Method and system for high-precision positioning using scene-adaptive detection technology

Abstract

The application provides a method and system for high-precision positioning using scene adaptive detection technology, the method comprising collecting multiple sets of static observation data and multiple sets of dynamic observation data of an observed scene, forming a data set, and labeling the scene where the data set is located, calculating the data set using a high-precision single-point positioning method to separate the environmental error in the observation value, using the labeled data set to calculate satellite elevation angle data, inputting the satellite elevation angle data and the environmental error into a support vector machine, Light GBM, random forest, or other model for training, selecting the model with the minimum error as the current model according to the model output error, saving the current model, and using the current model as a positioning algorithm to provide an environmental noise estimation value when real-time positioning is performed, so that the positioning algorithm uses the estimation value as the environmental noise variance in the random model, thereby improving the positioning accuracy of the algorithm.

Description

Technical Field

[0001] This application relates to the field of detection and positioning technology, and in particular to a method and system for high-precision positioning using scene adaptive detection technology. Background Technology

[0002] The positioning performance of a Global Navigation Satellite System (GNSS) is closely related to its environment. For example, when using GNSS for high-precision measurements, it is important to choose an open environment and avoid areas with severe multipath effects, such as tall buildings and water surfaces. Environmental factors can serve as scene information to enhance the performance of multi-sensor fusion positioning and are an indispensable component of next-generation navigation and positioning technologies.

[0003] In traditional methods, to reduce the impact of environmental factors, an empirical variance is typically set in the stochastic model to relax the constraints of the stochastic model corresponding to environmental factors, thereby reducing the load on the model's filtering process from drastically changing environmental noise. However, a fixed empirical variance cannot meet the needs of dynamically changing environments and cannot effectively improve positioning accuracy. Summary of the Invention

[0004] This application provides a method and system for high-precision positioning using scene adaptive detection technology. The invention performs statistical analysis on various satellite GNSS signal elements provided by the receiver and calculates the environmental noise estimate of the current location by combining it with a pre-trained model. This estimate is then applied to the GNSS high-precision positioning process to improve positioning accuracy. This solves the problem that traditional positioning methods cannot meet the needs of dynamic and changing environments and cannot effectively improve positioning accuracy.

[0005] In a first aspect, embodiments of this application provide a method for high-precision positioning using scene adaptive detection technology, comprising:

[0006] Collect multiple sets of static observation data and multiple sets of dynamic observation data of the observed scene to form a dataset, and label the scene in which the dataset is located;

[0007] The dataset is calculated using a high-precision single-point positioning method to separate environmental errors from the observed values;

[0008] Using the labeled dataset, the satellite elevation angle, azimuth angle, accuracy factor, satellite number change rate, signal-to-noise ratio change rate, and their statistical values ​​were calculated.

[0009] The satellite elevation angle, azimuth angle, accuracy factor, satellite number change rate, signal-to-noise ratio change rate, statistical values, and environmental error are input into models such as support vector machine, Light GBM, and random forest for training. The model with the smallest error is selected as the current model based on the model output error and the current model is saved.

[0010] During real-time positioning, the current model is used as the positioning algorithm to provide an estimated value of environmental noise. The positioning algorithm uses this estimated value as the variance of environmental noise in the stochastic model to improve the positioning accuracy of the algorithm.

[0011] In one feasible implementation, the method also includes:

[0012] During the user's real-time location process, the observed data is stored and then extracted and added to the dataset after the process ends to obtain an updated dataset. Based on the updated dataset, the model is updated, trained, and optimized to improve model accuracy.

[0013] In one feasible implementation, the high-precision single-point positioning method is used to calculate the dataset and separate the environmental error in the observations. The operation method is as follows:

[0014] First, based on the statically acquired data, a static long-term baseline calculation is performed by combining the known GNSS reference station data with the current data to obtain the accurate geographical location of the receiver data.

[0015] For the dynamically acquired data, dynamic baseline calculation is performed by combining the known GNSS reference station, the static acquired data, and the current dynamic data to obtain accurate dynamic geographic location coordinates;

[0016] Then, using the precise satellite orbits, precise satellite clock errors, high-precision ionospheric models, and tropospheric products and receivers released by the IGS organization, and accurate geographical locations, the remaining errors in the positioning process are removed, leaving only the errors caused by environmental noise in the model. The errors caused by environmental noise are then output according to different satellites and different epochs.

