Carrier phase jump detection method and device, and storage medium

Abstract

The application discloses a carrier phase jump detection method and device and a storage medium. The method comprises the following steps: acquiring a satellite observation sequence of a current epoch collected for any satellite; extracting first feature data of the satellite observation sequence, wherein the first feature data comprises a carrier phase original value; inputting the first feature data into a preset model which is trained, wherein the preset model comprises a convolutional neural network and a long short-term memory network; extracting local mutation features in a time sequence of the feature data through the convolutional neural network; performing time sequence modeling on the local mutation features through the long short-term memory network to output time-dependent features; and acquiring a jump type of the carrier phase of the satellite observation sequence and a jump probability corresponding to the jump type which are output by the preset model according to the local mutation features and the time-dependent features. The above scheme can automatically learn and adapt to jump modes in different scenarios, predict future jump trends in advance, and provide a more efficient early warning mechanism.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of satellite navigation signal processing, in particular to a carrier phase jump detection method and device and a storage medium. BACKGROUND

[0002] In the GNSS (Global Navigation Satellite System) receiving process, the carrier phase observation value is the key to realize high-precision positioning. However, in the environment of strong interference, shielding, multipath, etc., the carrier phase signal is prone to jump phenomenon, especially the cycle slip and half cycle slip, which seriously affects the accuracy of subsequent differential solution or precise point positioning. The commonly used methods such as TEQC or differential detection based method cannot respond to sudden changes in time or accurately classify multiple types of jumps, and the support for multi-star multi-frequency complex observation scene is limited. In addition, the existing methods are mostly static threshold rules, and lack the dynamic judgment ability of jump trend and data quality. SUMMARY

[0003] The purpose of the embodiments of the present application is to provide a carrier phase jump detection method, device and storage medium, to solve the technical problem that the prior art cannot accurately identify the carrier phase jump.

[0004] In order to achieve the above purpose, the first aspect of the present application provides a carrier phase jump detection method, the method comprising:

[0005] Obtaining a satellite observation sequence of a current epoch collected for any satellite;

[0006] Extracting first feature data of the satellite observation sequence, the first feature data including carrier phase raw values;

[0007] Inputting the first feature data into a preset model trained, the preset model including a convolutional neural network and a long short-term memory network, wherein the local mutation features in the time sequence of the feature data are extracted by the convolutional neural network, and the local mutation features are time series modeled by the long short-term memory network to output time-dependent features;

[0008] Obtaining the jump type of the carrier phase of the satellite observation sequence and the jump probability corresponding to the jump type output by the preset model according to the local mutation features and the time-dependent features.

[0009] In the embodiments of the present application, the method comprises: acquiring a plurality of satellite observation sequences collected by a plurality of satellites at different epochs; extracting second feature data of the first historical satellite observation sequence; dividing the second feature data into a plurality of sub-feature data based on a preset sliding window; acquiring a historical jump type corresponding to each sub-feature data; and training the plurality of sub-feature data and the corresponding historical jump type on a preset model to obtain a trained preset model.

[0010] In the embodiments of the present application, the preset model further comprises a task branch module, the task branch module comprises a first output head and a second output head, and the obtaining of the jump type of the carrier phase of the satellite observation sequence and the jump probability corresponding to the jump type output by the preset model according to the local mutation feature and the time-dependent feature comprises: fusing the local mutation feature and the time-dependent feature based on a fully connected layer of the first output head to generate a raw score of each preset jump type; converting the raw score of each preset jump type into a corresponding probability distribution based on a softmax function of the first output head to determine the jump type of the carrier phase of the satellite observation sequence; fusing the local mutation feature and the time-dependent feature based on a fully connected layer of the second output head to generate a raw score of each preset jump type; and converting the raw score of each preset jump type into a probability value based on a sigmoid function of the second output head to determine the jump probability of the jump type.

