Visual feature assisted enhanced GNSS PPP fast high-precision positioning method, system, storage medium and equipment
Patent Information
- Application Number
- CN202610973196.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-01
- Publication Date
- 2026-10-09
AI Technical Summary
[0005]本发明的目的是提供视觉特征辅助增强的GNSS PPP快速高精度定位方法，以解决现有技术收敛时间长、定位精度低、与视觉紧耦合定位复杂的技术问题
[0025]本发明的技术优点如下：本发明通过动态调整滑动窗口提取载体线性运动区间，构建视觉线性几何观测约束，充分发挥单目视觉不受卫星信号遮挡、独立解算载体相对运动轨迹的优势，而直接提取载体线性运动特征构建几何观测约束，为PPP提供额外观测约束，以提升定位精度与鲁棒性，缩短收敛时间，降低了定位复杂性，解决现有技术在复杂环境下的应用瓶颈，维持了动态场景下PPP连续高精度定位性能。
Smart Images
Figure CN122883216A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of integrated navigation technology, specifically relating to a visual feature-assisted enhanced GNSS PPP fast and high-precision positioning method, system, storage medium, and device. Background Technology
[0002] Precise Point Positioning (PPP) of Global Navigation Satellite System (GNSS) can achieve high-precision positioning globally using a single receiver, and has broad application prospects in surveying, navigation, and autonomous driving. Traditional GNSS PPP has inherent drawbacks such as long convergence time and low positioning accuracy in the initial stage. Especially in weak signal environments such as urban canyons and tree-shaded areas, it is susceptible to multipath effects and signal loss, leading to decreased positioning accuracy or even failure, making it difficult to meet the rapid positioning requirements of high-dynamic and high-reliability scenarios.
[0003] In recent years, visual sensors have been widely used in pose estimation of moving vehicles due to their advantages of low cost, high frequency, and immunity to electromagnetic interference. Visual-GNSS fusion technology has become an important direction for solving the problem of fast convergence of PPP (Positioning for Surface-to-Surface) models. However, existing solutions mostly employ tightly coupled filtering frameworks, resulting in high computational complexity, long convergence times, and difficulty in effectively utilizing the geometric constraints of visual features to construct direct PPP observation models. Furthermore, traditional visual SLAM relies on continuous feature tracking, making it susceptible to changes in illumination and texture loss, thus failing to provide stable and reliable auxiliary information for PPP.
[0004] Therefore, there is an urgent need for a fast and high-precision GNSS PPP positioning scheme based on visual information assistance. This scheme constructs a geometric constraint model of carrier motion by extracting visual features, providing additional observation constraints for PPP to improve positioning accuracy and robustness, shorten convergence time, and solve the problems of high computational complexity in existing technologies, difficulty in effectively utilizing the geometric constraints of visual features through fusion methods, and limitations in the performance of the visual front end in complex environments. As a result, there are technical problems such as low positioning accuracy and complex positioning with tight coupling with vision in complex environments. Summary of the Invention
[0005] The purpose of this invention is to provide a fast and high-precision GNSS PPP positioning method with visual feature-assisted enhancement, so as to solve the technical problems of long convergence time, low positioning accuracy and complex positioning with tight coupling with vision in the existing technology.
[0006] The aforementioned visual feature-assisted enhanced GNSS PPP fast and high-precision positioning method includes the following steps: S1. Calibrate the intrinsic parameters and distortion coefficients of a monocular camera based on multi-angle images acquired using a checkerboard pattern; S2. A linear window is extracted based on the computational relative trajectory of the camera, and then used to construct GNSS PPP observation constraints; In this step, the relative motion trajectory composed of each historical sampling point is linearly fitted, and a linear window that satisfies the preset fitting residual is extracted. Based on the extracted linear window, GNSS observation constraints within the same time period are constructed. The GNSS observation constraints are that the sum of the lengths of each segment within the linear window is equal to the total length. After linearizing the GNSS observation constraints, an error equation is obtained to represent the degree of deviation of the geometric constraints within the linear window, thus obtaining the constraint model. S3. Based on the constraint model and GNSS pseudorange / phase observation equations constructed in the previous step, perform a fast and high-precision solution for the GNSS receiver position.
