A high-precision time-frequency synchronization method based on a time reference station

Abstract

The present invention relates to the technical field of time-frequency synchronization based on a time reference station, and specifically to a high-precision time-frequency synchronization method based on a time reference station. The present invention screens out standard time-frequency synchronization data by combining historical time-frequency mapping data with a mapping time-frequency influence model and a clustering algorithm; then, based on the mapping time-frequency influence model, the abnormal features in the standard time-frequency synchronization data and the first time-frequency synchronization features are analyzed as the second time-frequency synchronization features, and based on the second time-frequency synchronization features, abnormal time-frequency synchronization interference targets causing multipath effects are further screened out, and the third time-frequency synchronization features are calculated and stored; by this method, abnormal data can be screened out and other interference factors can be reduced while ensuring the consistency of the data environment interference based on historical data in the same region; at the same time, the environmental features of the interference targets are determined based on the abnormal data, so as to avoid them when encountering the same environmental features subsequently, further reducing the influence of multipath effects on mapping.

Description

Technical Field

[0001] The present invention relates to the technical field of time-frequency synchronization based on a time reference station, and specifically to a high-precision time-frequency synchronization method based on a time reference station. Background Art

[0002] During the surveying and mapping process, high-precision and high-quality real-time terrain data has always been the key in surveying and mapping work; due to the large scope of terrain surveying and mapping, in order to obtain high-precision and accurate terrain data, it is necessary to achieve accurate time synchronization of multiple surveying and mapping devices. In this process, the Beidou CORS station (Continuously Operating Reference Station) plays an auxiliary role in time synchronization, enabling the surveying and mapping devices to achieve high-precision time synchronization based on the Beidou CORS station, so as to achieve the consistency of surveying and mapping data in terms of time.

[0003] However, there are still some problems during the synchronization of the Beidou CORS station; during the surveying and mapping process, there will inevitably be some buildings or trees, etc., and these plants and artificial facilities will reflect signals when the surveying and mapping devices receive time synchronization signals, causing multipath effects. The result of this problem is that some surveying and mapping devices will be interfered, resulting in the final surveying and mapping results being affected and the quality of surveying and mapping data being reduced.

[0004] In the prior art, some algorithms are used to reduce the multipath effect error, but the influence brought by the multipath effect cannot be completely eliminated; in order to further reduce the influence of the multipath effect on surveying and mapping, a high-precision time-frequency synchronization method based on a time reference station is proposed. Summary of the Invention

[0005] The purpose of the present invention is to provide a high-precision time-frequency synchronization method based on a time reference station. By combining historical time-frequency surveying and mapping data with a surveying and mapping time-frequency influence model and a clustering algorithm, standard time-frequency synchronization data is screened out as a reference standard for subsequent anomaly analysis; then, based on the surveying and mapping time-frequency influence model, the anomaly features in the standard time-frequency synchronization data and the first time-frequency synchronization features are analyzed as the second time-frequency synchronization features. Based on the second time-frequency synchronization features, the abnormal time-frequency synchronization interference targets that cause multipath effects are further screened out and the third time-frequency synchronization features are calculated for storage; through this method, based on historical data in the same region and ensuring the consistency of the data environment interference, abnormal data can be screened out and other interference factors can be reduced; at the same time, the environmental characteristics of the interference targets are determined based on the abnormal data, so as to avoid them when encountering the same environmental characteristics in the future, so as to further reduce the influence of the multipath effect on surveying and mapping.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A high-precision time-frequency synchronization method based on a time reference station, comprising:

[0008] Obtain a target surveying and mapping area, and obtain historical time-frequency surveying and mapping data according to the target surveying and mapping area;

[0009] Establish a surveying and mapping time-frequency influence model, preliminarily analyze time-frequency synchronization interference targets based on historical time-frequency surveying and mapping data, and set standard time-frequency synchronization data according to the time-frequency synchronization interference targets in combination with a clustering algorithm;

[0010] Determine the range of multipath interference targets according to the DBSCAN clustering algorithm, and determine the standard time-frequency synchronization data through the historical time-frequency surveying and mapping data in the minimum range of the multipath interference target range;

