GNSS (Global Navigation Satellite System) scene recognition method and device and GNSS data resolving method and device

Abstract

The invention provides a GNSS scene recognition method and device and a GNSS data resolving method and device.The GNSS scene recognition method and device comprises the steps that firstly, model training is conducted, and specifically, actual GNSS data corresponding to all predefined resolving scenes are obtained; labeling the actual GNSS data according to a first preset time period; for the actual GNSS data corresponding to each label, extracting model training parameters of the actual GNSS data corresponding to the label; and training the resolving scene recognition model by using the actual GNSS data and the model training parameters corresponding to the various labels, and evaluating the performance of the resolving scene recognition model by using the test set in the training process until the performance of the resolving scene recognition model reaches a preset requirement, thereby obtaining the trained resolving scene recognition model. And scene recognition is carried out after the resolving scene recognition model is obtained through training, specifically, GNSS data to be resolved in a second preset time period are obtained according to a preset frequency, and the resolving scene of the GNSS data to be resolved is recognized by using the resolving scene recognition model after training is completed to obtain target resolving scene information.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the field of GNSS positioning technology, and in particular to a GNSS scene recognition method and device, and a GNSS data solution method and device. BACKGROUND

[0002] In high-precision positioning solution, whether it is single point positioning (SPP), real-time kinematic positioning (RTK) or precise point positioning (PPP), scene judgment is a crucial prerequisite step. Accurate scene judgment helps to select the appropriate solution mode, optimize the observation strategy, and reasonably set the relevant parameters, thereby significantly improving the accuracy and reliability of the final positioning result.

[0003] For example, in RTK application, it is necessary to judge whether there are strong multipath effects, signal shielding or base station coverage range and other factors according to the working environment. For example, in complex environments such as urban canyons, forest areas or mountainous areas, satellite signals are easily shielded or reflected, affecting the stable transmission of differential data, at which time it may be necessary to combine multi-band observations or enhanced systems to improve the solution quality. In PPP solution, scene judgment is more concerned about the motion state of the receiver, the initial convergence time requirement, and the availability of external precise ephemeris and clock error products. For example, in dynamic navigation application, if the platform is not correctly identified in high-speed moving state, it may lead to unstable convergence process, and thus affect the positioning accuracy.

[0004] Therefore, only on the basis of fully understanding and judging the actual working scene, can the solution strategy be reasonably configured, various error sources be effectively dealt with, and the high-precision potential of the solution technology be ensured.

[0005] The current technical solution of scene judgment is prone to the following problems in dealing with complex and variable actual situations:

[0006] 1. Single threshold setting is not suitable for dynamic changing environment

[0007] For example, setting "number of satellites > 6" as the standard of good environment, but in urban canyons, even if there are 6 satellites, the data quality may be poor due to shielding and multipath effect.

[0008] The CN0 threshold is fixed and cannot distinguish the real availability of different frequency bands or different types of signals.

[0009] 2. Lack of overall perception ability of environmental characteristics

[0010] Traditional methods are difficult to identify the influence of complex ground objects such as building shielding, vegetation interference, water surface reflection on signal propagation.

[0011] It is difficult to distinguish static and dynamic multipath effects, which affects the accuracy of subsequent cycle slip detection and repair and ambiguity fixing.

[0012] 3. Ignoring the correlation between platform motion state and error

[0013] In high dynamic scenarios (such as fast flight of unmanned aerial vehicles), it is difficult to accurately evaluate the cumulative trend of observation errors caused by receiver dynamics only by the number of satellites and GDOP.

[0014] The coupling relationship between different error sources (such as the interaction between ionospheric delay and satellite geometric distribution) is ignored.

[0015] 4. No learning and adaptive ability

[0016] The traditional logic is based on pre-set rules and lacks generalization ability in the face of new environments or extreme weather, which is prone to "one-size-fits-all" misjudgment. SUMMARY

[0017] The present disclosure at least provides a GNSS scene recognition method and device, and a GNSS data solving method and device to solve at least one of the above technical problems.

