A navigation optimization-oriented space information visualization labeling method and system

Abstract

The application provides a navigation-optimization-oriented space information visualization labeling method and system, and relates to the technical field of navigation interaction. The method comprises the following steps: acquiring indoor map data and terminal start and end point positions and determining a navigation path; identifying a next scene switching node along a travel direction and constructing a corresponding decision corridor in combination with a current position; extracting candidate labeling objects intersecting the decision corridor or semantically associated with the scene switching node from the map data; determining respective decision contribution degrees of the candidate labeling objects based on their disambiguation effects on the traffic decisions at the scene switching node; mapping the candidate labeling objects to a display area corresponding to a current view, determining target labeling objects and display modes thereof based on the decision contribution degrees and the projection conflict relationships after the mapping, and outputting a visualization result. The application significantly improves the relevance of the displayed labeling information and the traffic decisions and effectively reduces visual interference caused by irrelevant labeling.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and more specifically, to a spatial information visualization annotation method and system optimized for navigation. Background Technology

[0002] With the development of mobile terminals and spatial computing technology, indoor spatial visualization navigation has been widely applied. In existing indoor navigation systems, such as 3D or augmented reality navigation, spatial annotations such as points of interest are usually overlaid on the screen view to help users identify their location. Existing technologies mostly use a fixed search radius based on the current physical coordinates to extract surrounding information and display it according to distance or static category priority. However, indoor environments are often characterized by dense road networks, severe obstruction, and complex spatial topology. In such environments, physically close annotations may not necessarily have practical navigational guidance value. Conversely, when users are at critical stages requiring decision-making, such as turning or crossing floors, a large number of invalid annotations piled up on the limited screen can easily lead to information overload, severely obstructing the view of the real environment or key landmarks.

[0003] Therefore, in the process of indoor navigation, how to improve the relevance of the displayed spatial labeling information to the current stage of the navigation decision within a limited display area, thereby reducing the interference of irrelevant labels on key guidance information, is a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this application provides a spatial information visualization annotation method and system optimized for navigation.

[0005] Firstly, this application provides a spatial information visualization annotation method for navigation optimization, including: Acquire indoor space map data, as well as the current location and target location of the navigation terminal; A navigation path is determined based on the current location and the target location, and the next scene switching node is identified along the direction of travel of the navigation path; a decision corridor corresponding to the current location is constructed based on the current location and the next scene switching node. Candidate annotation objects that intersect with the decision corridor or are semantically associated with the next scene switching node are extracted from the indoor space map data; for each candidate annotation object, the decision contribution degree corresponding to each candidate annotation object is determined based on its disambiguation effect on the passage decision at the next scene switching node; Each candidate labeled object is mapped to the display area corresponding to the current view. Based on the decision contribution and the projection conflict relationship after mapping, the target labeled object and the display method of the target labeled object are determined. The visualization labeling result of the target labeled object is output.

[0006] Optionally, constructing the decision corridor corresponding to the current location includes: A forward main guide strip is generated along the path centerline between the current position and the next scene switching node; At each sampling position of the forward main guide strip, the forward main guide strip is laterally expanded according to the passage boundary and the entrance of the accessible branch road at the corresponding position; The extended continuous spatial region is defined as the decision corridor.

[0007] Optionally, determining the decision contribution of each of the candidate labeled objects includes: Determine the first topological distance from each candidate labeled object to the target access branch, and the second topological distance to at least one other accessible access branch; Based on the difference between the first topological distance and the second topological distance, the degree of disambiguation of each candidate labeled object for the passage decision is determined; Based on the degree of disambiguation, the decision contribution of each candidate labeled object is determined.

[0008] Optionally, determining the target annotation object and the display method of the target annotation object includes: Determine the projection area of ​​each candidate annotation object after it is mapped to the current view; In response to the projection regions of at least two candidate annotation objects satisfying a preset conflict condition, the candidate annotation object with higher decision contribution is retained as the target annotation object, and at least one of the following processing methods is performed on the candidate annotation object with lower decision contribution: hiding, shrinking display, and offset display.

[0009] Optionally, in response to a change in the floor identifier corresponding to the next scene switching node, the method further includes: Extract the first object corresponding to the entrance of the cross-floor access facility on the current floor and the second object corresponding to the exit of the cross-floor access facility on the target floor from the candidate labeled objects; Establish a cross-layer association display relationship between the first object and the second object; In the visual annotation results, the first object and the second object are highlighted together.

[0010] Optionally, determining the decision contribution of each of the candidate labeled objects further includes: Obtain the current direction of travel of the navigation terminal; Determine the angle between each candidate annotation object and the current direction of travel, as well as the front-to-back position relationship of each candidate annotation object relative to the current position; In response to a candidate annotation object being located in front of the current position and the angle between the candidate annotation object and the current direction of travel being less than a preset angle threshold, the decision contribution of the candidate annotation object is increased; in response to a candidate annotation object being located behind the current position or the angle between the candidate annotation object and the current direction of travel being greater than a preset angle threshold, the decision contribution of the candidate annotation object is decreased.

[0011] Optional, also includes: Based on the remaining path distance from the current location to the next scene switching node, adjust the display salience of the target labeled object; As the remaining path distance decreases, the display salience of the target labeled object directly associated with the next scene switching node increases.

[0012] Optionally, determining the extended continuous spatial region as the decision corridor includes: For each sampling position of the forward main guide band, determine the line-of-sight ray pointing from the current position to the corresponding sampling position; In response to the intersection of the line-of-sight ray and the non-transparent barrier boundary, the occlusion truncation component of the corresponding sampling position is determined based on the difference between the visible distance from the current position to the first intersection position and the ray path distance to the corresponding sampling position. The decision corridor is determined by subtracting the occlusion truncation component from the lateral expansion amount corresponding to each sampling location.

[0013] Optionally, determining the decision corridor includes: Determine the directional transition amount between the first path direction of the navigation path before the next scene switching node and the second path direction after the next scene switching node; Based on the directional transition amount, the turning contour of the passage boundary in the neighborhood of the next scene switching node, and the positional relationship of the corresponding sampling position relative to the next scene switching node, the turning pre-display component of the corresponding sampling position is determined. The turning pre-display component is used to characterize the pre-guidance area that will be entered into the field of view after passing the next scene switching node. The decision corridor is determined by subtracting the occlusion truncation component from the lateral expansion amount corresponding to each sampling position and superimposing the steering pre-display component.

[0014] Secondly, this application provides a spatial information visualization and annotation system optimized for navigation, comprising: The data acquisition module is used to acquire indoor spatial map data as well as the current location and target location of the navigation terminal; The first processing module is used to determine a navigation path based on the current position and the target position, and identify the next scene switching node along the direction of travel of the navigation path; and to construct a decision corridor corresponding to the current position based on the current position and the next scene switching node. The second processing module is used to extract candidate annotation objects that intersect with the decision corridor or are semantically associated with the next scene switching node from the indoor space map data; and for each candidate annotation object, based on its disambiguation effect on the passage decision at the next scene switching node, determine the decision contribution degree corresponding to each candidate annotation object. The annotation module is used to map each of the candidate annotation objects to the display area corresponding to the current view, determine the target annotation object and the display method of the target annotation object based on the decision contribution degree and the projection conflict relationship after mapping, and output the visualization annotation result of the target annotation object.

[0015] Compared to existing technologies, this application does not simply display indoor space annotations based on object category, spatial proximity, or general display priority. Instead, it constructs a decision corridor corresponding to the current navigation stage around the next scene transition node in the navigation path, taking into account the current location, and extracts candidate annotation objects based on this corridor. Furthermore, this application determines the decision contribution of each candidate annotation object based on its disambiguation effect on the passage decision at the next scene transition node, and uses this decision contribution for subsequent annotation object selection and display control. This allows the limited display area to prioritize annotation information that is truly helpful for the current stage's passage decision, reducing the probability of annotation objects with low relevance to the current decision appearing in the display results.

[0016] Based on the above technical solution, this application enables the filtering logic of indoor navigation labels to shift from general map display guidance to decision-guided guidance, thereby improving the relevance of labels during critical path change phases, reducing the obscuring and interference of irrelevant labels on key guidance information, and enhancing the user's efficiency in identifying the next direction of travel in scenarios such as turning, intersections, and crossing levels. Simultaneously, this solution also helps improve the rationality of information organization and guidance effectiveness of the navigation interface in complex indoor topological environments, thus improving the navigation interaction experience. Attached Figure Description

[0017] Figure 1 A flowchart illustrating a navigation-optimized spatial information visualization annotation method provided in this application embodiment; Figure 2 A flowchart illustrating a method for constructing decision corridors according to an embodiment of this application; Figure 3A flowchart illustrating a method for determining decision contribution as provided in an embodiment of this application; Figure 4 This is a schematic diagram of a navigation-optimized spatial information visualization and annotation system provided in an embodiment of this application. Detailed Implementation

[0018] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0019] See Figure 1 The diagram shows a flowchart of a spatial information visualization annotation method for navigation optimization provided in an embodiment of this application, including steps S101 to S104, wherein: S101: Obtain indoor space map data and the current and target locations of the navigation terminal; S102: Determine a navigation path based on the current location and the target location, and identify the next scene switching node along the direction of travel of the navigation path; construct a decision corridor corresponding to the current location based on the current location and the next scene switching node; S103: Extract candidate annotation objects from the indoor space map data that intersect with the decision corridor or are semantically associated with the next scene switching node; for each candidate annotation object, determine the decision contribution degree corresponding to each candidate annotation object based on its disambiguation effect on the passage decision at the next scene switching node; S104: Map each of the candidate annotation objects to the display area corresponding to the current view; determine the target annotation object and the display method of the target annotation object based on the decision contribution degree and the projection conflict relationship after mapping; output the visualization annotation result of the target annotation object.

