AR navigation method and system based on fusion of GIS and VGIS

Abstract

The invention discloses an AR (Augmented Reality) navigation method and system based on GIS (Geographic Information System) and VGIS (Visual Geographic Information System) fusion, and relates to the technical field of radio navigation. Counting step intensity of path difference vectors of a GIS planning path and a VGIS fine marking path, and primarily screening candidate step boundary bands; establishing fingerprint vector spatial distribution, and calculating a fingerprint transition amount to screen and confirm a key step boundary band; segmenting the route into stable segments by using a key step boundary band, calculating an independent translation amount based on a displacement sample, and generating a segmented alignment route; and calculating a transition route coordinate in the boundary band, combining to generate a final continuous fusion route, and carrying out AR navigation. According to the method, a geometric morphology and fingerprint dual verification mechanism is adopted, systematic deviation caused by random noise and obstacles can be accurately distinguished, accurate space fusion of virtual guidance and a real road surface in a dynamic construction scene is achieved, and navigation mold penetrating and visual drift are avoided.

Description

Technical Field

[0001] This invention relates to the field of radio navigation technology, and more specifically to an AR navigation method and system based on the fusion of GIS and VGIS. Background Technology

[0002] In complex working environments such as construction sites, workers wearing augmented reality smart terminals not only need real-time path navigation and work instructions in intricate spaces, but also need to rush to safety hazard points or accident sites in case of emergencies. These construction sites are characterized by extremely high spatial complexity and dynamic time-varying nature. They are often densely packed with steel scaffolding, and contain metal barriers that frequently move with the construction schedule, temporarily parked large machinery, and building materials with unpredictable locations. This unstructured, dynamic environment requires the navigation system to not only possess macro-level road network planning capabilities, but also to accurately integrate virtual 3D navigation icons with the real road surface to ensure the efficiency and safety of workers when traversing different construction areas or narrow passages.

[0003] However, existing augmented reality navigation technologies face a core technical challenge in the aforementioned scenarios: the difficulty in real-time matching of environmental perception with virtual map data. The evolution of the physical environment at construction sites often outpaces the update cycle of high-precision map data, leading to a mismatch in the spatiotemporal dimensions between the pre-planned geographic information system (GIS) geometric path and the current feasible navigable space. Furthermore, newly added metal fencing or large steel structures at construction sites severely interfere with the propagation of radio ranging signals, triggering complex non-line-of-sight propagation phenomena and multipath effects. This results in the positioning data received by the terminal exhibiting drastic jumps and drifts with a non-Gaussian distribution. Existing positioning and navigation technologies often struggle to effectively distinguish between sudden changes in positioning coordinates caused by random signal noise and systematic steps caused by real obstacles. This lack of boundary awareness directly prevents the navigation system from promptly and correctly avoiding or disconnecting from real physical obstacles on-site. This can lead to visual deviations such as navigation guidance lines penetrating walls or exhibiting violent shaking at obstacle edges, severely reducing the reliability of navigation guidance and potentially misleading workers into dangerous areas. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention proposes an AR navigation method and system based on the fusion of GIS and VGIS. This solves the problem that in dynamically changing and complex construction site environments, existing navigation methods struggle to effectively distinguish whether sudden changes in positioning data are caused by random signal noise or occlusion by real obstacles. This leads to augmented reality navigation guidance failing to perceive boundaries, exhibiting wall penetration or false drift, and failing to achieve accurate spatial fusion of virtual information and the real road surface.

[0005] To achieve the above objectives, the present invention provides the following technical solution: The estimated terminal position is calculated based on the set of reference point coordinates and the terminal ranging observations. The observation residual components between the terminal ranging observations and the theoretical distance determined based on the estimated terminal position are calculated, and a ranging residual fingerprint vector reflecting the signal propagation environment is constructed. Calculate the path difference vector between the GIS planned route and the VGIS precise route at the same arc length, statistically analyze the variation characteristics of the path difference vector to obtain the step intensity, and extract candidate step boundary zones based on the distribution characteristics of the step intensity. Establish a spatial distribution dataset of ranging residual fingerprint vectors along the arc length of the route, obtain the ranging residual fingerprint vector sets of the left and right neighborhoods of the candidate step boundary zone respectively, calculate the statistical difference between the two to obtain the fingerprint transition amount, and filter the candidate step boundary zones based on the fingerprint transition amount to obtain the key step boundary zone. The VGIS precision-marked route is divided into stable segments using the critical step boundary zone. The independent translation amount of the stable segment is determined based on the coordinate difference between the estimated terminal position within the stable segment and the VGIS precision-marked route. The segmented aligned route is obtained after correcting the stable segment based on the independent translation amount. The transition route coordinates connecting adjacent segment alignment routes are calculated within the critical step boundary zone, and combined to obtain the final continuous fusion route for AR navigation.

