A mobile terminal positioning and deviation rectification method based on space-time constraint and orientation perception

Abstract

The application provides a mobile terminal positioning and deviation rectification method based on space-time constraints and orientation perception, and belongs to the technical field of mobile terminal positioning. The method comprises the following steps: firstly, a scenic area road network graph is constructed based on target area electronic map data; secondly, Bluetooth anchor point signal data and a mobile terminal continuous time sequence original positioning point sequence are obtained, and the direction sensor is used to obtain the travel direction corresponding to each original positioning point; then, an initial candidate road section is determined for each original positioning point, the effective candidate road section is screened through direction consistency and Bluetooth area constraints, and the spatial matching probability is calculated based on a graph neural network model; then, a time sequence feature sequence is input into a time sequence neural network model to obtain a time sequence consistency score; finally, the target road section is determined by fusing the scores, the original positioning point is rectified to the target road section based on the geometric information of the target road section, and the rectified position is obtained. The application improves the accuracy and stability of positioning and deviation rectification in a scenic area.

Description

Technical Field

[0001] This invention relates to the field of mobile terminal positioning technology, and more specifically to a mobile terminal positioning correction method based on spatiotemporal constraints and orientation awareness. Background Technology

[0002] In fields such as intelligent navigation, location services, and tourist behavior analysis, accurate mobile terminal positioning is a core technological foundation. Currently, positioning technology based on the Global Navigation Satellite System has achieved good results in open environments, and related auxiliary enhancement technologies mainly focus on optimizing urban road networks with regular structures and distinct road features. These technologies assume that the positioning target typically moves continuously along a road network with clear geometric features, and correct the original positioning point onto the road through geometric projection, motion model prediction, and other means.

[0003] However, the existing technologies based on urban road models face significant limitations in complex environments such as scenic spots, parks, and natural scenic areas. First, the dense, winding, and closely spaced trails within scenic areas create an irregular topology that makes traditional map matching algorithms, which rely on the "shortest path" principle, prone to mismatches, attaching the location point to incorrect adjacent trails. Second, the dense forests, mountains, and canyons common in scenic areas severely obstruct and reflect satellite signals, resulting in significant shifts, jitter, and jumps in the original positioning data. Filtering methods based on statistical noise models struggle to converge effectively under such large observation errors. Finally, existing methods generally lack deep integration and utilization of the scenic area's road network topology, tourist travel directions, and regional anchor point information such as Bluetooth. This makes them unable to handle uncertain behaviors such as frequent stops and turns by tourists, resulting in poor trajectory smoothness and insufficient continuity, failing to meet the high-precision requirements of applications such as intelligent navigation in scenic areas. Summary of the Invention

[0004] This invention provides a mobile terminal positioning correction method based on spatiotemporal constraints and orientation awareness, which solves the problems of mismatched positioning trajectories, severe jitter and poor continuity caused by the difficulty of adapting to the complex environment of scenic spots due to the design of urban road models.

[0005] To achieve the above objectives, this invention provides a mobile terminal positioning correction method based on spatiotemporal constraints and orientation awareness, comprising the following steps: Scenic area road network map construction step: constructing a scenic area road network map based on electronic map data of the target area; Positioning data and sensor data acquisition step: acquiring signal data from Bluetooth anchor points and the original positioning point sequence of the mobile terminal in continuous time sequence, and acquiring the travel direction corresponding to each of the original positioning points through the orientation sensor of the mobile terminal; Candidate road segment matching probability calculation step: for each original positioning point, determining initial candidate road segments from the scenic area road network map; based on the consistency of the travel direction with the direction of the initial candidate road segment and the... The Bluetooth anchor signal indicates the area constraint, and effective candidate road segments are selected from the initial candidate road segments. The spatial matching probability of each original positioning point matching each effective candidate road segment is calculated based on the graph neural network model. The temporal consistency score calculation step is as follows: the temporal feature sequence of the original positioning point is input into the temporal neural network model to obtain the temporal consistency score corresponding to each original positioning point. The correction position determination step is as follows: for each original positioning point, the spatial matching probability and the temporal consistency score are fused to determine the target road segment from its effective candidate road segments, and the original positioning point is corrected to the target road segment based on the geometric information of the target road segment to obtain the corrected positioning position.

[0006] Optionally, the scenic area road network map construction step includes: based on the electronic map data, taking the scenic spots, path intersections, viewing platforms and entrance facilities in the scenic area as graph nodes, and taking the walkways connecting the graph nodes as graph edges to form a scenic area road network map; and assigning semantic labels to the graph nodes to indicate the type of the nodes.