[0020] In one embodiment, the method of this application can be executed locally by the navigation terminal or jointly by the navigation terminal and a server. For example, the navigation terminal can be a smartphone, tablet computer, in-vehicle terminal, handheld navigation device, or augmented reality glasses, etc., and the server can be an indoor map service platform, a navigation business platform, or an edge computing node deployed in a local computer room in a park. The navigation terminal may include a processor, memory, positioning and communication module, display module, and user interaction module, and may also include a camera, inertial measurement unit, Bluetooth module, Wi-Fi module, ultra-wideband module, etc., if necessary. The memory may store navigation program, map rendering program, path planning program, and related configuration data for executing the method of this application. The above-mentioned functional programs can be uniformly carried by a single application, or they can be carried separately by the map engine, positioning engine, and navigation business module and completed collaboratively through interface calls.

[0021] Regarding the above S101: Indoor spatial map data can be understood as a digital data set used to represent the spatial structure, traffic relationships, and location semantics of an indoor scene. This indoor spatial map data can be pre-stored locally on the navigation terminal, or it can be distributed by the server on demand, or it can be obtained through a combination of local caching and remote updates. For example, in scenarios such as large shopping malls, airport terminals, hospital outpatient buildings, convention centers, or underground parking lots, indoor spatial map data can be organized by floor and logically include a planar structure layer, a traffic topology layer, and a semantic annotation layer. The planar structure layer can be used to describe the two-dimensional or two-and-a-half-dimensional spatial outlines such as wall boundaries, shop boundaries, corridor boundaries, stair areas, elevator lobbies, turnstile areas, and parking areas; the traffic topology layer can be used to describe the traffic network composed of nodes and edges, for example, setting turning points, intersections, lobby entrances, elevator entrances, escalator entrances, and stair entrances as nodes, and setting corridors, passages, doorway connecting segments, and ramp connecting segments as edges; the semantic annotation layer can be used to record semantic information such as area names, facility categories, shop names, parking zone numbers, floor numbers, entrance / exit numbers, restrooms, and service counters.

[0022] In some implementations, indoor space map data can be stored using a combination of graph structures and geometric objects. For example, adjacency lists or adjacency matrices can be used to store access topology relationships, while polylines, rectangles, polygons, or raster occupancy maps can be used to represent walls, access boundaries, and functional area outlines. Another example is that each floor can be organized as a floor data object, including a floor identifier, a set of accessible area boundaries, a set of path nodes, a set of path edges, and a set of spatial semantic objects. For cross-floor access scenarios, cross-floor connections can also be established between different floors to represent the correspondence between the upper and lower ends of the same elevator shaft, stairwell, or escalator. The above data formats are all exemplary manifestations of this application and do not constitute a limitation.

[0023] In one implementation, indoor space map data can be acquired from multiple sources. For example, it can be generated by preprocessing CAD floor plans, BIM models, architectural construction drawings, or manual surveying results using a map editing backend; or it can be generated by vectorizing collected point clouds, panoramic images, or laser scan results using map creation tools. For indoor navigation systems already in operation, indoor space map data can be continuously maintained and updated by a map management platform. For instance, when shopping malls adjust their fencing, hospital departments relocate, parking lots partially close, or temporary passages are added, the backend can update the corresponding boundary and semantic data, and the navigation terminal loads the updated map version the next time it connects to the network or enters the target building. This ensures that the data used for subsequent route planning and corridor construction remains consistent with the actual indoor environment.

[0024] In one implementation, the current location represents the indoor location of the navigation terminal or user, and the target location represents the destination the user wishes to reach. The current location can be obtained through indoor positioning technology, while the target location can be determined through user input, business system linkage, or historical task recovery. For example, the current location can be obtained by Bluetooth beacon positioning, Wi-Fi fingerprint positioning, ultra-wideband positioning, geomagnetic positioning, visual positioning, inertial dead reckoning, or a fusion of two or more of the above technologies. For scenarios requiring high stability, a fusion approach combining absolute positioning and relative reckoning can be preferred. For instance, an initial absolute location can be determined based on a Bluetooth beacon or UWB tag, and then the location can be continuously updated over a short period using gait information, heading information, and displacement estimation output by the inertial measurement unit. For indoor scenarios with good visual conditions, a mobile visual odometry module can also be invoked to match the image sequence captured by the camera with pre-stored environmental features to obtain the current location. The target location can be the store name, clinic name, parking space number, or boarding gate number entered by the user in the search box, or it can be a target location obtained through scanning, ticketing system, queuing system, parking system, or property management system linkage.

[0025] In some implementations, the current location and the target location can be uniformly mapped to an indoor map coordinate system. For example, the current location and the target location can be represented as floor number and planar coordinates, or as area identifier and local coordinates within the area.

[0026] For example, in a parking lot scenario, the target location can be specifically represented as "B2 level - A3 zone - parking space 126"; in a hospital scenario, the target location can be specifically represented as "3rd floor - outpatient east area - electrocardiogram room"; and in an airport scenario, the target location can be specifically represented as "2nd floor - domestic departure level - gate C18". When the positioning engine outputs raw coordinates while the business system returns semantic location, the system can first utilize the binding relationship between semantic location and object identifiers on the map to resolve the target location into a target node on the map, the center point of the target area, the entrance point of the target facility, or a access reference point associated with the target object.

[0027] In one implementation, the results of obtaining the current and target locations can be preprocessed before proceeding with subsequent path planning. For example, the current location can undergo floor verification, accessibility verification, and boundary snapping. Boundary snapping refers to correcting the current location to the nearest accessible centerline, accessible boundary point, or nearest path node when the positioning result falls near a passable boundary, the edge of a non-passable area, or the neighborhood of a node. This avoids instability in path planning results due to instantaneous positioning drift. Similarly, for the target location, the target object can be mapped to a reference point on its accessible side, rather than simply taking the object's geometric center, based on actual navigation needs. For example, for a store target, the target location can be a passable point near the store entrance; for a parking space target, it can be a nearby point on the side of the lane where the parking space is located; for a clinic target, it can be a stop point in the corridor outside the clinic entrance. This better meets the actual guidance needs of indoor navigation.

[0028] Regarding S102 above: The navigation path can be understood as a recommended route from the current location to the target location. This navigation path can be generated locally on the navigation terminal or generated by a server and then sent to the navigation terminal. For example, path search can be performed based on the access topology map in an indoor spatial map. Common implementation methods include Dijkstra's algorithm, A*, heuristic search, hierarchical path search, or a hybrid path planning method based on the linkage of regional and fine-grained maps. If the system considers factors such as accessibility, congestion, construction closures, staircase disabling, escalator direction, and elevator priority, these factors can be written into the edge weights or node costs and considered by the path planning module when generating the navigation path. For mobile implementations, the map SDK or navigation engine module can call the locally cached map for fast solution; for large-scale parks or complex commercial buildings, the server can uniformly plan the path and send the results to the terminal. The navigation path can be represented as an ordered sequence of nodes, an ordered sequence of edges, a polyline sequence, or a composite path object with floor switching indicators.

[0029] In one implementation, after obtaining the navigation path, the next scene transition node can be identified along the direction of travel on the navigation path. Here, a "scene transition node" can be understood as a location where the navigation focus will change as you continue from the current position. This location is not necessarily a traditional turning point, but can also be a location where the passage mode, connectivity structure, or floor relationship changes.

[0030] For example, scene switching nodes include: inflection points where the path direction changes significantly, intersections where the number of accessible side roads changes, entrances from straight passages to open lobbies, cross-floor entrances from the current floor to elevator lobbies or stairwells, passage nodes through turnstiles, and entrance nodes from driveways to parking zones. By identifying these types of nodes, the system can subsequently organize the navigation guidance range around the "location that most needs to be correctly perceived and decided by the user."