[0006] Furthermore, an AR navigation system based on the fusion of GIS and VGIS is proposed to implement the AR navigation method based on the fusion of GIS and VGIS as described above, including: The positioning fingerprint module is used to obtain the set of reference point coordinates and the terminal ranging observation values, calculate the terminal position estimate using the least squares method, calculate the observation residual components, and construct a ranging residual fingerprint vector that characterizes the current signal propagation environment. The candidate boundary module is used to acquire and align the GIS planned route and the VGIS precise marked route, calculate the path difference vector between the two at the same arc length, calculate the step intensity using a sliding window, and extract a set of candidate step boundary zones based on the step intensity. The critical boundary module is used to perform spatial mapping from terminal location sequence to route arc length, establish spatial distribution dataset, obtain fingerprint sets of the neighborhood on both sides of the candidate step boundary zone, calculate fingerprint transition amount, and screen and confirm the critical step boundary zone. The segment alignment module is used to divide the VGIS precision-marked route into stable segments through the key step boundary zone, and to generate displacement samples by statistically analyzing the difference between the estimated terminal position value falling within each stable segment and the route point, thereby obtaining independent translation amounts and generating segment alignment routes. The stitching navigation module is used to calculate the linearly weighted transition route coordinates based on the left and right segmented alignment routes within the critical step boundary zone, generate the final continuous fusion route based on the segmented alignment routes and transition routes, and perform AR navigation.

[0007] Compared with existing technologies, it has the following advantages: The proposed AR navigation method and system based on the fusion of GIS and VGIS effectively solves the technical challenges of mismatch between augmented reality navigation guidance and real road space, as well as the problem of penetrating physical structures, caused by complex obstacles in dynamic construction sites. By constructing a dual verification mechanism of geometric step strength and signal fingerprint transition, this invention can accurately identify the systematic deviation boundary caused by changes in the real environment from non-Gaussian distributed positioning data. Based on this, the route is divided into several stable segments with uniform properties and subjected to independent rigid correction based on median statistics. Combined with a linear weighted smooth transition algorithm within the boundary band, this eliminates false jitter caused by signal multipath effects and ensures the visual continuity and accuracy of navigation guidance when crossing different areas, achieving a close fit between virtual information and the actual scene of the work site. Attached Figure Description

[0008] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 This is a schematic diagram of the system framework of the present invention. Detailed Implementation

[0009] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0010] Please see Figure 1 This application provides an AR navigation method based on the fusion of GIS and VGIS; The method specifically includes the following steps: Step 1: Obtain the set of reference point coordinates using pre-set hardware. and terminal ranging observations Among them, the set of reference point coordinates The three-dimensional spatial coordinate data of pre-deployed positioning base stations or selected satellite reference points in the engineering coordinate system can establish a global geometric reference frame for ranging calculations, and each reference point has a unique number in the system, and the terminal ranging observation values... The straight-line scalar distance data between the AR navigation terminal and the i-th reference point is obtained by the AR navigation terminal using ultra-wideband or real-time dynamic differential radio ranging modules at the current sampling time. The AR navigation terminal is both the object being located and the carrier device for displaying the AR navigation screen to the user. The terminal ranging observation value is the basic observation value for subsequent position calculation. It should be noted that, to ensure the accuracy of subsequent calculations, the system must ensure that all terminal ranging observations are strictly synchronized with the corresponding reference point coordinates in the time dimension. Under dynamic navigation, if the terminal moves rapidly, time differences in ranging data from different base stations can cause distortion of the calculated position. Therefore, the system needs to perform time alignment processing on multiple ranging data to ensure that all terminal ranging observations are synchronized in time. They all belong to the same discrete sampling time; Based on the set of reference point coordinates and terminal ranging observations The overdetermined ranging equations are processed using a least squares optimization algorithm to obtain the optimal coordinate solution of the terminal in three-dimensional space. The specific calculation formula is as follows: ; In the formula, Here, m represents the estimated terminal location, and m is the total number of reference points. This equals the value of the position variable x that minimizes the objective function; specifically, the estimated terminal position. It represents the most likely location of the terminal in geometric space under the current observation noise level. It is a three-dimensional vector and serves as the geometric reference point for projecting the terminal onto the navigation route and rendering augmented reality images in subsequent steps. It should be noted that the least squares method used here is the optimal estimation based on the Gaussian white noise assumption. In actual working conditions, if the signal quality of different base stations varies greatly, this step can also use the weighted least squares algorithm, that is, assign different weight coefficients to each error term according to the signal strength or signal-to-noise ratio, so as to further improve the solution accuracy of geometric coordinates. Based on the set of reference point coordinates Terminal ranging observations and terminal location estimates The observation residual components are calculated and the ranging residual fingerprint vector is constructed. The specific calculation formula is as follows: ; In the formula, To observe the residual components, observe the residual components. Equal to the measured terminal distance observation value Subtract the estimated value based on the terminal location Coordinates of the i-th reference point The difference obtained from the calculated theoretical Euclidean distance, specifically, the observed residual components. It reflects the degree of non-line-of-sight obstruction or the intensity of multipath effects on the signal transmission path; Set of reference point coordinates The observation residual components corresponding to all reference points are arranged sequentially according to the reference point numbering order to form the ranging residual fingerprint vector e, which is: ranging residual fingerprint vector e = , where m indicates that the ranging residual fingerprint vector is an m-dimensional vector. Because there are m reference points, there will be m observation residual components. The ranging residual fingerprint vector e is a multi-dimensional vector, which is a digital fingerprint that characterizes the signal propagation quality of the current terminal's micro-environment. It is used in subsequent steps as a key basis for determining whether there are boundary (the boundary refers to the physical edge of the construction site that is spatially distributed in a linear or strip shape, including but not limited to temporary fence lines, steel plate road edges, deep foundation pit edges, and the obstruction boundary of metal scaffolding or large mechanical equipment. This boundary is the root cause of non-line-of-sight propagation, multipath effect abrupt changes, and step changes in the positioning error distribution of radio ranging signals) caused by signal abrupt changes. This helps the system effectively distinguish between pure geometric errors and systematic deviations caused by environmental abrupt changes. It should be noted that the observed residual components It has a clear directionality. When the signal is transmitted in a straight line without obstruction, this value usually fluctuates around zero. However, when the signal passes through a barrier or is reflected by metal, the measured distance will be greater than the actual geometric distance due to non-line-of-sight propagation, and this value will show a significant positive bias. Therefore, the ranging residual fingerprint vector actually records the obstruction distribution around the current location. At the same time, when constructing the ranging residual fingerprint vector, the fixed order of the reference point number must be followed to ensure that each dimension of the vector corresponds to the spatial features of the same base station when fingerprint comparison is performed in subsequent steps.

