Scene recognition method and device, terminal equipment and storage medium

By constructing feature vectors and using machine learning algorithm scene classification model, the problem of inaccurate GNSS data in outdoor scene recognition is solved, and the accurate identification of the scene where the terminal device is located is achieved, especially in occlusion scenarios.

CN119006992APending Publication Date: 2024-11-22GUANGDONG OPPO MOBILE TELECOMMUNICATIONS CORP LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202310583700.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-05-22
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

The prior art has problems with insufficient accuracy in outdoor scene recognition through GNSS data, and it is difficult to effectively distinguish the scene where the terminal device is located, especially the occlusion scene.

Method used

By obtaining the original observation information related to the terminal device, satellite information and pseudorange, building feature vectors, and using preset machine learning algorithms with data classification capabilities to train the scene classification model to identify the scene where the terminal device is located. Includes empty and obstructed scenes.

Benefits of technology

The terminal equipment's accurate discrimination ability of scenes is improved, especially in occlusion scenarios, and the accuracy and reliability of recognition are improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119006992A_ABST
    Figure CN119006992A_ABST
Patent Text Reader

Abstract

The invention discloses a scene recognition method, a scene recognition device, terminal equipment and a storage medium. The scene recognition method comprises the following steps: acquiring positioning information of a positioning device of terminal equipment, satellite information of a global satellite navigation system and original observation information related to a pseudo range; constructing a feature vector of the terminal equipment in the current scene according to the positioning information, the satellite information and the original observation information; according to the feature vector, a scene where the terminal device is located is distinguished through a scene classification model, the scene classification model is obtained through construction and training of a preset classification algorithm, the scene comprises one of an open scene and a sheltered scene, and the preset classification algorithm is a machine learning algorithm with the data classification capability. According to the scene recognition method provided by the embodiment of the invention, the feature vector of the terminal equipment constructed by the positioning information, the satellite information and the original observation information in the current scene is input into the scene classification model, so that the shielding feature and type of the scene can be recognized more accurately.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of image recognition, and in particular to a scene recognition method, a scene recognition device, a terminal device, and a computer-readable storage medium. Background Art

[0002] It is more difficult to identify outdoor scenes than indoor scenes. In related technologies, terminal devices can identify outdoor scenes through Global Navigation Satellite System (GNSS) data. However, there is currently a problem of insufficient accuracy in identifying outdoor scenes through GNSS data. Summary of the invention

[0003] In view of this, the present application provides a scene recognition method, a scene recognition apparatus, a terminal device, and a non-volatile computer-readable storage medium.

[0004] The scene recognition method of the embodiment of the present application includes:

[0005] Acquire positioning information of a positioning device of the terminal device, satellite information of a global satellite navigation system, and original observation information related to pseudorange;

[0006] Constructing a feature vector of the terminal device in the current scenario according to the positioning information, the satellite information and the original observation information;

[0007] The scene in which the terminal device is located is identified according to the feature vector through a scene classification model, wherein the scene classification model is constructed and trained by a preset classification algorithm, and the scene includes one of an open scene and an occluded scene, wherein the preset classification algorithm is a machine learning algorithm with data classification capabilities.

[0008] The scene recognition device of the embodiment of the present application includes:

[0009] An acquisition module, used to acquire positioning information of a positioning device of the terminal device, satellite information of a global satellite navigation system, and original observation information related to pseudorange;

[0010] A construction module, used to construct a feature vector of the terminal device in the current scenario according to the positioning information, satellite information and the original observation information;

[0011] A recognition module is used to recognize the scene in which the terminal device is located through a scene classification model according to the feature vector, wherein the scene classification model is constructed and trained by a preset classification algorithm, and the scene includes one of an open scene and an occluded scene, wherein the preset classification algorithm is a machine learning algorithm with data classification capabilities.

[0012] The terminal device of the embodiment of the present application includes a processor and a storage unit; a computer program is stored in the storage unit, and when the computer program is executed by the processor, the processor executes the above-mentioned scene recognition method. The scene recognition method includes: obtaining the positioning information of the positioning device of the terminal device, the satellite information of the global satellite navigation system, and the original observation information related to the pseudorange; constructing the feature vector of the terminal device in the current scene according to the positioning information, satellite information and the original observation information; according to the feature vector, the scene in which the terminal device is located is identified by a scene classification model, and the scene classification model is constructed and trained by a preset classification algorithm, and the scene includes one of an open scene and an occluded scene, wherein the preset classification algorithm is a machine learning algorithm with data classification capabilities.

[0013] The non-volatile computer-readable storage medium of the embodiment of the present application includes a computer program, and when the computer program is executed by a processor, the processor executes the scene recognition method described above. The scene recognition method includes: obtaining the positioning information of the positioning device of the terminal device, the satellite information of the global satellite navigation system, and the original observation information related to the pseudorange; constructing the feature vector of the terminal device in the current scene according to the positioning information, satellite information and the original observation information; according to the feature vector, the scene in which the terminal device is located is identified by a scene classification model, and the scene classification model is constructed and trained by a preset classification algorithm, and the scene includes one of an open scene and an occluded scene, wherein the preset classification algorithm is a machine learning algorithm with data classification capabilities.

[0014] In the scene recognition method, scene recognition device, terminal device and readable storage medium of the implementation mode of the present application, positioning information, satellite information and original observation information, etc., which can be used to judge and comprehensively identify the surrounding environment scene, are obtained from the positioning device of the terminal device, and then a feature vector of the terminal device in the current scene is constructed according to the positioning information, satellite information and original observation information, so that the feature vector contains multiple features that represent the scene of the terminal device. In this way, inputting the feature vector into a scene classification model trained by a preset classification algorithm with data analysis capabilities can more accurately identify the occlusion type and characteristics of the scene where the terminal device is located, thereby improving the accuracy of the terminal device in identifying the scene.