[0031] In some implementations, identifying the next scene switching node can be based on both geometric and topological changes of the path nodes. For example, candidate path nodes can be traversed sequentially along the navigation path after the current position. For each candidate path node, its associated node attributes can be calculated or read, such as the directional deflection between preceding and following path segments, the number of adjacent connected branches, the floor identifier of the node, and the node's access method identifier. Then, the node that first meets the preset switching conditions from the current position along the path's forward direction is selected as the next scene switching node. The directional deflection can be set based on the user's perceived need for a significant turn during indoor navigation. For example, in a narrow corridor, a small deflection often does not lead to a significant change in path decision, while at the junction of halls and corridors or at the connection of diagonal passages, a larger directional change is more likely to mean that the user needs to make a clear judgment. For example, the preset turning threshold can be set between 20 and 45 degrees, such as 30 or 35 degrees; when the directional deflection between preceding and following path segments reaches or exceeds this threshold, the node can be considered a candidate point for directional switching. Of course, this threshold can be adjusted according to the building type, path scale, and display terminal type. For example, the threshold can be appropriately increased in parking lot driveway scenarios and appropriately decreased in hospital outpatient dense corridor scenarios.

[0032] In some implementations, the determination of a change in the number of connected branches can be based on the number of accessible edges, entrances / exits, or branch sectors within the neighborhood of a path node. For example, when moving along a path from the current position and entering a three-way intersection, four-way intersection, or open lobby, although the direction of the preceding and following path segments may not change significantly, the increased number of alternative travel directions makes this location a key decision point requiring the user's attention. Therefore, the change in the number of connected branches can be used as an important basis for identifying scene transition nodes. Similarly, in elevator lobby entrances, escalator entrances, or stairwell areas, the geometric angle of the path may not be significant, but because the mode of travel changes from planar walking to cross-floor passage, this node can also be identified as a scene transition node.

[0033] In one implementation, after identifying the next scene transition node, a decision corridor corresponding to the current location can be constructed based on the current location and the next scene transition node. The decision corridor can be understood as a spatial region within the current navigation phase, from the current location to the next scene transition node, that is highly relevant to the user's upcoming path selection behavior. It differs from a simple fixed-radius area centered on the current location, and also from a fixed-width path buffer zone uniformly expanded along the entire navigation path. Instead, it is a phased guidance area constructed around "where the user is going from and where they will make a path decision in the current phase." By introducing a decision corridor, subsequent navigation processing can focus on the spatial range most relevant to the decision in the current phase.

[0034] In some implementations, constructing a decision corridor can begin by generating a forward main guidance strip along the path centerline from the current location to the next scene transition node. This forward main guidance strip can be viewed as a narrow strip-shaped path region laid out along the current main navigation direction, representing the user's basic trajectory at the current stage. This main guidance strip can be generated based on a path polyline, lines connecting path nodes, or a smoothed centerline. To ensure appropriate processing granularity, multiple sampling positions can be set along the forward main guidance strip at preset intervals. The subsequent corridor shape can be locally constructed around each sampling position and then spliced ​​together. The sampling interval is typically related to the scale of indoor path changes, the positioning update frequency, and the terminal's processing capabilities. For example, the sampling interval can be set from 0.5 meters to 2 meters, such as 1 meter. Smaller sampling intervals can be used when the scene is complex and corners are dense; larger sampling intervals can be used when the corridor is long and the route is relatively straight.

[0035] In one implementation, the forward main guide strip can be laterally extended at each sampling location based on the corresponding access boundaries and accessible branch entrances. Access boundaries can be boundary objects used to define the passable area, such as inner wall boundaries, guardrail boundaries, fence boundaries, turnstile boundaries, inaccessible shop boundaries, or parking space boundaries. "Accessible branch entrances" can be branch entrances, lobby entrances, elevator lobby openings, stairwell openings, escalator openings, or parking lane entrances adjacent to the current main path. Lateral extension is not a simple, uniform expansion to the left and right with a fixed width, but is primarily constrained by the passable structures within the period from the current location to the next scene transition node. For example, in a standard double-walled corridor, lateral extension can be limited to the space between the two walls; in a lobby connecting corridor with one open side and one closed side, lateral extension can extend along the open side to the vicinity of the accessible branch entrance associated with the next scene transition node; in a parking lane scenario, lateral extension can cover the current lane width and the range of branch lane entrances leading to the target zone near the transition node.

[0036] In some implementations, the lateral expansion width can be set to a fixed value or an adaptive value that changes with the scene. A fixed value is suitable for standardized scenarios with a high degree of regularity, such as a corridor on a standard floor of an office building; an adaptive value is more suitable for scenarios with large boundary changes, such as commercial complexes, hospital lobbies, and underground parking lots. The setting can be based on factors such as the remaining distance from the current location to the next scene switching node, the width of the passable area around the current location, the local connectivity of path nodes, user movement speed, or navigation update frequency. For example, in a scenario where the corridor width is 2.4 meters to 3.2 meters, the lateral expansion width can be limited within the corridor boundary; in a scenario where an open lobby leads to three branch passages, the lateral expansion range can be gradually widened at sampling positions near the next scene switching node, so that the corridor shape can naturally cover the neighboring area of ​​the branch entrance associated with that node as it approaches the switching node. The resulting corridor is no longer a simple rectangular or circular area, but a continuous area that conforms to the current stage's indoor passage topology and boundary structure.

[0037] For example, taking the scenario of navigating to a target store in a large shopping mall, the current location is on the outer corridor of the atrium on the third floor, and the target store is located in the east wing of the third floor. The system first loads the indoor space map data of the third floor, reads the user's current location and the reference point of the target store entrance, and generates a navigation path from the current location to the target store. Analyzing forward along this navigation path, it is found that after the user continues forward for about 18 meters, they will enter a three-way junction area, and there are three possible branches after this location: left, straight, and right. The target path requires turning right at this node to enter the east wing corridor, so this three-way junction area can be identified as the next scene switching node. Subsequently, the system generates a forward main guidance band along the centerline of the path from the current location to the three-way junction area, and expands laterally at sampling positions every 1 meter based on the corridor wall boundary, atrium railing boundary, and the opening range of the right branch entrance, thus forming a decision corridor related to the three-way node. This decision corridor can effectively cover the effective passage space of the user's current straight-ahead stage and the turning area that needs to be focused on when approaching the three-way junction, providing a spatial basis for subsequent navigation guidance around the current stage.

[0038] For example, in a reverse parking lot scenario, the user's current location is at the boundary between Zone A and Zone B on level B2, and the target location is near a parking space in Zone A3 on level B2. After acquiring the parking lot map, the system generates a navigation path based on the lane centerline network and identifies the next node where the user needs to turn from the main lane into the target zone's side road as the next scene switching node. Subsequently, a main guide strip is generated along the lane centerline between the current location and this node, and then laterally expanded based on lane boundaries, pillar enclosure boundaries, and zone entrance locations, thus forming a decision corridor that conforms to the geometric relationship between the lanes and zone entrances. Compared to a simple fixed-width buffer zone, this method can more accurately reflect the spatial range that the user truly needs to focus on at the current stage.

[0039] Regarding the above S103: Spatial objects are objects that have geometric locations and semantic attributes in indoor maps and can be used as sources of annotations for navigation interfaces. For example, spatial objects may include area name objects, facility entrance objects, service facility objects, directional sign objects, zone number objects, passage number objects, floor switching facility objects, parking area objects, shop entrance objects, or environmental signage objects directly related to the building's access structure.

[0040] To facilitate subsequent processing, each spatial object in the indoor spatial map data can be pre-recorded with its object identifier, floor level, object category, geometric outline, reference display point, semantic label, and its relationship with path nodes or area objects. This information can be written into the object attribute table during map creation or supplemented by the backend maintenance tools before the navigation service goes live.

[0041] In one implementation, to improve the efficiency of candidate object extraction, spatial and semantic indexes can be established for each spatial object in the map data. The spatial index can be used to support fast retrieval of objects intersecting with the decision corridor, and can exemplarily employ an R-tree, grid index, quadtree index, or floor-level bucket index. The semantic index can be used to support fast retrieval of objects semantically related to the next scene switching node, and can exemplarily employ an object category label table, node association table, regional semantic inverse index table, or facility-node binding table. Thus, when executed locally on the navigation terminal, the map engine can first quickly filter out a batch of coarse candidate objects based on the bounding box of the decision corridor, and then further perform precise intersection judgments; when executed on the server side, the spatial query service and semantic query service can also collaboratively complete the candidate object extraction.

[0042] In one implementation, objects intersecting the decision corridor are not limited to objects whose complete geometric outline falls within the corridor. For example, an object is considered to intersect the decision corridor if its reference display point, entrance association point, accessible side access point, or at least a portion of its geometric outline overlaps with the corridor. This is because some objects are large, such as shop outlines, waiting area outlines, or parking zone outlines, and their geometric centers may not be located within the decision corridor, but their actual navigation-significant entrance points, access points, or boundary adjacencies may already be within the corridor. Therefore, in some implementations, corresponding "navigation reference location types" can be configured for different categories of spatial objects.