[0011] Step Two: Obtain the GIS planned route and the VGIS precise mapping route, and preprocess and spatially synchronize the GIS planned route and the VGIS precise mapping route to establish a common benchmark for comparing the two routes. Specifically, the GIS planned route... This is a sequence of navigation path points extracted from a geographic information system database and generated based on road network topology planning. It represents the designed and planned route, a VGIS-marked route. To generate a high-precision pathpoint sequence reflecting the actual passable area of ​​the construction site using drone aerial photography or 3D laser scanning, the system resamples the two routes using an arc-length parameterization method for point-by-point comparison. and The parameter representing the cumulative arc length along the same path. The three-dimensional spatial coordinates of the sampling point are given, where j is the sequence number of the sampling point, and Where n is the total number of sampling points; GIS-based route planning and VGIS precision route The spatial displacement deviation between two routes at the same arc length is calculated using the following formula: ; In the formula, The path difference vector. It equals the spatial coordinate vector of the VGIS precision route at the j-th sampling point. Subtract the spatial coordinate vector of the GIS planning route The resulting difference vector, specifically, the path difference vector. The value reflects the geometric deviation of the actual passable path at the current location from the macro-planned path. Under ideal rigid registration, the path difference vector should remain constant or change slowly throughout the entire road segment. However, at the construction site, abrupt changes in the path difference vector often indicate the existence of a boundary. Quantizing path difference vectors using sliding window statistical method The degree of change in [the value of the step] is used to identify the step and obtain the step intensity. The calculation formula is as follows: ; In the formula, Step intensity, step intensity The median of the path difference vectors in the right window of the current sampling point minus the median of the path difference vectors in the left window is the Euclidean norm of the resulting difference vector. This represents the set of path difference vectors for a predetermined number of sampling points following sampling point j. This represents the set of path difference vectors for a predetermined number of sampling points preceding sampling point j. `med` represents the operation of taking the median of the vector components. Specifically, the step intensity... It can effectively suppress the influence of measurement noise and capture systematic jumps in displacement deviation. When vehicles or personnel cross boundaries such as barriers, the path deviation mode switches, causing a significant increase in the median difference between the left and right windows, thus increasing the step intensity. A peak occurred; Specifically, the size of the sliding window (i.e., the number of sampling points within the window) is not fixed and can be adjusted according to the sampling frequency and the scale of the construction site. In a preferred embodiment, the actual road section covered by the sliding window is preferably 2 to 5 meters. If the sampling interval of the VGIS route is 0.5 meters, the corresponding preset number of sampling points is preferably 4 to 10. If the sampling interval is denser (e.g., 0.1 meters), the corresponding preset number of sampling points can be increased to 20 to 50. If there are too few sampling points (e.g., less than 3), the median statistics cannot effectively eliminate occasional outliers, leading to false alarms in step detection. If there are too many sampling points (e.g., covering more than 10 meters), the step signal will be smoothed out, resulting in a decrease in the accuracy of the detected boundary position and an inability to accurately locate the edge of the fence. It should be noted that a step refers to a discontinuous abrupt change in the spatial positional deviation vector between the VGIS precision route and the GIS planned route when passing through a specific geographical location during the navigation process. Geometrically, this manifests as a significant discontinuous difference in the deviation distribution before and after the location rather than a smooth gradual change. The specific technique for identifying this step phenomenon involves setting up a sliding window along the route that covers a preset number of sampling points (e.g., the number of sampling points corresponding to an actual distance of 3 meters). The path difference vector sets in the left and right neighborhoods of the center point of the window are extracted respectively, and the statistical median vectors of these two sets are calculated to eliminate the influence of individual noise points. The step intensity is a physical quantity used to digitally characterize the severity of the above-mentioned abrupt change. The value of the step intensity is equal to the Euclidean norm of the difference vector obtained by subtracting the median vector on the left from the median vector on the right side of the window. The larger the value, the more significant the difference in systematic positioning deviation on both sides of the location, thus indicating that there is a signal propagation environment switch caused by boundaries such as fences or building structure edges. Based on step strength Based on the distribution characteristics, arc length intervals where anomalies may exist are delineated. Specifically, the system traverses the step intensity sequence of the entire road segment, identifies all local maxima, and for each local maxima, the system extends to the left and right sides along the arc length direction until the step intensity value falls back to the background median level of the entire road segment. The continuous arc length interval covering the peak determined in this way is a candidate step boundary zone. , where k is the candidate band number; Specifically, candidate step boundary zone This refers to a continuous interval intercepted on the arc-length coordinate axis along the navigation route. This interval spatially covers the transition area where the positional deviation between the GIS planned route and the VGIS precision-calibrated route experiences drastic fluctuations or jumps; it is a candidate step boundary zone. The extraction process is as follows: The system first identifies the local maxima in the step intensity sequence and regards them as the center of potential mutations. Then, starting from the center, it searches in the forward and backward directions of the route until the step intensity value decays to the background median level of the entire road segment or the preset noise floor level. The range from the starting arc length to the ending arc length determined at this time is the candidate step boundary zone. This area represents the road segment with the worst geometric data consistency in the navigation system. It not only contains the theoretical mutation breakpoints, but also accommodates the transition edges introduced by the data sampling discreteness and the smoothing effect of the sliding window calculation. It is an effective target area for the system to use the ranging residual fingerprint for consistency verification and to remove pure geometric noise interference in subsequent steps. It should be noted that the use of median difference rather than mean difference to define step intensity is to deal with common data interference in construction sites. Since VGIS data may have local modeling noise, the mean calculation is easily skewed by a single extreme outlier, thus generating false step signals. Median statistics are more robust and can ensure that the step intensity will only increase significantly when the deviation of a continuous area shifts as a whole. In another embodiment, to improve computational speed, the step intensity calculation can employ a simplified algorithm based on first-order difference or gradient. In this embodiment, the step intensity is defined as the Euclidean distance between the path difference vectors of two adjacent sampling points, or the rate of change of the path difference vector within a certain step size. Although this method is weaker than the median method in suppressing high-frequency noise, it can still eliminate false detections caused by noise through fingerprint verification in subsequent steps on low-power embedded terminals with extremely limited computing resources, thereby reducing system power consumption while ensuring a certain detection effect.