[0015] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0017] Figure 1 is a module schematic diagram of a scene recognition method in some embodiments of the present application;

[0018] Figure 2 is a schematic diagram of a module of a scene recognition device in some embodiments of the present application;

[0019] Figure 3 is a schematic diagram of a module of a terminal device in some embodiments of the present application;

[0020] Figure 4 is a schematic diagram of a scene of a scene recognition method in some embodiments of the present application;

[0021] Figure 5 is a schematic diagram of a scene of a scene recognition method in some embodiments of the present application;

[0022] Figure 6 is a flow chart of a scene recognition method in some embodiments of the present application;

[0023] Figure 7 is a schematic diagram of a module of a scene recognition device in some embodiments of the present application;

[0024] Figure 8 is a flow chart of a scene recognition method in some embodiments of the present application;

[0025] Fig. 9 is a flow chart of a scene recognition method in some embodiments of the present application;

[0026] Figure 10-17 It is a flowchart of a scene recognition method in certain embodiments of the present application. DETAILED DESCRIPTION

[0027] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0028] In the subsequent description, the suffixes such as "module", "component" or "unit" used to represent components are only used to facilitate the description of the present invention, and have no specific meanings. Therefore, "module", "component" or "unit" can be used in a mixed manner.

[0029] See also Figure 1 The present application embodiment provides a scene recognition method, the scene recognition method comprising:

[0030] 01. Obtain the positioning information of the positioning device of the terminal equipment, the satellite information of the global satellite navigation system and the original observation information related to the pseudorange;

[0031] 02. Construct the feature vector of the terminal device in the current scenario based on the positioning information, satellite information and original observation information;

[0032] 03. According to the feature vector, the scene classification model is used to identify the scene where the terminal device is located. The scene classification model is constructed and trained by a preset classification algorithm. The scene includes one of an open scene and an occluded scene. Among them, the preset classification algorithm is a machine learning algorithm with data classification capabilities.

[0033] See also Figure 2 The embodiment of the present application provides a scene recognition device 10. The above scene recognition method can be implemented by the scene recognition device 10, and the scene recognition device 10 includes an acquisition module 11, a construction module 12 and a recognition module 13.

[0034] Among them, 01 can be implemented by the acquisition module 11, 02 can be implemented by the construction module 12, and 03 can be implemented by the identification module 13. In other words, the acquisition module 11 can be used to obtain the positioning information of the positioning device of the terminal device, the satellite information of the global satellite navigation system, and the original observation information related to the pseudorange; the construction module 12 can be used to construct the feature vector of the terminal device in the current scene according to the positioning information, satellite information and original observation information. The identification module 13 can be used to identify the scene where the terminal device is located through the scene classification model according to the feature vector. The scene classification model is constructed and trained by a preset classification algorithm. The scene includes one of an open scene and an occluded scene. Among them, the preset classification algorithm is a machine learning algorithm with data classification capabilities.

[0035] Please combine Figure 3 The present application also provides a terminal device 100, which includes a processor 20, a memory 30 and a positioning device 40. The memory 30 stores a computer program. When the computer program is executed by the processor 20, the processor 20 implements the above-mentioned scene recognition method, that is, the processor 20 can be used to obtain the positioning information of the positioning device 40 of the terminal device, the satellite information of the global satellite navigation system and the original observation information related to the pseudorange, and construct a feature vector of the terminal device in the current scene according to the positioning information, the satellite information and the original observation information; and identify the scene where the terminal device is located through a scene classification model according to the feature vector. The scene classification model is constructed and trained by a preset classification algorithm. The scene includes one of an open scene and an occluded scene, wherein the preset classification algorithm is a machine learning algorithm with data classification capabilities.

[0036] In the scene recognition method, scene recognition device 10 and terminal device 100 of the present application, positioning information, satellite information, original observation information and other data that can be used to judge and comprehensively identify the surrounding environment scene are obtained from the positioning device 40 of the terminal device 100, and then a feature vector of the terminal device 100 in the current scene is constructed based on the positioning information, satellite information and original observation information, so that the feature vector contains multiple features that represent the scene of the terminal device 100. In this way, inputting the feature vector into a scene classification model trained by a preset classification algorithm with data analysis capabilities can more accurately identify the occlusion type and characteristics of the scene where the terminal device 100 is located, thereby improving the accuracy of the terminal device 100 in identifying the scene.

[0037] The terminal device 100 of this embodiment includes, but is not limited to, electronic devices such as mobile phones, laptops, tablets, smart cars, smart watches, or other smart wearable devices. For example, this embodiment can be described with the terminal device 100 as a mobile phone, that is, this application implements the above-mentioned scene recognition method through a mobile phone. The scene recognition device can be hardware or software pre-installed on the mobile phone, and the scene recognition method can be executed when it is started and run on the mobile phone. For example, the scene recognition device 10 can be a low-level software code segment of the mobile phone or a part of the operating system. In this way, when the mobile phone can determine the scene of the surrounding environment based on the positioning information of the positioning device 40, the satellite information of the global satellite navigation system, and the original observation information related to the pseudorange, the corresponding operation strategy is executed according to the scene where the terminal device 100 is located to improve the user experience.

[0038] It should be noted that the preset classification algorithm refers to a machine learning algorithm with data classification capability that is pre-set in the terminal device 100 or the scene recognition device. It can be understood that the machine learning algorithm is a method of training a model based on existing data and using this model to predict the future. It takes data as the object, and it extracts data features, discovers knowledge in the data, abstracts the model, and makes predictions about the data. Machine learning algorithms can be classified into clustering algorithms, classification algorithms, regression algorithms, and labeling algorithms according to the tasks completed. In this application, the preset classification algorithm can be a classification algorithm. The preset classification algorithm may include, but is not limited to, a support vector machine (SVM) algorithm, a feature tree classification algorithm, a logistic regression algorithm, or a naive Bayes algorithm.

[0039] The scene classification model is a classification model built and trained based on a preset classification algorithm. The scene classification model may include but is not limited to a support vector machine classification model (SVM), a feature tree classification model, etc. It can be understood that the vector machine classification model can be built and trained by a preset SVM algorithm, and the feature tree classification model can be built by a preset feature tree algorithm and trained through samples. The scene classification model takes a feature vector as input and different scene types as the scene classification model output.

[0040] Those skilled in the art can understand that SVM refers to a linear classifier with the largest interval in the feature space. The learning strategy of SVM is to maximize the interval, which can be formalized as a problem of solving a convex quadratic programming. The learning algorithm of SVM is an optimization algorithm for solving a convex quadratic programming.