[0017] In one feasible implementation, the labeled dataset is used to calculate the satellite elevation angle, azimuth angle, accuracy factor, satellite number change rate, signal-to-noise ratio change rate, and their statistical values. The operation method is as follows:

[0018] The formulas for calculating the satellite elevation angle E1 and azimuth angle Az are as follows:

[0019]

[0020]

[0021] Az = arctan(e e / en );

[0022] El = arcsin(e u );

[0023] Where r s and r r Let λ be the coordinates of the receiver and the satellite in the geocentric coordinate system. This refers to the longitude and latitude of the receiver's location;

[0024] The formula for calculating the precision factor GDOP is as follows:

[0025]

[0026] Where H is the normal equation matrix constructed using the least squares algorithm during the localization process;

[0027] The formulas for calculating the number of satellites nSat and its rate of change dnSat are as follows:

[0028]

[0029] Where nSat(t1) is the number of satellites at time t1, and nSat(t2) is the number of satellites at time t2;

[0030] The calculation methods for the signal-to-noise ratio change rate dSNR and its statistical values ​​are as follows:

[0031]

[0032]

[0033]

[0034] Where avg and std are the mean and standard deviation of the rate of change of signal-to-noise ratio.

[0035] In one feasible implementation, both the static observation data and the dynamic observation data are stored in RINEX 3.05 format and are jointly categorized and stored to form the dataset.

[0036] Secondly, embodiments of this application also provide a system for high-precision positioning using scene adaptive detection technology. The system executes the above-described method and includes:

[0037] The receiver is configured to acquire multiple sets of static observation data and multiple sets of dynamic observation data of the observed scene, form a dataset, and label the scene in which the dataset is located.

[0038] A GNSS reference station, electrically connected to the receiver, is configured to perform static long-term baseline calculation based on the static observation data to obtain the accurate geographical location of the receiver data; and is also configured to perform dynamic baseline calculation based on the dynamic observation data to obtain accurate dynamic geographical location coordinates.

[0039] A computer, communicatively connected to the receiver and the GNSS reference station, is configured to: separate environmental errors from observations based on the accurate geographic coordinates of the terminal products and receiver released by the IGS organization; the computer is also configured to: use the labeled dataset to calculate satellite elevation angle, azimuth angle, accuracy factor, satellite number change rate, signal-to-noise ratio change rate, and their statistical values, and input the calculation results into models such as support vector machine, LightGBM, and random forest for training, and select the model with the smallest error as the current model based on the model output error, and save the current model.

[0040] This application provides a method and system for high-precision positioning using scene adaptive detection technology. The positioning method collects multiple sets of static and dynamic observation data of the observed scene to form a dataset, labels the scene containing the dataset, and uses a high-precision single-point positioning method to calculate and separate environmental errors from the observation values. Using the labeled dataset, satellite elevation angle, azimuth angle, accuracy factor, satellite number change rate, signal-to-noise ratio change rate, and their statistical values ​​are calculated. The satellite elevation angle, azimuth angle, accuracy factor, satellite number change rate, signal-to-noise ratio change rate, statistical values, and the environmental errors are input into models such as Support Vector Machine (SVM), Light GBM, and Random Forest for training. The model with the smallest output error is selected as the current model and saved. During real-time positioning, the current model is used as the positioning algorithm to provide an environmental noise estimate. The positioning algorithm uses this estimate as the environmental noise variance in the random model, improving the positioning accuracy. This invention performs statistical analysis on various satellite GNSS signal elements provided by the receiver and calculates the estimated environmental noise at the current location using a pre-trained model. This estimated value is then applied to the GNSS high-precision positioning process to improve positioning accuracy. Attached Figure Description

[0041] Figure 1 This is a flowchart of the method for high-precision positioning using scene adaptive detection technology provided in this application;

[0042] Figure 2 This is a schematic diagram of a method for high-precision positioning using scene adaptive detection technology;

[0043] Figure 3This is a structural block diagram of the system for high-precision positioning using scene adaptive detection technology provided in this application. Detailed Implementation

[0044] To enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this application.

[0045] In traditional methods, to reduce the impact of environmental factors, an empirical variance is typically set in the stochastic model to relax the constraints of the stochastic model corresponding to environmental factors, thereby reducing the load on the model's filtering process due to drastically changing environmental noise. However, a fixed empirical variance cannot meet the requirements of dynamically changing environments and cannot effectively improve positioning accuracy. This invention performs statistical analysis on various satellite GNSS signal elements provided by the receiver and combines this with a pre-trained model to calculate an estimated environmental noise level for the current location. This estimated value is then applied to the GNSS high-precision positioning process to improve positioning accuracy, thus solving the problem that traditional positioning methods cannot meet the requirements of dynamically changing environments and cannot effectively improve positioning accuracy.