[0011] In the embodiments of the present application, the method further comprises: determining a confidence degree of the jump type based on the preset model; and performing weighted fusion according to the confidence degree, the jump probability and the first feature data to output an observation quality score of the first feature data; and generating a corresponding quality level and prompt information according to the observation quality score and the jump type.

[0012] In the embodiments of the present application, the first feature data is at least one of a carrier-to-noise ratio, a satellite elevation angle, a satellite relative speed, a receiver platform speed, a first-order phase difference and a second-order phase difference.

[0013] In the embodiments of the present application, the observation quality is calculated according to the following formula (1):

[0014] (1)

[0015] wherein, is an observation quality score, , , , , is a corresponding weight value, is a confidence degree corresponding to the jump type, is a jump probability, is a carrier-to-noise ratio, is referred to as satellite elevation angle, is referred to as sliding variance of phase first-order difference.

[0016] In the embodiment of the present application, the method further comprises: compressing the trained preset model to obtain a compressed preset model; converting the compressed preset model into data in ONNX format; converting and quantifying the data in ONNX format to export data in TFLite format; and deploying the data in TFLite format on an embedded platform with Cortex-A7 architecture, so that the output data of the preset model is structured data.

[0017] The second aspect of the present application provides a carrier phase jump detection device, comprising:

[0018] a memory configured to store instructions;

[0019] a processor configured to call the instructions from the memory and capable of implementing the carrier phase jump detection method according to the above when executing the instructions.

[0020] The third aspect of the present application provides a GNSS receiver comprising the carrier phase jump detection device according to the above.

[0021] The fourth aspect of the present application provides a machine-readable storage medium having instructions stored thereon for causing a machine to execute the carrier phase jump detection method according to the above.

[0022] Through the above technical solution, unlike the traditional rule-based jump detection method, the preset model built by the convolutional neural network and the long short-term memory network can automatically learn and adapt to the jump pattern in different scenarios, and can identify the tiny half-cycle jump of ±0.5 cycles. At the same time, the future jump trend is predicted in advance, a more efficient early warning mechanism is provided, and the intelligence and response ability of the system are enhanced.

[0023] Other features and advantages of the embodiments of the present application will be described in detail in the following specific implementation part. BRIEF DESCRIPTION OF DRAWINGS

[0024] The accompanying drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification, and are used together with the following specific implementation to explain the embodiments of the present application, but do not constitute a limitation on the embodiments of the present application. In the drawings:

[0025] Figure 1 schematically shows a flowchart of a carrier phase jump detection method according to an embodiment of the present application;

[0026] Figure 2A schematic diagram of a deployment system structure according to an embodiment of the present application is shown schematically;

[0027] Figure 3 A structural block diagram of a carrier phase jump detection device according to an embodiment of the present application is shown schematically.

[0028] Figure 4 A structural schematic diagram of a computer device according to an embodiment of the present application is shown schematically. DETAILED DESCRIPTION

[0029] To make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. It should be understood that the specific implementation manners described herein are merely used to explain and illustrate the embodiments of the present application, and should not be used to limit the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative effort fall within the scope of protection of the present application.

[0030] It should be noted that if the embodiments of the present application involve directional indications (such as up, down, left, right, front, back, etc.), the directional indications are merely used to explain the relative positional relationship, movement condition, etc. between components in a certain posture (as shown in the drawings), and if the certain posture changes, the directional indications also change accordingly.

[0031] In addition, if the embodiments of the present application involve descriptions such as “first”, “second”, etc., the descriptions of “first”, “second”, etc. are merely for description purposes, and should not be understood as indicating or implying the relative importance of the indicated technical features or implicitly indicating the number of the indicated technical features. Therefore, the features limited by “first”, “second” can explicitly or implicitly include at least one of the features. In addition, the technical solutions of the various embodiments can be combined with each other, but it must be based on the fact that a person of ordinary skill in the art can realize the combination, and when the combination of the technical solutions contradicts each other or cannot be realized, it should be considered that the combination of the technical solutions does not exist, and is also not within the scope of protection claimed by the present application.