[0012] Step 3: Obtain the estimated terminal position at the current moment. And the ranging residual fingerprint vector e, the VGIS precision route sampling point sequence that has been resampled. As a spatial reference benchmark, the VGIS precision route sampling point sequence is traversed. Calculate the estimated terminal position at the current moment. VGIS precision route sampling point sequence The Euclidean distance between each sampling point is used to select the sampling point corresponding to the minimum value of the Euclidean distance as the best matching point. The arc length parameter s corresponding to the best matching point on the VGIS precision route is read. The distance residual fingerprint vector e at the current time is marked as a spatial fingerprint sample associated with the arc length parameter s. In this way, a spatial distribution dataset is constructed with the arc length parameter of the route as the index and the distance residual fingerprint vector as the element. Specifically, the observation data in the discrete time dimension is anchored to the navigation route in the continuous spatial dimension through the geometric projection algorithm to obtain the spatial distribution dataset. Traversing candidate step boundary zones Set, for each candidate step boundary zone Define the region to the left of the starting arc length of the candidate step boundary zone by a predetermined length (e.g., extending 2 meters) as the left neighborhood, and define the region to the right of its ending arc length by a predetermined length as the right neighborhood. From the spatially distributed dataset, retrieve all ranging residual fingerprint vectors that fall within the left neighborhood to form the left neighborhood fingerprint set. Similarly, all ranging residual fingerprint vectors falling within the right neighborhood are retrieved to form the right neighborhood fingerprint set. ; It should be noted that the candidate step boundary zone As a continuous route segment experiencing severe geometric fluctuations, it has a clearly defined coverage area on the route arc length parameter axis. The initial arc length refers to the minimum cumulative arc length value corresponding to the candidate step boundary zone in the route's forward direction, while the final arc length is the corresponding maximum cumulative arc length value. To capture the stable signal propagation state before and after this abrupt change in the route segment, the system defines the route interval located before and immediately adjacent to the initial arc length as the left neighborhood, and the route interval located after and immediately adjacent to the final arc length as the right neighborhood. The left and right sides correspond to the upstream and downstream directions of the increasing cumulative arc length. The preset length refers to the actual distance span covered by these two neighborhood intervals on the arc length axis. The preset length is set by... The spatial scale of typical obstacles at the construction site (such as fences or intersections) and the minimum sampling distance required for the signal propagation state to stabilize are often referenced. For example, a range of 2 to 5 meters is set to accommodate enough samples for robust statistics. The neighborhood interval defined in this way actually delineates the sampling window for extracting background features. The construction of the left neighborhood fingerprint set and the right neighborhood fingerprint set is to obtain the distance measurement residual statistical samples at two different actual positions of the terminal before entering the suspected mutation area and after leaving the area. By comparing the differences between the two sets, the system can quantitatively determine whether the signal propagation environment has undergone a substantial step change with the change of geometric path, thereby effectively distinguishing the real obstacle boundary from accidental measurement noise. Based on each candidate step boundary zone Left neighborhood fingerprint set and the fingerprint set of the right neighboring area The specific formula for calculating fingerprint transitions is as follows: ; In the formula, For fingerprint transitions, fingerprint transitions The difference vector is equal to the Euclidean norm of the difference vector obtained by subtracting the median vector of the left neighborhood fingerprint set from the median vector of the right neighborhood fingerprint set. Specifically, the system first sorts the same-dimensional components of all vectors in the right neighborhood fingerprint set and takes the median value to synthesize a right median vector representing the right signal environment. Similarly, the left median vector is generated. The difference vector obtained by subtracting the two represents the systematic change magnitude of the signal environment when crossing the boundary band. The magnitude of this difference vector is the fingerprint transition amount. fingerprint transition amount It quantifies the severity of the systematic step change in the radio signal propagation state when the terminal crosses this boundary band; Get all candidate step boundary bands in the current batch processing Calculated fingerprint transition amount The transition values ​​are sorted in descending order of magnitude to generate a sorted sequence. The difference between adjacent elements in this sorted sequence is calculated to construct a first-order difference sequence of transition values. The position with the largest rate of change in the first-order difference sequence is identified, namely the elbow point of the sorting curve. This elbow point not only represents the turning point of the most drastic numerical decay in the sorted sequence, but also delineates the boundary between strong signal abrupt changes and weak background fluctuations. All high-fingerprint transition values ​​located at and before the elbow point in the sorted sequence are then identified. Corresponding candidate step boundary zone If a genuine signal occlusion abrupt change is identified, it is marked as valid, and all candidate step boundary bands marked as valid are included. Marked as critical step boundary zone Where q represents the q-th critical step boundary zone, and the critical step boundary zone To definitively confirm the existence of a specific route segment with abrupt changes in the actual signal propagation state, reflecting the dual attributes of severe geometric deviation jitter and substantial signal environment switching in this region at that moment, it is proven that the path offset at this location is not caused by random noise, but by a systematic deviation caused by environmental changes, while the key step boundary zone... This indicates the precise spatial location of physical obstacles such as the edge of metal fences, the obstruction line of large steel structures, or the boundary of deep foundation pits on the navigation route. It is the core anchor point for the system to segment and cut the entire route, independently correct the offset, and perform smooth splicing operations. Specifically, on the construction site, regardless of whether the overall signal environment is too small or too large, the fingerprint jump caused by the real boundary (such as metal fence) is always significantly different from ordinary noise, which is manifested as a few maximum values ​​at the head of the sorting curve. This relative filtering mechanism ensures the robustness of the system in different construction scenarios. In another embodiment, to further distinguish between static real boundaries and dynamic interference (such as temporarily parked trucks), a time dimension stability check can be added during the calculation of fingerprint transitions. In this embodiment, the system not only calculates the spatial fingerprint transitions, but also calculates the temporal variance of the fingerprint sets in the left and right neighborhoods. Only when the fingerprint transitions are significant and the temporal variances within the sets on both the left and right sides are lower than a preset stability threshold is the boundary confirmed as a critical step boundary zone. This implementation effectively filters out instantaneous signal fluctuations caused by people walking or vehicles passing by, ensuring that the navigation system only segments the path for long-standing facility boundaries.