[0043] For example, for a store object, the reference point outside its entrance / exit can be used; for an elevator lobby object, the reference point at its elevator lobby entrance can be used; for a parking zone object, the connecting point at the zone entrance can be used; and for a directional sign object, the ground projection point corresponding to its installation location or the observation reference point it is bound to can be used. Subsequently, as long as this navigation reference location enters the decision corridor, the corresponding spatial object can be included in the candidate object extraction range.

[0044] In one implementation, an object semantically associated with the next scene transition node refers to an object whose semantic content, functional attributes, or accessibility are directly related to the path selection at the next scene transition node, even if it does not necessarily intersect geometrically with the current decision corridor. For example, in a shopping mall intersection scenario, an object semantically associated with the next scene transition node could be a directional sign such as "East Wing," "Dining Area," or "Parking Exit"; in a hospital multi-floor scenario, an object semantically associated with the next scene transition node could be guidance information such as "Elevator Lobby," "Escalator Up," or "3rd Floor Outpatient East Area"; and in a parking lot scenario, an object semantically associated with the next scene transition node could be a zone semantic object such as "A3 Zone," "B Zone Exit," or "Visitor Parking Area." For such objects, even if their geometric location is outside the decision corridor, as long as they can significantly help the user identify the target branch to enter at the next scene transition node, they can still be included in the candidate annotation object set.

[0045] In some implementations, the semantic association extraction described above can be supported by pre-establishing "node-object" semantic relationships. For example, during map editing or data preprocessing, a set of associated objects can be maintained for each type of scene switching node. For instance, for a branching node, it can be pre-associated with the area name object, direction sign object, and entrance object corresponding to each branch direction; for a multi-level node, it can be pre-associated with elevator entrance objects, staircase entrance objects, escalator entrance objects, and the exit object corresponding to the target floor; for a parking zone entrance node, it can be pre-associated with zone number objects, lane guidance objects, and the target zone entrance object. Thus, after identifying the next scene switching node during the runtime phase, a set of semantically associated objects bound to that node can be directly retrieved from the node association table and merged with the spatial intersection query results to obtain the candidate annotation object set for the current stage.

[0046] In one implementation, to avoid an excessively large set of candidate objects, after extracting a batch of initial objects that intersect with the decision corridor or are semantically associated with the next scene switching node, basic filtering can be performed. For example, objects that are not on the same floor as the current location and do not involve cross-floor switching at the current stage can be filtered out first; objects with missing identification content, missing positioning anchors, or that are in a disabled state can be filtered out; objects marked as temporarily undisplayable by the background can be filtered out, such as construction closure area signs, inactive store signs, or out-of-service facility signs.

[0047] In one embodiment, after extracting candidate labeled objects, the decision contribution of each candidate labeled object can be further determined based on its disambiguation effect on the path selection at the next scene switching node. Here, "disambiguation effect" can be understood as: when a user arrives near the next scene switching node, whether a candidate labeled object can help the user more clearly distinguish the target path among multiple accessible paths. In other words, this application does not merely consider whether an object is important or close to the path, but rather focuses on whether the object can reduce path selection ambiguity for the user at the next scene switching node.

[0048] In some implementations, to quantify the disambiguation effect, an object reference point or object access point can be determined for each candidate labeled object, and the topological relationship between this reference point and each accessible branch after the next scene switching node can be analyzed. For example, if the candidate object is a store entrance object, the access point outside the store entrance can be taken as the object reference point; if the candidate object is a zone sign object, the branch access point corresponding to the zone it points to can be taken as the object reference point; if the candidate object is a parking zone number object, the lane entrance point of the zone corresponding to that number can be taken as the object reference point. Subsequently, a first topological distance from the candidate object to the target access branch and a second topological distance to at least one other accessible branch can be determined. The topological distance here does not necessarily have to be an Euclidean straight-line distance, but is more suitable to use path length, edge distance, node hop count, or weighted access distance calculated along the access network. This can more realistically reflect the reachability relationship between the object and different branches.

[0049] In one implementation, when the first topological distance of a candidate object to the target branch is significantly less than its second topological distance to other accessible branches, the candidate object is considered to have a strong specific cue effect on the target branch, thus its disambiguation effect is strong. Conversely, when a candidate object is close to multiple accessible branches simultaneously, or may be valid in multiple branch directions, it indicates that it cannot effectively help the user eliminate other branches, and its disambiguation effect is weak.

[0050] For example, at a shopping mall intersection, a "food court" sign that is strongly associated only with the right-hand branch has a strong disambiguation effect on right-turn decisions; while a "service counter" sign located in the central open area of ​​the node and visible from the left, right, and left directions often has a weaker disambiguation effect on specific branch choices. Similarly, near the entrances to two adjacent zones A and B in a parking lot, the "zone A3" sign has a strong disambiguation effect on the branch road leading to zone A, while the "exit" sign, if accessible to multiple lanes, has a relatively limited disambiguation effect on the current zone selection.

[0051] In some implementations, the difference between the first topological distance and the second topological distance can be used as the basis for determining the degree of disambiguation. This difference is set based on the principle that a larger difference indicates a candidate object is more inclined to support a specific target branch than others, thus demonstrating a stronger ability to differentiate between different paths. For example, a preset distance difference threshold can be used to distinguish between high-disambiguation and low-disambiguation objects. For instance, in a shopping mall or hospital pedestrian navigation scenario, if the first topological distance is at least 3 meters, 5 meters, or the length of a complete passageway segment smaller than the second topological distance, the candidate object can be considered to have a significant disambiguation effect on the target branch. In a parking lot lane navigation scenario, a threshold can also be set based on the number of lane segments or the number of hops at the zone entrance. For example, if the target branch passes through at least one fewer entrance node than other branches, the corresponding object can be considered to have a strong disambiguation effect. These thresholds can be configured according to building scale, branch road density, and navigation accuracy requirements, and are not intended to be limiting.

[0052] In one implementation, after obtaining the disambiguation level of the candidate objects, it can be further converted into a corresponding decision contribution score. The decision contribution score can be represented by a continuous value, a discrete level, or an interval score. For example, it can be divided into three levels: high, medium, and low, or it can be divided into a normalized contribution value between 0 and 1, or an integer score between 0 and 100. Its specific representation can be determined according to the system implementation conventions. This application does not limit the use of a certain scoring form. The key point is that the decision contribution score reflects the strength of the candidate object's support for the passage decision at the next scene switching node, rather than the object display priority in a general sense.

[0053] In some implementations, to improve stability, object category credibility, map data integrity markers, or object status markers can be superimposed on the disambiguation level to make a basic correction to the decision contribution. For example, for long-term stable facility entrance objects, area name objects, and partition number objects that are strongly bound to the target branch, their decision contribution can directly inherit the main ranking result of the disambiguation level; while for temporary promotional signs, event posters, and weakly bound semantic objects, although they may be included in the candidate set in terms of location, their weight can be appropriately reduced when converting them into decision contribution because their support for long-term passage decisions is weak. The corrections here reflect more on object stability and data credibility and do not change the technical concept of this application, which is "centered on the disambiguation function of passage decisions".

[0054] For example, in a hospital outpatient scenario, the user's current location is in the main corridor on the second floor, and the target location is in the ultrasound department on the third floor. The system has identified the next scene transition node as the elevator lobby entrance ahead. After constructing a decision corridor before this node, a batch of initial objects can be extracted from the map, including elevator lobby entrance objects, staircase entrance objects, escalator entrance objects, second-floor service desk objects, third-floor ultrasound department directional sign objects, and "Outpatient East Zone" area objects. Among them, the elevator lobby entrance object and the "third-floor ultrasound department directional sign object," although one directly intersects the decision corridor and the other may be located outside the corridor but is semantically related to the node, can both be included in the candidate object set. Subsequently, the system analyzes the topological relationships of these objects to the elevator access branch, staircase access branch, and escalator access branch, respectively. If the target path requires the user to enter the elevator lobby, the objects directly corresponding to the elevator lobby entrance and the elevator guidance objects bound to the third-floor ultrasound department will usually have a high degree of disambiguation and a high decision contribution. However, although the service desk object located on the side of the corridor has entered the decision corridor, it is less helpful in distinguishing between elevators, stairs and escalators, so its decision contribution is relatively low.

[0055] For example, in a reverse car-finding scenario in an underground parking lot, the user needs to drive from the current main lane to the entrance of zone A3. After constructing a decision corridor between the current location and the next scene transition node, the system can extract candidate objects such as the "Zone A3" numbered object, the "Zone A Entrance" object, the "Zone B Entrance" object, the "Exit" object, and the "Charging Space Area" object. For the target path, the "Zone A3" numbered object has the shortest first topological distance to the target route branch, while its second topological distance to the corresponding branch road in the direction of the Zone B entrance and exit is significantly longer, thus its disambiguation effect is considered strong. In contrast, although the "Exit" object is visually more eye-catching, its associated route is inconsistent with the current target branch, so its decision contribution is lower. Through the above processing, the system can prioritize candidate objects that truly help the user identify the target branch at the current stage.