[0007] Optionally, the process of determining the initial candidate road segments includes: retrieving the scenic area road network map of the target area based on the spatial index, representing the center line of each road segment in the scenic area road network map as a line segment; calculating the projection distance from the original positioning point to the center line of each road segment, and determining the road segments with a projection distance less than a preset distance threshold as the initial candidate road segments.

[0008] Optionally, the calculation process of the spatial matching probability includes: constructing a bipartite graph with each original location point as a location point node and each effective candidate road segment as a road segment node; performing multiple rounds of iterative processing on the bipartite graph through a graph neural network model, each round of iteration including neighborhood feature aggregation and node and edge feature updates; wherein, during neighborhood feature aggregation, the temporally adjacent location point node features of the location point node are aggregated, and the spatially adjacent road segment node features of the road segment node are aggregated; the edge features after iterative processing are processed through the prediction layer of the graph neural network model, and after normalization, the spatial matching probability of each original location point matching each effective candidate road segment is output.

[0009] Optionally, the calculation process of the temporal consistency score includes: extracting features from the original location point sequence in a continuous time series, constructing a temporal feature vector for each original location point; arranging the temporal feature vectors of each original location point in chronological order to form a temporal feature sequence corresponding to the original location point sequence; inputting the temporal feature sequence into a temporal neural network model in chronological order, wherein the temporal neural network model captures the motion pattern of the mobile terminal based on the historical temporal information and current temporal features of the temporal feature sequence; and, based on the correlation between the motion pattern and the temporal sequence, quantitatively evaluating the temporal rationality of the current original location point through the output layer of the temporal neural network model, and outputting the corresponding temporal consistency score.

[0010] Optionally, the process of determining the target road segment includes: calculating the geometric projection distance from the original positioning point to each valid candidate road segment; calculating the comprehensive score of each valid candidate road segment based on the spatial matching probability, temporal consistency score, geometric projection distance, and area constraints indicated by the Bluetooth anchor signal for each valid candidate road segment corresponding to each original positioning point; and determining the valid candidate road segment with the highest comprehensive score as the final matched target road segment.

[0011] Optionally, the step of determining the correction location further includes: assigning different path priority weights to different types of graph nodes; and using the path priority weight of the corresponding graph node type as a weighting factor when calculating the comprehensive score of each valid candidate road segment.

[0012] Optionally, the process of obtaining the corrected positioning position includes: obtaining the geometric information of the centerline of the target road segment, performing vertical projection processing on the original positioning point, and determining the position that falls on the centerline of the target road segment after projection as the corrected positioning position of the original positioning point.

[0013] Optionally, the process for determining directional consistency includes: calculating the angle between the travel direction vector and the candidate road segment direction vector; determining whether the angle is less than or equal to a preset angle threshold; if so, determining that the travel direction is consistent with the candidate road segment direction.

[0014] This invention provides a mobile terminal positioning correction method based on spatiotemporal constraints and orientation awareness. By introducing scenic area road network topology constraints, terminal orientation angle perception, and Bluetooth anchor point observation information, and based on the collaborative inference of graph neural networks and temporal neural networks, it achieves intelligent correction of the original positioning data of mobile terminals. Its technical advantages lie in significantly improving the accuracy and robustness of positioning trajectories in complex environments such as scenic areas: In the spatial dimension, graph neural networks are used to model the global topology of the scenic area road network, ensuring a high degree of fit between the corrected trajectory and the actual trail network, effectively avoiding mismatch problems caused by dense road networks and irregular structures; in the temporal dimension, temporal neural networks are used to learn the continuous behavioral patterns of user movement, effectively suppressing abnormal jitter and jumps in positioning points, ensuring the smoothness and temporal continuity of the output trajectory. Finally, this method overcomes the shortcomings of traditional technologies, such as dependence on urban road models and poor adaptability in weak signal areas, providing a stable and reliable location data foundation for applications such as scenic area navigation and guide triggering. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0016] Figure 1 This is a general flowchart of the mobile terminal positioning correction method provided in the embodiments of the present invention;

[0017] Figure 2 This is a flowchart of the scenic area road network map construction process provided in an embodiment of the present invention;

[0018] Figure 3 This is a flowchart of spatial matching probability calculation provided in an embodiment of the present invention;

[0019] Figure 4 This is a flowchart of the target road segment determination and correction location acquisition provided in the embodiments of the present invention. Detailed Implementation

[0020] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.

[0021] It should be noted that the acquisition, transmission, storage, use, and processing of data in the technical solution of this application all comply with the relevant provisions of national laws and regulations. In the embodiments of this application, certain existing industry solutions such as software, components, and models may be mentioned. These should be considered exemplary, intended only to illustrate the feasibility of implementing the technical solution of this application, and do not imply that the applicant has already used or necessarily used such solutions.