[0032] Figure 1 A flowchart of a carrier phase jump detection method according to an embodiment of the present application is shown schematically. As shown in Figure 1 The present application provides a carrier phase jump detection method, which can include the following steps.

[0033] S102, acquiring a satellite observation sequence of a current epoch collected for any satellite.

[0034] S104, extract first feature data of the satellite observation sequence, the first feature data including carrier phase raw values.

[0035] S106, input the first feature data into a preset model trained, the preset model including a convolutional neural network and a long short-term memory network, wherein the convolutional neural network is used to extract local mutation features in time series of the feature data, and the long short-term memory network is used to model the local mutation features in time series to output time-dependent features.

[0036] S108, obtain a jump type of the carrier phase of the satellite observation sequence and a jump probability corresponding to the jump type output by the preset model according to the local mutation features and the time-dependent features.

[0037] It can be understood that the satellite observation sequence refers to a data sequence obtained by observing a satellite at a specific time point (epoch). These data usually include information such as the position, velocity, and acceleration of the satellite, which are used for subsequent orbit calculation and positioning analysis. Through the epoch, the observation data at different time points can be associated to perform accurate orbit calculation and positioning analysis. The satellite observation sequence can be obtained by a GNSS receiver. The first feature data of the satellite observation sequence includes carrier phase raw values. In order to realize accurate identification and trend warning of the jump row in the GNSS carrier phase observation value, the preset model trained in the embodiment combines the convolutional neural network (CNN) to extract the local mutation features in the input feature data, and combines the long short-term memory network (LSTM) to model the time-dependent relationship. The ability can be applied to multi-type jump detection tasks including whole-week jump, half-week jump, and normal observation. The preset model can output the jump type of the carrier phase of the satellite observation sequence and the jump probability corresponding to the jump type according to the local mutation features and the time-dependent features. In a specific embodiment, the first feature data further includes, but is not limited to, carrier-to-noise ratio, satellite elevation angle, satellite relative speed, receiver platform speed, first-order phase difference, second-order phase difference, satellite number, sliding variance, sliding average, and other local fluctuation statistics. The first feature data can be extracted by a sliding window mechanism according to the epoch to form a fixed-length time sequence for model processing. Unlike traditional rule-based jump detection methods, the preset model built by CNN-LSTM can automatically learn and adapt to jump patterns in different scenarios, and can identify ±0.5-week such small half-week jumps, at the same time, predict the future jump trend in advance, provide a more efficient warning mechanism, and enhance the intelligence and response ability of the system.

[0038] In the embodiments of the present application, the method comprises: acquiring a plurality of satellite first historical satellite observation sequences collected at different epochs; extracting second feature data of the first historical satellite observation sequences; dividing the second feature data into a plurality of sub-feature data based on a preset sliding window; acquiring a historical jump type corresponding to each sub-feature data; and training the plurality of sub-feature data and the corresponding historical jump type on a preset model to obtain a trained preset model.

[0039] Specifically, the second feature data can be extracted by the sliding window mechanism according to the epoch, the second feature data is divided into a plurality of sub-feature data, and a fixed-length time sequence is formed for model processing. By acquiring a historical jump type corresponding to each sub-feature data, the plurality of sub-feature data and the corresponding historical jump type are trained on a preset model to obtain a trained preset model. It can be understood that the training data of the preset model comes from the measured GNSS data and the simulated jump data set, which can cover a variety of scenarios, including but not limited to open area, sheltered area, urban canyon, etc. environment, multi-system (such as GPS / BD) satellite trajectory, normal, whole-week jump, half-week jump are covered. The sliding window method is used to train the time sequence samples, and each window generates a set of sub-feature data and label pairs. The jump type in the data is based on artificial annotation or comparative error analysis to improve the accuracy of the label.