[0056] In one embodiment, the processing result of S103 can be represented as a set of candidate labeled objects and their corresponding decision contribution sets for the current stage. For example, the set of candidate labeled objects can be represented as a list of objects, where each object entry includes fields such as object identifier, object category, reference point coordinates, floor level, associated node identifier, disambiguation degree, and decision contribution. This result can be stored in the navigation terminal's memory for subsequent modules to call, or it can be returned to the terminal by the server, allowing the terminal to perform subsequent navigation interface processing based on this result.

[0057] Regarding S104 above: The current view can be understood as the display perspective currently used by the navigation terminal to present map, path, and annotation information. For a 2D map interface, the current view can be determined by the current center point, zoom level, rotation angle, and display window range. For pseudo-3D map interfaces, 3D indoor browsing interfaces, or augmented reality navigation interfaces, the current view can be further determined by parameters such as camera position, pitch angle, field of view, viewing direction, and near / far clipping range. The navigation terminal can obtain the above-mentioned current view parameters through the map rendering engine, graphical interface framework, or AR rendering module and use them for subsequent object mapping processing.

[0058] In one implementation, mapping candidate annotation objects to the display area corresponding to the current view can be understood as converting the spatial position or reference display position of the candidate annotation object in the indoor map coordinate system into the screen coordinate position or screen area occupied by the candidate annotation object in the current display interface. For two-dimensional map mode, the map engine can directly map the reference point of the candidate annotation object from the floor plane coordinate system to the screen pixel coordinate system based on the current zoom level, translation offset, and rotation angle. For three-dimensional navigation mode or AR mode, the spatial coordinates of the candidate annotation object can be transformed into the camera coordinate system first, then projected onto the screen plane, and the display position of the object on the screen can be further obtained. Here, the candidate annotation object can be directly mapped as a point using the object reference point, or it can be combined with the object's preset icon size, text box size, text length, or guide arrow length to further form a corresponding display bounding area. For example, after mapping, a candidate annotation object can be represented as a display candidate entry including fields such as object identifier, screen center point coordinates, screen bounding box size, object category, decision contribution, and preset display style.

[0059] In some implementations, to ensure the stability of the mapping results, display anchor points can be selected for the reference display positions of candidate labeled objects before mapping. For entrance-type objects, anchor points outside the entrance can be used preferentially; for area name objects, representative points within the area outline that are easy to observe can be used preferentially; for parking zone objects, access-side anchor points corresponding to the currently passable lane or pedestrian passage can be used preferentially; for directional sign objects, ground projection points corresponding to their installation locations or semantically pointed target branch entrance points can be used. This can avoid labels appearing in locations that are difficult for users to understand because the geometric center of the object is not suitable as a display location. For example, in a large shopping mall, the outline of a store may be long and narrow, and its geometric center may extend deep into the store, while the entrance point near the corridor side is more suitable for navigation display; in a parking lot, a zone number object may correspond to an entire parking space area, and its geometric center may not be directly related to the current lane, so the connected point on the zone entrance side is more suitable as a display anchor point.

[0060] In one implementation, after the candidate annotation objects are mapped, their projection conflict relationships in the current view can be determined. Projection conflict relationships can be understood as situations where, after different candidate annotation objects are projected onto the current display interface, their display areas overlap, obscure, are too close, or compete for display space with key navigation elements. Key navigation elements here may include navigation path lines, current location markers, target location markers, turning prompt arrows, floor switching prompts, or other high-priority guidance information. For example, in a 2D map mode, projection conflicts can be determined by comparing whether the screen bounding boxes corresponding to each candidate annotation object overlap or are less than a preset spacing threshold; in 3D or AR modes, depth order, occlusion relationships, and viewpoint visibility can also be combined to determine whether an object is partially or completely obscured by another object, wall, column, or fixed interface elements.

[0061] In some implementations, projection conflict determination can be achieved using methods such as object-by-object detection, grid occupancy detection, or sorted scan detection. For local real-time processing on mobile devices, to balance efficiency and effectiveness, a sorted scan combined with simplified bounding box detection can be prioritized. For example, candidate labeled objects can first be sorted according to their vertical position, horizontal position, or decision contribution on the screen, and then the display bounding area of ​​the current object can be sequentially checked to see if it overlaps with an already occupied display area. When overlap occurs, a projection conflict can be identified. For mobile interfaces with limited screen space, the minimum safe distance between bounding boxes can be set to 4 to 20 pixels, such as 8 pixels, 12 pixels, or 16 pixels, to avoid causing reading difficulties for users even when they are too close together, even if they do not completely overlap. The setting of this safe distance can be related to the terminal resolution, font size, icon size, and the user's expected reading distance. For example, the safe distance can be appropriately increased on high-resolution tablet devices, and appropriately decreased in modes that only display icons and not long text.

[0062] In one embodiment, after determining the projection conflict relationship after mapping, the target annotation object and its display method can be determined based on the decision contribution and the projection conflict relationship. The target annotation object can be understood as the object that is finally allowed to be actually displayed in the interface after the current stage of screening; the display method can be understood as the interface presentation form adopted by the target annotation object, such as icon display, text display, icon and text combination display, arrow guidance display, associated highlight display, zoomed-out display, offset display, partial transparent display, or only expanded display after interaction triggering, etc. It should be understood that this application is not limited to a specific display form. The key is whether the limited display space is ultimately allocated preferentially to objects that are more helpful to the current stage's decision-making.

[0063] In some implementations, decision contribution can be used as the core basis for selecting and configuring the display method of target annotation objects. For example, candidate annotation objects can be sorted from highest to lowest decision contribution, and then placed into the current view sequentially according to the sorting results. For a given object to be placed, if its projection area does not conflict with the projection areas of already retained objects, it can be directly identified as the target annotation object; if a conflict occurs, objects with higher decision contribution can be retained first, and objects with lower decision contribution can be processed by at least one of hiding, shrinking the display, or offsetting the display. Here, "hiding" can mean not displaying the object at the current stage; shrinking the display means only retaining the icon, shortening the text, reducing the font size, simplifying the border, or compressing the display area; offsetting the display means moving the annotation object away from its original mapping position by a preset distance and establishing a visual correspondence with the original object position through guide lines, guide arrows, or anchor points. This allows the current interface to maintain high readability and navigation relevance as much as possible without losing key guidance information.

[0064] In one implementation, the determination of the display method can consider not only whether to display, but also the display intensity and detail. For example, target objects can be divided into key display objects and auxiliary display objects based on their contribution to the decision. Key display objects can be displayed using icons with text, larger fonts, highlighted backgrounds, directional guide lines, or breathing animation prompts; auxiliary display objects can be displayed with simplified icons or secondary text, and detailed information can be expanded only after the user clicks, pauses, or zooms in on the map. This further enhances the hierarchy of on-screen annotations. For example, in a shopping mall intersection scenario, objects strongly associated with the target branch, such as the "East Wing" and "Dining Area," can be displayed with highlighted text boxes or icons with directional arrows, while objects weakly associated with the current decision but still located in the corridor, such as the "Service Desk," can be displayed only as small icons, or temporarily omitted when there is significant conflict.

[0065] In some implementations, for objects that are strongly related to the current traffic decision but are located at the edge of the screen after mapping, and whose original projection area may conflict with the path line or the current location marker, edge snapping display or offset guided display can be used. "Edge snapping display" refers to moving the target labeled object to a reserved display area at the edge of the interface, while simultaneously using a guide line to point to the object's original mapping position or the target branch direction. "Offset guided display" refers to moving the label box along a preset offset direction while keeping the object within the main display area, and pointing it back to the anchor point using short guide lines or connecting markers. The offset distance can be set according to the object size, surrounding congestion, and terminal screen size. For example, the offset distance can be set from 10 pixels to 60 pixels, such as 20 pixels, 30 pixels, or 40 pixels. This setting is based on the fact that an offset that is too small may still not effectively eliminate the conflict, while an offset that is too large may weaken the correspondence between the object and its original spatial position.

[0066] In one embodiment, to enhance the user's continuous perception of the current navigation stage, appropriate temporal smoothing processing can be applied to the display results after the target labeled object and its display method are determined. For example, when continuous updates to the current location cause a slight shift in the current view, if a target labeled object still meets the display conditions, its display state can be maintained unchanged for a short time window to avoid frequent appearances and disappearances of the object due to slight positional jitter or minor changes in viewing angle. This short time window can be set according to the navigation update cycle, positioning refresh rate, and terminal frame rate, for example, it can be 300 milliseconds to 1500 milliseconds. In this application, this type of smoothing processing can be understood as a stability enhancement measure for the display layer, the purpose of which is to reduce interface flicker and enhance the continuous readability of the navigation interface.