[0018] According to an aspect of the present disclosure, a GNSS scene recognition method is provided, comprising the following steps:

[0019] Respectively acquiring actual GNSS data corresponding to each pre-defined solving scene;

[0020] Labeling the actual GNSS data according to a first pre-set time period;

[0021] For each label corresponding to the actual GNSS data, extracting the model training parameters of the label corresponding to the actual GNSS data;

[0022] Using the actual GNSS data and the model training parameters corresponding to various labels to train a solving scene recognition model, and using a test set to evaluate the performance of the solving scene recognition model during the training process until the performance of the solving scene recognition model meets a pre-set requirement, obtaining a trained solving scene recognition model;

[0023] According to a predetermined frequency, acquiring GNSS data to be solved within a second pre-set time period, using the trained solving scene recognition model to recognize the solving scene of the GNSS data to be solved, and obtaining target solving scene information.

[0024] In a possible implementation, the target solving scene information includes the probability of each pre-defined solving scene; and the solving scene recognition model further includes a positioning branch.

[0025] The GNSS scene recognition method further comprises:

[0026] acquiring IMU data and geomagnetic data synchronously collected with the GNSS data to be solved;

[0027] delivering the IMU data and the geomagnetic data to a scene weight recognition branch in the positioning branch, processing the IMU data and the geomagnetic data by using the scene weight recognition branch, and outputting solving weight information corresponding to each predefined solving scene to a solving branch in the positioning branch;

[0028] sending the target solving scene information to the solving branch;

[0029] performing feature enhancement processing on the target solving scene information based on the solving weight information corresponding to each predefined solving scene by using the solving branch, and solving the GNSS data to be solved by using the information after the feature enhancement processing, to obtain positioning information.

[0030] In a possible implementation, the target solving scene information comprises a target solving scene.

[0031] The GNSS scene recognition method further comprises:

[0032] acquiring a solving strategy corresponding to each predefined solving scene;

[0033] selecting a solving strategy matching the target solving scene from the solving strategies, to obtain a target solving strategy;

[0034] solving the GNSS data to be solved by using the target solving strategy, to obtain positioning information.

[0035] In a possible implementation, the predefined solving scene comprises: an airborne scene; a handheld scene; a vehicle-mounted scene;

[0036] The airborne scene comprises at least one of the following sub-scenes: a take-off and landing stage, an in-flight stage, a canyon flying stage, and a city flying stage.

[0037] The handheld scene comprises at least one of the following sub-scenes: a city building, a city tree shade, a city open space, and a city single-side shelter.

[0038] The vehicle-mounted scene comprises at least one of the following sub-scenes: under a viaduct, on a viaduct, a city building, a city tree shade, a city open space, a city single-side shelter, and a city CBD.

[0039] In a possible implementation, the model training parameter comprises at least one of the following:

[0040] The number of satellite systems, the number of satellites in each satellite system, the average CN0 of each frequency point, the maximum CN0 of each frequency point, the number of cycle slips of each frequency point, the number of half cycle slips of each frequency point, the average continuous locking time of each frequency point, and the average elevation angle of each frequency point.

[0041] In a possible implementation, the training of the calculation scenario recognition model by using the actual GNSS data corresponding to various labels and the model training parameters comprises:

[0042] The actual GNSS data corresponding to various labels and the model training parameters are normalized.

[0043] The calculation scenario recognition model is trained by using the normalized actual GNSS data corresponding to various labels and the normalized model training parameters.

[0044] In a possible implementation, the calculation scenario recognition model comprises a BP neural network, the input layer of the BP neural network is the parameters of GNSS, the hidden layer is two layers, the output layer is a classification layer, and the activation function is sigmoid.

[0045] According to another aspect of the present disclosure, a GNSS data calculation method is provided, comprising:

[0046] At a predetermined frequency, GNSS data to be calculated in a second preset time period is acquired;

[0047] The calculation scenario of the GNSS data to be calculated is recognized by using the trained calculation scenario recognition model, and target calculation scenario information is obtained;

[0048] The GNSS data to be calculated is calculated based on the target calculation scenario information, and positioning information is obtained.

[0049] According to another aspect of the present disclosure, a GNSS scenario recognition device is provided, comprising:

[0050] A training data acquisition module is configured to acquire actual GNSS data corresponding to each predefined calculation scenario respectively.

[0051] A labeling module is configured to label the actual GNSS data according to a first preset time period.