[0011] According to the remaining historical time-frequency surveying and mapping data, obtain first time-frequency synchronization data; the time-frequency synchronization interference targets include artificial time-frequency synchronization interference targets and natural time-frequency synchronization interference targets; the number, data type, and data type quantity of the standard time-frequency synchronization data and the first time-frequency synchronization data are the same, and the data types of the standard time-frequency synchronization data and the first time-frequency synchronization data include time synchronization signals, GNSS observation data, and differential data;

[0012] Furthermore, the time-frequency synchronization interference targets are obtained according to the multipath surveying and mapping time-frequency interference analysis model of the surveying and mapping time-frequency influence model;

[0013] The multipath surveying and mapping time-frequency interference analysis model includes a multipath time-frequency interference processing unit, a multipath time-frequency interference feature analysis unit, a multipath time-frequency interference feature synthesis unit, and a multipath time-frequency interference output unit;

[0014] The multipath time-frequency interference processing unit performs image preprocessing on the historical time-frequency surveying and mapping data;

[0015] The multipath time-frequency interference feature analysis unit extracts interference features from the historical time-frequency surveying and mapping data after image preprocessing using multiple models, and marks multipath interference targets; the multiple models include the Yolov5 model, the Faster RCNN model, and the SSD model;

[0016] The multipath time-frequency interference feature synthesis unit performs comprehensive calculations based on the marked multipath interference targets to further screen the multipath interference targets; specifically:

[0017] ;

[0018] Among them, is the multipath interference target score, Let \(P\) be the detection probability of multi - path interference targets, \(yolov\) be the Yolov5 model, \(FRCNN\) be the Faster RCNN model, and \(SSD\) be the SSD model. 、 and be the score weights. ;

[0019] The multi - path time - frequency interference output unit outputs and stores the multi - path interference targets.

[0020] Based on the mapping time - frequency impact model, analyze the standard time - frequency synchronization data to obtain the reference time - frequency synchronization features, analyze the first time - frequency synchronization data to obtain the first time - frequency synchronization features. The mapping time - frequency impact model compares the reference time - frequency synchronization features and the first time - frequency synchronization features, obtains the abnormal time - frequency synchronization features of the first time - frequency synchronization features, and generates the second time - frequency synchronization features; according to the second time - frequency synchronization features corresponding to the historical time - frequency mapping data, obtain the abnormal historical time - frequency mapping data; the abnormal historical time - frequency mapping data is obtained according to the abnormal time - frequency analysis model of the mapping time - frequency impact model; the abnormal time - frequency analysis model includes a time - frequency data segmentation processing unit, a time - frequency data abnormal feature analysis unit, and a time - frequency data abnormal mapping data acquisition unit.

[0021] The time - frequency data segmentation processing unit pre - processes the standard time - frequency synchronization data and the first time - frequency synchronization data, and after pre - processing, performs time - window segmentation of the same length on the standard time - frequency synchronization data and the first time - frequency synchronization data.

[0022] The time - frequency data abnormal feature analysis unit learns the reference time - frequency synchronization features through the standard time - frequency synchronization data and stores them; learns the first time - frequency synchronization features through the first time - frequency synchronization data, compares the reference time - frequency synchronization features and the first time - frequency synchronization features, and obtains the second time - frequency synchronization features.

[0023] The time - frequency data abnormal feature analysis unit includes 1 layer of Attention mechanism, 8 layers of LSTM layers, 2 layers of residual convolution layers, and 2 layers of temporal convolution layers, which are used to extract the reference time - frequency synchronization features and the first time - frequency synchronization features; compare the reference time - frequency synchronization features and the first time - frequency synchronization features based on the cosine similarity algorithm.

[0024] The time - frequency data abnormal mapping data acquisition unit outputs the abnormal historical time - frequency mapping data according to the second time - frequency synchronization features corresponding to the historical time - frequency mapping data.

[0025] According to the second time - frequency synchronization features corresponding to the historical time - frequency mapping data, obtain the abnormal historical time - frequency mapping data.

[0026] Input the abnormal historical time-frequency mapping data into the mapping time-frequency influence model to obtain the abnormal time-frequency synchronization interference targets of the abnormal historical time-frequency mapping data;

[0027] Combine the clustering algorithm with the abnormal time-frequency synchronization interference targets to determine the same abnormal time-frequency synchronization interference targets, calculate the third time-frequency synchronization feature of the multipath effect, and store it; specifically:

[0028] ;

[0029] Among them, is the third time-frequency synchronization feature of the i-th type of the abnormal time-frequency interference target, , n is the total number of types of the abnormal time-frequency interference targets; min is the minimum value; is the distance of the j-th abnormal time-frequency interference target of the i-th type from the mapping instrument, , is the total number of synchronization interference coverage ranges of the i-th type of abnormal time-frequency interference target, is the synchronization interference coverage range of the j-th abnormal time-frequency interference target of the i-th type.