[0041] The scene classification model is used to detect according to the input feature vector, so as to determine which scene the current terminal device 100 is in. The scene can include two categories: occlusion scene and open scene. The occlusion scene can be divided into light occlusion, medium occlusion, heavy occlusion, etc. according to the degree of occlusion (for example, Figure 5 ), or, according to the occlusion characteristics, it is divided into small city environment, medium city environment, urban canyon environment, etc., or, according to the number of occlusions, it is divided into single-side occlusion, double-side occlusion, three-side occlusion or four-side occlusion, etc. Correspondingly, the classification degree of the scene by the scene classification model can be set according to actual needs, without specific limitation.

[0042] The Global Navigation Satellite System (GNSS) is an air-based radio navigation and positioning system that can provide users with all-weather 3D coordinates, speed and time information at any location on the Earth's surface or in near-Earth space. GNSS may include but is not limited to the BeiDou Navigation Satellite System (BDS), the Global Positioning System (GPS) and the Galileo Navigation Satellite System (GALILEO).

[0043] Please combine Figure 4 The positioning device 40 may be a GNSS chip, that is, in step 01, the positioning information, the satellite information of the global satellite navigation system, and the original observation information related to the pseudorange may be obtained from the GNSS chip of the terminal device 100. The positioning information may be obtained by the positioning device 40 according to the received observation information and the information provided by the base station, while the satellite information of the global satellite navigation system and the original observation information related to the pseudorange may be directly obtained by the positioning device 40.

[0044] The GNSS chip includes a Location interface, a GnssStatus interface, and a GnssMeasurement interface, wherein the Location interface is used to output the positioning information of the GNSS chip positioning solution, the GnssStatus interface is used to output the satellite information obtained by the GNSS chip, and the GnssMeasurement interface is used to output the observation information observed by the GNSS chip. The processor 20 or the acquisition module can obtain the positioning information from the Location interface, obtain the satellite information from the GnssStatus interface, and obtain the original observation information related to the pseudorange from the GnssMeasurement interface.

[0045] It should also be noted that the processor 20 or the acquisition module 11 can obtain the positioning information of the current epoch in the positioning device 40, the satellite information of the global satellite navigation system, and the original observation information related to the pseudorange, that is, obtain the positioning information, satellite information and original observation information of a single epoch to realize the identification of the scene where the terminal device 100 is located. The processor 20 or the acquisition module 11 can also obtain the positioning information, satellite information and original observation information related to the pseudorange of a predetermined number of epochs in the current time period, that is, obtain the positioning information, satellite information and original observation information of multiple epochs to realize the identification of the scene where the terminal device 100 is located.

[0046] Pseudorange refers to the distance between the measuring station and the satellite, which contains clock errors and atmospheric refraction delays, measured using the pseudo-random code broadcast by the satellite and the receiver replica code when navigating and positioning in the global satellite navigation system.

[0047] It is understandable that the original observation information related to the pseudorange refers to the observation data used to calculate the pseudorange. The original observation information can be the original observation information related to the multi-band pseudorange, for example, it can be the original observation information related to the pseudorange of the GPSL1 band and the GPSL5 band. It is understandable that the pseudorange residual of the GPSL1 band and the pseudorange residual of the GPSL5 band perform differently in a multipath environment. Therefore, the existence of multipath signals can be effectively identified, and multipath signals are the key factor in distinguishing between an open environment and an obstructed environment. In this way, the scene where the terminal device 100 is located can be identified to a certain extent by processing the original observation information related to the pseudorange.

[0048] The positioning information may include but is not limited to the latitude and longitude, altitude, horizontal accuracy estimate HACC, vertical accuracy estimate VACC, and spatial position precision factor PDOP of the GNSS chip positioning solution.

[0049] The horizontal accuracy estimate HACC, vertical accuracy estimate VACC, and spatial position precision factor PDOP parameters of the GNSS chip positioning solution can provide a reliability reference for the pseudorange residuals, thereby improving the reliability of the pseudorange residuals.

[0050] Satellite information may include but is not limited to satellite ephemeris, signal-to-noise ratio CN0, satellite azimuth and UsedInFix information, among which UsedInFix information is used to determine whether the satellite and the corresponding frequency band participate in positioning. If the UsedInFix information of the satellite and the corresponding frequency band is 1, the satellite and the corresponding frequency band participate in positioning. If the UsedInFix information of the satellite and the corresponding frequency band is 0, it means that the satellite and the corresponding frequency band do not participate in positioning.

[0051] See also Figure 6 In some embodiments, the scene recognition method further comprises the steps of:

[0052] 001, obtain the estimated horizontal accuracy of the terminal device;

[0053] 002. When the estimated horizontal accuracy is greater than or equal to the preset accuracy threshold, it is determined that the terminal device is in an open scene.

[0054] Please combine Figure 7 In some embodiments, the scene recognition device 10 further includes a determination module 14, step 001 can be implemented by the acquisition module 11, and step 002 can be implemented by the determination module 14. In other words, the acquisition module 11 can be used to obtain the estimated horizontal accuracy of the terminal device, and the determination module 14 can be used to determine that the terminal device is in an open scene when the estimated horizontal accuracy is greater than or equal to a preset accuracy threshold.

[0055] In some embodiments, the processor 20 may be configured to obtain a horizontal accuracy estimate of the terminal device, and determine that the terminal device is in an open scene when the horizontal accuracy estimate is greater than or equal to a preset accuracy threshold.

[0056] It should be noted that steps 001 and 002 may be implemented before step 01 or after step 01. The preset accuracy threshold may be 3 meters, 5 meters, 7 meters, or 10 meters or even a larger value. Since the current terminal device 100 can generally achieve a positioning error of less than 5 meters in an open environment, in this embodiment, the preset accuracy threshold may be 5 meters, that is, when the estimated horizontal accuracy value is greater than or equal to the preset accuracy threshold, it can be determined that the terminal device 100 is in an open scene.

[0057] For example, please combine Figure 8In some examples, steps 001 and 002 may be implemented before step 01, and the processor 20 may first obtain the estimated horizontal accuracy of the terminal device 100, and when the estimated horizontal accuracy is greater than or equal to the preset accuracy threshold, determine that the current scene is an open scene, and stop subsequent steps. When the estimated horizontal accuracy is less than the preset accuracy threshold, the positioning information of the positioning device 40, the satellite information of the global satellite navigation system, and the original observation information related to the pseudorange are obtained. In this way, when the estimated horizontal accuracy is greater than or equal to the preset accuracy threshold, the current scene is directly judged to be an open scene, and there is no need to perform steps 01 and 02, thereby saving power consumption of the terminal device 100.