[0040] Specifically, the convolutional neural network includes one or two one-dimensional convolutional layers, each layer uses a 3-5 length convolution kernel to extract local mutation features, and is matched with a normalization (BatchNorm) and an activation function (ReLU) to improve the expression ability, and a Dropout is selected to suppress overfitting. The long short-term memory network can adopt a bidirectional LSTM structure, and the number of hidden units can be 32-64. It can be used to model the bidirectional dependence of time sequence features, capture the evolution trend before and after the jump, and the output can be connected to the task branch module.

[0041] In the embodiments of the present application, the preset model further includes a task branch module, the task branch module includes a first output head and a second output head, and the obtaining of the jump type of the carrier phase of the satellite observation sequence and the jump probability corresponding to the jump type output by the preset model according to the local mutation feature and the time-dependent feature includes: fusing the local mutation feature and the time-dependent feature based on a fully connected layer of the first output head to generate a raw score of each preset jump type; converting the raw score of each preset jump type into a corresponding probability distribution based on a softmax function of the first output head to determine the jump type of the carrier phase of the satellite observation sequence; fusing the local mutation feature and the time-dependent feature based on a fully connected layer of the second output head to generate a raw score of each preset jump type; and converting the raw score of each preset jump type into a probability value based on a sigmoid function of the second output head to determine the jump probability of the jump type. The two output heads of the task branch module can be combined by using a weighted loss function, and the optimization direction takes into account the classification accuracy and the prediction confidence. Optionally, the standard sequence connection method can be used between the modules to form an end-to-end trainable structure.

[0042] In the embodiments of the present application, the model training is implemented by using a deep learning framework (PyTorch), the optimization method can use cross-entropy loss for classification of the model, and the binary cross-entropy loss is combined to predict the probability of the jump trend. The optimizer can use Adam, and the initial learning rate is recommended to be 0.001. Moreover, the EarlyStopping and learning rate adjustment mechanism can be used to improve the convergence efficiency, the batch size is 64-128, the training rounds are 30-50, and the data size is determined.

[0043] In the embodiments of the present application, the method further includes: determining the confidence of the jump type based on the preset model; performing weighted fusion according to the confidence, the jump probability and the first feature data to output an observation quality score of the satellite observation sequence; and generating a corresponding quality level and prompt information according to the observation quality score and the jump type.

[0044] Specifically, the preset model can output the jump type and the jump probability thereof, and the corresponding confidence. The observation quality of the satellite observation sequence at the current epoch can also be scored in combination with the external physical observation features of the first feature data. The first feature data further includes whether the current channel is continuously tracked, a phase-locked state and a GNSS system type. The GNSS system type includes GPS, BDS, GAL and GLO, which can be used as optional input data.

[0045] In the embodiments of the present application, the weighted fusion according to the confidence, the jump probability and the first feature data forms a linear scoring function, and the observation quality is calculated according to the following formula (1):

[0046] (1)

[0047] wherein, denotes the observed mass fraction, , , , , denote the corresponding weight value respectively, denotes the confidence corresponding to the jump type, denotes the jump probability, denotes the carrier-to-noise ratio, denotes the satellite elevation angle, denotes the sliding variance of the first-order difference of the phase. All weight values can be statically configured or fitted based on historical data. The score value ranges from 0 to 1, and the higher the value represents the better observation quality.

[0048] In another embodiment, the first feature data can be constituted into a joint feature vector as the quality evaluation input of the current epoch. A lightweight neural network score is used, which is a 1-layer or 2-layer perception machine structure (MLP) with no more than 16 nodes, an activation function of ReLU function combined with Sigmoid function, an input of the joint feature vector, and an output of Q value. The score supports running on the TFLite platform and is integrated with the preset model for deployment. The score has low delay, and the overall calculation time is less than 5 ms per satellite.

[0049] Further, according to the observation quality fraction and the jump type, the corresponding quality level and prompt information are generated, as shown in the following Table 1.