[0007] Preferably, in step S2, the optical flow method is used to perform motion detection on feature points in the image frame, and the relative motion trajectory of the moving vehicle is calculated based on the relatively moving feature points. The formula for extracting the linear window is expressed as follows:
[0008] in Indicates the sampling point. , Indicates the starting point. Indicates the current point; This represents the direction vector formed by the first and last points of the window.
[0009] Preferably, in step S2, the GNSS observation constraints are expressed as follows:
[0010] in Indicates spatial distance within different time windows. Indicates the GNSS sampling interval. Indicates the current moment. Indicates the time elapsed since the current time The time window length of each sampling interval; constraints on GNSS observations. , , After linearization, the error equations are obtained. :
[0011] in: This represents the combination of the approximate distances after linearization of the three segments. They represent The receiver's three-dimensional coordinate error vector at time t. This represents the coefficient matrix corresponding to the three-dimensional coordinate error. for The three-dimensional coordinates at time [time]. for Given the three-dimensional coordinates at time t, the resulting constraint model is:
[0012] in express Approximate coordinates calculated using precise single-point positioning by GNSS at a given time. Indicates the time window Approximate spatial distance within, Indicates the time window Approximate spatial distance within.
[0013] Preferably, in step S3, the GNSS pseudorange / phase observation equation is expressed as:
[0014] in and Indicates satellite to receiver exist Dual-frequency ionospheric carrier phase and pseudorange observation residuals at time t. Indicates GNSS satellite system, The receiver's three-dimensional position error vector is represented by the coefficient matrix. ; and These represent the receiver clock bias and the satellite clock bias, respectively. Indicates the ambiguity of the non-ionospheric combination. Indicates tropospheric delay, and These represent the carrier phase and pseudorange observations without ionospheric assemblies, respectively. and The noise consists of phase and pseudorange observations; satellite clock errors can be corrected using precision products, while other parameters are estimated as unknown parameters. Based on the GNSS pseudorange / phase observation equations and the constructed constraint model, the constraint observation equations for GNSS precise point positioning are expressed as follows:
[0015] in ; The tropospheric wet delay is to be estimated. The tropospheric dry delay is corrected by an empirical model. The least squares estimation method is used to make optimal estimates of parameters such as receiver three-dimensional coordinate error, receiver clock error, zenith tropospheric wet delay, and ambiguity. Finally, the position of the GNSS receiver is calculated quickly and with high accuracy.
[0016] Preferably, in step S1, the intrinsic parameters of the monocular camera are first calibrated. The chessboard used for the intrinsic parameter calibration of the monocular camera is a black and white rectangular chessboard. By taking pictures of the chessboard from different angles, the corner points of the chessboard are extracted from the pictures, and the intrinsic parameter matrix and distortion coefficient of the monocular camera are calculated.
[0017] Preferably, based on continuous frame images acquired by a monocular camera in a real-world working environment, the ORB operator is used to extract feature points from the images, RANSAC and Lowe's ratio test methods are used to match feature point pairs, and mismatched point pairs are eliminated. Then, the fundamental matrix is calculated based on epipolar geometric constraints, and the constraint conditions are expressed as follows:
[0018] in This represents the homogeneous coordinates of the corresponding point in the first frame of the image. This represents the homogeneous coordinates of the corresponding point in the second frame image. Represents the fundamental matrix.
[0019] Preferably, the monocular camera intrinsic parameter matrix, distortion coefficients, and fundamental matrix are calculated. Calculate the essential matrix The relative rotation and displacement of adjacent frames are decomposed into the following formula:
[0020] in This represents the rotation matrix, indicating the relative rotation of the camera. Let represent the rotation vector, which represents the relative displacement between adjacent frames, and be further expressed as:
[0021] in , and These represent the translation vector components.
[0022] This invention also provides a visual feature-assisted enhanced GNSS PPP fast and high-precision positioning system, comprising: The monocular camera intrinsic parameters and distortion coefficient calibration module calibrates the monocular camera intrinsic parameters and distortion coefficients based on multi-angle images acquired using a checkerboard pattern. The GNSS PPP observation constraint module extracts a linear window based on the calculated camera relative trajectory sliding, which is then used to construct GNSS PPP observation constraints. The fast and high-precision positioning module performs fast and high-precision calculation of the GNSS receiver position based on the constructed constraint model and GNSS pseudorange / phase observation equations.
