A DBSCAN-based GNSS positioning method and apparatus
By smoothing GNSS data using RTK and Kalman filtering, and selecting the average value of the most frequent baseline vector clusters using the DBSCAN clustering algorithm, the problems of high model dependence and large computational resource requirements in existing technologies are solved, and high-precision positioning of GNSS positioning systems in complex scenarios is achieved.
Patent Information
- Application Number
- CN202311210432.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-09-19
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2043-09-19
AI Technical Summary
Existing GNSS measurement methods suffer from high model dependence, large computational resource requirements, and stringent parameter setting requirements in complex scenarios, which limits the accuracy and reliability of gross error detection and removal.
RTK check and Kalman filtering are used to smooth the baseline vectors. The DBSCAN clustering algorithm is then used to cluster the baseline vectors. The average value of the cluster with the most baseline vectors is selected as the final baseline vector to improve positioning accuracy.
By using RTK and Kalman filtering for smoothing and DBSCAN clustering, the accuracy of GNSS positioning results is improved, gross errors are effectively eliminated, and measurement accuracy and reliability in complex scenarios are enhanced.
Smart Images

Figure CN117388889B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of GNSS navigation and positioning, and in particular to a GNSS positioning method and apparatus based on DBSCAN. Background Technology
[0002] Global Navigation Satellite Systems (GNSS) have become one of the most widely used technologies in modern surveying. They can be used to measure parameters such as position, velocity, and time, and are therefore widely adopted in many applications, including aviation, maritime, vehicle navigation, and geological exploration.
[0003] However, in GNSS measurements, due to various reasons (such as interference during signal transmission and errors in the receiver itself), abnormally large measurement values may occur, known as outliers. These outliers can seriously affect the accuracy and reliability of GNSS measurement results, thus requiring outlier detection and removal.
[0004] GNSS gross error detection and removal techniques can be divided into two categories: statistical methods and model-based methods. Statistical methods primarily rely on data analysis and utilize classic statistical techniques to detect gross errors. Examples include the Grubbs test, Dixon test, and mean-variance method. These methods determine the presence of gross errors by comparing measured values with parameters such as mean and variance. While suitable for simple measurement scenarios, their accuracy and reliability may be limited in complex scenarios.
[0005] Model-based methods involve building a physical model of the GNSS measurement data and identifying gross errors by fitting the model. Examples of methods include least squares fitting, Kalman filtering, and fluctuation analysis. These methods perform well in complex measurement scenarios and can effectively detect and eliminate gross errors. Furthermore, these methods can also be used to evaluate and improve model accuracy.
[0006] The main drawbacks of model-based methods include:
[0007] (1) Dependence on the model Model-based methods rely on modeling the physical data of the measurements, thus requiring high accuracy and reliability from the model. If the model is inaccurate or biased, model-based methods may misclassify normal values as gross errors or identify gross errors as normal values.
[0008] (2) Requires a large amount of computing resources Model-based methods typically require significant computational resources to complete the modeling and fitting process. Therefore, in practical applications, the limitations of computation time and resources need to be considered.
[0009] (3) High requirements for parameter settings Model-based methods typically require setting certain parameters, such as the noise variance of the filter. Inappropriate parameter settings can affect the accuracy and reliability of the model. Summary of the Invention
[0010] The purpose of this invention is to overcome the above-mentioned shortcomings in the prior art and to provide a GNSS positioning method and device based on DBSCAN.