[0050] Table 1 Observation quality fraction, quality level and prompt information

[0051]

[0052] The above table information can be used for quality level judgment, and can be directly used for satellite ambiguity management, observation selection, state weight adjustment and other strategy fusion, to improve the selection ability of observation data instructions and significantly reduce the calculation failure problem caused by jump errors.

[0053] In the embodiments of the present application, the method further comprises: compressing the trained preset model to obtain a compressed preset model; converting the compressed preset model into ONNX format data; converting and quantizing the ONNX format data to export TFLite format data; deploying the TFLite format data on a Cortex-A7 architecture embedded platform, so that the output data of the preset model is structured data.

[0054] Specifically, to achieve the real-time running capability of the preset model in a low-power GNSS terminal device, it is ensured that the preset model can complete the jump recognition and observation evaluation functions without relying on external computing resources. After the model training is completed, model compression is performed using technologies such as quantization, pruning, and structure reparameterization. The model is converted into the ONNX format and then into the TensorFlow Lite (TFLite) format to support low-power device deployment.

[0055] Specifically, the quantization of the model converts the original 32-bit floating-point weights and activation values in the model into 8-bit integer (INT8) representations. Using the Post-training quantization scheme, the model can be compressed to 1 / 4 of the original model size without additional training. At the same time, the computational path integrity is maintained, and the structure pruning of the model is performed on the insensitive channels in the CNN and LSTM layers. The pruning strategy is based on the weight amplitude or channel importance score, and the pruned model retains more than 85% of the progress with a 40% improvement in inference speed. The convolutional layer and the activation layer are fused into a single layer of computational expression, the redundant activation function is deleted, the linear layer is merged, the memory access frequency is reduced, the pipeline efficiency is improved, and structure reparameterization and fusion optimization are achieved. In this way, the size of the compressed preset model can be controlled between 100 and 150 KB, which is suitable for SoC deployment.

[0056] To adapt to the embedded end inference framework, the preset model uses the following format conversion process.

[0057] Step 1: The model is trained and exported in the ONNX (Open Neural Network Exchange) format. Specifically, the trained model is converted into ONNX using PyTorch / TensorFlow tools, and the model topology structure and weight file are preserved as a cross-platform model intermediate expression form.

[0058] Step 2: Use the TensorFlow official converter pnnx-tf→TfLiteConverter tool chain to perform INT8 quantization mapping during the conversion process, and export the TFLite FLatBuffer file. The converted model has high portability and is easy to load and execute on the embedded end. The final output model file is in the.tflite format, and the file size is generally not more than 200 KB.

[0059] Further, the model is run on an embedded platform with Cortex-A7 architecture, and the lightweight local deployment of the Cortex-A7 platform. The deployment platform targets the Cortex-A7 embedded processor architecture, with constraints such as running memory ≤ 128 MB, no GPU / NPU coprocessor, etc. Real-time inference and jump detection are realized, and the post-processing process or delayed calculation is not relied on. The output structured data format is JSON, CSV, binary format data, which is called by the RTK / PPP solving engine or remote monitoring platform.

[0060] The deployment system structure is shown in Figure 2 The TensorFlow Lite for Microcontrollers (TFLM) inference engine is used to optimize memory usage, support running only with static allocated memory, and be compatible with Cortex-A series bare metal or Linux embedded systems such as buildroot. During runtime, the model initialization stage loads the tflite file and maps it to memory. The inference engine processes the satellite observation sequence input of N satellites per epoch per second, and outputs the identification result of jump and the prediction probability value of jump trend for calling by the upper module. In the above scheme, the single needle model execution delay is less than 30 ms, the average power consumption increment is less than 10 mV, and the inference can be completed during the idle period of the GNSS master, without affecting the core business of the receiver.

[0061] The interface mode of the system supports connecting with the master navigation system through UART, SPI, I2C or network interface, and outputting the inference result in a structured manner. The output structured data format is JSON, CSV, binary format data, which is called by the RTK / PPP solving engine or remote monitoring platform or participates in the observation selection decision, so that it can be easily connected with different platforms and adapt to different GNSS application scenarios.