[0058] For example, please combine Fig. 9 In some examples, when the processor 20 obtains the positioning information of the positioning device 40, the satellite information of the global satellite navigation system, and the original observation information related to the pseudorange, the horizontal accuracy estimate in the positioning information is compared with the preset accuracy threshold. When the horizontal accuracy estimate is greater than or equal to the preset accuracy threshold, it is determined that the terminal device 100 is in an open scene. When the horizontal accuracy estimate is less than the preset accuracy threshold, a feature vector is constructed based on the positioning information, satellite information, and original observation information. In this way, when the horizontal accuracy estimate is greater than or equal to the preset accuracy threshold, there is no need to add an additional judgment process, which saves the power consumption of the terminal device 100.

[0059] See also Fig.10 In some embodiments, step 02 includes the sub-steps of:

[0060] 021, calculate the pseudo-range residual of each constellation and frequency band based on latitude and longitude, altitude, satellite information and original observation information;

[0061] 022, calculate satellite signal strength based on satellite information;

[0062] 023, calculate the satellite geometric distribution ratio based on satellite information;

[0063] 024, construct a feature vector based on the horizontal accuracy estimate, vertical accuracy estimate, spatial position precision factor, pseudorange residual, satellite signal strength and satellite geometric distribution ratio.

[0064] Please refer to further Figure 2In some implementations, sub-steps 021-024 may be completed by the construction module 12. In other words, the construction module 12 may be used to calculate the pseudo-range residuals of each constellation and frequency band according to the latitude and longitude, altitude, satellite information and original observation information, and calculate the satellite signal strength according to the satellite information. The construction module 12 may also be used to calculate the satellite geometric distribution ratio according to the satellite information, and to construct a feature vector according to the horizontal accuracy estimate, the vertical accuracy estimate, the spatial position precision factor, the pseudo-range residuals, the satellite signal strength and the satellite geometric distribution ratio.

[0065] In some embodiments, the processor 20 can be used to calculate the pseudorange residuals of each constellation and frequency band based on the longitude and latitude, altitude, satellite information and original observation information, and calculate the satellite signal strength based on the satellite information. The processor 20 can also be used to calculate the satellite geometric distribution ratio based on the satellite information, and construct a feature vector based on the horizontal accuracy estimate, the vertical accuracy estimate, the spatial position precision factor, the pseudorange residuals, the satellite signal strength and the satellite geometric distribution ratio.

[0066] It should be noted that the pseudorange residual refers to the residual of the pseudorange observation value. It can be understood that the residual can be considered as the remaining part after the useful information contained in the observation value is deducted. There may be a surplus (the residual is positive) or a shortage (the residual is negative). The residual is a relatively direct and objective standard for evaluating the quality of the observation value. When the quality of the observation value is good, the residual must be near zero and the jitter is small. Otherwise, the observation value may be poor or some deviations or error terms are not completely corrected. Therefore, the processor 20 solves the pseudorange residuals of each satellite and each frequency band, and constructs the pseudorange residuals as feature vectors and inputs them into the scene classification model. It can effectively identify multipath signals based on the pseudorange residuals, so that it can distinguish between open scenes and blocked scenes based on multipath signals in one step.

[0067] The satellite signal strength can be calculated from the signal-to-noise ratio CN0 of the satellite information. Generally, the more severe the scene occlusion of the terminal device 100 is, the weaker the satellite signal strength is due to the influence of obstacles. Conversely, the more spacious the scene of the terminal device 100 is, the stronger the satellite signal strength is. Therefore, the satellite signal strength is constructed as a feature vector and input into the scene classification model, which can improve the scene classification model's recognition accuracy of the scene.

[0068] Specifically, in sub-step 021, the number of satellites whose signal-to-noise ratio is greater than a preset threshold is counted, and a weighted sum is taken for all signal-to-noise ratios greater than the preset threshold to generate satellite signal strength.

[0069] The preset threshold may be 25 dB / Hz, that is, the number of satellites satisfying the signal-to-noise ratio CN0 greater than the threshold 25 dB / Hz is counted, and weighted summation is performed with the signal-to-noise ratio CN0 as the weight to obtain the satellite signal strength. The calculation formula for the satellite signal strength is:

[0070] CN0 25 [k] = ∑ i CN0 i [k]·I{CN0 i [k]>25dB / Hz}

[0071] Among them, CN0 25 [k] represents the satellite signal strength, I{x} is the indicator function, and outputs 1 if event x is true, otherwise outputs 0.

[0072] The satellite geometric distribution ratio can effectively identify the proportion of the blocked area and is an effective indicator for determining unilateral / bilateral occlusion. The processor 20 or the construction module 12 can calculate the satellite geometric distribution ratio based on the satellite azimuth. The calculation formula for the satellite geometric distribution ratio is:

[0073]

[0074] Among them, Azi prop [k] represents the geometric distribution ratio of satellites. The azimuth range from 0° to 360° is divided into 36 sectors with an interval of 10°. If there is a satellite signal in region j and the corresponding signal-to-noise ratio CN0 is greater than the threshold value of 25dB / Hz, then d j =1, otherwise 0.

[0075] The feature vector can be represented as:

[0076]

[0077] Among them, v[k] is the eigenvector, is the pseudorange residual of each satellite and each frequency band. Specifically, Indicates the pseudorange residual of GPS L1 band, Represents the pseudorange residual of GPSL5 band, represents the pseudorange residual of GLO G1 band, represents the pseudorange residual of BDS B1I band, represents the pseudorange residual of BDS B2a band, represents the pseudorange residual in the GAL E1 band, Indicates the pseudorange residual of the GAL E5A band, CN0 25 is the satellite signal strength, pdop is the spatial position precision factor, Azi propis the satellite geometric distribution ratio, hacc is the horizontal accuracy estimate, and vacc is the vertical accuracy estimate.

[0078] In this way, by inputting a feature vector containing multiple features into a scene classification model, the scene in which the terminal device 100 is located can be accurately identified.

[0079] In addition, the features included in the above feature vector are only used as an embodiment, and the specific features of the feature vector are not limited. For example, in some examples, the vector feature may only include a part of the above features (such as only pseudorange residuals, satellite signal strength and positioning information). For example, in some examples, not only all the features in the above feature vector are included, but other features may also be added. In this way, the accuracy of scene recognition of the terminal device 100 is further improved.