[0022] As the cultural tourism industry continuously improves its focus on visitor experience and digital management of scenic areas, existing mobile terminal positioning technologies suffer from several problems. These include poor accuracy and instability of GPS positioning in complex scenic environments, high mismatch rates due to dense, irregular trails caused by traditional map matching algorithms, insufficient adaptability of filtering-based trajectory smoothing methods to uncertain visitor movement patterns, and a lack of mechanisms for deep fusion of road network topology and multi-source sensing information to address instantaneous signal failures. Therefore, developing a positioning correction scheme that can deeply understand the spatial structure of scenic areas, integrate multi-source dynamic information, and possess spatiotemporal joint reasoning capabilities is crucial.

[0023] To address this, this invention proposes a mobile terminal positioning correction method based on spatiotemporal constraints and orientation awareness. It constructs a scenic area-specific road network map integrating semantic information such as attractions and intersections as a spatial reference. Utilizing a dual screening mechanism of directional consistency and Bluetooth area constraints, it quickly locates valid candidate road segments from geometrically adjacent road segments. A graph neural network is then used to infer spatial topological relationships between the positioning point and road segment bipartite graph to generate matching probabilities. Furthermore, a temporal neural network is introduced to model the user's continuous movement behavior to evaluate trajectory rationality. Finally, an optimal road segment matching and accurate projection correction are achieved through a weighted scoring model that integrates spatial matching probability, temporal consistency score, and geometric distance. This forms a closed-loop processing flow from road network modeling, multi-source perception, intelligent matching to decision correction, significantly improving the accuracy, smoothness, and robustness to weak signals and anomalies in complex scenic environments.

[0024] The following is combined Figures 1-4 This invention is described in detail.

[0025] like Figure 1As shown, this embodiment of the invention provides a mobile terminal positioning correction method based on spatiotemporal constraints and orientation awareness, including the following steps: Scenic area road network map construction step: constructing a scenic area road network map based on electronic map data of the target area; Positioning data and sensor data acquisition step: acquiring signal data from Bluetooth anchor points and the original positioning point sequence of the mobile terminal in continuous time sequence, and acquiring the travel direction corresponding to each of the original positioning points through the orientation sensor of the mobile terminal; Candidate road segment matching probability calculation step: for each original positioning point, determining initial candidate road segments from the scenic area road network map; based on the consistency of the travel direction with the direction of the initial candidate road segment and the Bluetooth anchor point... The system uses the area constraints indicated by the point signal to select effective candidate road segments from the initial candidate road segments; it calculates the spatial matching probability of each original positioning point matching each effective candidate road segment based on a graph neural network model; the temporal consistency score calculation step involves inputting the temporal feature sequence of the original positioning points into the temporal neural network model to obtain the temporal consistency score for each original positioning point; the correction position determination step involves fusing the spatial matching probability and the temporal consistency score for each original positioning point to determine the target road segment from its effective candidate road segments, and correcting the original positioning point to the target road segment based on the geometric information of the target road segment to obtain the corrected positioning position.

[0026] The scenic area road network map refers to a structured graph model constructed based on electronic map data of the target area. It represents the spatial relationships between passable paths and key facilities within the scenic area, consisting of graph nodes, graph edges, and semantic labels. Graph nodes include important facilities such as scenic spots, path intersections, viewing platforms, and entrances; graph edges connect the nodes to scenic trails and passable road segments. Bluetooth anchor point signal data refers to the wireless signal data emitted by Bluetooth devices deployed at fixed locations within the scenic area. Its core components include signal strength, signal presence, and area number, providing regional spatial constraints. The original positioning point sequence refers to a continuous temporal location data set obtained by the mobile terminal through a satellite positioning system. Each positioning point contains coordinates and a timestamp. The graph neural network model, built based on a graph convolutional neural network, is used to model the spatial topological relationship between the original positioning points and the scenic area road network, as well as the "positioning point-road segment" matching rules. Its core function is to output the spatial matching probability through neighborhood feature aggregation and iterative updates. The temporal feature sequence refers to a sequence formed by arranging the core features of continuous original positioning points in chronological order. Core features include the positioning coordinates, instantaneous speed, and direction of travel of the positioning points. Temporal neural network models refer to models built on long short-term memory networks, used to capture the temporal motion patterns of mobile terminals and output temporal consistency scores to constrain trajectory continuity. The geometric information of the target road segment refers to the geometric parameters of the centerline of the valid candidate road segments, specifically the coordinates of the endpoints of the line segments representing the centerline, and the direction vector, etc.