[0062] Compared with the traditional method, the above technical scheme, based on the deep learning model, can automatically learn the jump pattern in complex environment and improve the identification ability of micro-jumps such as half-cycle jump. A low-complexity model pruning and deployment scheme is proposed, which realizes the lightweight local deployment of CNN-LSTM model on Cortex-A7 platform for the first time. The observation quality scoring mechanism can dynamically evaluate the GNSS observation reliability, realize the filter weight adjustment, ambiguity automatic re-estimation and abnormal observation elimination. Moreover, the system as a whole supports modularization and cross-platform integration, and is suitable for various GNSS solving systems and receiver hardware.

[0063] The aforementioned intelligent assessment scheme and system for GNSS carrier phase hop detection and observation data quality based on a deep learning model are applicable to the identification and assessment of various types of hop anomalies in GNSS receivers. They are particularly suitable for intelligent navigation systems that improve solution stability and data reliability in complex scenarios such as interference, obstruction, and loss of lock. The method is deployed on low-power embedded platforms such as Cortex-A7, enabling real-time hop detection and data cleaning at the edge, effectively improving the stability and accuracy of RTK, PPP, and other solution links. It features lightweight deployment and rapid response.

[0064] The GNSS carrier phase jump detection process based on the CNN-LSTM deep learning model can automatically identify integer jumps, half-cycle jumps, and other abnormal states. It combines auxiliary features to construct a multi-task learning model and predict future jump trends. Based on the model output, an observation quality assessment engine is built to achieve Q-value scoring for each satellite per epoch and output quality level and processing suggestions, realizing low-latency and highly robust intelligent data processing.

[0065] Figure 1 This is a flowchart illustrating a carrier phase transition detection method in one embodiment. It should be understood that, although... Figure 1 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 1 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0066] Figure 3 The diagram schematically illustrates a structural block diagram of a carrier phase transition detection device according to an embodiment of this application. Figure 3 As shown in the figure, this application provides a carrier phase transition detection device, which may include:

[0067] The memory is configured to store instructions;

[0068] The processor is configured to retrieve instructions from memory and, when executing instructions, to implement the aforementioned carrier phase transition detection method.

[0069] Specifically, in this embodiment of the application, the processor can be configured as follows:

[0070] The satellite observation sequence of the current epoch collected by any satellite is acquired; first feature data of the satellite observation sequence is extracted, the first feature data including carrier phase original values; the first feature data is input into a preset model trained, the preset model including a convolutional neural network and a long short-term memory network, wherein the local mutation features in the time sequence of the feature data are extracted through the convolutional neural network, and the time sequence modeling is performed on the local mutation features through the long short-term memory network to output time-dependent features; a jump type of the carrier phase of the satellite observation sequence and a jump probability corresponding to the jump type output by the preset model according to the local mutation features and the time-dependent features are acquired.

[0071] In the embodiments of the present application, the processor can be further configured to:

[0072] The first historical satellite observation sequence collected by a plurality of satellites at different epochs is acquired; second feature data of the first historical satellite observation sequence is extracted; the second feature data is divided into a plurality of sub-feature data based on a preset sliding window; a historical jump type corresponding to each sub-feature data is acquired; the plurality of sub-feature data and the corresponding historical jump type are trained on the preset model to obtain the trained preset model.

[0073] In the embodiments of the present application, the processor can be further configured to:

[0074] The preset model further includes a task branch module, the task branch module including a first output head and a second output head, and the determination of the jump type of the carrier phase of the satellite observation sequence and the jump probability corresponding to the jump type by the preset model according to the local mutation features and the time-dependent features includes: the local mutation features and the time-dependent features are fused based on a fully connected layer of the first output head to generate original scores of each preset jump type; the original scores of each preset jump type are converted into corresponding probability distributions based on a softmax function of the first output head to determine the jump type of the carrier phase of the satellite observation sequence; the local mutation features and the time-dependent features are fused based on a fully connected layer of the second output head to generate original scores of each preset jump type; and the original scores of each preset jump type are converted into probability values based on a sigmoid function of the second output head to determine the jump probability of the jump type.