[0080] See also Fig.11 In some embodiments, step 021 further includes the sub-steps of:

[0081] 0211, calculate the pseudo-range observation value of each constellation and frequency band based on the original observation information;

[0082] 0212, calculate the pseudorange estimate for each constellation and frequency band based on the satellite ephemeris, longitude, latitude and altitude;

[0083] 0213, calculate the pseudorange residual of each constellation and frequency band based on the pseudorange observation value and the pseudorange estimation value.

[0084] In some implementations, sub-steps 0211-0213 may be completed by the construction module 12. In other words, the construction module 12 may be used to calculate the pseudorange observation value of each constellation and frequency band according to the original observation information, calculate the pseudorange estimation value of each constellation and frequency band according to the satellite ephemeris, longitude and latitude, and altitude, and calculate the pseudorange residual of each constellation and frequency band according to the pseudorange observation value and the pseudorange estimation value.

[0085] In some embodiments, the processor 20 can be used to calculate the pseudorange observation value of each constellation and frequency band based on the original observation information, and calculate the pseudorange estimation value of each constellation and frequency band based on the satellite ephemeris, longitude and latitude and altitude, and calculate the pseudorange residual of each constellation and frequency band based on the pseudo-observation value and the pseudorange estimation value.

[0086] Specifically, the calculation formula of the pseudorange residual (Res[k]) is:

[0087]

[0088] Among them, ρ m is the pseudorange observation value, ρ e is the pseudorange estimate, S FixThe satellite set that participates in the positioning solution (parameter UsedInFix = 1) in this constellation and frequency band. m It can be calculated from the original observation information, and the pseudorange estimate ρ e Including the estimated geometric distance r e and the receiver clock error estimate δt e , estimated value of the geometric distance between the satellite and the ground r e and the receiver clock error estimate δt e It is calculated from the positioning information and satellite information. Specifically, the estimated value of the satellite-to-ground geometric distance r e It can be calculated based on the satellite ephemeris of the satellite information and the longitude, latitude and altitude of the positioning information, δt e It can be obtained by the least squares positioning algorithm.

[0089] In this way, the multipath signal can be effectively identified according to the pseudo-range residual, so that the open scene and the blocked scene can be distinguished according to the multipath signal in one step.

[0090] See also Fig.12 In some embodiments, step 03 further includes:

[0091] 031, identifying the occlusion type of the scene where the terminal device is located according to the satellite geometric distribution ratio and spatial position precision factor in the feature vector;

[0092] 032, identifying the degree of occlusion of the scene where the terminal device is located according to the pseudo-range residual and the satellite signal strength in the feature vector;

[0093] 033, identify the scene where the terminal device is located according to the occlusion type, occlusion degree, horizontal accuracy estimate and vertical accuracy estimate.

[0094] In some embodiments, sub-steps 031 and 032 may be implemented by the identification module 13. In other words, the identification module 13 may be used to identify the occlusion type of the scene where the terminal device is located according to the satellite geometric distribution ratio and the spatial position precision factor in the feature vector, and the identification module 13 may also be used to identify the occlusion degree of the scene where the terminal device is located according to the pseudorange residual and the satellite signal strength in the feature vector, and to identify the scene where the terminal device is located according to the occlusion type, occlusion degree, horizontal precision estimate, and vertical precision estimate.

[0095] In some embodiments, the processor 20 can be used to identify the occlusion type of the scene where the terminal device is located based on the satellite geometric distribution ratio and the spatial position precision factor in the feature vector. The processor 20 can also be used to identify the occlusion degree of the scene where the terminal device is located based on the pseudorange residual and the satellite signal strength in the feature vector, and to identify the scene where the terminal device is located based on the occlusion type, occlusion degree, horizontal accuracy estimate, and vertical accuracy estimate.

[0096] It should be noted that the occlusion type can be no occlusion, unilateral occlusion, bilateral occlusion and multilateral occlusion, etc. It can be understood that the satellite geometric distribution ratio can characterize which areas of the 36 fan-shaped areas with an interval of 10° have satellite signals and the signal-to-noise ratio CN0 is greater than 25dB / Hz of the positioning satellite. Therefore, the satellite geometric distribution ratio can distinguish to a certain extent which areas are blocked or unblocked, that is, identify the occlusion type of the scene, and the spatial position precision factor is used to characterize the geometric distribution of satellites. The better the distribution of satellites in the sky, the higher the positioning accuracy of the spatial position precision factor (the smaller the value, the higher the accuracy). Therefore, the satellite geometric distribution ratio and the spatial position precision factor can identify the occlusion type of the scene where the terminal device is located, and the occlusion type of the scene where the terminal device is located can be obtained more accurately.

[0097] The degree of occlusion can be divided into no occlusion, light occlusion, moderate occlusion or heavy occlusion, etc. Of course, it can be understood that the degree of occlusion can be divided according to actual needs. For example, in some examples, the degree of occlusion can also be divided into no occlusion and occlusion.

[0098] It can be understood that the stronger the satellite signal strength, the lower the degree of obstruction. Conversely, the weaker the satellite signal strength, the more severe the degree of obstruction. The pseudorange residual can identify the existence of multipath signals, and multipath signals distinguish between open environments and obstructed environments. Therefore, the satellite signal strength and pseudorange residual can be used to identify the degree of obstruction in the scene where the terminal device is located.

[0099] In this way, the scene in which the terminal device 100 is located can be accurately identified through the occlusion type, occlusion degree, horizontal accuracy estimate and vertical accuracy estimate.

[0100] See also Fig.13 In some embodiments, the characteristic vector includes multiple, and step 03 further includes sub-steps:

[0101] 034, calculate the average eigenvector based on multiple eigenvectors;

[0102] 035, the average feature vector is input into the scene classification model to identify the scene where the terminal device is located.

[0103] In some implementations, sub-steps 034 and 035 may be implemented by the identification module 13. In other words, the identification module 13 may be used to calculate an average feature vector based on multiple feature vectors, and input the average feature vector into a scene classification model to identify the scene where the terminal device is located.

[0104] In some embodiments, the processor 20 may be configured to calculate an average feature vector based on multiple feature vectors, and input the average feature vector into a scene classification model to identify the scene in which the terminal device is located.

[0105] It should be noted that the processor 20 or the acquisition module 11 acquires positioning information, satellite information and original observation information of multiple epochs, and the positioning information, satellite information and original observation information of each epoch construct a feature vector. The specific number of epochs is not limited, and can be 2, 3, 5, 8, 10 or even more.