[0027] The mobile terminal positioning correction method provided in this invention first constructs a scenic road network map adapted to the walking scenario of the scenic area using electronic map data of the target area as a spatial constraint basis. Then, it simultaneously acquires Bluetooth anchor point signal data, the continuous temporal sequence of original positioning points of the mobile terminal, and the travel direction corresponding to each original positioning point. Subsequently, for each original positioning point, it first determines initial candidate road segments from the scenic road network map and filters effective candidate road segments by combining travel direction consistency and Bluetooth anchor point area constraints. Then, it calculates the spatial matching probability between the original positioning point and each effective candidate road segment using a graph neural network model. Simultaneously, it inputs the temporal feature sequence of the original positioning point into a temporal neural network model to obtain a temporal consistency score. Finally, it fuses the spatial matching probability and the temporal consistency score to determine the target road segment and corrects the original positioning point to the road segment based on the geometric information of the target road segment. This process utilizes "road network spatial constraints + orientation perception + dual neural network temporal-" The innovative architecture of "spatial collaborative assessment" effectively solves the problems of mismatched positioning trajectories, severe jitter and poor continuity caused by the design of existing technologies for urban road models, which are difficult to adapt to the dense trails, terrain obstruction and complex road network topology in scenic areas. It significantly improves the accuracy, stability and trajectory continuity of mobile terminal positioning in scenic areas.

[0028] like Figure 2 As shown, preferably, the scenic area road network map construction step includes: based on the electronic map data, taking the scenic spots, path intersections, viewing platforms and entrance facilities in the scenic area as graph nodes, and taking the walkways connecting the graph nodes as graph edges to form a scenic area road network map; assigning semantic labels to the graph nodes to indicate the type of the nodes.

[0029] Semantic labels refer to the label information assigned to each graph node to uniquely identify its functional type. The labels strictly correspond to the actual use of the graph node, such as "scenic spot label", "intersection label", "viewing platform label", and "entrance label", with no ambiguous or duplicate labels.

[0030] In a preferred embodiment of the present invention, a road network map is constructed by using the core tourist facilities in the scenic area as graph nodes and walkable trails as graph edges, and semantic labels are assigned to the nodes. This accurately adapts to the walking scenario in the scenic area, provides precise spatial constraints for subsequent candidate route selection, supports the setting of path priority weights, and reduces the computational complexity of subsequent retrieval, laying the foundation for the accuracy and stability of overall positioning correction.

[0031] Preferably, the process of determining the initial candidate road segments includes: retrieving the scenic area road network map of the target area based on the spatial index, representing the centerline of each road segment in the scenic area road network map as a line segment; calculating the projection distance from the original positioning point to the centerline of each road segment, and determining the road segments with a projection distance less than a preset distance threshold as the initial candidate road segments.

[0032] Spatial indexing refers to an efficient spatial retrieval structure, such as an R-tree index, adapted to the scenic area's road network map. This allows for rapid location of road segments surrounding the original location point, avoiding a full traversal of the road network and improving retrieval efficiency. The centerline of a road segment is a virtual line segment representing the core passageway of the scenic area's trails. Its direction aligns with the actual extension direction of the trail, and the coordinates of its endpoints are extracted from the geometric center of the trail outline on the electronic map. Vertical projection refers to the geometric operation of mapping the original location point along its perpendicular direction to the centerline of the road segment onto the line containing the centerline. The projection result must satisfy the condition that "the foot of the perpendicular lies on the centerline segment or its extension." Projection distance refers to the straight-line distance between the original location point and its foot on the centerline of the road segment; it is a core indicator for measuring the spatial correlation between the location point and the road segment. Projection parameters include the foot coordinates and the positional proportion of the projected point on the line segment. The distance threshold is set based on the average width of the scenic area's trails.

[0033] Specifically, the formula for calculating the projection distance is:

[0034]

[0035] in, The original coordinates of the positioning point. The equation of the straight line for the centerline of the road segment is... This is the vertical projection distance from the original positioning point to the center line. The formula for calculating the perpendicular coordinates is:

[0036]

[0037] in, Let the coordinates be the foot of the perpendicular. The original coordinates of the positioning point. Let be the equation of the straight line representing the centerline of the target road segment.

[0038] For example, a mobile terminal obtains the original location coordinates as (118.1234°, 30.5678°). Based on the R-tree spatial index, it retrieves the scenic area road network map and represents the center line of a certain trail in the map as a line segment with a starting point (118.1230°, 30.5670°) and an ending point (118.1240°, 30.5680°). Through the planar geometric projection algorithm, the perpendicular foot coordinates of the original location point to the center line are calculated to be (118.1233°, 30.5676°), and the projection distance is 8 meters. The distance threshold is 15 meters. Since 8 meters is less than 15 meters, the trail is determined as the initial candidate road segment for the original location point.