[0075] In the embodiments of the present application, the processor can be further configured to:

[0076] The confidence of the jump type is determined based on the preset model; the observation quality score of the first feature data is output by weighted fusion according to the confidence, the jump probability and the first feature data; and the corresponding quality level and the prompt information are generated according to the observation quality score and the jump type.

[0077] In the embodiments of the present application, the first feature data is at least one of carrier-to-noise ratio, satellite elevation angle, satellite relative speed, receiver platform speed, first-order phase difference, and second-order phase difference.

[0078] In the embodiments of the present application, the observation quality is calculated according to formula (1) as follows:

[0079] (1)

[0080] wherein, is an observation quality score, , , , , respectively refer to corresponding weight values, refers to a confidence degree corresponding to a jump type, refers to a jump probability, refers to a carrier-to-noise ratio, refers to a satellite elevation angle, refers to a sliding variance of first-order phase difference.

[0081] In the embodiments of the present application, the processor can be further configured to:

[0082] compress the trained preset model to obtain a compressed preset model; convert the compressed preset model into data in an ONNX format; convert the data in the ONNX format and perform quantization mapping to export data in a TFLite format; and deploy the data in the TFLite format on an embedded platform with a Cortex-A7 architecture, so that output data of the preset model is structured data.

[0083] The embodiments of the present application also provide a GNSS receiver, which can include a carrier phase jump detection device.

[0084] The embodiments of the present application also provide a machine readable storage medium, which stores instructions for causing a machine to execute the above-mentioned carrier phase jump detection method.

[0085] In one embodiment, a computer device, which can be a server, is provided, and an internal structure diagram of the computer device can be as shown in FIG. 1. Figure 4As shown in the figure. The computer device includes a processor A01, a network interface A02, a memory (not shown in the figure) and a database (not shown in the figure) connected through a system bus. Among them, the processor A01 of the computer device is used to provide computing and control capabilities. The memory of the computer device includes an internal memory A03 and a non-volatile storage medium A04. The non-volatile storage medium A04 stores an operating system B01, a computer program B02 and a database (not shown in the figure). The internal memory A03 provides an environment for the operating system B01 and the computer program B02 in the non-volatile storage medium A04 to run. The database of the computer device is used to store the carrier phase jump detection data. The network interface A02 of the computer device is used to communicate with the external terminal through the network connection. The computer program B02 is executed by the processor A01 to realize a carrier phase jump detection method.

[0086] Those skilled in the art can understand that, Figure 4 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0087] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0088] The present application is described with reference to the flowcharts and / or block diagrams according to the method, device (system), computer program product of the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the computer or other programmable data processing device produce a device for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one flow or multiple flows and / or blocks Figure 1 The device for implementing the functions specified in one block or multiple blocks.

[0089] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the Figure 1 function specified in the flow or flows and / or blocks Figure 1 of the block or blocks.

[0090] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the Figure 1 function specified in the flow or flows and / or blocks Figure 1 of the block or blocks.

[0091] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0092] The memory can include non-persistent memory and / or volatile memory, such as a random access memory (RAM) including a cache area for the temporary storage of data. The memory can also include non-volatile memory, such as read only memory (ROM), electrically programmable read only memory (EPROM), electrically erasable programmable read only memory (EEPROM), flash memory, or a combination of non-volatile memories in different forms. The memory is an example of computer readable storage media.

[0093] Computer readable media includes permanent and non-permanent, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read only memory (ROM), electrically programmable read only memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital versatile disc (DVD), or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible to a computing device. According to the definition herein, computer readable media does not include transitory media, such as modulated data signals and carrier waves.