[0067] In one embodiment, outputting the visual annotation result of the target annotation object can be understood as submitting the finally determined target annotation object and its display method to the map rendering module, navigation interface module, or AR rendering module, and completing the drawing on the display interface of the navigation terminal. For two-dimensional maps, the target annotation object can be superimposed on the floor plan, navigation path line, current location marker, and target location marker; for three-dimensional maps, the target annotation object can be superimposed on the three-dimensional scene in the form of a floating information board, ground label, or arrow guide; for AR mode, the target annotation object can be registered with the environmental structure in the real-time image of the camera and then displayed on the screen. The output result can be a static display result of a single frame, or a series of continuously displayed results that are dynamically refreshed as the location updates. In other words, as long as the annotation interface finally presented to the user at the current stage can be obtained based on the decision contribution and projection conflict relationship, it can be regarded as the output of the visual annotation result in this application.

[0068] For example, in a shopping mall navigation scenario, the user's current location is in the outer corridor of the third-floor atrium. The system has already identified a set of candidate labels in stage S103, including objects such as "East Wing," "Dining Area," "Service Desk," "Restaurant," and "Activity Area Entrance," and assigned them decision contribution scores. The current view is a mobile 2D map interface, with the current location displayed in the center of the screen and the three-way connection area corresponding to the next scene switching node on the right front. The system first maps the above candidate objects to the current screen coordinate system and finds that the projection areas of the "East Wing" and "Dining Area" objects partially overlap on the right front, while the projection area of ​​the "Service Desk" object conflicts with the path line turning prompt area. Since the "East Wing" and "Dining Area" have a high decision contribution score for turning right into the target branch, while the "Service Desk" is of little help in the current passage decision, the system can prioritize retaining the "East Wing" and "Dining Area" as target labels, slightly offsetting one of the objects in the display, while hiding the "Service Desk" object. The final output of the visualization annotation will highlight guidance information that is strongly related to the right-turn target branch on the screen, while downplaying service facility information that is not directly related to the current stage.

[0069] For example, in a reverse parking lot scenario, a user walks along the main lane towards the entrance of Zone A. Candidate objects include "Zone A3," "Zone A Entrance," "Exit," "Charging Parking Area," and "Zone B Entrance." The current view uses a pseudo-3D navigation interface, with the path line extending forward from the bottom of the screen. After mapping these objects, the system finds that "Zone A3" and "Zone A Entrance" are both located in the slightly right-hand area ahead and are adjacent to the path line, while the "Exit" object, although located in the center of the upper part of the screen, has no direct correspondence with the target branch. Since "Zone A3" and "Zone A Entrance" have a higher decision-making contribution, the system can identify them as target labels. "Zone A3" is highlighted with an icon and text, while "Zone A Entrance" is displayed as a secondary guide label. The "Exit" object is not displayed at the current stage or is displayed after the user zooms in on the map. This allows users to more quickly establish spatial awareness that they should "turn into Zone A and continue to Zone A3" when approaching a zone entrance.

[0070] In one embodiment, the processing result of S104 can be represented as a set of target annotation objects and their corresponding display configuration set for the current stage. For example, the display configuration of each target annotation object may include at least the following fields: screen display position, display level, text content, icon identifier, font size, color style, offset parameter, whether to display connecting guide lines, and whether to enable highlighting. This display configuration set can be generated locally by the navigation terminal and directly submitted to the rendering thread, or it can be calculated by the server and sent to the navigation terminal for display.

[0071] Optional, seeFigure 2 The flowchart of a method for constructing decision corridors provided in an embodiment of this application includes: S201: Generate a forward main guide strip along the path centerline between the current position and the next scene switching node; S202: At each sampling position of the forward main guide strip, the forward main guide strip is laterally expanded according to the passage boundary and the entrance of the accessible branch road at the corresponding position; S203: The expanded continuous spatial region is determined as the decision corridor.

[0072] Furthermore, to avoid the decision corridor remaining merely at the abstract level of "relevant area" and failing to generate stably in a real system, some implementations can further refine the construction process of the decision corridor as follows: A forward main guidance band is generated along the path centerline between the current location and the next scene switching node. At each sampling position of the forward main guidance band, the area is laterally expanded based on the corresponding access boundary and accessible branch entrances. The expanded continuous spatial area is then defined as the decision corridor. In indoor navigation scenarios, if the guidance area for the current stage is defined solely based on the line connecting the current location and the next scene switching node, a fixed-width rectangular area, or a fixed-radius neighborhood, two types of problems can easily arise: one is poor fit to narrow spaces such as corridors, driveways, and lobbies, leading to the unnecessary inclusion of irrelevant areas; the other is the inability to naturally cover the neighboring areas of branch entrances related to the target path when near intersections, corridor boundaries, or zone entrances, thus affecting the accuracy of subsequent candidate labeling.

[0073] Based on this, in some implementations, a forward main guidance band can be generated preferentially along the path centerline between the current position and the next scene switching node. The path centerline here can be directly taken from the polyline center trajectory in the path planning results, or it can be obtained by smoothly connecting adjacent path segments. The forward main guidance band can be understood as a narrow strip-shaped area extending along the main direction of travel in the current stage, used to represent the user's most basic and stable forward trajectory in the current stage.

[0074] In some implementations, multiple sampling positions can be set along the forward main guide strip at preset sampling intervals. The sampling interval should not be too large, otherwise it will be difficult to reflect local traffic structure changes in a timely manner near path bends, lobby openings, or lane forks; the sampling interval should also not be too small, otherwise it will increase the computational burden and lead to high repetition of adjacent sampling results. Considering the processing requirements of real-time navigation on mobile devices, the sampling interval can be set to 0.5 meters to 2 meters, for example, 0.8 meters, 1 meter, 1.2 meters, or 1.5 meters. Smaller sampling intervals can be used in environments with rapid topological changes, such as hospital outpatient corridors or densely populated shopping mall areas; larger sampling intervals can be used in environments with gradual changes, such as long straight lanes in underground parking lots or straight corridors in standard office buildings. The above sampling positions can be represented as a set of points arranged in an orderly manner along the path centerline. Each sampling point can further record its path segment number, relative path distance, and adjacent boundary number, etc., for subsequent lateral expansion.

[0075] In some implementations, at each sampling location, the forward main guide strip can be laterally expanded based on the corresponding access boundaries and accessible side road entrances. The access boundaries can be inner wall boundaries, guardrail boundaries, fence boundaries, turnstile boundaries, parking space boundaries, or other boundaries used to define the accessible area; accessible side road entrances can be adjacent to the current main path and permitted access openings such as passageway openings, lobby openings, elevator lobby entrances, stairwell entrances, escalator entrances, driveway branch entrances, or area zone entrances. Lateral expansion is not simply an equidistant expansion to the left and right of the sampling location, but rather the endpoints on both sides are determined based on the actual local traffic patterns at that sampling location. For example, in a standard double-walled corridor, lateral expansion can be constrained between the two walls; in a scenario where one side is an atrium railing and the other side is a shop facade, the expanded area can cover the open, visible passageway on the railing side, while being restricted by the shop's inaccessible boundary; in a parking lane scenario, the lateral expansion can cover the current lane width and extend appropriately towards the target branch at the sampling position near the target branch entrance to include the neighborhood of the subsequent zone entrances. Thus, the decision corridor is not a simple geometric expansion result, but a continuous spatial region that can change with the actual traffic structure at the current stage.

[0076] For example, a user is located in a corridor on the third floor of a large shopping mall. About ten meters ahead, they will reach a connecting node where they can enter the atrium lounge area to the left, the main passageway to the straight path, or the east wing dining area to the right. At this point, a forward main guidance strip can be generated along the recommended path from the current location to the connecting node. Then, at sampling points approximately every 1 meter, the strip can be expanded laterally based on the walls on both sides of the corridor, the atrium railing, and the entrance to the dining area on the right. The resulting continuous area maintains the narrow strip shape of the main corridor while naturally expanding towards the right-hand entrance as the user approaches the connecting node, thus better aligning with the user's upcoming branch decision-making stage. Similarly, in an underground parking garage, as a user walks along the main driveway towards the entrance to Zone A, a forward main guidance strip can be generated along the centerline of the main driveway. At each sampling point, the strip can be expanded laterally based on the driveway boundaries, column barriers, and the location of the Zone A entrance, thus forming a decision corridor that covers the current driveway's effective area and incorporates the target zone's entrance space as the user approaches the entrance.

[0077] Furthermore, in order to provide an executable quantitative basis for judging the "disambiguation effect", some implementations can determine the first topological distance from each candidate labeled object to the target passage branch and the second topological distance to at least one other accessible passage branch; based on the difference between the first topological distance and the second topological distance, determine the degree of disambiguation of each candidate labeled object for the passage decision; and then, based on the degree of disambiguation, determine the decision contribution of each candidate labeled object.

[0078] In indoor navigation, when a user arrives at the next scene transition node, the decision is not usually made simply by "seeing an object" in an abstract sense. Instead, it requires understanding the relative correspondence between that object and different accessible branches to eliminate incorrect directions and identify the target direction. Therefore, simply ranking candidate objects based on their distance from the current location, their category, salience, or proximity to the path often fails to accurately reflect their contribution to the "decision for the current branch." Based on this, some implementations can perform "branch discrimination" analysis on candidate labeled objects, focusing on the branch structure after the next scene transition node.