[0030] Compared with the prior art, the beneficial effects of the present invention are:

[0031] 1. The present invention determines the mapping area, and based on multiple models, determines the interference targets in the historical time-frequency mapping data that are likely to cause the multipath effect, and further confirms the multipath interference targets through multiple models; through this method, the position information of the multipath interference targets in the current mapping area can be more accurately determined based on the existing historical time-frequency mapping data, and multiple models are used for confirmation to reduce the problem of single model recognition error; it helps to subsequently screen out the historical time-frequency mapping data with the least multipath interference targets in the mapping area, provides a data basis and data basis for subsequent abnormal time-frequency data analysis, and further reduces the interference of the multipath effect on the mapping data.

[0032] 2. The present invention analyzes the reference time-frequency synchronization features of the current mapping area based on the selected reference time-frequency synchronization features and the first time-frequency synchronization features, and determines the abnormal features in the first time-frequency synchronization features; through this method, the data affected by the multipath interference in the current data can be screened out; at the same time, based on the data in the same area, the environmental interference factors of the data can also be unified, avoiding the influence of other interference factors on the analysis of abnormal data, improving the accuracy of abnormal time-frequency data analysis, providing an abnormal time-frequency synchronization interference target data basis for subsequent screening of abnormal time-frequency data, and further reducing the interference of the multipath effect on the mapping data.

[0033] 3. The present invention identifies abnormal time-frequency synchronization interference targets based on the mapping time-frequency influence model, determines the interference target coverage range under the same abnormal time-frequency synchronization interference target in combination with the clustering algorithm, and determines the third time-frequency synchronization feature in combination with the distance between the mapping instrument and the synchronization interference target. Through this method, it is possible to analyze the environmental characteristics of interference caused by each interference target based on the data analysis of multipath interference in the mapping area, identify and store these characteristics, so as to avoid the mapping instrument from detecting in this environmental characteristic to the greatest extent when performing mapping subsequently, reduce the occurrence of multipath interference, and further reduce the interference of the multipath effect on the mapping data. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 is the flowchart of the method of the present invention;

[0035] Figure 2 is the virtual device diagram of the multipath time-frequency interference processing unit, multipath time-frequency interference feature analysis unit, multipath time-frequency interference feature synthesis unit, multipath time-frequency interference output unit of the multipath mapping time-frequency interference analysis model of the mapping time-frequency influence model of the present invention, and the time-frequency data segmentation processing unit, time-frequency data abnormal feature analysis unit, and time-frequency data abnormal mapping data acquisition unit of the abnormal time-frequency analysis model;

[0036] Figure 3 is the network structure diagram of the feature extraction of the time-frequency data segmentation processing unit of the present invention;

[0037] Figure 4 is the network structure diagram of the residual convolution of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0038] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0039] In the actual mapping process, in order to ensure the time synchronization of real-time mapping data, it is necessary to synchronize the time of each mapping instrument. Currently, the high-precision synchronization method is to perform time synchronization assistance through the Beidou CORS station based on Beidou satellites. However, due to the buildings or trees around the mapping instrument in this process, the multipath effect is likely to occur, reducing the quality of mapping data; although there are already some algorithms in the prior art that can reduce the multipath effect error, the effects of these algorithms are still limited. Therefore, the present invention provides a high-precision time-frequency synchronization method based on a time reference station. Refer to Figure 1 as shown, the technical solution is as follows:

[0040] Obtain a target surveying and mapping area, and obtain historical time-frequency surveying and mapping data based on the target surveying and mapping area;

[0041] Establish a surveying and mapping time-frequency influence model, preliminarily analyze time-frequency synchronization interference targets based on historical time-frequency surveying and mapping data, and set standard time-frequency synchronization data according to the time-frequency synchronization interference targets in combination with a clustering algorithm; obtain first time-frequency synchronization data according to the remaining historical time-frequency surveying and mapping data;