[0106] The processor 20 performs filtering processing on the input features of the multi-epoch time domain. It is understandable that when the processor 20 performs scene determination based on the data of a single epoch, the rapid changes in the data of the positioning device 40 (such as the signal-to-noise ratio CN0) between epochs may cause instability in the scene determination results. Therefore, by acquiring data of multiple epochs, the deviation can be reduced and the accuracy of scene determination can be improved.

[0107] In addition, in some examples, the processor 20 can also filter multiple feature vectors to remove abnormal feature vectors, and calculate an average feature vector based on the filtered feature vectors, and then input the average feature vector into the scene classification model to identify the scene where the terminal device 100 is located. In this way, the scene judgment can be avoided from being affected by abnormal feature vectors, thereby further improving the accuracy of scene judgment.

[0108] It can be understood that since each feature vector is obtained from the positioning information, satellite information and original observation information related to the pseudorange of a single epoch, in some examples, the processor 20 can also pre-process the positioning information, satellite information and original observation information of pseudoranges of multiple epochs, remove abnormal epoch data, and then calculate the feature vector, so as to ensure the robustness of scene judgment based on the positioning information, satellite information and original observation information of pseudoranges.

[0109] See also Fig.14 In some embodiments, step 035 includes the sub-steps of:

[0110] 0351, calculate pseudorange noise based on original observation information;

[0111] 0352, calculate the change rate of satellite signal strength based on satellite information;

[0112] 0353, the average feature vector, pseudorange noise and / or the rate of change of satellite signal strength are input into the scene classification model to identify the scene where the terminal device is located.

[0113] In some embodiments, sub-steps 0351-0353 may be implemented by the identification module 13. In other words, the identification module 13 may be used to calculate the pseudorange noise based on the original observation information, calculate the rate of change of the satellite signal strength based on the satellite information, and input the average feature vector, the pseudorange noise and / or the rate of change of the satellite signal strength into the scene classification model to identify the scene where the terminal device is located.

[0114] In some embodiments, the processor 20 can be used to calculate pseudorange noise based on the original observation information, calculate the rate of change of satellite signal strength based on the satellite information, and input the average eigenvector, pseudorange noise and / or the rate of change of satellite signal strength into a scene classification model to identify the scene where the terminal device is located.

[0115] Pseudorange noise refers to the pseudorange measurement error caused by the tropospheric and ionospheric delay errors that are not eliminated by the GNSS receiver, as well as the orbital errors and clock error residuals that are not modeled. Pseudorange noise can be calculated by zero baseline (short baseline) station-to-station single difference, epoch-to-epoch triple difference, or multipath combination. The rate of change of satellite signal strength can characterize the degree of change of satellite signal strength over multiple epochs.

[0116] It can be understood that the rate of change of pseudorange noise and satellite signal strength can increase the dimension of the input features, thereby making the scene classification model more accurate in determining the scene where the terminal device 100 is located, further improving the robustness of the recognition result.

[0117] See also Fig.15 In some embodiments, step 03 further includes the sub-steps of:

[0118] 036, input multiple feature vectors into the scene classification model to identify the scene where the terminal device is located based on multiple output results of the scene classification model.

[0119] In some embodiments, sub-step 036 may be implemented by the identification module 13. In other words, the identification module 13 may be used to input multiple feature vectors into the scene classification model to identify the scene where the terminal device is located according to multiple output results of the scene classification model.

[0120] In some embodiments, the processor 20 can be used to input multiple feature vectors into the scene classification model to identify the scene in which the terminal device is located based on multiple output results of the scene classification model.

[0121] It should be noted that, in this embodiment, the processor 20 performs filtering processing on the output features of the multi-epoch time domain, thereby improving the accuracy of scene recognition.

[0122] For example, in some examples, the feature vectors include 10. If the 10 feature vectors are input into the scene classification model in sequence and 10 output results are obtained respectively, among which 9 output results are occlusion scenes and 1 output result is an open scene, then it can be considered that the scene where the terminal device 100 is located is an occlusion scene.

[0123] See also Fig.16 In some embodiments, the feature vector includes multiple, and the scene recognition method further includes:

[0124] 05. Perform an estimated calculation on the positioning information of the positioning device according to the identification result of the scene classification model to generate an estimated result, wherein the estimated result is used to characterize the current positioning accuracy of the positioning device;

[0125] 06. Maintain the current positioning information of the positioning device according to the current positioning accuracy represented by the estimated result or start a preset positioning algorithm corresponding to the current positioning accuracy and the scene where the terminal device is located to process the current positioning information to improve the positioning accuracy of the positioning device.

[0126] Please combine Figure 7 In some embodiments, the scene recognition device 10 further includes a calculation module 15 and a start module 16, and step 05 can be implemented by the calculation module 15, and step 06 can be implemented by the start module 16. In other words, the calculation module 15 can be used to perform an estimated calculation on the positioning information of the positioning device according to the identification result of the scene classification model to generate an estimated result, wherein the estimated result is used to characterize the current positioning accuracy of the positioning device, and the start module 16 can be used to maintain the current positioning information of the positioning device according to the current positioning accuracy characterized by the estimated result, or to start a preset positioning algorithm corresponding to the current positioning accuracy and the scene where the terminal device is located to process the current positioning information to improve the positioning accuracy of the positioning device.

[0127] In some embodiments, the processor 20 can be used to perform an estimated calculation on the positioning information of the positioning device based on the identification results of the scene classification model to generate an estimated result, wherein the estimated result is used to characterize the current positioning accuracy of the positioning device, and maintain the current positioning information of the positioning device based on the current positioning accuracy represented by the estimated result, or start a preset positioning algorithm corresponding to the current positioning accuracy and the scene where the terminal device is located to process the current positioning information to improve the positioning accuracy of the positioning device.

[0128] There may be multiple preset positioning algorithms, and each preset positioning algorithm may be applicable to a corresponding scenario. For example, the preset positioning algorithm may include but is not limited to a moderate occlusion positioning algorithm, a heavy occlusion positioning algorithm, an urban canyon positioning algorithm, etc. Among them, the moderate occlusion positioning algorithm is applicable to moderate occlusion scenes, the heavy occlusion positioning algorithm is applicable to heavy occlusion scenes, and the urban canyon positioning algorithm is applicable to urban canyon scenes.