[0023] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the corresponding steps according to the vision-assisted enhanced GNSS PPP fast high-precision positioning method described above.
[0024] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor. When the processor executes the computer program, it performs the corresponding steps in accordance with the vision-assisted enhancement-based GNSS PPP fast and high-precision positioning method described above.
[0025] The technical advantages of this invention are as follows: This invention extracts the linear motion range of the carrier by dynamically adjusting the sliding window, constructs visual linear geometric observation constraints, and fully leverages the advantages of monocular vision, which is not blocked by satellite signals and can independently solve the relative motion trajectory of the carrier. It directly extracts the linear motion features of the carrier to construct geometric observation constraints, providing additional observation constraints for PPP, thereby improving positioning accuracy and robustness, shortening convergence time, reducing positioning complexity, solving the application bottleneck of existing technologies in complex environments, and maintaining the continuous high-precision positioning performance of PPP in dynamic scenarios. Attached Figure Description
[0026] Figure 1 This is a flowchart of the visual feature-assisted enhancement GNSS PPP fast and high-precision positioning method of the present invention; Figure 2 A comparison of the root mean square error of the linear fitting residuals between existing techniques that do not use dynamically adjusted sliding windows and the sliding window trajectories under different thresholds in this invention; Figure 3 A comparison chart of the positioning error sequence between the present invention and existing technologies under normal satellite observation conditions and under obstruction conditions; Figure 4 This is a comparison diagram of the positioning trajectories of the present invention and the prior art. Detailed Implementation
[0027] The following detailed description of the embodiments, with reference to the accompanying drawings, will further illustrate the specific implementation of the present invention, in order to help those skilled in the art to have a more complete, accurate, and in-depth understanding of the inventive concept and technical solution of the present invention.
[0028] Example 1 like Figures 1-4 As shown, this invention provides a fast and high-precision GNSS PPP positioning method based on vision-assisted enhancement, comprising the following steps: S1. Calibrate the intrinsic parameters and distortion coefficients of a monocular camera based on multi-angle images acquired using a checkerboard pattern.
[0029] In this step, the intrinsic parameters of the monocular camera are first calibrated, ORB feature points are extracted and matched based on continuous frame images, the fundamental matrix is calculated through epipolar geometric constraints, and then the relative rotation and displacement of adjacent frames of the camera are solved.
[0030] Specifically, the intrinsic parameter calibration of the monocular camera was implemented using Zhang Zhengyou's checkerboard calibration method. During the calibration process, the monocular camera position was kept fixed, and the black and white checkerboard calibration board (3.5 cm on each side, 11×8 corners) was repeatedly moved within the camera's field of view. The pitch angle, yaw angle, and shooting distance of the calibration board were adjusted, and 26 sets of non-repeating multi-angle checkerboard images were acquired, ensuring the integrity of the checkerboard corners. A sub-pixel corner detection algorithm was used to extract the pixel coordinates of all checkerboard corners within the images. Combined with the known physical dimensions of the checkerboard, a homography matrix was constructed between the image pixel coordinate system and the calibration board's world coordinate system. The camera's intrinsic parameters and distortion coefficients were initially solved using the least squares method. The calibration results are shown in Table 1. Table 1 Camera Calibration Results
[0031] The process of calculating the fundamental matrix includes: extracting feature points from continuous frame images acquired by a monocular camera in a real-world working environment using the ORB operator; matching feature point pairs using RANSAC and Lowe's ratio test methods and eliminating mismatched point pairs; and then calculating the fundamental matrix based on epipolar geometric constraints, the constraints of which are expressed as follows: (1) in This represents the homogeneous coordinates of the corresponding point in the first frame of the image. This represents the homogeneous coordinates of the corresponding point in the second frame image. Represents the fundamental matrix.