[0042] Analyze the standard time-frequency synchronization data based on the surveying and mapping time-frequency influence model to obtain reference time-frequency synchronization features, analyze the first time-frequency synchronization data to obtain the first time-frequency synchronization features, compare the reference time-frequency synchronization features and the first time-frequency synchronization features by the surveying and mapping time-frequency influence model, obtain abnormal time-frequency synchronization features of the first time-frequency synchronization features, and generate second time-frequency synchronization features;

[0043] Obtain abnormal historical time-frequency surveying and mapping data according to the historical time-frequency surveying and mapping data corresponding to the second time-frequency synchronization features;

[0044] Input the abnormal historical time-frequency surveying and mapping data into the surveying and mapping time-frequency influence model to obtain abnormal time-frequency synchronization interference targets of the abnormal historical time-frequency surveying and mapping data;

[0045] Combine the abnormal time-frequency synchronization interference targets with a clustering algorithm to determine the same abnormal time-frequency synchronization interference targets, calculate the third time-frequency synchronization features of the multipath effect, and store them.

[0046] In order to be able to avoid the interference of the multipath effect on the surveying and mapping data to the greatest extent, based on the historical time-frequency surveying and mapping data of the same surveying and mapping area, identify the targets that cause multipath interference through the surveying and mapping time-frequency influence model, screen out the standard time-frequency synchronization data as the reference data according to these targets. Based on this method, the interference options of environmental data can be reduced to the greatest extent, and at the same time, the interference data features of all data can be unified, which is also convenient for subsequent processing work; at the same time, based on the existing results of the multipath interference algorithm, the standard time-frequency data with the least influence can be screened out, providing data support for subsequent abnormal analysis and processing to further reduce the interference of the multipath effect on the surveying and mapping data.

[0047] Based on the reference time-frequency synchronization features and the first time-frequency synchronization features of the selected standard time-frequency synchronization data, analyze the abnormal data according to the mapping time-frequency influence model, screen out the abnormal time-frequency synchronization interference targets that cause these abnormal data through the analyzed abnormal data, and combine the clustering algorithm and the minimum distance between the mapping instrument and the abnormal time-frequency synchronization interference target to analyze and store its environmental features; screen out the environmental features of the same abnormal time-frequency synchronization interference target from the historical data through this method, and at the same time store the environmental features in a digital form, so that when the same environmental features are encountered during the subsequent mapping process, the mapping instrument can be placed in the same environmental features to the greatest extent to avoid, so as to further reduce the impact of the multipath effect on the mapping work.

[0048] Embodiment 1

[0049] For specific illustration, the identification of multipath effect interference targets is described in combination with the following content:

[0050] Obtain the target mapping area, and obtain historical time-frequency mapping data based on the target mapping area; the historical time-frequency mapping data includes mapping data and time synchronization data;

[0051] Establish a mapping time-frequency influence model, initially analyze the time-frequency synchronization interference target based on the historical time-frequency mapping data, and set the standard time-frequency synchronization data based on the time-frequency synchronization interference target in combination with the clustering algorithm;

[0052] Determine the multipath interference target range according to the DBSCAN clustering algorithm, and determine the standard time-frequency synchronization data through the historical time-frequency mapping data of the minimum range of the multipath interference target range; perform clustering based on the density recognition result of the interference target of the multipath effect, and determine the mapping instrument data with the least multipath effect in the mapping instrument as the standard data for analysis through the smallest range. At the same time, this method considers and combines the existing error elimination algorithms for the multipath effect to improve the mapping data result screened after clustering to the greatest extent. Reduce the impact of the multipath effect on it to improve the data quality of the reference data, and provide an accurate regional reference data basis for further reducing the multipath effect interference in the future;

[0053] According to the remaining historical time-frequency mapping data, obtain the first time-frequency synchronization data; the time synchronization data includes standard time-frequency synchronization data and first time-frequency synchronization data;

[0054] The time-frequency synchronization interference targets include artificial time-frequency synchronization interference targets (such as, but not limited to, buildings, factories, etc.) and natural time-frequency synchronization interference targets (such as, but not limited to, trees, etc.); the number of data, data types, and the number of data types of the standard time-frequency synchronization data and the first time-frequency synchronization data are the same, and the data types of the standard time-frequency synchronization data and the first time-frequency synchronization data include time synchronization signals, GNSS observation data, and differential data; in order to ensure the comprehensiveness of data analysis, considering the multipath effect interference caused by vegetation and artificial buildings existing in the real scene, all possible interference targets are considered for marking and identification. At the same time, determining the standard time-frequency synchronization data based on the target mapping area also takes into account the different noises and multipath effect interferences generated in different regions, enabling unified data processing, reducing the workload, and conducting personalized analysis of the multipath effect in the current region, providing more accurate data for subsequent analysis to further reduce the impact of the multipath effect on the mapping data;