[0013] Step 4: Obtain the critical step boundary zone The VGIS precision route sampling point sequence after aggregation and resampling Using the starting and ending arc lengths of each critical step boundary zone as the dividing points, the following segmentation operation is performed: In the VGIS precision-marked route, a continuous route interval located between two adjacent critical step boundary zones is defined as a stable segment. The interval between the starting point of the route and the starting arc length of the first critical step boundary zone, and the interval between the ending arc length of the last critical step boundary zone and the end point of the route, are also defined as independent stable segments, where p is the p-th stable segment. It should be noted that by dividing the geometric space, the continuous VGIS precision route is discretized into several road segments with independent error attributes, namely stable segments. A stable segment refers to a route segment in the construction site that is located between two obstacles (such as fences) and within the same area. Within this segment, the propagation environment of the radio ranging signal (line-of-sight and non-line-of-sight states) remains relatively consistent and does not undergo abrupt changes. Therefore, the positioning error in this segment is a systematic bias of a single mode, which is suitable for correction using uniform smoothing parameters. The purpose of this step is to decompose the complex nonlinear full-segment error into several simple linear local error problems. In addition, the arc length interval covered by the key step boundary zone itself (i.e., from the starting arc length to the ending arc length) does not belong to any stable segment. As a transition area for signal environment switching, it will be processed separately for smooth splicing in subsequent steps. For each stable segment The systemic geometric bias specific to this region is estimated using real-time positioning data from the terminal. Specifically, this involves traversing each stable segment. Retrieve time-stamped terminal location estimates sequence Filter out all those that fall into this stable segment For each selected terminal position estimate (corresponding to time t) within the arc length range, calculate the terminal position estimate vector for that time. Subtract the VGIS precise route point coordinate vector at that moment. The resulting difference vector is the displacement sample. Specifically, displacement samples The object represents the deviation vector between the measured position of the terminal and the theoretical position on the map at time t; A stable segment All displacement samples within The median vector of a vector set is labeled as an independent translation. Independent translation amount This represents the average systematic positioning error within the stable segment due to specific environmental occlusion or multipath effects. Using median statistics effectively filters out instantaneous random noise caused by rapid personnel movement or equipment vibration, extracting the inherent environmental bias of the area, and then using independent translation amounts... It can quantify the degree of misalignment in each region; For those belonging to the stable phase Each VGIS precision route sampling point within Add the original coordinate vector to the independent translation corresponding to the stable segment. The corrected coordinate vector is obtained by adding the independent translation amount of the stable segment to the coordinates of the VGIS precision-marked route at arc length s, thus obtaining the segmented aligned route of the stable segment. traverse all stable segments Repeat the above steps to generate a set of segmented aligned routes containing all corrected road segment data. Specifically, the segmented aligned route set is a set of discontinuous high-precision route segments, in which each route segment is precisely aligned with the real space within its respective region, eliminating systematic biases in that region. Although these segments may have geometric discontinuities (i.e. breakpoints) at the boundary zone, they are accurate reference data for constructing the final continuous fused route. This step corrects the error of each segment separately. It should be noted that if the number of displacement samples collected in a certain stable segment is insufficient (e.g., less than 10, possibly due to the terminal passing through the segment quickly), the system will use the translation amount of the adjacent road segment or the translation amount of global coarse alignment as a substitute to avoid estimation bias caused by sparse samples. In addition, the calculation of displacement samples depends on the accurate arc length mapping established in the previous steps, ensuring that each positioning point is compared with the correct position on the route.

