Cycle slip detection and repair method, system and medium

Through preprocessing of satellite carrier phase observations and neural network analysis, the statistical characteristics of the perimeter jump anomaly data were extracted, and the problem of low perimeter jump detection accuracy in the existing technology was solved, efficient repair of perimeter jump and improved positioning Accuracy.

CN119024379APending Publication Date: 2024-11-26GUANGDONG ELECTRIC POWER SCI RES INST ENERGY TECH CO LTD
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202411136113.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-19
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

The prior art has low detection accuracy for the cycle jump, which leads to inability to effectively repair, affecting the carrier phase measurement accuracy and positioning accuracy.

Method used

By obtaining the satellite carrier phase observations for preprocessing, the statistical characteristics of the perimeter jump anomaly data are extracted, and the perimeter jump points are determined by using neural networks for repair, including second-order differential processing, sampling, sliding windows and attention mechanisms.

Benefits of technology

It improves the accuracy and robustness of the weekly jump detection, and can complete repairs without the need for inter-station or inter-star differences, which enhances positioning accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119024379A_ABST
    Figure CN119024379A_ABST
Patent Text Reader

Abstract

The invention discloses a cycle slip detection and repair method and system and a medium, and the method comprises the steps: obtaining all satellite carrier phase observation values over the years, carrying out the preprocessing of all satellite carrier phase observation values, and obtaining cycle slip abnormal data according to a preprocessing result; sampling the multi-dimensional original time sequence corresponding to the cycle slip abnormal data according to a preset sampling interval, generating an updated time sequence according to norm differences corresponding to a plurality of time windows obtained by sampling, and calculating statistical characteristics of each time window of the updated time sequence; and splicing the statistical features, inputting a splicing result into a preset neural network for solving, determining a plurality of cycle slip points according to a solving result, and repairing each cycle slip point. According to the method, detection and repair of the cycle slip can be completed without inter-station difference or inter-satellite difference, and the robustness is high. The technical problem that the cycle slip cannot be repaired due to low detection precision of the cycle slip in the prior art is solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of cycle slip detection and repair, and in particular to a cycle slip detection and repair method, system and medium. Background Art

[0002] With the continuous development of navigation technology, people's requirements for positioning accuracy are constantly increasing. In the process of satellite navigation, the preprocessing of satellite data is particularly important for the reliability and accuracy of positioning. Among them, correct and efficient cycle slip detection and repair is also a key issue in high-precision GNSS data processing. Generally speaking, the causes of cycle slips are generally divided into several aspects: the first is related to the satellite. When the satellite elevation angle is too low or the ionosphere is too active, the signal-to-noise ratio of the satellite signal is low; the second is related to the signal propagation environment. When there are obstructions such as tall buildings or mountains in the environment, the signal will be interrupted, resulting in a phase jump; the third is related to the receiver. When the receiver fails, such as when the internal clock is wrong, the satellite signal is processed incorrectly; the fourth is related to external interference. When the oscillator of the satellite equipment fails, or the receiver fails due to external conditions, the signal error occurs. If the cycle slip is not detected and repaired, it will inevitably affect the accuracy of the carrier phase measurement, thereby affecting the positioning accuracy.

[0003] Therefore, there are many methods for detecting cycle slips, such as the high-order difference method and wavelet analysis method under single-frequency carrier, the geometric distance combination method and MW combination method under dual-frequency carrier, and the additional introduction of INS to assist cycle slip detection, etc. However, the above methods have low detection accuracy for cycle slips, resulting in the inability to repair cycle slips. Summary of the invention

[0004] The present invention provides a cycle slip detection and repair method, system and medium, which solves the technical problem that the prior art has low cycle slip detection accuracy, resulting in the inability to repair the cycle slip.

[0005] A first aspect of the present invention provides a cycle slip detection and repair method, comprising:

[0006] Obtaining satellite carrier phase observation values ​​of all previous years, preprocessing all the satellite carrier phase observation values, and obtaining cycle slip anomaly data according to the preprocessing results;

[0007] Sampling the multidimensional original time series corresponding to the cycle slip abnormal data according to a preset sampling interval, generating an updated time series according to the norm differences corresponding to the multiple time windows obtained by sampling, and calculating the statistical characteristics of each time window of the updated time series;

[0008] The statistical features are spliced ​​together, the splicing results are input into a preset neural network for solving, a plurality of cycle slip points are determined according to the solution results, and each of the cycle slip points is repaired.