[0011] To achieve the above-mentioned objectives, the present invention provides the following technical solution: A DBSCAN-based GNSS positioning method includes the following steps: S1. The base station receiver and the rover receiver receive the raw data from the GNSS satellites; S2. Determine the phase center coordinates of the base station receiver antenna and label these coordinates as T; S3. Use the static solution method to process the raw data of GNSS satellites received by T, the base station receiver and the rover receiver, and calculate the baseline vector set A from the base station to the rover. S4. Use RTK to check the accuracy of the baseline vectors obtained in different time periods in the baseline vector set A. If the accuracy is lower than the set threshold, it is determined that the baseline vectors obtained in that time period are incorrect, and the incorrect baseline vectors obtained in different time periods are set into set B. RTK is a dynamic difference technique. S5. Use Kalman filtering to smooth the baseline vectors in B to obtain set C; S6. Arbitrarily select a time period, and use the DBSCAN clustering algorithm to cluster the baseline vectors C within that selected time period. sub Clustering is performed to obtain different clustered baseline vector clusters D1, D2, ..., D n And in D1, D2, ..., D n The baseline vector cluster D with the most selected baseline vectors max Baseline vector cluster C sub It is a subset of set C; S7. Find D max The average value is used to obtain the final baseline vector BF.
[0012] Preferably, in step S1, the raw data of the GNSS satellite includes carrier phase, ephemeris, and pseudorange.
[0013] Preferably, in step S4, the accuracy of the baseline vectors obtained in different time periods in the baseline vector set A is checked using RTK. If the accuracy is lower than a set threshold, the baseline vectors obtained in that time period are determined to be incorrect, and the incorrect baseline vectors obtained in different time periods are collected into set B. The specific process is as follows: The raw data received by the GNSS satellites from the base station receiver and the rover receiver are processed using RTK. The baseline vector set A from the base station to the rover is calculated. The accuracy of the baseline vectors in A for each time period is calculated using the baseline vectors in A as the standard.
[0014] Preferably, in step S6, a time period is arbitrarily selected, and the DBSCAN clustering algorithm is used to cluster the baseline vector clusters C within the selected time period. sub Clustering is performed to obtain different clustered baseline vector clusters D1, D2, ..., D n And in D1, D2, ..., D n The baseline vector cluster D with the most selected baseline vectors max Baseline vector cluster C sub To create a subset of set C, the following steps are included: S61. Define the neighborhood size, using Eps to represent the radius of the neighborhood, and MinPts to represent the baseline vector clusters D1, D2, ..., D n The minimum number of baseline vectors; S62. Baseline vector cluster C within a selected time range sub Choose any baseline vector P and determine whether P has been visited; S63. If it is determined that the baseline vector P has been visited, skip the baseline vector P and repeat step S62; if it is determined that the baseline vector P has not been visited, determine the neighborhood Q of the baseline vector P, find all the baseline vectors in Q and determine the number, and determine whether the number is less than MinPts. S64. When the number of Q is less than MinPts, mark the baseline vector P as noise and mark it as visited; when the number of Q is not less than MinPts, create a new baseline vector cluster M, and add the baseline vector P and all baseline vectors in its neighborhood to the baseline vector cluster D. n In the middle, the baseline vector cluster D n All baseline vectors in the array are marked as visited; S65. In baseline vector cluster D n Choose any baseline vector X, query the number of baseline vectors in the neighborhood W of X, and determine whether the number of baseline vectors in W is less than MinPts. If it is less, add all baseline vectors in W to D. nIn the process, mark all baseline vectors in W as visited; if they are not less than, return to step S62. S66. Determine if X is added to D. n If a cluster does not belong to any cluster, then X is retained in D. n If the condition is met, return to step S62; if it is met, then remove X from D. n Remove from the list and return to steps S65 and S66; S67. Repeat steps S62 to S66 until C is achieved. sub All baseline vectors are clustered into D1, D2, ..., D n Or it may be marked as noise; S68. In D1, D2, ..., D n The baseline vector cluster D with the most selected baseline vectors max .
[0015] A DBSCAN-based GNSS positioning device includes at least one processor and at least one memory communicatively connected to the processor. The memory stores instructions that are executed by the at least one processor to enable the at least one processor to perform any step of the method.