[0046] The following detailed description, in conjunction with the accompanying drawings, illustrates the specific structure of the high-precision positioning method using scene adaptive detection technology provided in this application.

[0047] Example 1:

[0048] Reference Figure 1-2 As shown in the figure, this application provides a method for high-precision positioning using scene adaptive detection technology, including:

[0049] Step 1: Utilize a full-band, full-system receiver to conduct continuous observations in various scenarios, including but not limited to open outdoor scenes, outdoor roadside tree-covered scenes, outdoor tall building-only-side-covered scenes, urban canyon scenes, under urban overpasses, and under urban interchanges. Collect static observation data for more than 24 hours and multiple sets of dynamic observation data from the above scenarios to form a dataset, and label the scenarios in which the dataset is located.

[0050] Step 2: Use a high-precision single-point positioning method to calculate the dataset and separate the environmental error in the observations.

[0051] Step 3: Using the satellite elevation angle, azimuth angle, accuracy factor, satellite number change rate, signal-to-noise ratio change rate and their statistical values ​​calculated from the labeled dataset in Step 1, and the corresponding environmental error from the dataset in Step 2 as input, train the model using models such as Support Vector Machine, Light GBM, and Random Forest to obtain the relationship between the above factors and the environmental error, and store the model.

[0052] Step 4: When the user is performing real-time positioning, the model trained in Step 3 is used to provide an environmental noise estimate for the positioning algorithm. The positioning algorithm uses this estimate as the environmental noise variance in the stochastic model to improve the positioning accuracy of the algorithm.

[0053] Step 5: Store the observation data during the user's real-time positioning process and extract the data after the process ends, adding it to the dataset constructed in Step 1. Repeat Step 2 and Step 3 to update, train, and optimize the model, thereby improving the model accuracy.

[0054] In some embodiments, the operation method described in step one of the above technical solutions, which involves using a full-band full-system receiver to continuously observe in various scenarios including but not limited to open outdoor scenes, outdoor roadside tree-covered scenes, outdoor tall building-only-side-covered scenes, urban canyon scenes, under urban overpasses, and under urban interchanges, collecting static observation data for more than 24 hours and multiple sets of dynamic observation data to form a dataset, and labeling the scenes in which the dataset is located, is as follows:

[0055] First, prepare multiple geodesic receivers covering the entire frequency range and system. After connecting the antennas and power supplies, place them in the aforementioned scenario. Set the receivers to store observation data in RINEX 3.05 format. After storing the data for 24 hours, retrieve the receivers. Copy the stored observation data from the receivers and store them separately according to the location where the receivers were placed. Manually note the detailed observation environment of the receiver placement location in the file.

[0056] The receiver was then placed on a mobile platform, such as a car or a handcart, and continuously moved around the location and its vicinity to collect and store observations for as long as possible. After the collection was completed, the stored observation data was copied and stored in RINEX 3.05 format, and classified and stored together with the static acquisition files to form a dataset.

[0057] In some embodiments, the method described in step two, which uses a high-precision single-point positioning method to calculate the dataset and separate the environmental error in the observations, is as follows:

[0058] First, for statically acquired data, static long-term baseline calculation is performed by combining known GNSS reference stations with the current data to obtain the accurate geographical location of the receiver data;

[0059] For dynamically acquired data, dynamic baseline calculation is performed by combining known GNSS reference stations, static data acquired in step one, and current dynamic data to obtain accurate dynamic geographic location coordinates;

[0060] Then, using the precise satellite orbits, precise satellite clock errors, high-precision ionospheric models, and tropospheric products and receivers released by the IGS organization, and accurate geographical locations, the remaining errors in the positioning process are removed, leaving only the errors caused by environmental noise in the model. The errors caused by environmental noise are then output according to different satellites and different epochs.

[0061] In some embodiments, step three involves using the satellite elevation angle, azimuth angle, accuracy factor, satellite number change rate, signal-to-noise ratio change rate, and their statistical values ​​calculated from the labeled dataset in step one, along with the environmental error corresponding to the dataset in step two, as input. This is then used to train models such as Support Vector Machine, Light GBM, and Random Forest to obtain the relationship between these factors and the environmental error. The model is then stored. The operation method is as follows:

[0062] First, using the RINEX data from step one, calculate the following parameters: satellite elevation angle, azimuth angle, accuracy factor, number of satellites and their rate of change, signal-to-noise ratio rate of change and its statistical value:

[0063] The formulas for calculating the satellite elevation angle F1 and azimuth angle Az are as follows:

[0064]

[0065]