[0009] Optionally, the step of obtaining satellite carrier phase observation values ​​of all previous years, preprocessing all the satellite carrier phase observation values, and obtaining cycle slip anomaly data according to the preprocessing results includes:

[0010] Get satellite carrier phase observations for all epochs;

[0011] Calculating the elevation angle of each satellite carrier phase observation value;

[0012] Determining whether each elevation angle is less than a preset elevation angle threshold;

[0013] If it is less than a preset elevation angle threshold, the satellite carrier phase observation value corresponding to the elevation angle is removed;

[0014] If it is greater than or equal to the preset elevation angle threshold, the satellite carrier phase observation value corresponding to the elevation angle is retained, and the retained satellite carrier phase observation value is processed by second-order difference processing. The satellite carrier phase observation value corresponding to the second-order difference processing result is used to generate cycle slip anomaly data.

[0015] Optionally, the step of sampling the multidimensional original time series corresponding to the cycle slip abnormal data according to a preset sampling interval, generating an updated time series according to the norm differences corresponding to the multiple time windows obtained by sampling, and calculating the statistical features of each time window of the updated time series includes:

[0016] Sampling the multidimensional original time series corresponding to the cycle slip abnormal data according to a preset sampling interval to generate multiple time windows;

[0017] Calculating the norm difference of each of the time windows, and determining an update time series according to the norm difference result;

[0018] Sliding the update time sequence according to a preset step length to generate multiple time windows of the update time sequence;

[0019] Calculate the statistical characteristics of each time window of the update time series.

[0020] Optionally, the step of calculating the norm difference of each of the time windows and determining to update the time series according to the norm difference result includes:

[0021] Calculating the norm of each of the time windows;

[0022] Calculating the norm difference between two adjacent norms;

[0023] The norm difference results are used to generate an updated time series.

[0024] Optionally, the step of splicing the statistical features, inputting the splicing result into a preset neural network for solving, determining a plurality of cycle slip points according to the solution result, and repairing each of the cycle slip points includes:

[0025] Connect the statistical features to generate a target time series;

[0026] Connect each row of each two-dimensional matrix of the target time series head to tail to generate a vector set;

[0027] Each vector of the vector set is input into a preset neural network for solving, a plurality of cycle slip points are determined according to the solving results, and each of the cycle slip points is repaired.

[0028] Optionally, the step of inputting each vector of the vector set into a preset neural network for solving, determining a plurality of cycle slip points according to the solution results, and repairing each of the cycle slip points comprises:

[0029] Inputting each vector of the vector set into an encoder of a preset neural network respectively, and determining a hidden state of each vector at each time step;

[0030] Introduce the attention mechanism to calculate the similarity between the current time step and the remaining time steps and generate the context vector;

[0031] The hidden state and context vector of each time step are respectively input into the decoder of the neural network, and the target vector of each time step is output;

[0032] Calculating the reconstruction error of each of the target vectors to generate a plurality of error values;

[0033] Sorting all the error values ​​according to a preset order, and determining a plurality of cycle slip points according to the sorting result;

[0034] Each of the cycle slip points is repaired.

[0035] Optionally, the step of sorting all the error values ​​in a preset order and determining a plurality of cycle slip points according to the sorting result includes:

[0036] Sorting all the error values ​​according to a preset order;

[0037] A preset number of error values ​​are obtained according to the sorting result, and the preset number of error values ​​are used as a plurality of cycle slip points.

[0038] Optionally, the step of repairing each of the cycle slip points includes:

[0039] Determine whether the epoch value corresponding to each of the week slip points is greater than the epoch value at the previous moment;

[0040] If it is greater than the epoch value of the previous moment, all satellite carrier phase observation values ​​after the cycle slip point are reduced by a preset absolute value to repair the cycle slip point;

[0041] If it is less than or equal to the epoch value of the previous moment, all satellite carrier phase observation values ​​after the cycle slip point are increased by the preset absolute value to repair the cycle slip point.

[0042] A second aspect of the present invention provides a cycle slip detection and repair system, comprising:

[0043] An acquisition module is used to acquire satellite carrier phase observation values ​​of all previous years, preprocess all the satellite carrier phase observation values, and acquire cycle slip anomaly data according to the preprocessing results;

[0044] A calculation module, used to sample the multidimensional original time series corresponding to the cycle slip abnormal data according to a preset sampling interval, generate an updated time series according to the norm differences corresponding to the multiple time windows obtained by sampling, and calculate the statistical characteristics of each time window of the updated time series;

[0045] The repair module is used to splice the statistical features, input the splicing results into a preset neural network for solving, determine multiple cycle slip points according to the solution results, and repair each of the cycle slip points.