[0052] A parameter extraction module is configured to extract, for the actual GNSS data corresponding to each label, a model training parameter of the actual GNSS data corresponding to the label.

[0053] a training module configured to train a calculation scenario recognition model by using actual GNSS data corresponding to various labels and model training parameters, and to evaluate the performance of the calculation scenario recognition model by using a test set during the training until the performance of the calculation scenario recognition model reaches a preset requirement, so as to obtain a trained calculation scenario recognition model;

[0054] a scenario determination module configured to obtain GNSS data to be calculated within a second preset time period at a predetermined frequency, and to recognize a calculation scenario of the GNSS data to be calculated by using the trained calculation scenario recognition model, so as to obtain target calculation scenario information.

[0055] According to another aspect of the present disclosure, a GNSS data calculation device is provided, which comprises:

[0056] a calculation data obtaining module configured to obtain GNSS data to be calculated within a second preset time period at a predetermined frequency;

[0057] a scenario recognition module configured to recognize a calculation scenario of the GNSS data to be calculated by using the trained calculation scenario recognition model according to any one of the above, so as to obtain target calculation scenario information;

[0058] a positioning module configured to calculate the GNSS data to be calculated based on the target calculation scenario information, so as to obtain positioning information.

[0059] The GNSS scenario recognition method and device of the present disclosure first perform model training. Specifically, actual GNSS data corresponding to each predefined calculation scenario is obtained; the actual GNSS data is labeled according to a first preset time period; for the actual GNSS data corresponding to each label, model training parameters of the actual GNSS data corresponding to the label are extracted; the calculation scenario recognition model is trained by using the actual GNSS data corresponding to various labels and the model training parameters, and the performance of the calculation scenario recognition model is evaluated by using a test set during the training until the performance of the calculation scenario recognition model reaches a preset requirement, so as to obtain a trained calculation scenario recognition model. After the calculation scenario recognition model is trained, scenario recognition is performed. Specifically, GNSS data to be calculated within a second preset time period is obtained at a predetermined frequency, and the calculation scenario of the GNSS data to be calculated is recognized by using the trained calculation scenario recognition model, so as to obtain target calculation scenario information. After the target scenario is recognized, an appropriate calculation strategy is matched based on the target calculation scenario information, or the GNSS data to be calculated is directly calculated by using a neural network combined with the scenario information, so as to obtain positioning information.

[0060] The scene judgment manner of the present disclosure using the neural network can overcome the defects that the prior art cannot adapt to the overall environment, lacks overall perception ability, and does not have learning and self-adaptive ability, can accurately identify the scene by comprehensively integrating overall information, and provides support for subsequent high-precision positioning.

[0061] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present disclosure, nor to limit the scope of the present disclosure. Other features of the present disclosure will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0062] The accompanying drawings are used to better understand the present scheme and do not constitute a limitation on the present disclosure. Among them:

[0063] Figure 1 is a flowchart of a GNSS scene recognition method in an embodiment of the present disclosure;

[0064] Figure 2 is a flowchart of a GNSS scene solving method in an embodiment of the present disclosure;

[0065] Figure 3 is a structural schematic diagram of a GNSS scene recognition device in an embodiment of the present disclosure;

[0066] Figure 4 is a structural schematic diagram of a GNSS scene solving device in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0067] Exemplary embodiments of the present disclosure are described below with reference to the accompanying drawings, which include various details of the embodiments of the present disclosure to help understanding, and should be considered as merely exemplary. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Also, in order to be clear and concise, the description below omits the description of well-known functions and structures.

[0068] The present disclosure provides a GNSS scene recognition method and device, and a GNSS data solving method and device, in order to overcome the large determination error defect of the prior art that only uses simple satellite number, CN0, etc. combination for scene judgment. Based on a large amount of measured data, model training can more accurately identify the corresponding scene, and using the results of scene recognition to complete subsequent solving can effectively improve the positioning quality.

[0069] The technical scheme of the present disclosure will be described below through specific embodiments.

[0070] As Figure 1As shown, it is a flowchart of the GNSS scene recognition method of the embodiment. The execution subject of the embodiment is a computing device or component with data processing capability. Specifically, the method of the embodiment can include the following steps:

[0071] S110, respectively acquiring actual GNSS data corresponding to each predefined calculation scene.