[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention uses RTK to check the accuracy of the baseline vector set A of the base station and rover calculated using the static solution method. Baseline vectors with accuracy below the accuracy threshold are grouped into a set, and the baseline vectors in the set are smoothed using the Kalman filter algorithm to obtain a new set. Then, the baseline vectors in the new set are clustered using the DBSCAN clustering algorithm. The average value of the cluster category containing the most baseline vectors is calculated and used as the final baseline vector, thus improving the accuracy of the baseline vector, which in turn improves the accuracy of the positioning results of the GNSS positioning system. Attached Figure Description
[0017] Figure 1 This is a flowchart of the method. Figure 2 This is a schematic diagram of the gross error data results after Kalman filtering; Figure 3 This is the flowchart for the first part of the DBSCAN clustering algorithm; Figure 4 This is the flowchart for the second part of the DBSCAN clustering algorithm. Detailed Implementation
[0018] The present invention will be further described in detail below with reference to experimental examples and specific embodiments. However, this should not be construed as limiting the scope of the above-mentioned subject matter of the present invention to the following embodiments; all technologies implemented based on the content of the present invention fall within the scope of the present invention.
[0019] Example 1 like Figure 1 As shown, a GNSS positioning method based on DBSCAN includes the following steps: S1. The base station receiver and the rover receiver receive the raw data from the GNSS satellites; S2. Determine the phase center coordinates of the base station receiver antenna and label these coordinates as T; S3. Use the static solution method to process the raw data of GNSS satellites received by T, the base station receiver and the rover receiver, and calculate the baseline vector set A from the base station to the rover. S4. Use RTK to check the accuracy of the baseline vectors obtained in different time periods in the baseline vector set A. If the accuracy is lower than the set threshold, it is determined that the baseline vectors obtained in that time period are incorrect, and the incorrect baseline vectors obtained in different time periods are set into set B. RTK is a dynamic difference technique. S5. Use Kalman filtering to smooth the baseline vectors in B to obtain set C; S6. Arbitrarily select a time period, and use the DBSCAN clustering algorithm to cluster the baseline vectors C within that selected time period. sub Clustering is performed to obtain different clustered baseline vector clusters D1, D2, ..., D n And in D1, D2, ..., D n The baseline vector cluster D with the most selected baseline vectors max Baseline vector cluster C sub It is a subset of set C; S7. Find D max The average value is used to obtain the final baseline vector BF.
[0020] In step S1, the raw data of the GNSS satellite includes carrier phase, ephemeris, and pseudorange.
[0021] In step S4, the raw data received by the GNSS satellites from the T, base station receiver and rover receiver are processed using RTK to calculate the baseline vector set A from the base station to the rover. The accuracy of the baseline vectors in A for each time period is calculated using the baseline vectors in A as the standard.
[0022] In step S5, Kalman filtering is an algorithm for estimating system state; it is a recursive filtering algorithm based on Bayes' theorem. It updates the prior estimate of the state using Bayes' theorem and fuses the observed data with the prior estimate to obtain the posterior estimate of the state. The system model is a key part of Kalman filtering, used to describe the dynamic behavior and state change patterns of the system.
[0023] A system model typically consists of two parts: a state transition model and an observation model.
[0024] State transition model: It describes the evolution of the system's state over time. It is usually represented using linear dynamic equations, and can be expressed in matrix form as follows:
[0025] Where t(k) represents the state of the system at time k, F(k) is the state transition matrix, which describes the evolution of the system state from time k-1 to time k, B(k) is the control input matrix, which describes the influence of external input on the state, u(k) represents the control input, and w(k) represents the process noise.
[0026] Observation Model: This describes the relationship between the system state and observed data. It is typically represented using a linear relationship and can be expressed in matrix form:
[0027] Where z(k) represents the data observed at time k, H(k) is the observation matrix, which describes the relationship between the observed data and the system state, and v(k) represents the observation noise (measurement error).