[0046] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed, implements the cycle slip detection and repair method as described in any one of the above items.

[0047] It can be seen from the above technical solutions that the present invention has the following advantages:

[0048] The present invention pre-processes satellite carrier phase observation values, determines cycle slip anomaly data, extracts statistical features of cycle slip anomaly data, and inputs neural network to extract each cycle slip point, and repairs all cycle slip points. The present invention can complete the detection and repair of cycle slips without inter-station or inter-satellite difference, and has strong robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0050] Figure 1 A flowchart of the steps of a cycle slip detection and repair method provided in Embodiment 1 of the present invention;

[0051] Figure 2 A process diagram with rich original time series features provided in the first embodiment of the present invention;

[0052] Figure 3 A schematic diagram of a two-dimensional matrix splicing method provided in the first embodiment of the present invention;

[0053] Figure 4 A schematic diagram of a long short-term memory autoencoder network provided in Embodiment 1 of the present invention;

[0054] Figure 5 A schematic diagram of a long short-term memory autoencoder network introducing an attention mechanism provided in Embodiment 1 of the present invention;

[0055] Figure 6 A structural block diagram of a cycle slip detection and repair system provided in Embodiment 2 of the present invention. DETAILED DESCRIPTION

[0056] The embodiments of the present invention provide a cycle slip detection and repair method, system and medium, which are used to solve the technical problem that the prior art has low cycle slip detection accuracy, resulting in the inability to repair the cycle slip.

[0057] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0058] Embodiment 1

[0059] See also Figures 1 to 5 , Figure 1 A flowchart of the steps of a cycle slip detection and repair method provided in Embodiment 1 of the present invention.

[0060] The present invention provides a cycle slip detection and repair method, comprising the following steps:

[0061] Step 101: Obtain all satellite carrier phase observation values ​​of all previous years, preprocess all satellite carrier phase observation values, and obtain cycle slip anomaly data based on the preprocessing results.

[0062] Optionally, step 101 includes the following steps S11-S15:

[0063] S11, obtaining satellite carrier phase observation values ​​of all epochs;

[0064] S12, calculating the elevation angle of each satellite carrier phase observation value;

[0065] S13, determining whether each elevation angle is less than a preset elevation angle threshold;

[0066] S14. If the elevation angle is less than the preset elevation angle threshold, the satellite carrier phase observation value corresponding to the elevation angle is removed;

[0067] S15. If it is greater than or equal to the preset elevation angle threshold, the satellite carrier phase observation value corresponding to the elevation angle is retained, and the retained satellite carrier phase observation value is processed by second-order difference processing. The satellite carrier phase observation value corresponding to the second-order difference processing result is used to generate cycle slip anomaly data.

[0068] It should be noted that the satellite carrier phase observation values ​​of all epochs are obtained. The elevation angles of the satellites in all time periods are calculated, and satellites with too low elevation angles are removed. The preset elevation angle threshold z is set based on experience to be 10° to 20°. If it is lower than this value, the satellite signal is considered unavailable. The satellite carrier phase observation values ​​above the preset elevation angle threshold z are retained, and the retained satellite carrier phase observation values ​​are processed by second-order difference, so that the abnormal value part, that is, the error part of the cycle slip data is exposed, that is, the cycle slip abnormal data is obtained. Specifically, since the second-order difference is actually an estimate of the rate of change of the differential carrier, this method can emphasize the signal characteristics that change rapidly in a short period of time. Relatively small but fast-changing errors or anomalies will become more obvious in the second-order difference results, making it easier to identify and eliminate them. The cycle slip signal is such an error.

[0069] Specifically, the second-order difference processing is performed on the satellite's carrier phase observation value. For example, if the carrier phase observation value of a certain satellite is known to be The expression for a difference is:

[0070]

[0071] The expression of its quadratic difference is:

[0072]

[0073] Among them, k represents a certain moment, D i Represents the carrier difference during the first differential, D i ' represents the carrier difference during the second difference.

[0074] Step 102: Sample the multidimensional original time series corresponding to the cycle slip abnormal data according to a preset sampling interval, generate an updated time series according to the norm differences corresponding to the multiple time windows obtained by sampling, and calculate the statistical characteristics of each time window of the updated time series.