[0129] The activation of the preset positioning algorithm can be determined by the current positioning accuracy represented by the estimated result. For example, when the recognition result of the scene classification model is an open scene, and the accuracy of the current positioning information represented by the estimated result is accurate, the positioning information of the positioning device 40 is maintained. When the recognition result of the scene classification model is a heavily occluded scene, and the positioning accuracy of the positioning information represented by the estimated result is poor, the heavily occluded positioning algorithm can be activated to improve the positioning accuracy of the positioning device 40 in the heavily occluded scene. When the recognition result of the scene classification model is an urban canyon environment, and the positioning accuracy of the estimated positioning information represented by the estimated result is poor, the urban canyon positioning algorithm can be activated to improve the positioning accuracy of the positioning device 40 in the urban canyon.

[0130] In this way, the positioning information of the positioning device 40 is estimated by the identification result generated by the scene classification model, so as to determine whether to maintain the positioning information of the positioning device 40 or start the preset positioning algorithm according to the estimation result. In this way, the accuracy of the positioning precision of the terminal device 100 in the current scene can be guaranteed.

[0131] Please combine Fig.17 In some embodiments, before step 01, the scene recognition method further includes:

[0132] 001, construct an initial classification model according to the preset classification algorithm;

[0133] 002, obtaining training samples of the initial classification model, where the training samples include multiple feature vectors and multiple scene types corresponding to different positioning information, satellite information, and original observation information;

[0134] 003, taking multiple feature vectors as inputs of a scene classification model, and taking multiple scene types as outputs of an initial classification model, so as to train the initial classification model;

[0135] 004, use the trained initial classification model as the scene classification model.

[0136] Please further combine Figure 7In some embodiments, the scene recognition device 10 may further include a training module 17, step 001 may be implemented by the construction module 12, step 002 may be implemented by the acquisition module 11, step 003 may be implemented by the training module 17, and step 004 may be implemented by the determination module 14. In other words, the construction module 12 is also used to construct an initial classification model according to a preset classification algorithm; the acquisition module 11 is also used to obtain training samples of the initial classification model, and the training samples include multiple feature vectors and multiple scene types corresponding to different positioning information, satellite information, and original observation information; the training module 17 can be used to use multiple feature vectors as inputs of the initial classification model and multiple scene types as outputs of the initial classification model to train the initial classification model, and the determination module 14 is also used to use the trained initial classification model as a scene classification model.

[0137] In some embodiments, the processor can also be used to construct an initial classification model according to a preset classification algorithm and obtain training samples for the initial classification model, the training samples including multiple feature vectors and multiple scene types corresponding to different positioning information, satellite information, and original observation information, and use multiple feature vectors as inputs to the initial classification model and use multiple scene types as outputs of the initial classification model to train the initial classification model, preferably using the trained initial classification model as a scene classification model.

[0138] In this implementation, the preset classification algorithm may be a support vector machine, that is, an initial classification model is constructed according to the support vector machine.

[0139] Multiple feature vectors can be divided into multiple categories, each category of feature vectors represents a scene type. The multiple categories of feature vectors can be input into the initial classification model respectively to obtain output results, and the initial classification model for the scene can be optimized according to the output results and the corresponding scene types, so that the accuracy of the output results of the initial classification model reaches a set threshold, so that the initial classification model can be used as a scene classification model.

[0140] For example, in some examples, the training samples are obtained in moderate occlusion scenes and open scenes, then the output of the initial classification model may include moderate occlusion scenes and open scenes, if multiple feature vectors are respectively input into the initial classification model, if the accuracy of the output result of the initial classification model is lower than the set threshold (for example, multiple feature vectors obtained in the moderate occlusion scene are input into the initial classification model and most of the output results are not moderate occlusion scenes), then the parameters of the initial classification model are adjusted to improve the accuracy of the output result of the initial classification model, that is, the output result of the initial classification model is basically consistent with the actual scene type (the accuracy reaches the set threshold), then the initial classification model is used as the scene classification model.

[0141] In this way, the current scene of the terminal device can be identified based on the trained scene classification model.

[0142] The embodiment of the present application further provides a non-volatile computer-readable storage medium, which stores a computer program. When the computer program is executed by the processor 20, the processor 20 executes the above-mentioned scene recognition method.

[0143] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (digital subscriber line, DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a digital video disc (DVD)), or a semiconductor medium (eg, a solid state disk (SSD)).

[0144] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0145] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0146] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0147] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A scene recognition method, used in a terminal device, characterized in that: The scene recognition method comprises: Acquire positioning information of a positioning device of the terminal device, satellite information of a global satellite navigation system, and original observation information related to pseudorange; Constructing a feature vector of the terminal device in the current scenario according to the positioning information, satellite information and the original observation information; The scene in which the terminal device is located is identified according to the feature vector through a scene classification model, wherein the scene classification model is constructed and trained by a preset classification algorithm, and the scene includes one of an open scene and an occluded scene, wherein the preset classification algorithm is a machine learning algorithm with data classification capabilities.

2. The scene recognition method according to claim 1, characterized in that: The positioning information includes a horizontal accuracy estimate, and the scene recognition method further includes: When the horizontal accuracy estimate is greater than or equal to a preset accuracy threshold, it is determined that the terminal device is in an open scene.

3. The scene recognition method according to claim 1, characterized in that: The positioning information includes a horizontal accuracy estimate, a vertical accuracy estimate, a spatial position precision factor, longitude and latitude, and altitude. Constructing a feature vector of the terminal device in the current scenario according to the positioning information, satellite information, and the original observation information includes: Calculate the pseudorange residual of each constellation and frequency band according to the latitude and longitude, the altitude, the satellite information and the original observation information; Calculate satellite signal strength according to the satellite information; Calculate the satellite geometric distribution ratio according to the satellite information; The feature vector is constructed according to the horizontal accuracy estimate, the vertical accuracy estimate, the spatial position precision factor, the pseudorange residual, the satellite signal strength and the satellite geometric distribution ratio.