[0072] The above-mentioned predefined calculation scenes include: airborne scene; handheld scene; vehicle-mounted scene. Wherein, the airborne scene includes at least one of the following sub-scenes: take-off and landing stage, in-flight, canyon flight, city flight; the handheld scene includes at least one of the following sub-scenes: city building, city tree shade, city open space, city single-sided shielding; the vehicle-mounted scene includes at least one of the following sub-scenes: under viaduct, on viaduct, city building, city tree shade, city open space, city single-sided shielding, city CBD.

[0073] Before model training, a large amount of data needs to be collected for each calculation scene (such as handheld, vehicle-mounted, airborne), which should cover as many environmental variables as possible to ensure that the model has sufficient generalization ability.

[0074] The handheld scene is generally applicable to handheld devices, which are usually used for pedestrian navigation or portable measurement. Data collection should consider diverse terrains and environments such as urban walking, parks, and forests.

[0075] The vehicle-mounted scene is generally applicable to vehicle-mounted devices, which are mainly used for vehicle navigation and monitoring. Data collection should consider various road conditions such as highways, urban streets, tunnels, and underground parking lots.

[0076] The airborne scene is generally applicable to airborne devices such as unmanned aerial vehicles (UAVs), airplanes, and other aviation platforms. Data collection should consider different scenarios such as open airspace, mountainous areas, and urban airspace.

[0077] For each case, in addition to GNSS raw observation data (such as pseudorange, carrier phase), relevant environmental parameters (such as temperature, humidity), sensor data (such as accelerometer, gyroscope readings), and image data (if applicable) should also be recorded. At the same time, ensure that there is accurate reference position data for training and verifying the model.

[0078] In some embodiments, the following calculation scenes can also be defined:

[0079] Open sky: no obstruction, high and stable signal strength.

[0080] Urban canyon: high-rise buildings, severe multipath effect and partial satellite signal loss.

[0081] Forest coverage: signal attenuation and scattering caused by trees.

[0082] Tunnel / indoor: almost no direct line of sight signal, possibly relying on inertial measurement units (IMU) or other assisted positioning technologies.

[0083] According to the different characteristics of handheld, vehicle-mounted and airborne, these scene modes are refined, and the classification criteria are adjusted according to actual needs.

[0084] S120, tagging the actual GNSS data according to the first preset time period.

[0085] The above tag is specifically each of the above predefined calculation scenes.

[0086] Specifically, first, the relative devices of each calculation scene are formulated N, and each device collects m groups of data on average, wherein the overall collection is performed according to the different scenes set in advance, and all are covered, the data is tagged according to the corresponding time period, and is determined as the corresponding calculation scene.

[0087] S130, for each tag corresponding actual GNSS data, extracting the model training parameter of the tag corresponding actual GNSS data.

[0088] The model training parameter includes at least one of the following:

[0089] The number of satellite systems, the number of satellites of each satellite system, the average CN0 of each frequency point, the maximum CN0 of each frequency point, the number of each frequency point cycle slip, the number of each frequency point half cycle slip, the average continuous locking time of each frequency point satellite, and the average elevation angle of each frequency point.

[0090] The above model training parameter is helpful to distinguish the characteristics of different calculation scenes.

[0091] S140, using the actual GNSS data and the model training parameter corresponding to each tag to train the calculation scene recognition model, and using the test set to evaluate the performance of the calculation scene recognition model in the training process until the performance of the calculation scene recognition model reaches the preset requirement, obtaining the trained calculation scene recognition model.

[0092] In the training process, the actual GNSS data and the model training parameter corresponding to each tag are normalized; the normalized actual GNSS data and the normalized model training parameter corresponding to each tag are used to train the calculation scene recognition model.

[0093] The data used for the above training is normalized, the data is scaled to a specific range (such as between 0 and 1), or the distribution is adjusted to adapt to the needs of the model.

[0094] Put the training data under these corresponding labels into the neural network for training, and use the check set for accuracy verification, and perform corresponding iterations, and finally get the parameter model with the highest accuracy.