[0039] The preferred embodiment of the present invention uses spatial indexing for rapid retrieval, centerline segmentation, and projection distance threshold filtering to accurately locate the initial candidate road segments of the original positioning point. This not only improves the efficiency of road segment retrieval but also avoids interference from irrelevant road segments, laying a solid foundation for the subsequent selection of effective candidate road segments.

[0040] like Figure 3 As shown, preferably, the calculation process of the spatial matching probability includes: constructing a bipartite graph with each original location point as a location point node and each effective candidate road segment as a road segment node; performing multiple rounds of iterative processing on the bipartite graph through a graph neural network model, each round of iteration including neighborhood feature aggregation and node and edge feature updates; wherein, during neighborhood feature aggregation, the temporally adjacent location point node features of the location point node are aggregated, and the spatially adjacent road segment node features of the road segment node are aggregated; the edge features after iterative processing are processed through the prediction layer of the graph neural network model, and after normalization, the spatial matching probability of each original location point matching each effective candidate road segment is output.

[0041] The bipartite graph is a bipartite undirected graph consisting of a set of local point nodes and a set of valid candidate road segment nodes. Neighborhood feature aggregation refers to the fusion operation of node-related features during GNN iterations. Local point nodes aggregate features from their temporally neighboring local point nodes, and road segment nodes aggregate features from their spatially adjacent road segment nodes. Node and edge feature updates refer to updating the embedding vectors of local point nodes, road segment nodes, and edges between nodes based on the aggregated neighborhood features through the GNN's message passing mechanism, resulting in more accurate feature representation after each iteration. The prediction layer is the last fully connected layer of the GNN, used to map the iteratively updated edge features to "local point-road segment" matching scores, with the output dimension consistent with the number of valid candidate road segments. Normalization involves performing a softmax operation on the matching scores output by the prediction layer, converting the scores into probability values ​​in the 0-1 range, ensuring that the sum of the matching probabilities of all valid candidate road segments is 1.

[0042] Specifically, the bipartite graph construction rules are as follows: each original location point node only establishes connections with its own valid candidate road segment nodes; there are no connections between road segment nodes of different location points, avoiding invalid associations across location points. The default number of GNN iterations is set to 3-5 rounds. Too few iterations will result in insufficient feature aggregation, while too many iterations will easily lead to overfitting. The number of iterations can be adaptively adjusted according to the complexity of the scenic area's road network. Feature aggregation uses weighted average aggregation, with the weights positively correlated with the spatiotemporal correlation between nodes.

[0043] Specifically, the formula for fusing initial values ​​of edge features is as follows:

[0044]

[0045] in, This is the projected distance between the original location point and the centerline of the road segment. Let cosine be the angle between the direction of travel and the direction of the road segment. =0.6、 =0.4 is the feature fusion weight. The formula for temporal neighborhood aggregation of localization point nodes is:

[0046]

[0047] in, Let i be the feature vector of the localized point node i after the t-th iteration. , Features of temporally adjacent location points This is the weight matrix. For bias terms, The ReLU activation function is used. The formula for spatial neighborhood aggregation of road segment nodes is:

[0048]

[0049] in, Let j be the feature vector of road segment node j after the t-th iteration. Let j be the set of spatially adjacent road segments. This is the weight matrix. This is a bias term.

[0050] For example, taking an original location point P5 as the location point node and its effective candidate road segments L2, L5, and L8 as road segment nodes, a bipartite graph is constructed. The GNN model is used for three rounds of iterative processing: In the first round of iteration, the location point node P5 aggregates the features of its temporally neighboring location points P4 and P6, and the road segment node L2 aggregates the features of its spatially adjacent road segments L1 and L3, and the node embedding vector and edge features are updated simultaneously. After three rounds of iteration, the model prediction layer processes the final edge features to obtain matching scores of P5 with L2, L5, and L8 of 1.2, 3.5, and 0.8, respectively. After softmax normalization, the spatial matching probabilities of P5 matching L2, L5, and L8 are output as 0.15, 0.70, and 0.15, respectively.

[0051] In a preferred embodiment of the present invention, a bipartite graph is constructed to model the association between the location point and the effective candidate road segment. The GNN is used to aggregate spatiotemporal neighborhood features and update node and edge features through multiple rounds of iteration. Finally, the accurate spatial matching probability is output after normalization. This effectively captures the spatiotemporal association pattern between the location point and the road segment, greatly improves the accuracy of the matching result, and provides a reliable spatial dimension basis for the subsequent determination of the target road segment.