[0075] Optionally, step 102 includes the following steps S21-S24:

[0076] S21, sampling the multidimensional original time series corresponding to the cycle slip abnormal data according to a preset interval to generate multiple time windows;

[0077] S22, calculating the norm difference of each time window, and determining the update time series according to the norm difference result;

[0078] S23, sliding the update time sequence according to a preset step length to generate multiple time windows of the update time sequence;

[0079] S24. Calculate and update the statistical characteristics of each time window of the time series.

[0080] It should be noted that, assuming that the sampling interval of the satellite signal is C, the k-dimensional original time series T is T = (S1, S2, ... S C ),in, Representation characteristics: For the present invention, it is the satellite carrier phase observation value of k satellites at time during the observation period.

[0081] In the specific implementation, a time window of length b is used to slide along the original time series along the time axis with a step size of b / 2. Therefore, for a time series of length C, we can get window.

[0082] In order to maintain the correlation between data, the first phase value in each window is set to And for each window of data, Make all data count from zero.

[0083] For each variable j in each window, d derived features are constructed. Consider the derived features with d=2: norm (NOR) and norm difference (DON). Among them, the NOR feature describes the change of sample values ​​in the window within a certain period of time, while the DON feature describes the change of sparse relationship.

[0084] Specifically, the norm difference of each time window is calculated, and the update time series T′= <G1,G2,...,G C >.

[0085] Among them, G i The dimension of is d*k, where d=2.

[0086] Set a new length of f, and slide along T′ with a step length of f / 2, and we can also get A window is opened to continue the characteristic calculation of statistical properties.

[0087] For each window, e = 8 statistical features are calculated for each derived feature in T' calculated in the first step. They are: mean (MEA), minimum (MIN), maximum (MAX), first quintile (20Q), second quintile (40Q), third quintile (60Q), fourth quintile (80Q), standard deviation (STD).

[0088] Among them, quintiles can be used to describe the distribution of data, such as sparsity, central tendency, etc.; mean value features can be used to fill missing values ​​or feature scaling; minimum and maximum features can be used for feature selection, and distinguishing features can be selected based on the difference between the minimum and maximum values; standard deviation features can be used to check the distribution of data and select features, etc. The purpose of calculating these statistical features is to further capture the degree to which derived features change over time.

[0089] Optionally, step S22 includes the following steps S221-S223:

[0090] Calculate the norm of each time window;

[0091] Calculate the norm difference between two adjacent norms;

[0092] The norm difference results are used to generate an updated time series.

[0093] It should be noted that, assuming that the window wd i Represents the interval [t i ,t i+b-1 ], containing the following b vectors from the original time series. The window wd i The norm of the j-th feature of NOR is defined as the Euclidean length of the vector:

[0094]

[0095] Window wd i The jth DON is defined as window wd i The norm of the previous window wd i-1 The difference in the norm of :

[0096] DON j (wd i )=norm j (wd i )-norm j (wdi-1 )(4)

[0097] After calculation, we obtain a new time series:

[0098] T′= <G1,G2,...,G C > (5)

[0099] Among them, G i The dimension of is d*k, where d=2.

[0100] Step 103: splice the statistical features, input the splicing results into a preset neural network for solving, determine multiple cycle slip points according to the solution results, and repair each cycle slip point.

[0101] Optionally, step 103 includes the following steps S31-S33:

[0102] S31, splicing various statistical features to generate a target time series;

[0103] S32, concatenating each row of each two-dimensional matrix of the target time series from beginning to end to generate a vector set;

[0104] S33, inputting each vector of the vector set into a preset neural network for solving, determining a plurality of cycle slip points according to the solution results, and repairing each cycle slip point.

[0105] It should be noted that all statistical features are spliced ​​together to obtain a new time series (i.e., target time series) of T″= C >, where H i The dimension is 8d*k, d=2.

[0106] Specifically, Figure 2 As shown in the figure, the overall process of feature enrichment of the original time series is as follows Figure 2 As shown, due to H i is a two-dimensional matrix, so H cannot be directly i Enter the long short-term memory neural network. i By splicing each row at the beginning and the end, we can get the vector L of (d*e*k,1) i , for specific splicing methods, please refer to Figure 3 shown.