[0079] In some implementations, a reference point or access point can be determined for each candidate labeled object first, and then the topological distance from the object to different accessible branches can be calculated based on the indoor traffic topology map. Here, the target traffic branch can be understood as the branch the recommended navigation path enters after the next scene switching node; the other accessible traffic branches can be branches that the user can still physically enter at that node, excluding the target traffic branch. For the selection of the object reference point, the reference position that is closer to the traffic relationship can be used preferentially. For example, a store entrance object can use the outer connecting point of the doorway, an area number object can use the corresponding partition entrance point, and a directional sign object can use the branch access point of the area it points to. The topological distance can be obtained by accumulating the edge weights in the navigation map, which can be the length of the traffic path, the number of edge segments, the number of node hops, or a weighted distance obtained after considering congestion and traffic level. Any distance that reflects the actual traffic relationship between the candidate object and different traffic branches can be used in this application.

[0080] In some implementations, the difference between the first topological distance and the second topological distance can be used as the core calculation basis for the degree of disambiguation. If the first topological distance of a candidate label object to the target branch is significantly less than its second topological distance to other accessible branches, it indicates that the object has a stronger binding relationship with the target branch, which is more conducive to the user identifying the correct branch in multiple directions, and its degree of disambiguation can be judged as high. Conversely, if the topological distance between a candidate label object and the target branch and other branches is close, or can be true in multiple branches, it indicates that the object is not enough to effectively help the user eliminate other directions, and its degree of disambiguation can be judged as low. To facilitate system implementation, the degree of disambiguation can be graded according to a preset difference threshold or difference range. For example, in the pedestrian navigation scenarios of shopping malls and hospitals, the first topological distance being at least 3 meters, 5 meters, or the length of a passageway shorter than the second topological distance can be used as an exemplary judgment condition for "high disambiguation"; in the parking lot lane scenario, it can also be judged by the number of lane segments or the number of entrance hops, for example, the target branch passing through at least one fewer entrance node than other branches can be considered a significant difference.

[0081] The aforementioned thresholds can be preset by the system based on building scale, path density, and navigation accuracy requirements, or they can be configured separately by the backend under different scenario templates.

[0082] In some implementations, after obtaining the disambiguation level, it can be further mapped to a decision contribution level. This decision contribution level can be understood as a quantitative result used for subsequent display sorting and conflict resolution. Its representation can be a discrete level, such as high, medium, and low; or a continuous value, such as a decimal between 0 and 1 or a score between 0 and 100. To improve operational stability, object reliability or stability correction information can be superimposed on the disambiguation level. For example, area name objects, partition number objects, and entry objects that are stably bound to the target branch for a long time can directly retain a higher contribution level, while event promotional signs, temporary promotional signs, and short-term change signs, although spatially close to the target branch, have limited long-term navigation guidance function and their contribution level can be appropriately reduced. In this way, the obtained decision contribution level reflects both the object's ability to distinguish the current branch and the reliability of the object as a navigation basis.

[0083] For example, if a user is located in the main corridor of a hospital outpatient department, and the next scene transition node ahead is the elevator lobby entrance, the recommended path requires the user to enter the elevator lobby and go upstairs at this node. In this case, the system can extract candidate objects such as "elevator lobby," "staircase entrance," "escalator entrance," "outpatient department east area," and "service desk," and calculate the topological distances of these objects to the elevator, staircase, and escalator access branches, respectively. If the first topological distance from the "elevator lobby" object to the target elevator branch is the shortest and significantly less than its second topological distance to the staircase and escalator branches, its disambiguation level is high. However, while the "service desk" object is also located within the current corridor, its distances to multiple access branches are close, offering limited help in determining whether the user should enter the elevator lobby, resulting in a lower disambiguation level. For example, in an underground parking lot, when a user should turn from the main lane into the branch road of area A to reach area A3, the first topological distance from the "area A3" object to the target branch road is usually significantly smaller than the second topological distance to the branch road in the direction of the entrance or exit of area B. Therefore, its decision contribution can be set to higher. Conversely, although the "exit" object may be more visually prominent, its contribution can be set to lower because it does not support the selection of the current target branch.

[0084] Furthermore, in order to make the display results more executable and stable in a limited screen space, in some implementations, the projection area of ​​each candidate annotation object mapped to the current view can be determined first; when the projection areas of at least two candidate annotation objects meet the preset conflict conditions, the candidate annotation object with higher decision contribution is retained as the target annotation object, and at least one of the following processes, namely hiding, shrinking display and offset display, is performed on the candidate annotation object with lower decision contribution.

[0085] In indoor navigation scenarios, candidate objects are often densely distributed near decision corridors, especially at intersections, lobby connection areas, elevator entrances, and parking zone entrances. Multiple objects related to or partially related to the current stage may simultaneously fall into the current view. If all candidate objects are simply drawn directly onto the screen, problems such as overlapping text, icons obscuring each other, and annotation boxes pressing on path lines or current location markers can easily occur, which weakens the readability of the navigation interface.

[0086] Therefore, in some implementations, the mapped projection area can be determined for each candidate annotation object first. This projection area can be a rectangular bounding box corresponding to the object's reference display point, or it can be a composite bounding box containing the icon area, text area, and guide line starting point area. For objects that only display icons, their projection area can be smaller; for objects that display both icons and text, their projection area can be expanded according to the text length and font size.

[0087] In some implementations, "meeting the preset conflict conditions" can be understood as at least two projected areas overlapping, with a distance between them less than a preset minimum safety distance, or one of the projected areas engaging in unacceptable spatial competition with critical interface areas such as the path line turning prompt area, the current location indicator area, or the target location prompt area. To facilitate implementation on mobile devices, a simplified rectangle intersection judgment combined with minimum margin detection can be used for conflict determination. For example, the minimum safety distance can be set to 4 to 20 pixels, such as 8 pixels, 12 pixels, or 16 pixels. Its setting is typically related to the terminal screen size, resolution, display scaling ratio, and font size: when the screen is small and there is a lot of text, the safety distance can be appropriately increased; when only icons are displayed or when displayed on a tablet terminal, it can be appropriately decreased. This avoids affecting reading and clicking operations even if objects do not completely overlap but are too close together.

[0088] In some implementations, when a projection conflict occurs, candidate annotation objects with higher decision-making contribution can be prioritized and identified as target annotation objects. Candidate annotation objects with lower contribution can be processed using at least one of the following methods: hiding, shrinking display, and offset display. Hiding can mean completely concealing the object at the current stage; shrinking display can mean retaining only the icon, reducing the icon size, shortening the text content, reducing the font size, or simplifying the border display; offset display can mean moving the object's display position a certain distance from its original mapping position along a preset direction without changing its semantic meaning, and pointing back to the original object reference position via connecting guide lines, anchor lines, or arrows. Offset display is generally suitable for situations where the object has some retention value but the current original display position has a minor conflict; hiding is more suitable for situations where the object's decision-making contribution is low, and even if retained, it is difficult to significantly improve the current prevailing decision; shrinking display is suitable for situations where the object still needs to be retained, but a more simplified presentation is acceptable.

[0089] Furthermore, in some implementations, when the floor identifier corresponding to the next scene switching node changes, a first object corresponding to the entrance of the cross-floor access facility on the current floor and a second object corresponding to the exit of the corresponding cross-floor access facility on the target floor can be extracted from the candidate annotation objects. A cross-floor association display relationship is established between the two, and the first and second objects are highlighted in the visualization annotation results. In cross-floor navigation scenarios, the decision faced by the user on the current floor is not only "whether to enter a certain entrance," but also implicitly includes the continuous guidance requirement of "where to exit from the target floor after entering the entrance, and in which direction to continue." If only the elevator entrance, staircase entrance, or escalator entrance is displayed on the current floor without associating it with the corresponding exit on the target floor, the user may complete the short-term decision of "entering the entrance," but the understanding of the continuity of the subsequent cross-floor path remains insufficient, especially in scenarios with complex cross-floor spatial structures such as hospitals, shopping malls, and transportation hubs, where directional interruptions are more likely to occur. Based on this, in some implementations, when a floor switching is detected in the current navigation stage, the entrance and exit objects of the cross-floor access facilities can be processed in pairs.

[0090] In some implementations, based on the type of cross-floor facility corresponding to the next scene switching node, a first object corresponding to the entrance of the cross-floor access facility on the current floor can be extracted from the candidate labeled objects. This could be an elevator lobby entrance, an escalator entrance, or a stairwell entrance on the current floor. Then, based on the cross-floor connection relationship, a second object corresponding to the first object can be determined from the target floor. This could be an exit of the same elevator shaft on the target floor, an exit of the same escalator on the floor above, or an exit of the same stairwell on the target floor. Subsequently, a cross-floor association display relationship can be established between the two. This association display relationship can be a logical association or further reflected in the interface as paired colors, paired numbers, continuous arrow styles, upper and lower floor comparison labels, or entrance-exit linkage prompts. For example, the current floor entrance object can be displayed as "Enter this elevator," while simultaneously displaying "Leave from the east exit after reaching the 3rd floor" on the side of the interface or in the floor switching prompt box. Alternatively, the escalator entrance object and the target floor exit object can be highlighted with the same color to help users establish a continuous spatial perception before and after crossing floors.