4. The scene recognition method according to claim 3, characterized in that: The eigenvector is realized by the following conditions: CN0 25 ,dop,Azi prop [this, this, this] Among them, v[k] represents the feature vector, Indicates the pseudorange residual of GPS L1 band, represents the pseudorange residual of GPS L5 band, represents the pseudorange residual of GLO G1 band, represents the pseudorange residual of BDS B1I band, represents the pseudorange residual of BDS B2a band, represents the pseudorange residual in the GAL E1 band, Indicates the pseudorange residual of the GALE5A band, CN0 25 Indicates the satellite signal strength, pdop indicates the spatial position precision factor, Azi prop represents the satellite geometric distribution ratio, h represents the horizontal accuracy estimate, and vacc represents the vertical accuracy estimate.

5. The scene recognition method according to claim 3, characterized in that: The satellite information includes satellite ephemeris, and the pseudorange residual of each constellation and frequency band is calculated according to the longitude and latitude, the altitude, the satellite information and the original observation information, including: Calculate the pseudorange observation value of each constellation and frequency band according to the original observation information; Calculate a pseudorange estimation value for each constellation and frequency band according to the satellite ephemeris, the longitude and latitude, and the altitude; The pseudorange residual of each constellation and frequency band is calculated according to the pseudorange observation value and the pseudorange estimation value.

6. The scene recognition method according to claim 5, characterized in that: The pseudorange residual is achieved by the following conditions: Where Res[k] represents the pseudorange residual, ρ m is the pseudorange observation value, calculated from the original observation information, ρ e is a pseudorange estimation value, which is calculated from the satellite ephemeris, the longitude and latitude, and the altitude, S Fix It is the set of satellites involved in positioning solution.

7. The scene recognition method according to claim 3, characterized in that: The satellite information includes a signal-to-noise ratio, and calculating the satellite signal strength according to the satellite information includes: Counting the number of satellites whose signal carrier-to-noise ratio is greater than a preset threshold; The satellite signal strength is generated by performing a weighted summation on all signal-to-noise ratios greater than a preset threshold.

8. The scene recognition method according to claim 7, characterized in that: The satellite signal strength is achieved by the following conditions: Among them, CN0 25 [k] represents the satellite signal strength, CN0 is the signal-to-noise ratio, I{x} is the indicator function, and outputs 1 if event x is true, otherwise outputs 0.

9. The scene recognition method according to claim 3, characterized in that: The positioning information includes a spatial position precision factor, the satellite information includes a satellite azimuth, and calculating the satellite geometric distribution ratio according to the satellite information includes: The satellite geometric distribution ratio is calculated according to the satellite azimuth angle.

10. The scene recognition method according to claim 9, characterized in that: The satellite geometric distribution ratio is achieved by the following conditions: Among them, Azi prop [k] represents the geometric distribution ratio of satellites. The azimuth range from 0° to 360° is divided into 36 sectors with an interval of 10°. If there is a satellite signal in region j and the corresponding CN0 is greater than the threshold of 25dB / Hz, then d j =1, otherwise 0.

11. The scene recognition method according to claim 3, characterized in that: Identifying the scene where the terminal device is located according to the feature vector using a scene classification model includes: Identify the occlusion type of the scene where the terminal device is located according to the satellite geometric distribution ratio in the feature vector; Identify the degree of occlusion of the scene where the terminal device is located according to the pseudorange residual and the satellite signal strength in the feature vector; The scene where the terminal device is located is identified according to the occlusion type, the occlusion degree, the horizontal accuracy estimate and the vertical accuracy estimate.

12. The scene recognition method according to claim 1, characterized in that: The feature vectors include a plurality of feature vectors, and the scene where the terminal device is located is identified according to the feature vectors by a scene classification model, including: Calculate an average feature vector based on a plurality of the feature vectors; The average feature vector is input into the scene classification model to identify the scene where the terminal device is located.

13. The scene recognition method according to claim 12, characterized in that: The step of inputting the average feature vector into the scene classification model to identify the scene where the terminal device is located includes: Calculating pseudorange noise according to the original observation information; Calculate the rate of change of satellite signal strength according to the satellite information; The average feature vector, the pseudorange noise and / or the rate of change of the satellite signal strength are input into the scene classification model to identify the scene where the terminal device is located.

14. The scene recognition method according to claim 1, characterized in that: The feature vectors include a plurality of feature vectors, and the scene where the terminal device is located is identified according to the feature vectors by a scene classification model, including: Input a plurality of the feature vectors into the scene classification model to identify the scene in which the terminal device is located according to a plurality of output results of the scene classification model.

15. The scene recognition method according to claim 1, characterized in that: The scene recognition method further includes: Performing an estimation calculation on the positioning information of the positioning device according to the identification result of the scene classification model to generate an estimation result, wherein the estimation result is used to characterize the current positioning accuracy of the positioning device; Maintain the current positioning information of the positioning device according to the current positioning accuracy represented by the estimated result, or start a preset positioning algorithm corresponding to the current positioning accuracy and the scene where the terminal device is located to process the current positioning information to improve the positioning accuracy of the positioning device.

16. The scene recognition method according to claim 1, characterized in that: The scene recognition method further includes: Constructing an initial classification model according to the preset classification algorithm; Acquire training samples of the initial classification model, wherein the training samples include multiple feature vectors and multiple scene types corresponding to different positioning information, satellite information, and original observation information; Using the multiple feature vectors as inputs of the initial classification model and using the multiple scene types as outputs of the initial classification model to train the initial classification model; The trained initial classification model is used as the scene classification model.

17. A scene recognition device, used in a terminal device, characterized in that: The scene recognition device comprises: An acquisition module, used to acquire positioning information of a positioning device of the terminal device, satellite information of a global satellite navigation system, and original observation information related to pseudorange; A construction module, used to construct a feature vector of the terminal device in the current scenario according to the positioning information, satellite information and the original observation information; A recognition module is used to recognize the scene in which the terminal device is located through a scene classification model according to the feature vector, wherein the scene classification model is constructed and trained by a preset classification algorithm, and the scene includes one of an open scene and an occluded scene, wherein the preset classification algorithm is a machine learning algorithm with data classification capabilities.

18. A terminal device, characterized in that: The method comprises a processor and a storage unit, wherein the storage unit stores at least one computer program, and when the computer program is executed by the processor, the processor executes the scene recognition method according to any one of claims 1 to 16.

19. A non-volatile computer-readable storage medium containing a computer program, characterized in that: When the computer program is executed by a processor, the processor is caused to execute the scene recognition method according to any one of claims 1 to 16.

Citation Information

Cited By

  • Intelligent receiving method and device of communication signal

    CN120238404A