[0032] The process of determining the relative rotation and displacement between adjacent camera frames includes: calculating the monocular camera intrinsic parameter matrix, distortion coefficients, and fundamental matrix. Calculate the essential matrix The relative rotation and displacement of adjacent frames are decomposed into the following formula: (2) in This represents the rotation matrix, which indicates the relative rotation of the camera. Let represent the rotation vector, which represents the relative displacement between adjacent frames, and be further expressed as: (3) in , and These represent the translation vector components.
[0033] S2. A linear window is extracted based on the computational relative trajectory of the camera, and then used to construct GNSS PPP observation constraints.
[0034] In this step, the process of extracting a linear window based on the relative trajectory of the camera through calculation includes: using optical flow to perform motion detection on feature points in the image frame, and calculating the relative motion trajectory of the moving vehicle based on the feature points of relative motion.
[0035] Linear fitting is performed on the relative motion trajectory composed of various historical sampling points, and a linear window that satisfies the preset fitting residual T (T can be set to 0.01 meters) is extracted. The judgment formula is expressed as follows: (4) in Indicates the sampling point. , Indicates the starting point. Indicates the current point. This represents the direction vector formed by the first and last points of the window.
[0036] The process of constructing GNSS PPP observation constraints includes: based on the extracted linear window, constructing GNSS observation constraints within the same time period, that is, the sum of the lengths of each segment within the linear window equals the total length, and this constraint is expressed as: (5) in This indicates the spatial distance within different time windows. Indicates the GNSS sampling interval. Indicates the current moment. Indicates the time elapsed since the current time The length of the time window.
[0037] Regarding the above constraints , , After linearization, the error equations are obtained. : (6) in: This indicates the degree of deviation from the geometric constraints within the linear window. This represents the combination of the approximate distances after linearization of the three segments. They represent The receiver's three-dimensional coordinate error vector at time t. This represents the coefficient matrix corresponding to the three-dimensional coordinate error.
[0038] definition for The three-dimensional coordinates at time [time]. for Given the three-dimensional coordinates at time t, the resulting constraint model is: (7) in express Approximate coordinates calculated using precise single-point positioning by GNSS at a given time. Indicates the time window Approximate spatial distance within, Indicates the time window Approximate spatial distance within. Parameters marked with a superscript "0" in the formula represent approximate distances calculated using approximate coordinates.
[0039] S3. Based on the constructed constraint model and GNSS pseudorange / phase observation equation, perform fast and high-precision calculation of the GNSS receiver position.
[0040] In this step, the visual constraint model is combined with the GNSS pseudorange / phase observation equation to construct a unified precise single-point positioning solution model to calculate the GNSS receiver position.
[0041] Specifically, the equation for GNSS dual-frequency ionospheric pseudorange / phase observation is expressed as: (8) in and Indicates satellite to receiver exist Dual-frequency ionospheric carrier phase and pseudorange observation residuals at time t. Indicates GNSS satellite system, The receiver's three-dimensional position error vector is represented by the coefficient matrix. . and These represent the receiver clock bias and the satellite clock bias, respectively. Indicates the ambiguity of the non-ionospheric combination. Indicates tropospheric delay, and These represent the carrier phase and pseudorange observations for the ionospheric combination, respectively. and This refers to phase and pseudorange observation noise. Satellite clock errors can be corrected using precision products, while other parameters are estimated as unknowns.
[0042] The process of constructing and solving a unified precise point positioning solution model includes: based on the GNSS pseudorange / phase observation equations and the constructed constraint model, the constraint observation equations for GNSS precise point positioning are expressed as follows: (9) in . The tropospheric wet delay is to be estimated. The tropospheric dry delay is corrected by an empirical model. The least squares estimation method is used to make optimal estimates of parameters such as receiver three-dimensional coordinate error, receiver clock error, zenith tropospheric wet delay, and ambiguity. Finally, the position of the GNSS receiver is calculated quickly and with high accuracy.
[0043] The following section will further verify and explain the execution process and actual performance of the above-mentioned GNSS PPP fast and high-precision positioning method based on visual assistance enhancement through specific scenario experiments.
