate. The corresponding node is marked as the second deviation set. At the same time, the deviation classification probability at this time is marked as 2, assigned a secondary character, and 2 is combined with the secondary character to generate a secondary synchronization state. If the deviation classification probability is in the third comparison interval, it is determined that the temporal topology network circle order is high and the impact of the deviation is serious, and the corresponding node is marked as the third deviation set; at the same time, the deviation classification probability at this time is marked as 3, assigned a third-level character, and 3 is combined with the third-level character to generate a third-level synchronization state; The first comparison interval is smaller than the second comparison interval, which is smaller than the third comparison interval; Based on the first deviation set, the second deviation set, and the third deviation set, combined with the prediction index table, quickly map to the time partition where the deviation occurs; In this embodiment, the first comparison interval, the second comparison interval, and the third comparison interval are defined as follows: the historical deviation classification probability values are sorted in order, and the three tertile intervals are used for separation. The first tertile is the 33.33% position of the data set, and the second tertile is the 66.67% position of the data set, thereby dividing the first comparison interval, the second comparison interval, and the third comparison interval; Among them, the machine learning model uses the decision tree algorithm; For each initial time partition, the corresponding drift parameters and state matrix are extracted. These drift parameters and state matrix are used as input features of the machine learning model. Each decision tree makes a prediction based on the input features to obtain a prediction result. Based on all the prediction results, the final deviation classification probability is output. The final deviation classification probability is the average of all the prediction results. The data set corresponding to the input features is divided into a training set, a validation set, and a test set; the decision tree model is fitted through the training set, including: initializing the splitting criterion, maximum depth, minimum number of samples required for node splitting, minimum number of samples for leaf nodes, and category weights of the decision tree; the performance is evaluated by calculating indicators such as accuracy, precision, recall, and F1 score through the validation set; at the same time, the optimization range of hyperparameters such as maximum depth, minimum number of sample splits, and minimum number of leaf node samples is determined, and a grid search combined with 5-fold cross-validation is used. With the F1 score as the optimization target, the hyperparameter combinations are traversed, and the parameters with the best performance on the validation set are selected. The model is retrained using the optimal parameters and evaluated on the validation set.

[0021] The time synchronization module, based on the next time deviation partition, marks a time deviation event and synchronously processes the detected time deviation event in the Syntalos synchronization interface; An event dashboard is set up at the top of the Syntalos synchronization interface. If a deviation occurs, the time partition synchronization status is displayed in the form of a heat map. Red marks indicate the third-level synchronization status, orange marks the second-level synchronization status, and green marks the first-level synchronization status. At the same time, corresponding time synchronization processing is performed based on the time deviation time marked with different colors.

[0022] Example 2: The embodiment of the present invention provides a high-precision time synchronization acquisition method for multimodal data based on Syntalos; Figure 2 is a schematic diagram showing the steps of a method for high-precision time synchronization acquisition of multimodal data according to an exemplary embodiment; Figure 2 , the method comprises the following steps: Determine a first master clock device, several second master clock devices, and several slave clock devices, and the first master clock device sends a time synchronization message to the several second master clock devices and the several slave clock devices; wherein the time synchronization message includes at least a time granularity, and the slave clock device is used to collect multimodal data; Divide time partitions based on the time granularity of time synchronization messages; Establish a corresponding time partition index based on each time partition, and output the prediction index table through the synchronization topology network; Determine the initial time partition of multimodal data based on the prediction index table, introduce a machine learning model, build a rule engine, analyze the impact, and predict the next time partition where deviation will occur; A time deviation event is marked based on the next time partition where the deviation occurs, and the detected time deviation event is synchronously processed in the Syntalos synchronization interface.

[0023] Example 3: Embodiment 3 of the present invention provides a computer-readable storage medium having a program stored thereon. When the program is executed by a processor, the Syntalos-based multimodal data high-precision time synchronization acquisition method provided in Embodiment 2 is implemented.

[0024] The readable storage medium may include, but is not limited to, a portable disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0025] In a possible embodiment, the present invention can also be implemented in the form of a program product, which includes program code. When the program product is run on a terminal device, the program code is used to enable the terminal device to execute the steps of the Syntalos-based multimodal data high-precision time synchronization acquisition method in Example 2.

[0026] In the application, the several formulas involved are all calculated by removing the dimensions and taking their numerical values. The formula is a formula obtained by collecting a large amount of data and performing software simulation to obtain the latest real situation. The formula is set by technical personnel in this field according to actual conditions.

[0027] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art will appreciate that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution.

[0028] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, and may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment as needed.

[0029] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. The Syntalos-based multimodal data high-precision time synchronization acquisition system is characterized by: The system comprises: a target determination module, configured to determine a first master clock device, several second master clock devices, and several slave clock devices, wherein the first master clock device sends a time synchronization message to the several second master clock devices and the several slave clock devices; wherein the time synchronization message includes at least a time granularity, and the slave clock devices are configured to collect multimodal data; A time partitioning module, configured to divide time partitions based on the time granularity of the time synchronization message; An index establishment module, configured to establish a corresponding time partition index based on each of the time partitions, and output a prediction index table through a synchronous topology network; A state assessment module is used to determine the initial time partition of the multimodal data based on the prediction index table, introduce a machine learning model, build a rule engine, analyze the impact, and predict the next time partition where the deviation will occur; The time synchronization module is used to mark a time deviation event based on the next time partition where the deviation occurs, and synchronously process the detected time deviation event in the Syntalos synchronization interface.