[0052] Preferably, the calculation process of the temporal consistency score includes: extracting features from the original location point sequence in a continuous time series, constructing a temporal feature vector for each original location point; arranging the temporal feature vectors of each original location point in chronological order to form a temporal feature sequence corresponding to the original location point sequence; inputting the temporal feature sequence into a temporal neural network model in chronological order, wherein the temporal neural network model captures the motion pattern of the mobile terminal based on the historical temporal information and current temporal features of the temporal feature sequence; and, based on the correlation between the motion pattern and the temporal sequence, quantitatively evaluating the temporal rationality of the current original location point through the output layer of the temporal neural network model, and outputting the corresponding temporal consistency score.

[0053] Among them, the temporal feature vector refers to the multi-dimensional feature vector constructed for a single original positioning point. Its dimensions include core temporal-related features such as positioning point coordinates, instantaneous velocity, direction of travel, and positioning accuracy, serving as the basic unit characterizing the temporal attributes of the positioning point. Motion pattern refers to the typical behavioral characteristics of the mobile terminal in continuous time, such as constant-speed walking, standing still, turning, and variable-speed movement, which are learned and identified by a temporal neural network model from historical and current features. Temporal rationality refers to the degree of matching between the features of the current original positioning point and the historical trajectory motion pattern. A higher matching degree indicates stronger temporal rationality, with a score closer to 1; conversely, a lower matching degree indicates weaker rationality, with a score closer to 0.

[0054] Specifically, the temporal neural network model employs a long short-term memory network, utilizing its gating mechanism to capture long-term temporal dependencies, effectively solving the gradient vanishing problem of ordinary recurrent neural networks. The output layer of the temporal neural network model maps the evaluation results to the 0-1 range through the sigmoid activation function; a higher value indicates a stronger coherence between the current location point and the historical trajectory, and better temporal consistency.

[0055] The formula for time series feature vectors is:

[0056]

[0057] in, The coordinates of the original positioning point i are... instantaneous velocity , The direction angle is the direction of travel. For positioning accuracy, () is the normalization function.

[0058] In a preferred embodiment of the present invention, the temporal motion pattern of the mobile terminal is captured by an LSTM model, and an accurate temporal consistency score is output. This effectively constrains the temporal continuity of the positioning trajectory, avoids trajectory jumps caused by single-point positioning deviations, and provides a reliable temporal dimension basis for determining the target road segment.

[0059] like Figure 4 As shown, preferably, the process of determining the target road segment includes: calculating the geometric projection distance from the original positioning point to each effective candidate road segment; calculating the comprehensive score value of each effective candidate road segment based on the spatial matching probability, temporal consistency score, geometric projection distance, and area constraints indicated by the Bluetooth anchor signal for each effective candidate road segment corresponding to each original positioning point; and determining the effective candidate road segment with the highest comprehensive score value as the final matched target road segment.

[0060] More preferably, the step of determining the correction location further includes: assigning different path priority weights to different types of graph nodes; and using the path priority weight of the corresponding graph node type as a weighting factor when calculating the comprehensive score of each valid candidate road segment.

[0061] The comprehensive score is a quantitative score calculated based on four core factors: spatial matching probability, temporal consistency score, geometric projection distance, and Bluetooth anchor point area constraint, weighted by a path priority weight. This score is the sole criterion for selecting target road segments; a higher score indicates a stronger match between the road segment and the location point. The weighting factor, or path priority weight, plays a role in the comprehensive score calculation. By multiplying it with the core matching factors, it amplifies the matching weight of high-priority paths, making the score result more closely reflect the actual tourist route. The area constraint indicated by the Bluetooth anchor point signal refers to determining the physical area of ​​the scenic area to which the original location point belongs based on the Bluetooth anchor point's area number and signal strength, forming a binary constraint of "whether the road segment is located within this area," used to filter out invalid candidate road segments outside the designated area.

[0062] Specifically, the comprehensive score adopts a weighted summation model, with the core formula logic as follows: Comprehensive Score = (Spatial Matching Probability × 0.3 + Temporal Consistency Score × 0.25 + Bluetooth Regional Constraint Adaptation Value × 0.2) × Path Priority Weight + (1 - Normalized Geometric Projection Distance) × 0.25. The path priority weight is set based on the semantic labels of the nodes in the association graph. For example: entrance node / scenic spot node associated road segment weight = 1.5, intersection node associated road segment weight = 1.2, viewing platform / rest area node associated road segment weight = 1.0, ordinary road segment weight = 0.8. The weight values ​​are determined based on the popularity statistics of scenic area tour paths. Bluetooth regional constraint adaptation rules: If the spatial range of a valid candidate road segment is consistent with the area of ​​the positioning point indicated by the Bluetooth anchor point, the adaptation value is 1; if there is partial overlap, it is 0.5; if there is no overlap, it is 0, directly reducing the comprehensive score of the road segment to achieve regional-level screening. The comprehensive score formula is:

[0063]

[0064] in, Let j be the comprehensive score of the j-th valid candidate road segment. For spatial matching probability, For time series consistency scoring, The Bluetooth region constraint adaptation value is (0 / 0.5 / 1). For path priority weights, The geometric projection distance. For fixed weighting coefficients, This is the normalized result of the projected distance.