[0044] Firstly, as Figure 2 As shown in the upper part of the figure, with the movement of the carrier, each newly added sampling point will cause a change in the linear fitting residual of the corresponding sliding window, and the overall motion trajectory of the carrier exhibits a curved characteristic. A threshold is set. If the residual does not exceed the threshold, the constraint equation (7) is used to fuse with GNSS PPP positioning; if the residual exceeds the threshold at any position, the sliding window is reset and the process is repeated. The cases of different thresholds (0.01 meters and 0.02 meters) are compared. Figure 2 The middle and bottom plots show the changes in linear fitting residuals under different sliding window thresholds. It can be seen that if the vehicle's trajectory is always assumed to conform to linear characteristics, the fitting residuals will gradually increase, significantly deviating from reality and becoming unusable for PPP positioning constraints. When the fitting residual threshold is set to 0.02 meters, the linear constraint window is immediately reset when the vehicle exhibits curvilinear motion to ensure that the linear fitting residuals do not exceed the threshold. When the threshold is reduced to 0.01 meters, the fitting residuals are even smaller. Under these two threshold conditions, 41 and 142 linear windows, respectively, are used to constrain PPP positioning throughout the entire movement process.
[0045] Actual GPS observation data from a railway inspection trolley was selected, totaling 3835 seconds, with approximately 2600 seconds of effective dynamic observation. For the first 1225 seconds, the trolley remained stationary, with an average of more than 8 observable satellites and an average geometric position accuracy of less than 2.5. Furthermore, the multipath errors at C1C and C2W frequencies were analyzed for both stationary and moving trolley states. The average multipath errors for both frequencies were within 0.4 meters. Figure 3 As shown, this experiment compares the performance of PPP positioning based on visual information under normal satellite observation conditions and occlusion conditions. Under normal satellite observation conditions, PPP positioning is assisted by two schemes: linear geometric feature constraint and no constraint. The positioning sequences correspond to... Figure 3Schemes 1# and 2# were selected. Parameters such as receiver position, receiver clock error, ambiguity, and tropospheric delay were calculated using the least squares method, and the root mean square of the dynamic positioning error was used as the accuracy evaluation index. The statistical results of the three-dimensional positioning errors for Schemes 1# and 2# are as follows: Scheme 1#: 0.134m north, 0.237m east, and 0.121m azimuth; 2D and 3D positioning errors: 0.273m and 0.298m, respectively. Scheme 2#: 0.273m north, 0.288m east, and 0.238m azimuth; 2D and 3D positioning errors: 0.396m and 0.462m, respectively. This indicates that the linear constraints extracted by vision can significantly suppress the initial position error fluctuations in dynamic PPP, shortening the PPP convergence time to within a few minutes, far superior to the nearly 20-minute convergence time of traditional PPP.
[0046] By artificially reducing the number of available satellites for positioning, the number of available satellites under normal observation conditions is approximately 9-10. Under simulated conditions of weak satellite observation due to obstruction, the number of available satellites is less than 6. Schemes 3# and 4# represent the error sequences for PPP positioning with visual information assistance and PPP positioning alone. Scheme 3# has an error of 0.168m northward, 0.269m eastward, and 0.132m celestial, with 2D and 3D positioning errors of 0.318m and 0.344m, respectively. Scheme 4# has an error of 0.306m northward, 0.390m eastward, and 0.292m celestial, with 2D and 3D positioning errors of 0.496m and 0.575m, respectively. This indicates that with a decrease in the number of satellite observations, if positioning is based solely on PPP, the error will fluctuate significantly and it will be difficult to converge to centimeter-level positioning accuracy. Using visual information to assist constrain PPP positioning can effectively compensate for insufficient satellite observations, significantly suppress positioning divergence, and accelerate positioning convergence.
[0047] like Figure 4 As shown, the diagram includes a PPP positioning trajectory (1#) assisted by visual information, a PPP positioning trajectory (2#) without visual information assistance, and a reference trajectory. The reference trajectory is a high-precision positioning sequence obtained through the establishment of a static reference station and RTK positioning technology. The PPP positioning trajectory of 1# in the diagram shows the vehicle's northward and eastward trajectories in the initial stages of movement. Although the trajectory constrained by visual information still deviates from the reference trajectory by several centimeters, its overall shape and direction are completely consistent. In contrast, the trajectory (2#) obtained solely through PPP positioning already shows significant deviations from the reference trajectory in the initial stages of movement and appears more discrete than the 1# scheme.