[0095] Specifically, all data is divided into training set, validation set and test set, usually in the ratio of 7:2:1 or 8:1:1. The training set is used for model training, the validation set is used for adjusting hyperparameters, and the test set is used for evaluating the performance of the final model. The final purpose of the data set is to train the parameters, and the obtained training parameters are applied to the final scene recognition.

[0096] Finally, the trained model is strictly verified to evaluate its performance on unknown data. Once the model performance meets the expected target, it is deployed to the corresponding device, which monitors in real time and automatically switches to the working mode most suitable for the current scene.

[0097] The neural network-based pattern recognition method not only improves the accuracy of scene judgment, but also flexibly copes with various complex real-world challenges; with the accumulation of more high-quality data and the progress of technology, this model can become more efficient and reliable.

[0098] The solving scene recognition model includes a BP neural network, the input layer of the BP neural network is the parameters of GNSS, the hidden layer is 2 layers, the output layer is a classification layer, and the activation function is sigmoid.

[0099] S150, according to the predetermined frequency, obtain the GNSS data to be solved within the second preset time period, use the trained solving scene recognition model to identify the solving scene of the GNSS data to be solved, and obtain target solving scene information.

[0100] Put the trained model into the solution to provide the corresponding scene more accurately.

[0101] In some embodiments, the solving scene recognition model further includes a positioning branch for solving GNSS data to obtain positioning information. Specifically, after obtaining the target solving scene information, the positioning can be performed in the following way:

[0102] Step one, obtain IMU data and geomagnetic data collected synchronously with the GNSS data to be solved.

[0103] Step two, deliver the IMU data and geomagnetic data to the scene weight recognition branch in the positioning branch, so as to process the IMU data and geomagnetic data by using the scene weight recognition branch, and output the solving weight information corresponding to each predefined solving scene to the solving branch in the positioning branch.

[0104] Step three, sending the target solving scene information to the solving branch.

[0105] The target solving scene information includes the probability of each predefined solving scene. Here, the probability of each predefined solving scene is transmitted to the solving branch, so that the solving branch comprehensively uses the predefined solving scene and its probability, can determine a solving strategy including multiple modes, to improve the positioning accuracy.

[0106] Step four, using the solving branch to perform feature enhancement processing on the target solving scene information based on the solving weight information corresponding to each predefined solving scene, and using the information after the feature enhancement processing to solve the GNSS data to be solved, to obtain the positioning information.

[0107] This embodiment not only combines multiple predefined solving scenes and their probabilities, but also uses IMU data and geomagnetic data to perform enhancement processing on different solving scenes, which can greatly improve the solving accuracy.

[0108] In some embodiments, after obtaining the target solving scene information, the following method can also be used for positioning:

[0109] Step one, obtaining the solving strategy corresponding to each predefined solving scene.

[0110] Different solving strategies are set for different solving scenes before this step is performed.

[0111] Step two, selecting a solving strategy matching the target solving scene from the solving strategies, to obtain the target solving strategy; the target solving scene information includes the target solving scene.

[0112] Step three, using the target solving strategy to solve the GNSS data to be solved, to obtain the positioning information.

[0113] In this embodiment, the GNSS data is solved using the solving strategy matching the target solving scene, which can effectively improve the solving accuracy.

[0114] The neural network-based pattern recognition method of the present application is researched for overall scene judgment, and the neural network is used for scene judgment, especially for three different application modes of handheld, vehicle-mounted and airborne, which is indeed a very effective solving strategy. This method can improve the adaptability and accuracy of different environmental conditions, and provide support for subsequent RTK / PPP solving. The accuracy of scene judgment will affect the subsequent high-precision positioning.

[0115] Based on the same inventive concept, as shown in Figure 2 The present disclosure also provides a GNSS data solving method, including the following steps:

[0116] S210, acquiring GNSS data to be solved in a second preset time period according to a predetermined frequency.

[0117] S220, identifying a solution scene of the GNSS data to be solved by using the trained solution scene identification model of any of the above embodiments, to obtain target solution scene information.

[0118] S230, solving the GNSS data to be solved based on the target solution scene information, to obtain positioning information.

[0119] The solution process here can be completed by using the positioning branch in the above embodiments, or by using a solution strategy matched with the target solution scene, which will not be repeated here.