[0055] Further, the time-frequency synchronization interference target is obtained according to the multipath mapping time-frequency interference analysis model of the mapping time-frequency influence model;

[0056] Refer to Figure 2 As shown in the multipath mapping time-frequency interference analysis model, the multipath mapping time-frequency interference analysis model includes a multipath time-frequency interference processing unit, a multipath time-frequency interference feature analysis unit, a multipath time-frequency interference feature synthesis unit, and a multipath time-frequency interference output unit;

[0057] The multipath time-frequency interference processing unit performs image preprocessing on the historical time-frequency mapping data; the image preprocessing process includes histogram image enhancement, Gaussian filtering, Sobel operator edge sharpening, and image grayscale;

[0058] The multipath time-frequency interference feature analysis unit uses multiple models to extract interference features from the historical time-frequency mapping data after image preprocessing and marks the multipath interference targets; the multiple models include the Yolov5 model, the Faster RCNN model, and the SSD model;

[0059] Through the object detection algorithm, the interference targets can be quickly identified based on the existing mapping data, which is also convenient for the data tagging process of the early data training. At the same time, based on the output results of the object detection, the accuracy of the data detection can be further quantified to ensure the data recognition accuracy and the subsequent analysis work, providing more accurate image data to further reduce the impact of the multipath effect on the mapping data;

[0060] In this embodiment, the experimental results of the process of identifying multipath effect interference targets are presented. During the training phase, experts label the known multipath interference targets in the surveying and mapping data to train the model. After training, we selected 5 groups of data that did not participate in the test for simulation testing in combination with the suggestions of relevant technical personnel and experts. The 5 groups of data are different data in the same area. We compared 3 target detection algorithms and conducted multiple tests to obtain the highest recognition rate. The results of the recognition accuracy are as follows:

[0061] Table 1 Multiple recognition accuracies of 5 groups of data for multiple models

[0062]

[0063] As can be seen from Table 1, the recognition accuracies of the three models for the data are high and low. Combining the suggestions of experts, different models have their own advantages in different scenarios. Therefore, after experiments, it was decided to use three models for the joint recognition of multipath effect interference targets; based on the above experimental results, through the comparison and demonstration of the recognition results of multiple models, while ensuring the accurate recognition of multipath interference targets, it is also possible to aggregate the recognition results of each model for comprehensive analysis to ensure the high-precision recognition of multipath interference targets, providing high-precision recognition results of multipath interference targets to further reduce the impact of multipath effects on surveying and mapping data;

[0064] The multipath time-frequency interference feature integration unit performs comprehensive calculations based on the labeled multipath interference targets to further screen the multipath interference targets; specifically:

[0065] ;

[0066] Among them, is the score of the multipath interference target, is the detection probability of the multipath interference target, yolov is the Yolov5 model, FRCNN is the Faster RCNN model, SSD is the SSD model, 、 and are the score weights, ;

[0067] The multipath time-frequency interference output unit outputs and stores the multipath interference targets;

[0068] Based on the target detection algorithm, it can quickly learn and train for known targets. At the same time, it can also analyze and quantify the existing historical surveying and mapping data to enhance the robustness of these surveying and mapping data. Meanwhile, it can also reduce the work of data analysis related to multipath effects, provide a more efficient data basis for subsequent abnormal time-frequency synchronization data analysis work, and provide a fast and stable data basis for further reducing the impact of multipath effects on surveying and mapping data.