[0048] In the above example, the GNSS PPP fast and high-precision positioning system with visual information assistance fully leverages the advantages of monocular vision, which is not obstructed by satellite signals and can independently calculate the relative motion trajectory of the carrier. It directly extracts the linear motion features of the carrier to construct geometric observation constraints, and provides a low-cost and highly feasible solution to the problems of slow PPP positioning initialization convergence and poor positioning accuracy in conventional observation and satellite obstruction scenarios. It also maintains the continuous high-precision positioning performance of PPP in dynamic scenarios.
[0049] Example 2 This invention also provides a GNSS PPP fast and high-precision positioning system based on vision-assisted enhancement, comprising: The monocular camera intrinsic parameters and distortion coefficient calibration module calibrates the monocular camera intrinsic parameters and distortion coefficients based on multi-angle images acquired using a checkerboard pattern. The GNSS PPP observation constraint module extracts a linear window based on the calculated camera relative trajectory sliding, which is then used to construct GNSS PPP observation constraints. The fast and high-precision positioning module performs fast and high-precision calculation of the GNSS receiver position based on the constructed constraint model and GNSS pseudorange / phase observation equations.
[0050] Example 3 The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the corresponding steps in accordance with the method of Embodiment 1.
[0051] The aforementioned storage media include: USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), optical discs, and other media capable of storing program code. The specific limitations regarding the steps implemented after program execution in a computer-readable storage medium can be found in Embodiment 1, and will not be described in detail here.
[0052] Example 4 The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the computer program to implement corresponding steps in accordance with the method of Embodiment 1.
[0053] The specific limitations regarding the implementation steps of the computer device can be found in Example 1, and will not be described in detail here. It should be noted that each block in the block diagrams and / or flowcharts in the accompanying drawings of this invention, as well as combinations of blocks in the block diagrams and / or flowcharts, can be implemented by a dedicated hardware-based system that performs the specified function or action, or by a combination of dedicated hardware and computer instructions.
[0054] The present invention has been described above by way of example with reference to the accompanying drawings. Obviously, the specific implementation of the present invention is not limited to the above-described manner. Any non-substantial improvements made using the inventive concept and technical solution of the present invention, or the direct application of the inventive concept and technical solution of the present invention to other occasions without modification, are all within the protection scope of the present invention.
Claims
1. A fast and high-precision GNSS PPP positioning method based on vision-assisted enhancement, characterized in that, Includes the following steps: S1. Calibrate the intrinsic parameters and distortion coefficients of a monocular camera based on multi-angle images acquired using a checkerboard pattern; S2. A linear window is extracted based on the computational relative trajectory of the camera, and then used to construct GNSS PPP observation constraints; In this step, the relative motion trajectory composed of each historical sampling point is linearly fitted, and a linear window that satisfies the preset fitting residual is extracted. Based on the extracted linear window, GNSS observation constraints within the same time period are constructed. The GNSS observation constraints are that the sum of the lengths of each segment within the linear window is equal to the total length. After linearizing the GNSS observation constraints, an error equation is obtained to represent the degree of deviation of the geometric constraints within the linear window, thus obtaining the constraint model. S3. Based on the constraint model and GNSS pseudorange / phase observation equations constructed in the previous step, perform a fast and high-precision solution for the GNSS receiver position.
2. The GNSS PPP fast and high-precision positioning method based on vision-assisted enhancement according to claim 1, characterized in that, In step S2, optical flow is used to detect motion at feature points in the image frame, and the relative motion trajectory of the moving vehicle is calculated based on the relatively moving feature points. The formula for extracting the linear window is expressed as follows: in Indicates the sampling point. , Indicates the starting point. Indicates the current point; This represents the direction vector formed by the first and last points of the window.
3. The GNSS PPP fast and high-precision positioning method based on vision-assisted enhancement according to claim 1, characterized in that, In step S2, the GNSS observation constraints are expressed as follows: in Indicates spatial distance within different time windows. Indicates the GNSS sampling interval. Indicates the current moment. Indicates the time elapsed since the current time The length of the time window for each sampling interval; GNSS observation constraints , , After linearization, the error equations are obtained. : in: This represents the combination of the approximate distances after linearization of the three segments. They represent The receiver's three-dimensional coordinate error vector at time t. This represents the coefficient matrix corresponding to the three-dimensional coordinate error. for The three-dimensional coordinates at time [time]. for Given the three-dimensional coordinates at time t, the resulting constraint model is: in express Approximate coordinates calculated using precise single-point positioning by GNSS at a given time. Indicates the time window Approximate spatial distance within, Indicates the time window Approximate spatial distance within.