[0107] T″= C >Reconstruct as U = (L1, L2, ... L C ​​), pass all subsequences in U into LSTM-AE (Long Short-term Memory Neural Network-Auto Encoder), the Chinese name is Long Short-term Memory Auto Encoder Network for training. The schematic diagram of the Long Short-term Memory Auto Encoder Network is shown in the figure. Figure 4 As shown in Figure 2, the relu activation function is used. To prevent overfitting, the dropout rate of the autoencoder is set to 0.1.

[0108] See also Figure 5 As shown in the figure, the autoencoder can reconstruct normal data very well, but it is often difficult to reconstruct abnormal data, that is, outliers. Therefore, when outliers are reconstructed by the autoencoder, the reconstruction error is often large, and due to the introduction of the attention mechanism, the outliers will be more prominent, and outliers (i.e., week jumps) can be judged based on this feature.

[0109] The long short-term memory autoencoder network analyzes the statistical features in the rich time series: the first quintile (20Q), the second quintile (40Q), the third quintile (60Q), and the fourth quintile (80Q), and based on the correlation between the previous five epochs and the current epoch, it analyzes the outliers at the corresponding moments in the outliers and considers them as week jump points.

[0110] The long short-term memory autoencoder network further analyzes the statistical characteristics: mean value (MEA), minimum value (MIN), maximum value (MAX), standard deviation (STD), and determines the preset absolute value of the cycle slip δ C The method is similar to the table lookup method. Each δ C Corresponding to a statistical feature, the size of the cycle slip is determined and output through the network, so as to repair the cycle slip.

[0111] Optionally, step S32 includes the following steps S321-S326:

[0112] S321, inputting each vector of the vector set into an encoder of a preset neural network respectively, and determining that each vector is in a hidden state at each time step;

[0113] S322, introduce the attention mechanism, calculate the similarity between the current time step and the remaining time steps, and generate a context vector;

[0114] S323, using the hidden state and context vector of each time step to input the decoder of the neural network respectively, and outputting the target vector of each time step;

[0115] S324, calculating the reconstruction error of each target vector to generate multiple error values;

[0116] S325, sorting all error values ​​according to a preset order, and determining a plurality of cycle slip points according to the sorting result;

[0117] S326. Repair each week jump point.

[0118] It should be noted that the hidden layers and vector reconstruction in the encoder and decoder follow the traditional LSTM structure. i For example, the time step of the LSTM network is set to 5, which means that the current data is considered to be related to the time state of the previous 5 epochs. When obtaining the information at the previous moment, the weight parameter vector W in the unit is used. i After the dot product, the hidden layer is obtained is copied into the decoder, i.e. In the decoding phase, each hidden state Generate the reconstructed vector L′ k Hidden Layer Depends on the next moment The hidden layer and the next moment L′ I+1 The reconstruction vector of .

[0119] At the same time, an attention mechanism is introduced at each time step of the encoding and decoding ends of the network to calculate the similarity between the current time step and all encoding time steps, thereby obtaining a weighted context vector c t , which is used as the input of the decoding end at the current time step to generate the prediction result of the next time step.

[0120]

[0121] In the formula, h i represents the hidden state of the i-th time step; h t represents the hidden state of the current time step; c t represents the context vector of the current time step; T represents the sequence length; α t,i Represents the attention weight of the current time step to the i-th time step; Indicates the score for calculating the attention weight. v, W h , W i and b t Represents the parameters of the attention mechanism; where v is a scalar, W h and W i is a matrix, b t is a bias term, which together determine how to calculate the attention score

[0122] By introducing the attention mechanism, we can not only improve the accuracy of model detection, but also make the model more interpretable, and obtain a neural network model that can align the input sequence and the output sequence, which is convenient for subsequent analysis of how the model processes the input sequence.

[0123] In the LSTM-AE network, the mapping parts of the encoder and decoder can be defined as φ and ψ, as shown in the following formula:

[0124] Encoder:φR m →R n (9)

[0125] Decoder:ψR n →R m (10)

[0126] The encoder will input H i is mapped into an intermediate n-dimensional vector F, and the decoder maps F to i The same dimension Usually, m>n. The process of network learning is to determine the appropriate φ and ψ, and continuously update the weight matrix W in LSTM through the back propagation algorithm. i , K i , so that the input L i With output Minimize the difference between them.

[0127] Autoencoders can reconstruct normal data very well, but it is often difficult to reconstruct abnormal data, i.e. outliers. Therefore, when outliers are reconstructed by autoencoders, the reconstruction error is often large, and due to the introduction of the attention mechanism, outliers will be more prominent, and outliers can be judged based on this feature.