[0069] Embodiment 2

[0070] The following content is combined to illustrate how to identify abnormal time-frequency synchronization data and abnormal multi-effect interference targets:

[0071] Based on the surveying and mapping time-frequency influence model, analyze the standard time-frequency synchronization data to obtain the reference time-frequency synchronization features, and analyze the first time-frequency synchronization data to obtain the first time-frequency synchronization features. The surveying and mapping time-frequency influence model compares the reference time-frequency synchronization features and the first time-frequency synchronization features, obtains the abnormal time-frequency synchronization features of the first time-frequency synchronization features, and generates the second time-frequency synchronization features; according to the second time-frequency synchronization features corresponding to the historical time-frequency surveying and mapping data, obtain the abnormal historical time-frequency surveying and mapping data; the abnormal historical time-frequency surveying and mapping data is obtained according to the abnormal time-frequency analysis model; referring to Figure 2 As shown in the surveying and mapping time-frequency influence model, the abnormal time-frequency analysis model includes a time-frequency data segmentation processing unit, a time-frequency data abnormal feature analysis unit, and a time-frequency data abnormal surveying and mapping data acquisition unit;

[0072] The time-frequency data segmentation processing unit preprocesses the standard time-frequency synchronization data and the first time-frequency synchronization data, and after preprocessing, performs time window segmentation of the same length on the standard time-frequency synchronization data and the first time-frequency synchronization data; the preprocessing method includes Fourier transform denoising and Z-score standardization;

[0073] The time-frequency data abnormal feature analysis unit learns and stores the reference time-frequency synchronization features through the standard time-frequency synchronization data; learns the first time-frequency synchronization features through the first time-frequency synchronization data, compares the reference time-frequency synchronization features and the first time-frequency synchronization features, and obtains the second time-frequency synchronization features;

[0074] The time-frequency data abnormal feature analysis unit includes 1 layer of Attention mechanism, 8 layers of LSTM layers, 2 layers of residual convolutional layers, and 2 layers of temporal convolutional layers for extracting the reference time-frequency synchronization features and the first time-frequency synchronization features; compares the reference time-frequency synchronization features and the first time-frequency synchronization features based on the cosine similarity algorithm; referring to Figure 3 shown, the Attention mechanism corresponds to Figure 3 the attention mechanism, and the LSTM layer corresponds toFigure 3 In the long short-term memory layer, the residual convolutional layer uses one-dimensional convolution, referring to Figure 4 As shown, the temporal convolutional layer also uses a one-dimensional convolutional layer, which are two 1*3 temporal convolutional layers respectively;

[0075] In this embodiment, one group of untrained abnormal data is selected in combination with expert experience to compare the recognition accuracy with the existing model. The recognition accuracy results of other abnormal synchronous data are shown in Table 2:

[0076] Table 2 Comparison results of the recognition accuracy of abnormal synchronous data of each model

[0077]

[0078] It can be seen from Table 2 that the model in this paper has a higher recognition accuracy compared with several conventional abnormal time series detection models;

[0079] Based on the Attention mechanism, it can strengthen the marking of abnormal features, increase abnormal data features. The LSTM layer can capture the temporal features in the data to further improve the recognition accuracy. The residual convolutional layer and the temporal convolutional layer capture from the morphology of the abnormal time series data. Through the recognition of data anomalies from three perspectives, the recognition accuracy of abnormal synchronous data can be further improved to achieve the purpose of confirming the multipath effect interference target through abnormal synchronous data, providing a data basis for subsequent environmental feature analysis and further avoiding multipath effect interference;

[0080] The time-frequency data abnormal mapping data acquisition unit outputs abnormal historical time-frequency mapping data according to the second time-frequency synchronization feature corresponding to the historical time-frequency mapping data; the time-frequency data abnormal mapping data acquisition unit includes a fully connected layer and a Softmax classification layer;

[0081] By comparing the time-frequency synchronization data of the standard reference data with other time-frequency synchronization data, extracting the data features based on the method of time series analysis, and labeling abnormal data based on the standard reference data and other time-frequency data synchronization data, this method can minimize the remaining interference factors based on the current regional features and the time features of the collected data, so as to achieve the purpose of enhancing the abnormal data of the multipath interference effect, improving the recognition accuracy of abnormal synchronous data, providing an abnormal data basis for further avoiding multipath effect interference later, and ensuring the accuracy of the interference target data;

[0082] According to the second time-frequency synchronization feature corresponding to the historical time-frequency mapping data, abnormal historical time-frequency mapping data is obtained; specifically, based on expert experience combined with the second time-frequency synchronization feature, the abnormal environmental features in the historical time-frequency mapping data are determined and data segmentation is performed to obtain abnormal historical time-frequency mapping data;

[0083] Input the abnormal historical time-frequency mapping data into the mapping time-frequency influence model to obtain the abnormal time-frequency synchronization interference target of the abnormal historical time-frequency mapping data;