4. The GNSS PPP fast and high-precision positioning method based on vision-assisted enhancement according to claim 1, characterized in that, In step S3, the GNSS pseudorange / phase observation equation is expressed as: in and Indicates satellite to receiver exist Dual-frequency ionospheric carrier phase and pseudorange observation residuals at time t. Indicates GNSS satellite system, The receiver's three-dimensional position error vector is represented by the coefficient matrix. ; and These represent the receiver clock bias and the satellite clock bias, respectively. Indicates the ambiguity of the non-ionospheric combination. Indicates tropospheric delay, and These represent the carrier phase and pseudorange observations without ionospheric assemblies, respectively. and The noise consists of phase and pseudorange observations; satellite clock errors can be corrected using precision products, while other parameters are estimated as unknown parameters. Based on the GNSS pseudorange / phase observation equations and the constructed constraint model, the constraint observation equations for GNSS precise point positioning are expressed as follows: in ; The tropospheric wet delay is to be estimated. The tropospheric dry delay is corrected by an empirical model. The least squares estimation method is used to make optimal estimates of parameters such as receiver three-dimensional coordinate error, receiver clock error, zenith tropospheric wet delay, and ambiguity. Finally, the position of the GNSS receiver is calculated quickly and with high accuracy.
5. The GNSS PPP fast and high-precision positioning method based on vision-assisted enhancement according to claim 1, characterized in that, In step S1, the intrinsic parameters of the monocular camera are first calibrated. The chessboard used for the intrinsic parameter calibration of the monocular camera is a black and white rectangular chessboard. By taking pictures of the chessboard from different angles, the corner points of the chessboard are extracted from the pictures, and the intrinsic parameter matrix and distortion coefficient of the monocular camera are calculated.
6. The GNSS PPP fast and high-precision positioning method based on vision-assisted enhancement according to claim 5, characterized in that, In step S1, based on continuous frame images acquired by a monocular camera in an actual working environment, the ORB operator is used to extract feature points from the images. RANSAC and Lowe's ratio test methods are used to match feature point pairs, and mismatched point pairs are removed. Then, the fundamental matrix is calculated based on epipolar geometric constraints, which are expressed as follows: in This represents the homogeneous coordinates of the corresponding point in the first frame of the image. This represents the homogeneous coordinates of the corresponding point in the second frame image. Represents the fundamental matrix.
7. The GNSS PPP fast and high-precision positioning method based on vision-assisted enhancement according to claim 6, characterized in that, In step S1, the calculated intrinsic parameter matrix, distortion coefficients, and fundamental matrix of the monocular camera are used. Calculate the essential matrix The relative rotation and displacement of adjacent frames are decomposed into the following formula: in This represents the rotation matrix, indicating the relative rotation of the camera. Let represent the rotation vector, which represents the relative displacement between adjacent frames, and be further expressed as: in , and These represent the translation vector components.
8. A visual feature-assisted enhanced GNSS PPP fast and high-precision positioning system, characterized in that, include: The monocular camera intrinsic parameters and distortion coefficient calibration module calibrates the monocular camera intrinsic parameters and distortion coefficients based on multi-angle images acquired using a checkerboard pattern. The GNSS PPP observation constraint module extracts a linear window based on the calculated camera relative trajectory sliding, which is then used to construct GNSS PPP observation constraints. The fast and high-precision positioning module performs fast and high-precision calculation of the GNSS receiver position based on the constructed constraint model and GNSS pseudorange / phase observation equations.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the corresponding steps according to the visual-assisted enhanced GNSS PPP fast high-precision positioning method as described in any one of claims 1-7.
10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that, When the processor executes the computer program, it implements the corresponding steps according to the visual-assisted enhanced GNSS PPP fast and high-precision positioning method as described in any one of claims 1-7.