[0014] Step 5: For each critical step boundary zone The transition geometry connecting the two stable segments is calculated as follows: Obtain the set of critical step boundary zones and segmented alignment route set For the q-th critical step boundary zone, identify its starting arc length s1 and ending arc length s2. At the same time, lock the left segment alignment path corresponding to the stable segment located before the critical step boundary zone. (i.e., the correction results of the preceding road segment), and the right segment alignment route corresponding to a stable segment located after this critical step boundary zone. (i.e., the correction results for subsequent road segments); Based on the left-side segment alignment route Align the route with the right-side segment Calculate the coordinates of the transition path within the interval [s1, s2]. The calculation formula is as follows: ; In the formula, The transition path coordinates are the transition path coordinates at any arc length s inside the critical step boundary zone. It equals the coordinate vector of the left segment alignment route extending at that point multiplied by the left weight coefficient, plus the coordinate vector of the right segment alignment route extending in the opposite direction at that point multiplied by the right weight coefficient. The right weight coefficient is equal to the difference between the current arc length s and the initial arc length s1 divided by the total length of the critical step boundary zone (i.e., s2 minus s1), and the left weight coefficient is equal to 1 minus the right weight coefficient. Specifically, the left-side segmented alignment route and the right-side segmented alignment route refer to the geometric trajectories after rigid correction of the original VGIS route using independent systematic deviation parameters within their respective stable regions, before entering the critical step boundary zone and after leaving the boundary zone. Due to the drastically different radio signal propagation environments (e.g., line-of-sight and non-line-of-sight) on these two sides, there are significant spatial breaks or misalignments at the start and end positions of these two routes within the boundary zone. The transition route coordinates are essentially a virtual bridging path constructed within this misalignment interval. This simulates the dynamic gradual transition of navigation guidance from the error state of the preceding segment to the error state of the subsequent segment, preventing frame skipping or arrow teleportation in the AR image due to sudden changes in the positioning reference. The calculation process of the transition route coordinates is based on linear... The interpolation principle is applied to solve the AR navigation discontinuity problem caused by ranging fingerprint transitions. The right-hand weight coefficient introduced in the formula quantifies the normalized proportion of the current position relative to the total length of the boundary band (i.e., the closer to the end point, the larger the proportion), while the left-hand weight coefficient is the corresponding complementary value. Through the weight distribution that changes linearly with the arc length, it can force the coordinate values ​​of the transition route at the starting point s1 to be completely equal to the left-hand route, and the coordinate values ​​at the ending point s2 to be completely equal to the right-hand route. In the middle area, the geometric features of the two are mixed according to the distance ratio, thereby strictly ensuring the continuity of the synthesized route in three-dimensional space. This processing eliminates the visual frame skipping or breakage caused by the sudden change of the positioning coordinate system due to obstacle occlusion, and realizes seamless connection of navigation guidance when crossing physical obstacles. Iterate through the range of arc lengths along the entire navigation route. If the current arc length s falls into a certain stable segment... Within the specified range, the system directly calls the segment alignment route. As the final coordinates, if the current arc length s falls into a certain critical step boundary zone Within the range, system call transition route coordinates As the final coordinates, the final continuous fusion route is formed by combining these coordinates. It is a continuous curve in three-dimensional space, which retains the high-precision correction characteristics of each local area and ensures global connectivity. Obtain the estimated terminal position at the current moment. And the terminal's attitude angle data (obtained by the built-in IMU) are used to calculate the terminal's position. In the final continuous fusion route The nearest neighbor point is used as the rendering anchor point. Based on the rendering anchor point and its tangent direction, a virtual navigation arrow or guide light strip model is generated using a computer graphics perspective projection algorithm. Finally, the system overlays and renders the virtual model onto the display module of the AR terminal (such as smartphones, AR glasses, etc.) so that it visually fits perfectly with the real road scene at the construction site. In another embodiment, to achieve better visual smoothness, especially in high-speed moving scenarios (such as engineering vehicle navigation), the weight coefficient calculation in the transition route coordinate calculation process can adopt a non-linear Sigmoid function or cubic spline interpolation. In this embodiment, the rate of change of the weight coefficient approaches zero at both ends of the boundary band and is fastest in the middle. This implementation makes the transition route continuous not only in position coordinates but also in the tangential direction, thereby making the turning action of the AR navigation arrow smoother and more natural.