[0084] Combine the clustering algorithm with the abnormal time-frequency synchronization interference target to determine the same abnormal time-frequency synchronization interference target, calculate the third time-frequency synchronization feature of the multipath effect, and store it; specifically:

[0085] ;

[0086] Among them, is the third time-frequency synchronization feature of the i-th type of the abnormal time-frequency interference target, , n is the total number of types of the abnormal time-frequency interference target; min is the minimum value; is the distance between the i-th and j-th abnormal time-frequency interference target and the mapping instrument, , is the total number of synchronization interference coverage ranges of the i-th type of abnormal time-frequency interference target, is the synchronization interference coverage range of the i-th and j-th abnormal time-frequency interference target;

[0087] Based on the environmental characteristics in the mapping data confirmed by the abnormal synchronization data, combine the density clustering algorithm to determine the coverage ranges of these multipath interference targets, and at the same time quantify the data based on the distance and coverage range between the mapping instrument and the multi-target interference; through this method, the actual environmental interference data can be data-indexed, which is convenient for subsequent comparison of the quantization indexes through formula calculation during the mapping process, so as to avoid being placed in the same environment and minimize the interference of the multipath effect to the greatest extent;

[0088] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirits of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A high-precision time-frequency synchronization method based on a time reference station, characterized in that Including: Obtain a target surveying and mapping area, and obtain historical time-frequency surveying and mapping data based on the target surveying and mapping area; Establish a surveying and mapping time-frequency influence model, preliminarily analyze time-frequency synchronization interference targets based on historical time-frequency surveying and mapping data, set standard time-frequency synchronization data according to the time-frequency synchronization interference targets in combination with a clustering algorithm; obtain first time-frequency synchronization data according to the remaining historical time-frequency surveying and mapping data; Analyze the standard time-frequency synchronization data based on the surveying and mapping time-frequency influence model to obtain reference time-frequency synchronization features, analyze the first time-frequency synchronization data to obtain first time-frequency synchronization features, the surveying and mapping time-frequency influence model compares the reference time-frequency synchronization features and the first time-frequency synchronization features, obtain abnormal time-frequency synchronization features of the first time-frequency synchronization features, and generate second time-frequency synchronization features; Obtain abnormal historical time-frequency surveying and mapping data according to the historical time-frequency surveying and mapping data corresponding to the second time-frequency synchronization features; Input the abnormal historical time-frequency surveying and mapping data into the surveying and mapping time-frequency influence model, and obtain abnormal time-frequency synchronization interference targets of the abnormal historical time-frequency surveying and mapping data, including: The time-frequency synchronization interference target is obtained according to the multipath surveying and mapping time-frequency interference analysis model of the surveying and mapping time-frequency influence model; the multipath surveying and mapping time-frequency interference analysis model includes a multipath time-frequency interference processing unit, a multipath time-frequency interference feature analysis unit, a multipath time-frequency interference feature synthesis unit, and a multipath time-frequency interference output unit; The multipath time-frequency interference processing unit performs image preprocessing on the historical time-frequency surveying and mapping data; The multipath time-frequency interference feature analysis unit extracts interference features from the historical time-frequency surveying and mapping data after image preprocessing using multiple models, and marks multipath interference targets; the multipath time-frequency interference feature synthesis unit performs comprehensive calculation based on the marked multipath interference targets to further screen the multipath interference targets; The multipath time-frequency interference output unit outputs and stores the multipath interference targets; Combine the abnormal time-frequency synchronization interference targets with a clustering algorithm to determine the same abnormal time-frequency synchronization interference targets, calculate the third time-frequency synchronization features of the multipath effect, and store them.

2. The high-precision time-frequency synchronization method based on a time reference station according to claim 1, wherein The time-frequency synchronization interference targets include artificial time-frequency synchronization interference targets and natural time-frequency synchronization interference targets; the number of data, data type, and data type quantity of the standard time-frequency synchronization data and the first time-frequency synchronization data are the same, and the data types of the standard time-frequency synchronization data and the first time-frequency synchronization data include time synchronization signals, GNSS observation data, and differential data.

3. A high-precision time-frequency synchronization method based on a time reference station according to claim 1, characterized in that The multipath time-frequency interference feature analysis unit extracts interference features from the historical time-frequency surveying and mapping data after image preprocessing using multiple models, including: the multiple models include the Yolov5 model, the Faster RCNN model, and the SSD model.