[0028] Kalman filtering predicts the current state by fusing system model and observation data. It calculates the posterior estimate (posterior state) of the system state based on the difference between the prior estimate (prior state) at the current time and the observed data. The specific steps are as follows: Predict step: Based on the state transition model, use prior states and control inputs to predict the state of the system at the next time step.
[0029]
[0030] Update step: Based on the observation model, compare the predicted state with the observed data, and calculate the posterior estimate of the system state and the covariance of the estimation error.
[0031]
[0032]
[0033]
[0034] Where x'(k) represents the prior state, x(k) represents the posterior state, P'(k) represents the state error covariance of the prior estimate, P(k) represents the state error covariance of the posterior estimate, K(k) represents the Kalman gain, and R(k) represents the covariance matrix of the observation noise.
[0035] By continuously performing prediction and update steps, the Kalman filter algorithm can progressively estimate the true state of the system and provide state estimation results with higher accuracy.
[0036] In GNSS calculations, Kalman filtering can be used to process parameters such as receiver position, velocity, and acceleration. By continuously updating and optimizing the calculation results, Kalman filtering can improve the accuracy and reliability of the calculation in complex environments, such as those with signal interference or multipath effects. However, the data after Kalman filtering still contains gross errors, such as... Figure 2 As shown.
[0037] like Figure 3 and Figure 4 As shown, the specific steps to achieve step S6 include the following: S61. Define the neighborhood size, using Eps to represent the radius of the neighborhood, and MinPts to represent the baseline vector clusters D1, D2, ..., D n The minimum number of baseline vectors; S62. Baseline vector cluster C within a selected time range sub Choose any baseline vector P and determine whether P has been visited; S63. If it is determined that the baseline vector P has been visited, skip the baseline vector P and repeat step S62; if it is determined that the baseline vector P has not been visited, determine the neighborhood Q of the baseline vector P, find all the baseline vectors in Q and determine the number, and determine whether the number is less than MinPts. S64. When the number of Q is less than MinPts, mark the baseline vector P as noise and mark it as visited; when the number of Q is not less than MinPts, create a new baseline vector cluster M, and add the baseline vector P and all baseline vectors in its neighborhood to the baseline vector cluster D. n In the middle, the baseline vector cluster D n All baseline vectors in the array are marked as visited; S65. In baseline vector cluster D nChoose any baseline vector X, query the number of baseline vectors in the neighborhood W of X, and determine whether the number of baseline vectors in W is less than MinPts. If it is less, add all baseline vectors in W to D. n In the process, mark all baseline vectors in W as visited; if they are not less than, return to step S62. S66. Determine if X is added to D. n If a cluster does not belong to any cluster, then X is retained in D. n If the condition is met, return to step S62; if it is met, then remove X from D. n Remove from the list and return to steps S65 and S66; S67. Repeat steps S62 to S66 until C is achieved. sub All baseline vectors are clustered into D1, D2, ..., D n Or it may be marked as noise; S68. In D1, D2, ..., D n The baseline vector cluster D with the most selected baseline vectors max .
[0038] A DBSCAN-based GNSS positioning device uses a Core i7-12700 processor and a Samsung 980PRO 1TB solid-state drive for memory.
[0039] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A GNSS positioning method based on DBSCAN, characterized in that, Includes the following steps: S1. The base station receiver and the rover receiver receive the raw data from the GNSS satellites; S2. Determine the phase center coordinates of the base station receiver antenna and label these coordinates as T; S3. Use the static solution method to process the raw data of GNSS satellites received by T, the base station receiver and the rover receiver, and calculate the baseline vector set A from the base station to the rover. S4. Use RTK to check the accuracy of the baseline vectors obtained in different time periods in the baseline vector set A. If the accuracy is lower than the set threshold, it is determined that the baseline vectors obtained in that time period are incorrect, and the incorrect baseline vectors obtained in different time periods are collected into set B. S5. Use Kalman filtering to smooth the baseline vectors in B to obtain set C; S6. Arbitrarily select a time period, and use the DBSCAN clustering algorithm to cluster the baseline vectors C within that selected time period. sub Clustering is performed to obtain different clustered baseline vector clusters D1, D2, ..., D n And in D1, D2, ..., D n The baseline vector cluster D with the most selected baseline vectors max Baseline vector cluster C sub It is a subset of set C; S7. Find D max The average value is used to obtain the final baseline vector BF.