[0065] The preferred embodiment of this invention calculates a comprehensive score by integrating multiple factors and superimposing path priority weights. This approach takes into account the accuracy of spatial matching, the continuity of time sequence, and the effectiveness of regional constraints, while also conforming to the visiting patterns of scenic spots. It effectively avoids mismatches caused by single-factor judgments, significantly improves the accuracy and rationality of target road segment selection, and provides core decision-making basis that conforms to the scenic spot scenario for final positioning correction.

[0066] Preferably, the process of obtaining the corrected positioning position includes: obtaining the geometric information of the centerline of the target road segment, performing vertical projection processing on the original positioning point, and determining the position that falls on the centerline of the target road segment after projection as the corrected positioning position of the original positioning point.

[0067] Among them, the centerline geometric information of the target road segment specifically refers to the core geometric parameters of the target road segment, including the starting coordinates, ending coordinates, and equation of the centerline segment. Vertical projection processing refers to the geometric operation based on a planar analytical geometric algorithm, mapping the original positioning point along its perpendicular direction to the centerline of the target road segment to the line containing the centerline. The projection process strictly follows the "shortest perpendicular line" principle, ensuring that the projected point is the closest point from the positioning point to the centerline.

[0068] A preferred embodiment of the present invention provides a method to perform precise vertical projection by reusing the geometric information of the centerline of the target road segment, thereby correcting the offset and jittery original positioning point to the centerline of the trail. This not only effectively eliminates positioning deviations but also ensures that the correction position conforms to the actual traffic area, ultimately achieving accurate and stable positioning in scenic areas.

[0069] Preferably, the process for determining directional consistency includes: calculating the angle between the travel direction vector and the candidate road segment direction vector; determining whether the angle is less than or equal to a preset angle threshold; if so, determining that the travel direction is consistent with the candidate road segment direction.

[0070] Here, the travel direction vector refers to the azimuth angle collected by the orientation sensor of the mobile terminal, such as 0° for true north and increasing clockwise, resulting in a two-dimensional unit vector. It accurately represents the actual direction and trend of the user's movement. This refers to the azimuth angle acquired by the sensor. The candidate road segment direction vector refers to the starting point based on the centerline of the candidate road segment. and the finish line Unit vector for coordinate calculation , This is the azimuth angle extending from the starting point to the end point of the centerline, with the vector direction consistent with the actual travel direction of the trail. The preset angle threshold is adapted to the direction matching threshold set for scenic walking scenarios. The default value is 30°. For straight trails and open areas in scenic areas, it can be lowered to 20°, and for curves and intersections, it can be raised to 45° to avoid mis-selection of valid road segments due to normal turning.

[0071] Specifically, the angle is calculated using the dot product formula to find the minimum angle between two vectors. The formula is:

[0072]

[0073] Among them, Let be the angle between the two vectors. Since both are unit vectors, it simplifies to... This ensures efficient and accurate calculations.

[0074] In a preferred embodiment of the present invention, the angle between the travel direction and the candidate road segment direction is accurately calculated and combined with a preset angle threshold to determine consistency. This effectively filters out invalid road segments with mismatched directions, reduces interference in subsequent matching, provides reliable directional constraints for determining valid candidate road segments, further improves the accuracy of target road segment matching, and conforms to the actual travel path patterns of tourists.

[0075] In summary, the mobile terminal positioning correction method based on spatiotemporal constraints and orientation awareness provided by this invention lays a precise spatial constraint foundation by constructing a road network map adapted to the walking scenario in scenic areas. Initial candidate road segments are screened using spatial indexing and projection thresholds, and effective candidate road segments are screened by determining the directional angle. The spatial matching probability is output by the GNN model, and the temporal consistency score is output by the LSTM model. The target road segment is determined by superimposing path priority weights and multi-factor fusion calculations. Finally, positioning correction is completed through vertical projection. The entire process forms a technical closed loop of "spatiotemporal constraints - multi-dimensional screening - spatiotemporal collaborative evaluation - weighted matching - precise correction," effectively solving the problems of mismatched positioning trajectories, severe jitter, and poor continuity caused by the difficulty of adapting to the dense walking paths, terrain occlusion, and complex road network topology in scenic areas. This significantly improves the accuracy, stability, and matching rationality of mobile terminal positioning in scenic areas, aligning with the actual travel patterns of tourists.