[0128] Using the following formula Distance is used to calculate the reconstruction error for each vector in the enriched time series:

[0129]

[0130] For the vectors in the time series, sort them from large to small based on this error value, so that the outlier vector is sorted higher. Get multiple week slip points according to the sorting results.

[0131] Optionally, step S325 includes the following steps S3251-S3252:

[0132] S3251, sorting all error values ​​according to a preset order;

[0133] S3252. Obtain a preset number of error values ​​according to the sorting result, and use the preset number of error values ​​as multiple cycle slip points.

[0134] It should be noted that for the vectors in the time series, the error value is sorted from large to small, so that the outlier vector is sorted higher. The first α% (e.g., 5%) is initially set as an outlier, and the subsequent network can adaptively learn and adjust the threshold.

[0135] The network analyzes the statistical features in the rich time series: the first quintile (20Q), the second quintile (40Q), the third quintile (60Q), and the fourth quintile (80Q), and based on the correlation between the previous five epochs and the current epoch, it analyzes the outliers at the corresponding moments in the outliers and considers them as week jump points.

[0136] Optionally, step S326 includes the following steps S3261-S3263:

[0137] S3261, judging whether the epoch value corresponding to each week jump point is greater than the epoch value at the previous moment;

[0138] S3262: If it is greater than the epoch value of the previous moment, all satellite carrier phase observation values ​​after the week slip point are reduced by a preset absolute value to repair the week slip point;

[0139] S3263. If it is less than or equal to the epoch value of the previous moment, all satellite carrier phase observation values ​​after the week slip point are increased by a preset absolute value to repair the week slip point.

[0140] In the specific implementation, starting from the detected week slip point, for it, check whether the epoch value of the week slip point is greater than the epoch value of the previous moment. If so, all satellite carrier phase values ​​after the week slip point are subtracted by the preset absolute value δ C , otherwise all satellite carrier phase values ​​after the cycle slip point are added with the preset absolute value δ C , to achieve cycle slip repair.

[0141] The present invention realizes unsupervised learning based on long short-term memory neural network and autoencoder. When changing the signal, since it only involves the change of the input layer, it is only necessary to fine-tune the internal parameters of the network. There is no need to train from scratch. Instead, it can start from the basic model that has been trained and continuously update the parameters in the network model.

[0142] The present invention utilizes the advantages of time series anomaly detection of autoencoders, determines the sample set related to satellite signals, designs the network structure and activation function suitable for cycle slip detection, constructs and trains a deep neural network model based on the long short-term memory neural autoencoder network, and introduces the attention mechanism, which can judge and process satellite signals, thereby realizing the detection and repair of cycle slips.

[0143] Embodiment 2

[0144] See also Figure 6 , Figure 6 A structural block diagram of a cycle slip detection and repair system provided in Embodiment 2 of the present invention.

[0145] The present invention provides a cycle slip detection and repair system, comprising:

[0146] The acquisition module 201 is used to acquire all satellite carrier phase observation values ​​of all previous years, pre-process all satellite carrier phase observation values, and acquire cycle slip anomaly data according to the pre-processing results;

[0147] The calculation module 202 is used to sample the multidimensional original time series corresponding to the cycle slip abnormal data according to a preset sampling interval, generate an updated time series according to the norm difference corresponding to the multiple time windows obtained by sampling, and calculate the statistical characteristics of each time window of the updated time series;

[0148] The repair module 203 is used to splice the statistical features, input the splicing results into a preset neural network for solving, determine multiple cycle slip points according to the solution results, and repair each cycle slip point.

[0149] Optionally, the acquisition module 201 includes:

[0150] The acquisition submodule is used to obtain the satellite carrier phase observation values ​​of all epochs;

[0151] The elevation submodule is used to calculate the elevation angle of each satellite carrier phase observation value;

[0152] The elevation angle threshold submodule is used to determine whether each elevation angle is less than a preset elevation angle threshold;

[0153] A removal submodule is used to remove the satellite carrier phase observation value corresponding to the elevation angle if it is less than a preset elevation angle threshold;

[0154] The retention submodule is used to retain the satellite carrier phase observation value corresponding to the elevation angle if it is greater than or equal to the preset elevation angle threshold, and perform second-order difference processing on the retained satellite carrier phase observation value, and use the satellite carrier phase observation value corresponding to the second-order difference processing result to generate cycle slip anomaly data.