[0015] Furthermore, refer to Figure 2 As shown, an AR navigation system based on the fusion of GIS and VGIS is proposed to implement the AR navigation method based on the fusion of GIS and VGIS as described above, including: The positioning fingerprint module is used to obtain the set of reference point coordinates and the terminal ranging observation values, calculate the terminal position estimate using the least squares method, calculate the observation residual components, and construct a ranging residual fingerprint vector that characterizes the current signal propagation environment. The candidate boundary module is used to acquire and align the GIS planned route and the VGIS precise marked route, calculate the path difference vector between the two at the same arc length, calculate the step intensity using a sliding window, and extract a set of candidate step boundary zones based on the step intensity. The critical boundary module is used to perform spatial mapping from terminal location sequence to route arc length, establish spatial distribution dataset, obtain fingerprint sets of the neighborhood on both sides of the candidate step boundary zone, calculate fingerprint transition amount, and screen and confirm the critical step boundary zone. The segment alignment module is used to divide the VGIS precision-marked route into stable segments through the key step boundary zone, and to generate displacement samples by statistically analyzing the difference between the estimated terminal position value falling within each stable segment and the route point, thereby obtaining independent translation amounts and generating segment alignment routes. The stitching navigation module is used to calculate the linearly weighted transition route coordinates based on the left and right segmented alignment routes within the critical step boundary zone, generate the final continuous fusion route based on the segmented alignment routes and transition routes, and perform AR navigation.

[0016] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. An AR navigation method based on the fusion of GIS and VGIS, characterized in that, include: The estimated terminal position is calculated based on the set of reference point coordinates and the terminal ranging observations. The observation residual components between the terminal ranging observations and the theoretical distance determined based on the estimated terminal position are calculated, and a ranging residual fingerprint vector reflecting the signal propagation environment is constructed. Calculate the path difference vector between the GIS planned route and the VGIS precise route at the same arc length, statistically analyze the variation characteristics of the path difference vector to obtain the step intensity, and extract candidate step boundary zones based on the distribution characteristics of the step intensity. Establish a spatial distribution dataset of ranging residual fingerprint vectors along the arc length of the route, obtain the ranging residual fingerprint vector sets of the left and right neighborhoods of the candidate step boundary zone respectively, calculate the statistical difference between the two to obtain the fingerprint transition amount, and filter the candidate step boundary zones based on the fingerprint transition amount to obtain the key step boundary zone. The VGIS precision-marked route is divided into stable segments using the critical step boundary zone. The independent translation amount of the stable segment is determined based on the coordinate difference between the estimated terminal position within the stable segment and the VGIS precision-marked route. The segmented aligned route is obtained after correcting the stable segment based on the independent translation amount. The transition route coordinates connecting adjacent segment alignment routes are calculated within the critical step boundary zone, and combined to obtain the final continuous fusion route for augmented reality navigation.

2. The AR navigation method based on the fusion of GIS and VGIS according to claim 1, characterized in that, include: Based on the set of reference point coordinates and the terminal ranging observations, the terminal position estimate is determined by the least squares optimization algorithm. The terminal position estimate is the three-dimensional coordinates of the terminal in geometric space under the current observation noise level. For each reference point in the set of reference point coordinates, the theoretical Euclidean distance between the coordinates of the reference point and the estimated terminal position is calculated. The measured terminal distance observation value corresponding to the reference point is subtracted from the theoretical Euclidean distance, and the difference is the observation residual component corresponding to the reference point. The observation residual component is used to characterize the degree of non-line-of-sight occlusion on the signal transmission path. The observation residual components corresponding to all benchmark points in the benchmark point coordinate set are arranged sequentially according to the preset numbering order of the benchmark points to form a multi-dimensional vector as the ranging residual fingerprint vector.

3. The AR navigation method based on the fusion of GIS and VGIS according to claim 1, characterized in that, include: Arc length parameterized resampling is performed on the GIS planned route and the VGIS precise marked route. The difference vector obtained by subtracting the spatial coordinate vector of the GIS planned route from the spatial coordinate vector of the VGIS precise marked route at the same arc length sampling point is used as the path difference vector. The step intensity is obtained by analyzing the changes in the statistical path difference vector, as follows: Centered on the current sampling point, define a left sliding window and a right sliding window along the route that cover a preset number of sampling points; Obtain the set of path difference vectors that fall into the left sliding window and the set of path difference vectors that fall into the right sliding window respectively; Calculate the first median vector of the path difference vector set within the left sliding window and the second median vector of the path difference vector set within the right sliding window, respectively. The Euclidean norm of the difference vector between the second median vector and the first median vector is calculated to obtain the step intensity. The step intensity reflects the severity of the discontinuous change in the spatial position deviation between the GIS planned route and the VGIS precise route at the current location.

4. The AR navigation method based on the fusion of GIS and VGIS according to claim 3, characterized in that, The specific process for extracting candidate step boundary zones includes: Traverse the sequence of step intensities distributed along the arc length of the route to identify all local maxima. Each local maximum point is taken as the center of a potential mutation, and the search is performed from the center point in both the forward and backward directions of the route. The search stops when the detected step intensity value decays to the background median level of the entire road segment, and the starting arc length and ending arc length corresponding to the search stop position are determined respectively. A continuous interval from the starting arc length to the ending arc length is defined as a candidate step boundary zone. The candidate step boundary zone is a transitional area that spatially covers the positional deviation between the GIS planned route and the VGIS precision-marked route, where there is a jitter or jump.

5. The AR navigation method based on the fusion of GIS and VGIS according to claim 4, characterized in that, include: Project each terminal location estimate onto the VGIS precision route to obtain the corresponding route arc length. Establish a one-to-one mapping relationship between the calculated ranging residual fingerprint vector and the route arc length, and collect all mapping relationships into a spatial distribution dataset sorted by arc length. The region extending a first preset length in the opposite direction of the route from the starting position of the candidate step boundary zone is taken as the left neighborhood, and the region extending a second preset length in the positive direction of the route from the ending position of the candidate step boundary zone is taken as the right neighborhood. The ranging residual fingerprint vectors falling into the left neighborhood and the right neighborhood are extracted from the spatial distribution dataset respectively to obtain the left neighborhood fingerprint set and the right neighborhood fingerprint set. The Euclidean norm of the difference vector obtained by subtracting the median vector of the left neighborhood fingerprint set from the median vector of the right neighborhood fingerprint set is used as the fingerprint transition quantity. The fingerprint transition quantity is used to quantify the degree of abrupt change in the signal propagation environment on both sides of the candidate step boundary band.