[0076] The above description is merely a preferred embodiment of the technical solution of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for positioning and rectifying a mobile terminal based on space-time constraints and orientation perception, characterized in that, Includes the following steps: Steps for constructing a scenic area road network map: Construct a scenic area road network map based on electronic map data of the target area; Location data and sensor data acquisition steps: acquire the signal data of the Bluetooth anchor point and the original location point sequence of the mobile terminal in a continuous time sequence, and acquire the travel direction corresponding to each of the original location points through the orientation sensor of the mobile terminal; Candidate road segment matching probability calculation steps: For each original positioning point, determine initial candidate road segments from the scenic area road network map; based on the directional consistency between the travel direction and the direction of the initial candidate road segments and the regional constraints indicated by the Bluetooth anchor point signal, select effective candidate road segments from the initial candidate road segments; calculate the spatial matching probability of each original positioning point matching each effective candidate road segment based on the graph neural network model; The steps for calculating the temporal consistency score are as follows: Input the temporal feature sequence of the original localization points into the temporal neural network model to obtain the temporal consistency score for each original localization point; Correction location determination steps: For each original positioning point, the spatial matching probability and the temporal consistency score are fused to determine the target road segment from its valid candidate road segments, and based on the geometric information of the target road segment, the original positioning point is corrected to the target road segment to obtain the corrected positioning location.

2. The mobile terminal positioning and deviation correction method of claim 1, wherein, The steps for constructing the scenic area road network map include: Based on the electronic map data, scenic spots, path intersections, viewing platforms, and entrance facilities within the scenic area are designated as graph nodes, and the walkways connecting these graph nodes are designated as graph edges to form a scenic area road network map. Assign semantic labels to the graph nodes to indicate the type of the nodes.

3. The mobile terminal positioning and deviation correction method of claim 1, wherein, The process of determining the initial candidate road segments includes: retrieving the scenic area road network map based on the spatial index, representing the centerline of each road segment in the scenic area road network map as a line segment; calculating the projection distance from the original positioning point to the centerline of each road segment, and determining the road segments with a projection distance less than a preset distance threshold as the initial candidate road segments.

4. The mobile terminal positioning and deviation correction method of claim 1, wherein, The calculation process for the spatial matching probability includes: A bipartite graph is constructed using each original location point as a location point node and each valid candidate road segment as a road segment node. The bipartite graph is processed through multiple iterations using a graph neural network model. Each iteration includes neighborhood feature aggregation and node and edge feature updates. During neighborhood feature aggregation, the temporal neighboring features of the location point node are aggregated, and the spatial adjacent road segment node features of the road segment node are also aggregated. The prediction layer of the graph neural network model processes the edge features after iterative processing, and after normalization, outputs the spatial matching probability of each original location point matching each effective candidate road segment.

5. The mobile terminal positioning and deviation correction method of claim 1, wherein, The calculation process for the time series consistency score includes: Feature extraction is performed on the original location point sequence in a continuous time series, and a time series feature vector is constructed for each original location point; Arrange the temporal feature vectors of each original location point in chronological order to form a temporal feature sequence corresponding to the original location point sequence; The time-series feature sequence is input into the time-series neural network model in chronological order. The time-series neural network model captures the motion pattern of the mobile terminal based on the historical time-series information and current time-series features of the time-series feature sequence. Based on the motion pattern and temporal correlation, the temporal rationality of the current original positioning point is quantitatively evaluated through the output layer of the temporal neural network model, and the corresponding temporal consistency score is output.

6. The mobile terminal positioning and deviation correction method of claim 1, wherein, The process of determining the target road segment includes: Calculate the geometric projection distance from the original location point to each valid candidate road segment; Based on the spatial matching probability, temporal consistency score, geometric projection distance, and area constraints indicated by the Bluetooth anchor signal for each valid candidate road segment corresponding to each original positioning point, a comprehensive score value for each valid candidate road segment is calculated. The valid candidate road segment with the highest comprehensive score is determined as the final target road segment for matching.

7. The mobile terminal positioning and deviation correction method of claim 6, wherein, The step of determining the correction location further includes: assigning different path priority weights to different types of graph nodes; and using the path priority weight of the corresponding graph node type as a weighting factor when calculating the comprehensive score of each valid candidate road segment.

8. The mobile terminal positioning and deviation correction method of claim 1, wherein, The process of obtaining the corrected positioning position includes: Obtain the geometric information of the centerline of the target road segment, perform vertical projection processing on the original positioning point, and determine the position of the projected point on the centerline of the target road segment as the corrected positioning position of the original positioning point.

9. The mobile terminal positioning and deviation correction method of claim 1, wherein, The process for determining directional consistency includes: calculating the angle between the travel direction vector and the candidate road segment direction vector; determining whether the angle is less than or equal to a preset angle threshold; if so, determining that the travel direction is consistent with the candidate road segment direction.

Images