[0155] Optionally, the calculation module 202 includes:

[0156] A sampling submodule is used to sample the multi-dimensional original time series corresponding to the cycle slip abnormal data according to a preset sampling interval to generate multiple time windows;

[0157] The norm difference submodule is used to calculate the norm difference of each time window and determine the update time series based on the norm difference result;

[0158] A sliding submodule is used to slide the update time series according to a preset step length to generate multiple time windows of the update time series;

[0159] The statistical feature submodule is used to calculate the statistical features of each time window of the update time series.

[0160] Optionally, the norm difference submodule includes:

[0161] The norm submodule is used to calculate the norm of each time window;

[0162] A norm difference calculation submodule is used to calculate the norm difference between two adjacent norms;

[0163] The update time series submodule is used to generate an update time series using each norm difference result.

[0164] Optionally, the repair module 203 includes:

[0165] The splicing submodule is used to splice various statistical features to generate the target time series;

[0166] The vector set submodule is used to concatenate the head and tail of each row of each two-dimensional matrix of the target time series to generate a vector set;

[0167] The repair submodule is used to input each vector of the vector set into a preset neural network for solving, determine multiple cycle slip points according to the solution results, and repair each cycle slip point.

[0168] Optionally, the repair submodule includes:

[0169] A hidden state submodule, used to input each vector of the vector set into an encoder of a preset neural network to determine the hidden state of each vector at each time step;

[0170] The similarity submodule is used to introduce the attention mechanism, calculate the similarity between the current time step and the remaining time steps, and generate the context vector;

[0171] An output submodule is used to input the hidden state and context vector of each time step into the decoder of the neural network respectively, and output the target vector of each time step;

[0172] The error value submodule is used to calculate the reconstruction error of each target vector and generate multiple error values;

[0173] A cycle slip point submodule is used to sort all error values ​​according to a preset order and determine multiple cycle slip points according to the sorting results;

[0174] The cycle slip repair submodule is used to repair each cycle slip point.

[0175] Optionally, the cycle slip point submodule includes:

[0176] A sorting submodule, used to sort all error values ​​according to a preset order;

[0177] The multiple cycle slip point submodules are used to obtain a preset number of error values ​​according to the sorting results, and use the preset number of error values ​​as multiple cycle slip points.

[0178] Optionally, the cycle slip repair submodule includes:

[0179] The judgment submodule is used to judge whether the epoch value corresponding to each week jump point is greater than the epoch value of the previous moment;

[0180] A reduction submodule is used to reduce all satellite carrier phase observation values ​​after the week slip point by a preset absolute value if the value is greater than the epoch value of the previous moment, so as to repair the week slip point;

[0181] The adding submodule is used to increase all satellite carrier phase observation values ​​after the week slip point by a preset absolute value if it is less than or equal to the epoch value of the previous moment, so as to repair the week slip point.

[0182] Embodiment 3

[0183] Embodiment 3 of the present invention provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed, a cycle slip detection and repair method as described in any embodiment of the present invention is implemented.

[0184] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0185] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0186] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0187] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0188] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0189] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A cycle slip detection and repair method, It is characterized in that include: Obtaining satellite carrier phase observation values ​​of all previous years, preprocessing all the satellite carrier phase observation values, and obtaining cycle slip anomaly data according to the preprocessing results; Sampling the multidimensional original time series corresponding to the cycle slip abnormal data according to a preset sampling interval, generating an updated time series according to the norm differences corresponding to the multiple time windows obtained by sampling, and calculating the statistical characteristics of each time window of the updated time series; The statistical features are spliced ​​together, the splicing results are input into a preset neural network for solving, a plurality of cycle slip points are determined according to the solution results, and each of the cycle slip points is repaired.

2. The cycle slip detection and repair method according to claim 1, It is characterized in that The step of obtaining all satellite carrier phase observation values ​​over the years, preprocessing all the satellite carrier phase observation values, and obtaining cycle slip anomaly data according to the preprocessing results includes: Get satellite carrier phase observations for all epochs; Calculating the elevation angle of each satellite carrier phase observation value; Determining whether each elevation angle is less than a preset elevation angle threshold; If it is less than a preset elevation angle threshold, the satellite carrier phase observation value corresponding to the elevation angle is removed; If it is greater than or equal to the preset elevation angle threshold, the satellite carrier phase observation value corresponding to the elevation angle is retained, and the retained satellite carrier phase observation value is processed by second-order difference processing. The satellite carrier phase observation value corresponding to the second-order difference processing result is used to generate cycle slip anomaly data.