2. The GNSS positioning method based on DBSCAN according to claim 1, characterized in that, In step S1, the raw data of the GNSS satellite includes carrier phase, ephemeris, and pseudorange.
3. The GNSS positioning method based on DBSCAN according to claim 1, characterized in that, In step S4, RTK is used to check the accuracy of baseline vectors obtained in different time periods in baseline vector set A. If the accuracy is lower than a set threshold, the baseline vectors obtained in that time period are determined to be incorrect, and the incorrect baseline vectors obtained in different time periods are collected into set B. The specific process is as follows: The raw data received by the GNSS satellites from the base station receiver and the rover receiver are processed using RTK. The baseline vector set A from the base station to the rover is calculated. The accuracy of the baseline vectors in A for each time period is calculated using the baseline vectors in A as the standard.
4. The GNSS positioning method based on DBSCAN according to claim 1, characterized in that, In step S6, an arbitrary time period is selected, and the DBSCAN clustering algorithm is used to cluster the baseline vectors C within the selected time period. sub Clustering is performed to obtain different clustered baseline vector clusters D1, D2, ..., D n And in D1, D2, ..., D n The baseline vector cluster D with the most selected baseline vectors max Baseline vector cluster C sub To create a subset of set C, the following steps are included: S61. Define the neighborhood size, using Eps to represent the radius of the neighborhood, and MinPts to represent the baseline vector clusters D1, D2, ..., D n The minimum number of baseline vectors; S62. Baseline vector cluster C within a selected time range sub Choose any baseline vector P and determine whether P has been visited; S63. If it is determined that the baseline vector P has been visited, skip the baseline vector P and repeat step S62; if it is determined that the baseline vector P has not been visited, determine the neighborhood Q of the baseline vector P, find all the baseline vectors in Q and determine the number, and determine whether the number is less than MinPts. S64. When the number of Q is less than MinPts, mark the baseline vector P as noise and mark it as visited; when the number of Q is not less than MinPts, create a new cluster of baseline vectors D. n And add the baseline vector P and all baseline vectors in its neighborhood to the baseline vector cluster D. n In the middle, the baseline vector cluster D n All baseline vectors in the array are marked as visited; S65. In baseline vector cluster D n Arbitrarily select a baseline vector X, query the number of baseline vectors in the neighborhood W of X, and determine whether the number of baseline vectors in W is less than MinPts. If it is less, add all baseline vectors in W to D. n In the process, mark all baseline vectors in W as visited; if they are not less than, return to step S62. S66. Determine if X is added to D. n If a cluster does not belong to any cluster, then X is retained in D. n If the condition is met, return to step S62; if it is met, then remove X from D. n Remove from the list and return to steps S65 and S66; S67. Repeat steps S62 to S66 until C is achieved. sub All baseline vectors are clustered into D1, D2, ..., D n Or it may be marked as noise; S68. In D1, D2, ..., D n The baseline vector cluster D with the most selected baseline vectors max .
5. A DBSCAN-based GNSS positioning device implementing the method of any one of claims 1 to 4, characterized in that, The method includes at least one processor and at least one memory communicatively connected to the processor, the memory storing instructions executable by the at least one processor to enable the at least one processor to perform any step of the method.
Citation Information
Patent Citations
Beidou and micro-motion landslide early warning method based on K-means clustering algorithm
CN113781745A
Real-time fast decimeter-level GNSS positioning
US20090135057A1