4. A high-precision time-frequency synchronization method based on a time reference station according to claim 1, characterized in that The multipath time-frequency interference feature synthesis unit performs comprehensive calculation based on the marked multipath interference targets to further screen the multipath interference targets, including: ； Among them, is the multi-path interference target score, is the multi-path interference target detection probability, yolov is the Yolov5 model, FRCNN is the Faster RCNN model, and SSD is the SSD model, , and are the score weights, .

5. A high-precision time-frequency synchronization method based on a time reference station according to claim 1, characterized in that Setting standard time-frequency synchronization data according to the time-frequency synchronization interference target in combination with the clustering algorithm includes: determining the multipath interference target range according to the DBSCAN clustering algorithm, and determining the standard time-frequency synchronization data through the historical time-frequency mapping data of the minimum range of the multipath interference target range.

6. A high-precision time-frequency synchronization method based on a time reference station according to claim 1, characterized in that Analyzing the standard time-frequency synchronization data based on the mapping time-frequency influence model to obtain the reference time-frequency synchronization features, analyzing the first time-frequency synchronization data to obtain the first time-frequency synchronization features, comparing the reference time-frequency synchronization features and the first time-frequency synchronization features by the mapping time-frequency influence model, obtaining the abnormal time-frequency synchronization features of the first time-frequency synchronization features, generating the second time-frequency synchronization features, and obtaining abnormal historical time-frequency mapping data according to the historical time-frequency mapping data corresponding to the second time-frequency synchronization features; the abnormal historical time-frequency mapping data is obtained according to the abnormal time-frequency analysis model of the mapping time-frequency influence model; the abnormal time-frequency analysis model includes a time-frequency data segmentation processing unit, a time-frequency data abnormal feature analysis unit, and a time-frequency data abnormal mapping data acquisition unit; The time-frequency data segmentation processing unit preprocesses the standard time-frequency synchronization data and the first time-frequency synchronization data, and after preprocessing, performs time window segmentation of the same length on the standard time-frequency synchronization data and the first time-frequency synchronization data. The time-frequency data abnormal feature analysis unit learns and stores the reference time-frequency synchronization features through the standard time-frequency synchronization data; learns the first time-frequency synchronization features through the first time-frequency synchronization data, compares the reference time-frequency synchronization features and the first time-frequency synchronization features, and obtains the second time-frequency synchronization features. The time-frequency data abnormal mapping data acquisition unit outputs abnormal historical time-frequency mapping data according to the historical time-frequency mapping data corresponding to the second time-frequency synchronization features.

7. A high-precision time-frequency synchronization method based on a time reference station according to claim 6, characterized in that, The time-frequency data abnormal feature analysis unit learns and stores the reference time-frequency synchronization features through the standard time-frequency synchronization data; learns the first time-frequency synchronization features through the first time-frequency synchronization data, compares the reference time-frequency synchronization features and the first time-frequency synchronization features, and obtains the second time-frequency synchronization features; the time-frequency data abnormal feature analysis unit includes 1 layer of Attention mechanism, 8 layers of LSTM layers, 2 layers of residual convolution layers, and 2 layers of temporal convolution layers for extracting the reference time-frequency synchronization features and the first time-frequency synchronization features. Comparing the reference time-frequency synchronization features and the first time-frequency synchronization features based on the cosine similarity algorithm.

8. A high-precision time-frequency synchronization method based on a time reference station according to claim 1, characterized in that Combining the clustering algorithm with the abnormal time-frequency synchronization interference target, determining the same abnormal time-frequency synchronization interference target, calculating the third time-frequency synchronization feature of the multipath effect, and storing it, including: obtaining the synchronization interference coverage range of the abnormal time-frequency synchronization interference target corresponding to the second time-frequency synchronization features according to the DBSCAN clustering algorithm, and calculating the third time-frequency synchronization feature based on the synchronization interference coverage range: ； Among them, is the third time-frequency synchronization feature of the i-th abnormal time-frequency interference target, , where n is the total number of types of the abnormal time-frequency interference targets; min is the minimum value; is the distance of the j-th target of the i-th abnormal time-frequency interference from the mapping instrument, , is the total number of synchronization interference coverage ranges of the i-th abnormal time-frequency interference target, is the synchronization interference coverage range of the j-th target of the i-th abnormal time-frequency interference target.

Images