[0094] It should also be noted that the terms "comprising", "comprises" or other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0095] The above embodiments are only used to illustrate the present application, but not to limit it. Instead of the above, various modifications and changes can be made to the application by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the application shall fall into the scope of the claims of the application.

Claims

1. A method for detecting carrier phase transitions, characterized in that, The method includes: Obtain the satellite observation sequence for the current epoch of any satellite; Extract the first feature data and the third feature data of the satellite observation sequence. The first feature data includes the original value of the carrier phase. The third feature data is at least one of the following: carrier-to-noise ratio, satellite elevation angle, satellite relative velocity, receiver platform velocity, first-order phase difference, and second-order phase difference. The first feature data is input into a pre-trained model, which includes a convolutional neural network and a long short-term memory network. The convolutional neural network extracts local mutation features in the time series of the first feature data, and the long short-term memory network performs time series modeling on the local mutation features to output time-dependent features. Obtain the carrier phase jump type and the jump probability corresponding to the jump type, output by the preset model for the satellite observation sequence based on the local mutation feature and the time dependence feature; The method further includes: The confidence level of the jump type is determined based on the preset model; The confidence level, the jump probability, and the third feature data are weighted and fused to output an observation quality score relative to the satellite observation sequence; Based on the observed quality score and the jump type, a corresponding quality level and prompt information are generated.

2. The carrier phase transition detection method according to claim 1, characterized in that, The method further includes: Acquire the first historical satellite observation sequences from multiple satellites at different epochs; Extract the second feature data from the first historical satellite observation sequence; The second feature data is divided into multiple sub-feature data based on a preset sliding window; Obtain the historical transition type corresponding to each sub-feature data; The preset model is trained using the multiple sub-feature data and their corresponding historical jump types to obtain the trained preset model.

3. The carrier phase transition detection method according to claim 1, characterized in that, The preset model further includes a task branch module, which includes a first output head and a second output head. The step of obtaining the carrier phase jump type and the corresponding jump probability of the jump type for the satellite observation sequence, output by the preset model based on the local mutation features and the time-dependent features, includes: The fully connected layer based on the first output head fuses the local mutation features and the time-dependent features to generate the original score for each preset jump type; The softmax function based on the first output head transforms the original score of each preset jump type into a corresponding probability distribution to determine the jump type of the carrier phase of the satellite observation sequence; The fully connected layer based on the second output head fuses the local mutation features and the time-dependent features to generate the original score for each preset jump type; The sigmoid function based on the second output head converts the original score of each preset jump type into a probability value to determine the jump probability of the jump type.

4. The carrier phase transition detection method according to claim 1, characterized in that, The observation quality is calculated according to the following formula (1): （1） in, This refers to the observed quality score. , , , , These refer to the corresponding weight values. This refers to the confidence level corresponding to the aforementioned transition type. This refers to the transition probability. This refers to the carrier-to-noise ratio. This refers to the satellite's elevation angle. It refers to the sliding variance of the first-order phase difference.

5. The carrier phase transition detection method according to claim 1, characterized in that, The method further includes: The trained preset model is compressed to obtain the compressed preset model; Convert the compressed preset model into ONNX format data; The ONNX format data is converted and quantized to export TFLite format data; The TFLite format data is deployed on an embedded platform with a Cortex-A7 architecture so that the output data of the preset model is structured data.

6. A carrier phase transition detection device, characterized in that, include: The memory is configured to store instructions; A processor is configured to retrieve the instructions from the memory and, when executing the instructions, to implement the carrier phase transition detection method according to any one of claims 1 to 5.

7. A GNSS receiver, characterized in that, Includes the carrier phase transition detection device according to claim 6.

8. A machine-readable storage medium, characterized in that, The machine-readable storage medium stores instructions for causing the machine to perform the carrier phase transition detection method according to any one of claims 1 to 5.

Images