[0120] Based on the same inventive concept, the present disclosure provides a GNSS scene identification device, the components of which perform the same or similar steps as the above-mentioned GNSS scene identification method, and therefore similar places will not be repeated. As shown in Figure 3 The GNSS scene identification device of the present embodiment comprises:

[0121] The training data acquisition module 310 is configured to acquire actual GNSS data corresponding to each predefined solution scene respectively.

[0122] The labeling module 320 is configured to label the actual GNSS data according to a first preset time period.

[0123] The parameter extraction module 330 is configured to extract model training parameters of the actual GNSS data corresponding to each label for the actual GNSS data corresponding to the label.

[0124] The training module 340 is configured to train a solution scene identification model by using the actual GNSS data and the model training parameters corresponding to various labels, and to evaluate the performance of the solution scene identification model by using a test set during the training process until the performance of the solution scene identification model reaches a preset requirement, to obtain a trained solution scene identification model.

[0125] The scene determination module 350 is configured to acquire GNSS data to be solved in a second preset time period according to a predetermined frequency, and to identify a solution scene of the GNSS data to be solved by using the trained solution scene identification model, to obtain target solution scene information.

[0126] Based on the same inventive concept, the present disclosure provides a GNSS data solution device, the components of which perform the same or similar steps as the above-mentioned GNSS data solution method, and therefore similar places will not be repeated. As shown in Figure 4As shown, the GNSS data solving device of the embodiment includes:

[0127] The solving data acquisition module 410 is configured to acquire the GNSS data to be solved in a second preset time period according to a predetermined frequency.

[0128] The scene recognition module 420 is configured to recognize the solving scene of the GNSS data to be solved by using the trained solving scene recognition model of any embodiment to obtain target solving scene information.

[0129] The positioning module 430 is configured to solve the GNSS data to be solved based on the target solving scene information to obtain positioning information.

[0130] Various embodiments of the techniques described above in this document can be implemented in digital electronic circuitry, integrated circuitry, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0131] Program code for carrying out methods of the present disclosure can be written in any combination of one or more programming languages. The program code can be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the program code, when executed by the processor or controller, produces the functions / operations specified in the flowcharts and / or the block diagrams. The program code can be executed entirely on a machine, partially on a machine, partially on a machine as a stand-alone software package, partially on a machine and partially on a remote machine or entirely on a remote machine or server.

[0132] In the context of this disclosure, a machine-readable medium can be a tangible medium that contains or stores a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include but is not limited to an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0133] To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0134] The systems and techniques described here can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0135] The computer system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0136] It should be understood that the various forms of flow shown above can be used to reorder, add, or remove steps. For example, the steps described in the present disclosure can be performed in parallel, in series, or in a different order, without limitation herein, so long as the desired results of the technology described in the present disclosure are achieved.

[0137] The specific implementation described above does not constitute a limitation on the protection scope of the present disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A GNSS scene recognition method, characterized in that, The method comprises the following steps: respectively acquiring actual GNSS data corresponding to each predefined calculation scenario; labeling the actual GNSS data according to a first preset time period; for the actual GNSS data corresponding to each label, extracting model training parameters of the actual GNSS data corresponding to the label; training a calculation scenario recognition model by using the actual GNSS data corresponding to each label and the model training parameters, and evaluating the performance of the calculation scenario recognition model by using a test set during the training process until the performance of the calculation scenario recognition model reaches a preset requirement, thereby obtaining a trained calculation scenario recognition model; acquiring GNSS data to be calculated within a second preset time period according to a predetermined frequency, and identifying the calculation scenario of the GNSS data to be calculated by using the trained calculation scenario recognition model, thereby obtaining target calculation scenario information. 2.The GNSS scene recognition method of claim 1, wherein, The target calculation scenario information comprises probabilities of each predefined calculation scenario. The calculation scenario recognition model further comprises a positioning branch. The method further comprises the following steps: acquiring IMU data and geomagnetic data collected synchronously with the GNSS data to be calculated; feeding the IMU data and the geomagnetic data to a scenario weight recognition branch in the positioning branch, so as to process the IMU data and the geomagnetic data by using the scenario weight recognition branch, and output calculation weight information corresponding to each predefined calculation scenario to a calculation branch in the positioning branch; sending the target calculation scenario information to the calculation branch; performing feature enhancement processing on the target calculation scenario information based on the calculation weight information corresponding to each predefined calculation scenario by using the calculation branch, and performing calculation on the GNSS data to be calculated by using the information after the feature enhancement processing, thereby obtaining positioning information.