3. The cycle slip detection and repair method according to claim 1, It is characterized in that The step of sampling the multidimensional original time series corresponding to the cycle slip abnormal data according to a preset sampling interval, generating an updated time series according to the norm differences corresponding to the multiple time windows obtained by sampling, and calculating the statistical features of each time window of the updated time series includes: Sampling the multidimensional original time series corresponding to the cycle slip abnormal data according to a preset sampling interval to generate multiple time windows; Calculating the norm difference of each of the time windows, and determining an update time series according to the norm difference result; Sliding the update time sequence according to a preset step length to generate multiple time windows of the update time sequence; Calculate the statistical characteristics of each time window of the update time series.

4. The cycle slip detection and repair method according to claim 3, It is characterized in that The step of calculating the norm difference of each time window and determining to update the time series according to the norm difference result comprises: Calculating the norm of each of the time windows; Calculating the norm difference between two adjacent norms; The norm difference results are used to generate an updated time series.

5. The cycle slip detection and repair method according to claim 1, It is characterized in that The step of splicing the statistical features, inputting the splicing results into a preset neural network for solving, determining a plurality of cycle slip points according to the solution results, and repairing each of the cycle slip points comprises: Connect the statistical features to generate a target time series; Connect each row of each two-dimensional matrix of the target time series head to tail to generate a vector set; Each vector of the vector set is input into a preset neural network for solving, a plurality of cycle slip points are determined according to the solving results, and each of the cycle slip points is repaired.

6. The cycle slip detection and repair method according to claim 5, It is characterized in that The step of inputting each vector of the vector set into a preset neural network for solving, determining a plurality of cycle slip points according to the solution results, and repairing each of the cycle slip points comprises: Inputting each vector of the vector set into an encoder of a preset neural network respectively, and determining a hidden state of each vector at each time step; Introduce the attention mechanism to calculate the similarity between the current time step and the remaining time steps and generate the context vector; The hidden state and context vector of each time step are respectively input into the decoder of the neural network, and the target vector of each time step is output; Calculating the reconstruction error of each of the target vectors to generate a plurality of error values; Sorting all the error values ​​according to a preset order, and determining a plurality of cycle slip points according to the sorting result; Each of the cycle slip points is repaired.

7. The cycle slip detection and repair method according to claim 6, It is characterized in that The step of sorting all the error values ​​in a preset order and determining a plurality of cycle slip points according to the sorting result comprises: Sorting all the error values ​​according to a preset order; A preset number of error values ​​are obtained according to the sorting result, and the preset number of error values ​​are used as a plurality of cycle slip points.

8. The cycle slip detection and repair method according to claim 6, It is characterized in that The step of repairing each of the cycle slip points comprises: Determine whether the epoch value corresponding to each of the week slip points is greater than the epoch value at the previous moment; If it is greater than the epoch value of the previous moment, all satellite carrier phase observation values ​​after the cycle slip point are reduced by a preset absolute value to repair the cycle slip point; If it is less than or equal to the epoch value of the previous moment, all satellite carrier phase observation values ​​after the cycle slip point are increased by the preset absolute value to repair the cycle slip point.

9. A cycle slip detection and repair system, It is characterized in that include: An acquisition module is used to acquire satellite carrier phase observation values ​​of all previous years, preprocess all the satellite carrier phase observation values, and acquire cycle slip anomaly data according to the preprocessing results; A calculation module, used to sample the multidimensional original time series corresponding to the cycle slip abnormal data according to a preset sampling interval, generate an updated time series according to the norm differences corresponding to the multiple time windows obtained by sampling, and calculate the statistical characteristics of each time window of the updated time series; The repair module is used to splice the statistical features, input the splicing results into a preset neural network for solving, determine multiple cycle slip points according to the solution results, and repair each of the cycle slip points.

10. A computer-readable storage medium, It is characterized in that A computer program is stored thereon, and when the computer program is executed, the cycle slip detection and repair method according to any one of claims 1 to 8 is implemented.

Citation Information

Cited By

  • Sensor jump point detection method, device, equipment and medium

    CN119936926A

  • Carrier phase hopping detection method and device and storage medium

    CN120742359A