[0066] Az = arctan(e e / e n );

[0067] El = arcsin(e u );

[0068] Where r s and r r Let λ be the coordinates of the receiver and the satellite in the geocentric coordinate system. This refers to the longitude and latitude of the receiver's location;

[0069] The formula for calculating the precision factor GDOP is as follows:

[0070]

[0071] Where H is the normal equation matrix constructed using the least squares algorithm during the localization process;

[0072] The formulas for calculating the number of satellites nSat and its rate of change dnSat are as follows:

[0073]

[0074] Where nSat(t1) is the number of satellites at time t1, and nSat(t2) is the number of satellites at time t2;

[0075] The calculation methods for the signal-to-noise ratio change rate dSNR and its statistical values ​​are as follows:

[0076]

[0077]

[0078]

[0079] Where avg and std are the mean and standard deviation of the rate of change of signal-to-noise ratio.

[0080] Next, the above results are used to train models such as environmental noise input support vector machine, Light GBM, and random forest obtained in step two. The model with the smallest error is selected as the current model based on the model output error, and the current model is saved.

[0081] In some embodiments, when the user performs real-time positioning, the model trained in step three is used to provide an environmental noise estimate for the positioning algorithm. The positioning algorithm uses this estimate as the environmental noise variance in the stochastic model to improve the positioning accuracy. The specific implementation method is as follows:

[0082] First, save the model from step three as a parameter configuration file. Then, during real-time positioning, use the formula from step three to calculate parameters such as satellite elevation angle, azimuth angle, accuracy factor, number of satellites and their rate of change, signal-to-noise ratio rate of change and its statistical value. Input these parameters into the model to obtain the environmental noise model output. Use this value as the environmental noise variance in the real-time positioning process and add it to the stochastic model of the positioning process to improve positioning accuracy.

[0083] In some embodiments, the specific implementation method for storing observation data during the user's real-time positioning process in step five and extracting the data after the process is completed and adding it to the dataset constructed in step one, and repeating steps two and three to update, train and optimize the model to improve the model accuracy is as follows:

[0084] In step four, the user stores observations during real-time positioning and exports the stored observations after the real-time positioning process ends. After adding the environmental information corresponding to the observation time period, the data is merged with the dataset collected in step one. Steps two and three are repeated to train a new environmental noise model, thereby improving the positioning accuracy in subsequent use.

[0085] Example 2:

[0086] Reference Figure 3 As shown, this application provides a system for high-precision positioning using scene adaptive detection technology. The system executes the above-described method and includes:

[0087] Receiver 1 is configured to acquire multiple sets of static observation data and multiple sets of dynamic observation data of the observed scene, form a dataset, and label the scene in which the dataset is located.

[0088] GNSS reference station 2 is electrically connected to receiver 1. GNSS reference station 2 is configured to perform static long-term baseline calculation based on static observation data to obtain the accurate geographical location of the receiver data; it is also configured to perform dynamic baseline calculation based on dynamic observation data to obtain the accurate dynamic geographical location coordinates.

[0089] Computer 3 is connected to receiver 1 and GNSS reference station 2. Computer 3 is configured to: separate environmental errors in observations based on the accurate geographic coordinates of the terminal products and receivers released by the IGS organization; Computer 3 is also configured to: use labeled datasets to calculate satellite elevation angle, azimuth angle, accuracy factor, satellite number change rate, signal-to-noise ratio change rate and their statistical values, and input the calculation results into models such as support vector machine, Light GBM, and random forest for training, and select the model with the smallest error as the current model based on the model output error, and save the current model.

[0090] The end products released by the IGS organization include precision satellite orbits, precision satellite clock errors, high-precision ionospheric models, and tropospheric products.

[0091] When a user performs real-time positioning, the current model is used to provide an environmental noise estimate for the positioning algorithm. The positioning algorithm uses this estimate as the environmental noise variance in the stochastic model to improve the positioning accuracy of the algorithm.

[0092] During real-time positioning, users simultaneously store observations and export them after the real-time positioning process ends. After adding the environmental information corresponding to the observation time period, the data is merged with the collected dataset and repeatedly trained to obtain a new environmental noise model, thereby improving the positioning accuracy in subsequent use.

[0093] It is readily understood that, based on the several embodiments provided in this application, those skilled in the art can combine, split, or reorganize the embodiments of this application to obtain other embodiments, none of which exceed the protection scope of this application.