6. The AR navigation method based on the fusion of GIS and VGIS according to claim 5, characterized in that, include: Obtain the fingerprint transition values ​​corresponding to all candidate step boundary bands, and sort them in descending order of numerical value to generate a transition value sorting sequence; Calculate the difference between adjacent elements in the transition quantity sorting sequence, construct a first-order difference sequence of transition quantities, identify the position with the largest rate of change in this first-order difference sequence, and mark it as the elbow point. Candidate step boundary bands corresponding to all fingerprint transitions located at and before the elbow point in the transition sorting sequence are identified as having real signal occlusion abrupt changes and are marked as valid. All candidate step boundary bands marked as valid are marked as critical step boundary bands. The critical step boundary bands reflect the environmental boundary points that cause systematic abrupt changes in the distribution characteristics of terminal positioning errors.

7. The AR navigation method based on the fusion of GIS and VGIS according to claim 6, characterized in that, include: The route sections located between two adjacent critical step boundary zones, the route sections located between the starting point of the route and the first critical step boundary zone, and the route sections located between the last critical step boundary zone and the end point of the route are respectively marked as stable segments with uniform system error characteristics. Obtain the estimated terminal positions for all projected locations falling within the current stable segment. Calculate the difference vector between the estimated terminal position vector and the corresponding VGIS-marked route point coordinate vector. Mark this difference vector as a displacement sample. The median vector of the set of all displacement sample vectors within a stable segment is marked as the independent translation. The independent translation represents the average systematic positioning error in the region where the stable segment is located due to specific environmental occlusion or multipath effects. Adding the independent translation amount to the arc length coordinates of the VGIS-marked route to obtain the segmented alignment route of the stable segment.

8. The AR navigation method based on the fusion of GIS and VGIS according to claim 7, characterized in that, include: For each critical step boundary zone, determine the left segment alignment path corresponding to the stable segment immediately preceding the critical step boundary zone, and the right segment alignment path corresponding to the stable segment immediately following the critical step boundary zone. Obtain the arc length position of the current calculation point within the critical step boundary zone, calculate the ratio of the distance between this arc length position and the starting arc length of the critical step boundary zone to the total length of the critical step boundary zone, determine this ratio as the right-side weight coefficient, and determine the difference between 1 and the right-side weight coefficient as the left-side weight coefficient. The left-side weighting coefficient is used to weight the extended coordinates of the left-side segmented alignment route at the current position, and the right-side weighting coefficient is used to weight the extended coordinates of the right-side segmented alignment route at the current position. The weighted results of the two are added together to obtain the transition route coordinates. Construct a continuous virtual guide curve in three-dimensional space. The curve uses the coordinates of the corresponding segmented alignment route in each stable segment and the coordinates of the transition route in each critical step boundary zone. Render continuous navigation guide icons in the augmented reality view based on the virtual guide curve.

9. The AR navigation method based on the fusion of GIS and VGIS according to claim 3, characterized in that, The specific process of arc-length parameterized resampling includes: By utilizing the road network topology, the GIS planned routes and the VGIS precise marked routes are projected onto the same engineering coordinate system, establishing a unified reference framework for spatial geometric comparison between the two. Based on the common starting point of the GIS planned route and the VGIS precise standard route, the cumulative arc length along the route direction is calculated respectively, and their respective arc length parameter models are constructed. By setting a uniform sampling interval, the two routes are resampled at equal intervals using the arc length parameter model, generating a GIS resampled coordinate set and a VGIS resampled coordinate set with the same arc length index sequence.

10. An AR navigation system based on the fusion of GIS and VGIS, used to implement the AR navigation method based on the fusion of GIS and VGIS as described in any one of claims 1-9, characterized in that, include: The positioning fingerprint module is used to obtain the set of reference point coordinates and the terminal ranging observation values, calculate the terminal position estimate using the least squares method, calculate the observation residual components, and construct a ranging residual fingerprint vector that characterizes the current signal propagation environment. The candidate boundary module is used to acquire and align the GIS planned route and the VGIS precise marked route, calculate the path difference vector between the two at the same arc length, calculate the step intensity using a sliding window, and extract a set of candidate step boundary zones based on the step intensity. The critical boundary module is used to perform spatial mapping from terminal location sequence to route arc length, establish spatial distribution dataset, obtain fingerprint sets of the neighborhood on both sides of the candidate step boundary zone, calculate fingerprint transition amount, and screen and confirm the critical step boundary zone. The segment alignment module is used to divide the VGIS precision-marked route into stable segments through the key step boundary zone, and to generate displacement samples by statistically analyzing the difference between the estimated terminal position value falling within each stable segment and the route point, thereby obtaining independent translation amounts and generating segment alignment routes. The stitching navigation module is used to calculate the linearly weighted transition route coordinates based on the left and right segmented alignment routes within the critical step boundary zone, generate the final continuous fusion route based on the segmented alignment routes and transition routes, and perform AR navigation.

Images