3. The GNSS scene recognition method of claim 1, wherein, The target calculation scenario information comprises a target calculation scenario. The method further comprises the following steps: acquiring calculation strategies corresponding to each predefined calculation scenario; selecting a calculation strategy matching the target calculation scenario from the calculation strategies, thereby obtaining a target calculation strategy; performing calculation on the GNSS data to be calculated by using the target calculation strategy, thereby obtaining positioning information.

4. The GNSS scene recognition method of claim 1, wherein, The predefined calculation scenarios comprise the following scenarios: an airborne scenario; a handheld scenario; and a vehicle-mounted scenario.

5. The GNSS scene recognition method of claim 1, wherein, The airborne scenario comprises at least one of the following sub-scenarios: a take-off and landing stage, an in-flight stage, a canyon flying stage, and a city flying stage. The handheld scenario comprises at least one of the following sub-scenarios: a city building, a city tree shade, a city open space, and a city single-side shelter. The vehicle-mounted scenario comprises at least one of the following sub-scenarios: under a viaduct, on a viaduct, a city building, a city tree shade, a city open space, a city single-side shelter, and a city CBD. The model training parameters comprise at least one of the following parameters: a number of satellite systems, a number of satellites in each satellite system, an average CN0 of each frequency point, a maximum CN0 of each frequency point, a number of cycle slips of each frequency point, a number of half-cycle slips of each frequency point, and an average continuous locking time of each frequency point; an average elevation angle of each frequency point.

6. The GNSS scene recognition method of claim 1, wherein, The actual GNSS data corresponding to various labels and model training parameters are used to train the calculation scene recognition model, including: The actual GNSS data corresponding to various labels and model training parameters are normalized; The normalized actual GNSS data corresponding to various labels and the normalized model training parameters are used to train the calculation scene recognition model.

7. The GNSS scene recognition method according to claim 1, characterized in that, The calculation scene recognition model comprises a BP neural network, the input layer of the BP neural network is the parameters of GNSS, the hidden layer is 2 layers, the output layer is a classification layer, and the activation function is sigmoid.

8. A GNSS data resolution method, characterized by, It includes: According to a predetermined frequency, the GNSS data to be calculated in a second preset time period is obtained; The calculation scene of the GNSS data to be calculated is identified by using the calculation scene recognition model according to any one of claims 1 to 7, and target calculation scene information is obtained; Based on the target calculation scene information, the GNSS data to be calculated is calculated, and positioning information is obtained.

9. A GNSS scene recognition apparatus, characterized by It includes: The training data acquisition module is used to acquire actual GNSS data corresponding to each predefined calculation scene respectively; The labeling module is used to label the actual GNSS data according to a first preset time period; The parameter extraction module is used to extract model training parameters of the actual GNSS data corresponding to each label for the actual GNSS data corresponding to each label; The training module is used to train the calculation scene recognition model by using the actual GNSS data corresponding to various labels and the model training parameters, and the performance of the calculation scene recognition model is evaluated by using a test set during the training process until the performance of the calculation scene recognition model reaches a preset requirement, and a trained calculation scene recognition model is obtained; The scene determination module is used to acquire GNSS data to be calculated in a second preset time period according to a predetermined frequency, identify the calculation scene of the GNSS data to be calculated by using the trained calculation scene recognition model, and obtain target calculation scene information.

10. A GNSS data processing apparatus, characterized by comprising: It includes: The calculation data acquisition module is used to acquire GNSS data to be calculated in a second preset time period according to a predetermined frequency; The scene recognition module is used to identify the calculation scene of the GNSS data to be calculated by using the calculation scene recognition model according to any one of claims 1 to 7, and target calculation scene information is obtained; The positioning module is used to calculate the GNSS data to be calculated based on the target calculation scene information, and positioning information is obtained.

Images