[0094] The above detailed embodiments further illustrate the purpose, technical solution, and beneficial effects of the embodiments of this application. It should be understood that the above are merely specific embodiments of the embodiments of this application and are not intended to limit the protection scope of the embodiments of this application. Any modifications, equivalent substitutions, improvements, etc., made on the basis of the technical solutions of the embodiments of this application should be included within the protection scope of the embodiments of this application.

Claims

1. A method for high-precision positioning using scene adaptive detection technology, characterized in that: include; Collect multiple sets of static observation data and multiple sets of dynamic observation data of the observed scene to form a dataset, and label the scene in which the dataset is located; The dataset is calculated using a high-precision single-point positioning method to separate environmental errors from the observed values; Using the labeled dataset, the satellite elevation angle, azimuth angle, accuracy factor, satellite number change rate, signal-to-noise ratio change rate, and their statistical values ​​were calculated. The satellite elevation angle, azimuth angle, accuracy factor, satellite number change rate, signal-to-noise ratio change rate, statistical value, and environmental error are input into the support vector machine, Light GBM, and random forest models for training. The model with the smallest error is selected as the current model based on the model output error and the current model is saved. When performing real-time positioning, the current model is used as the positioning algorithm to provide an estimated value of environmental noise. The positioning algorithm uses this estimated value as the variance of environmental noise in the random model to improve the positioning accuracy of the algorithm. The method of using a high-precision single-point positioning method to calculate and separate environmental errors in the observed values ​​is as follows: For statically acquired data, a static long-term baseline calculation is performed by combining known GNSS reference stations with the current data to obtain the accurate geographical location of the receiver data; For dynamically acquired data, dynamic baseline calculation is performed by combining known GNSS reference stations, the statically acquired data, and the current dynamic data to obtain accurate dynamic geographic location coordinates; Using precise satellite orbits, precise satellite clock errors, high-precision ionospheric models, tropospheric products, and accurate receiver geolocation data released by the IGS organization, the remaining errors in the positioning process are removed, leaving only the errors caused by environmental noise in the model. The errors caused by environmental noise are then output according to different satellites and different epochs.

2. The method for high-precision positioning using scene adaptive detection technology according to claim 1, characterized in that: The method also includes: During the user's real-time location process, the observed data is stored and then extracted and added to the dataset after the process ends to obtain an updated dataset. Based on the updated dataset, the model is updated, trained, and optimized to improve model accuracy.

3. The method for high-precision positioning using scene adaptive detection technology according to claim 1, characterized in that: The method for calculating satellite elevation angle, azimuth angle, accuracy factor, satellite number change rate, signal-to-noise ratio change rate, and their statistical values ​​using the labeled dataset is as follows: The formulas for calculating the satellite elevation angle E1 and azimuth angle A2 are as follows: in and The coordinates of the receiver and the satellite in the geocentric coordinate system. and This refers to the longitude and latitude of the receiver's location; The formula for calculating the precision factor GDOP is as follows: ； Where H is the normal equation matrix constructed using the least squares algorithm during the localization process; The formulas for calculating the number of satellites nSat and its rate of change dnSat are as follows: ； in for Number of satellites at any given time for Number of satellites at any given time; The calculation methods for the signal-to-noise ratio change rate dSNR and its statistical values ​​are as follows: ； ； ； Where avg and std are the mean and standard deviation of the rate of change of signal-to-noise ratio.

4. The method for high-precision positioning using scene adaptive detection technology according to claim 1, characterized in that: Both the static and dynamic observation data are stored in RINEX 3.05 format and are collectively categorized and stored to form the dataset.

5. A system for high-precision positioning using scene adaptive detection technology, characterized in that: The system performs the method according to any one of claims 1-4, the system comprising: Receiver (1), the receiver (1) is configured to acquire multiple sets of static observation data and multiple sets of dynamic observation data of the observed scene, form a dataset, and label the scene in which the dataset is located; A GNSS reference station (2) is electrically connected to the receiver (1). The GNSS reference station (2) is configured to perform static long-term baseline calculation based on the static observation data to obtain the accurate geographical location of the receiver data; it is also configured to perform dynamic baseline calculation based on the dynamic observation data to obtain the accurate dynamic geographical location coordinates. Computer (3) is communicatively connected to receiver (1) and GNSS reference station (2). Computer (3) is configured to: separate environmental errors in observations based on the accurate geographic coordinates of terminal products and receivers released by IGS organization; computer (3) is also configured to: use the labeled dataset to calculate satellite elevation angle, azimuth angle, accuracy factor, satellite number change rate, signal-to-noise ratio change rate and their statistical values, and input the calculation results into support vector machine, Light GBM and random forest models for training, and select the model with the smallest error as the current model based on the model output error, and save the current model.

Images