2. The Syntalos-based multimodal data high-precision time synchronization acquisition system according to claim 1, characterized in that: The determining of the first master clock device, the plurality of second master clock devices, and the plurality of slave clock devices includes: When receiving the power-on signal, immediately obtain the device's MAC address and clock code; A unique hash value is formed based on the MAC address and the clock code, and the device type is determined based on the hash value, including a first master clock device, a second master clock device, and a slave clock device.

3. The Syntalos-based multimodal data high-precision time synchronization acquisition system according to claim 1, characterized in that: The time partitioning includes: Based on the multimodal data, obtaining time information including the full acquisition period of each slave clock device, including at least the acquisition timestamp of the slave clock device; Based on the slave clock device and the corresponding time information, combined with the time granularity of the first master clock device, the multimodal data is divided into multiple time partitions, and the division is performed according to the sampling frequency range determined by the slave clock device type and the time granularity of the first master clock device.

4. The Syntalos-based multimodal data high-precision time synchronization acquisition system according to claim 1, characterized in that: The construction of the synchronous topology network includes: Mapping a time partition into a topological node, where the topological node includes a first master clock node, a second master clock node, and a slave clock node; extracting the hop count information and communication link information of the time partition, and constructing a topological edge between the topological nodes of adjacent time partitions, where the topological edge carries at least a link delay parameter; wherein the communication link information includes at least the reception and transmission timestamps of the time synchronization message; At the same time, the clock is configured for the first master clock node, and the time synchronization message is broadcast to the second master clock node and the slave clock node through the topological edge, and a synchronization signal is generated. The second master clock node receives the synchronization signal of the first master clock node and synchronizes in real time.

5. The Syntalos-based multimodal data high-precision time synchronization acquisition system according to claim 1, characterized in that: The step of outputting the prediction index table includes: Traverse all nodes in the synchronous topology network to obtain the device type and connection relationship of each node, and the connection relationship is obtained through a preset graph model; Through each time partition, based on the node connection relationship, determine the time partitions where all nodes connected to the nodes in the partition are located, and form a set of associated time partitions; The device type set, device quantity, and associated time partition set are integrated into a prediction index table.

6. The Syntalos-based multimodal data high-precision time synchronization acquisition system according to claim 1, characterized in that: The introduction of machine learning models, building a rule engine, analyzing the impact, and predicting the time partition where the next deviation will occur include: Mark the association between nodes and links in a time topology network; a node is a device involved in time synchronization in a time topology network, and a link is a path connecting two nodes corresponding to an actual communication line, including but not limited to transmission channels composed of communication media such as wired Ethernet, wireless Wi-Fi, and optical fiber, and each link carries at least a synchronization signal delay parameter; Setting a set of equations for node synchronization status: a first set of equations generated by analyzing node time deviations and a second set of equations generated by analyzing link delays; obtaining solutions to the first and second sets of equations by obtaining the timestamps of all clocks on the nodes and the receiving and sending timestamps of time synchronization messages; The feature matrix is obtained by performing de-joining processing on the corresponding nodes, and the feature matrix corresponding to each node is compared and analyzed with the preset feature threshold matrix. When the feature matrix is less than the preset feature threshold matrix, the corresponding position is marked as 1, otherwise it is marked as 0, and the state matrix of the corresponding node combination is formed in sequence. Based on the first set of equations, the time deviation sequence is obtained; based on the time deviation sequence, the drift parameter is obtained, and combined with the state matrix, both are used as input features of the machine learning model, and the deviation classification probability is output. It is compared with the preset comparison interval in the model to obtain the corresponding deviation set and impact degree, and combined with the prediction index table, it is quickly mapped to the time partition where the deviation occurs; among them, the impact degree includes first-level impact, second-level impact and third-level impact.

7. The Syntalos-based multimodal data high-precision time synchronization acquisition system according to claim 6, characterized in that: The machine learning model adopts a decision tree algorithm.

8. The Syntalos-based multimodal data high-precision time synchronization acquisition system according to claim 4, characterized in that: The step of configuring a clock for the first master clock node includes: The main link and redundant link are introduced through the time topology network. If the main link has a secondary or tertiary impact, the slave clock node automatically switches to the redundant link to receive the synchronization signal.

9. A high-precision time synchronization acquisition method for multimodal data based on Syntalos, characterized in that: The method comprises: Determine a first master clock device, several second master clock devices, and several slave clock devices, wherein the first master clock device sends a time synchronization message to the several second master clock devices and the several slave clock devices; wherein the time synchronization message includes at least a time granularity, and the slave clock devices are used to collect multimodal data; Dividing time partitions based on the time granularity of the time synchronization message; Establishing a corresponding time partition index based on each of the time partitions, and outputting a prediction index table through a synchronous topology network; Determine the initial time partition of the multimodal data based on the prediction index table, introduce a machine learning model, build a rule engine, analyze the impact, and predict the next time partition where the deviation will occur; A time deviation event is marked based on the next time partition where the deviation occurs, and the detected time deviation event is synchronously processed in the Syntalos synchronization interface.

10. A computer-readable storage medium, characterized in that A program is stored thereon, and when the program is executed by the processor, the method for high-precision time synchronization acquisition of multimodal data based on Syntalos as described in claim 9 is implemented.

Images