[0091] This further enhances the continuity of annotation information in cross-floor navigation scenarios, allowing users to not only receive entrance selection prompts at the current floor level but also simultaneously establish an expectation of the exit direction on the target floor, thereby reducing cognitive interruptions caused by re-finding directions after crossing floors. The extraction results of the first and second objects mentioned above, as well as the cross-floor association display relationship, can be generated locally by the navigation terminal or calculated and distributed by the server based on the cross-floor connection table and object association table.

[0092] Optional, see Figure 3 A flowchart of a method for determining decision contribution provided in an embodiment of this application includes: S301: Obtain the current direction of travel of the navigation terminal; S302: Determine the angle between each candidate annotation object and the current direction of travel, and the front-back position relationship of each candidate annotation object relative to the current position; S303: In response to a candidate annotation object being located in front of the current position and the angle between it and the current direction of travel being less than a preset angle threshold, increase the decision contribution of the candidate annotation object; in response to a candidate annotation object being located behind the current position or the angle between it and the current direction of travel being greater than a preset angle threshold, decrease the decision contribution of the candidate annotation object.

[0093] Furthermore, in some implementations, the decision contribution is determined solely based on the topological relationship between the object and the target branch, as well as other accessible branches. While this reflects the object's ability to differentiate between branches, the following situations may still exist: some objects, although topologically related to the target branch, may be located behind the user's current direction of travel, or at a large angle behind the user's current line of sight, making them difficult for the user to perceive directly at the current stage; conversely, some objects, although topologically less different, may be located directly in front of or slightly to the side of the user within a visually observable area, thus providing greater immediate guidance value at the current stage. Therefore, in some implementations, the current direction of travel of the navigation terminal and the relative positions of candidate markers to the current location can be further incorporated into the decision contribution determination process to correct the original decision contribution based on direction perception.

[0094] In some implementations, the current orientation of the navigation terminal can be obtained from the attitude angle output by the inertial measurement unit, the heading information output by the magnetic sensor, the direction of travel estimated by the visual odometry, or the local tangent direction of the path centerline near the current position. To avoid sudden changes in orientation caused by short-term jitter, human turning, or positioning drift, the raw orientation data can be smoothed by a short time window, for example, by using the sliding average, weighted average, or median result of the orientation information of the most recent few frames as the current orientation. For mobile pedestrian navigation, the smoothing window can be set to 0.3 seconds to 1.5 seconds; for vehicle terminals or slow-moving navigation scenarios in parking lots, the smoothing window can be appropriately increased to enhance the stability of orientation estimation.

[0095] In some implementations, after obtaining the current direction of travel, the angle between each candidate annotation object and the current direction of travel, as well as the front-back position relationship of each candidate annotation object relative to the current position, can be further determined. The angle can be understood as the deflection angle between the direction of the candidate annotation object's reference point and the reference direction, with the current position as the starting point and the current direction of travel as the reference direction. The front-back position relationship can be understood as comparing the projection of the candidate annotation object's reference point onto the current direction of travel with the current position. If it is located in the forward half-space of the current direction of travel, it can be considered to be in front of the current position; if it is located in the backward half-space, it can be considered to be behind the current position. To improve processing stability, the determination of the front-back position relationship can be based either directly on the direction of the line connecting the current position and the object's reference point, or it can be based on the projection judgment of the path centerline direction near the current position. For more regular corridor and driveway scenarios, using the path centerline direction is often more stable; for open hall or lobby connection areas, using the actual direction of travel of the current terminal better reflects the user's real-time observation direction.

[0096] In some implementations, when a candidate label is located in front of the current position and the angle between it and the current direction of travel is less than a preset angle threshold, the decision contribution of the candidate label can be increased; when a candidate label is located behind the current position, or the angle between it and the current direction of travel is greater than the preset angle threshold, the decision contribution of the candidate label can be decreased. The angle threshold setting here is typically related to the user's natural observation range while moving. For example, in a pedestrian navigation scenario, the threshold can be set between 30 and 90 degrees, such as 45, 60, or 75 degrees; in parking lot driveway scenarios or cart walking scenarios, where the user's line of sight is usually more forward, the threshold can be appropriately reduced; in open shopping mall lobbies or open airport waiting areas, where the user's field of vision is relatively wider, the threshold can be appropriately increased. Increasing or decreasing the decision contribution can be achieved through methods such as weighted multiplier adjustments, fixed score increases or decreases, or grade increases or decreases. For example, for objects located at the front with a small angle, the disambiguation level can be increased by one level or a preset contribution value can be added; for objects located at the rear or with a large angle, the level can be decreased by one level or the contribution value can be reduced by a preset ratio. As long as the corrective effect of "directional perceptibility" on the navigation value at the current stage can be reflected, it is a feasible implementation method of this application.

[0097] In a specific scenario, such as a user located in the main corridor of a hospital outpatient department, about 8 meters ahead, there is a node where a right turn is needed to enter the elevator lobby. Both the "elevator lobby" and "service desk" objects may be related to this node. However, if the "elevator lobby" object is located within a 30-degree angle to the right of the user's current position, while the "service desk" object is located to the left and rear of the current position and requires the user to turn around to see it clearly, then even if their basic topological disambiguation levels are similar, the "elevator lobby" object can receive a higher decision contribution through the aforementioned direction perception correction. Similarly, in an underground parking lot, when a user moves towards the entrance of area A, if the "A3 zone" object is located in a small-angle area to the right of the user's current position, while the "exit" object is located diagonally behind the user, then the former is more suitable as the priority guide for the current stage.

[0098] In the foregoing embodiments, it has been explained that target labeled objects and their display methods can be determined based on decision contribution and projection conflict relationships. Furthermore, in some embodiments, even if the target labeled objects have been determined, the display intensity of different target labeled objects in the current stage should not remain fixed. This is because the remaining path distance between the user and the next scene switching node will continuously shorten. Some objects directly associated with the next scene switching node only need to maintain basic display in the long-distance stage, while they should be more prominent in the near-node stage to help the user complete the final path confirmation. Therefore, in some embodiments, the display salience of target labeled objects can also be dynamically adjusted based on the remaining path distance from the current location to the next scene switching node.

[0099] In some implementations, the remaining path distance can be obtained by projecting the current position onto the current navigation path and accumulating it along the path to the next scene transition node. For more regular polyline paths, the sum of the remaining length of the current path segment and the lengths of several subsequent segments can be used directly; for smoothed centerline paths, the remaining distance can be obtained by accumulating it along the arc length of the centerline. To avoid frequent fluctuations in the remaining distance due to positioning jitter, the results of several consecutive updates can be smoothed. For example, a short window average or a limit on the maximum change in a single update can be used to make the remaining path distance decrease more stably as the user moves forward.

[0100] In some implementations, display saliency can be understood as the degree to which an labeled object stands out in the interface. This saliency can be reflected by one or more display parameters such as font size, icon size, color saturation, border thickness, background brightness, transparency, animation tooltip intensity, guide line brightness, or hierarchy priority. As the remaining path distance decreases, the target labeled object directly associated with the next scene transition node can gradually increase its display saliency, while auxiliary objects with a weaker relationship to the current node but still retained maintain a low or stable display intensity.

[0101] For example, multiple distance intervals can be set based on the remaining path distance: in the long-distance stage, the target node's associated object is displayed as a regular icon or simplified text; when entering the mid-distance stage, the font size can be increased, color contrast enhanced, or a light background added; when entering the short-distance stage, further highlighting, breathing animations, enhanced directional arrows, or a linked top prompt bar can be used. For pedestrian navigation scenarios, the long-distance, mid-distance, and short-distance stages can be set to, for example, greater than 15 meters, 5 to 15 meters, and less than 5 meters, respectively; for parking lot lane scenarios, these can be appropriately relaxed based on user speed and path scale, for example, set to greater than 25 meters, 8 to 25 meters, and less than 8 meters. The above values ​​are only examples, and the specific configuration can be made by the system according to the scenario template.

[0102] For example, a user is in a shopping mall corridor, about 18 meters ahead, when they will enter a three-way intersection. The target path requires the user to turn right into the east wing dining area. In this case, the objects "East Wing" and "Dining Area" can be used as target labels directly associated with the next scene transition node. When the remaining path distance is greater than 15 meters, the system can display both objects as regular labels. When the remaining path distance is reduced to about 10 meters, the font size of the "East Wing" object can be increased and the background contrast enhanced. When the remaining path distance continues to shrink to within 5 meters, the "East Wing" or "Dining Area" objects can be further highlighted, guided by enhanced arrows, or accompanied by breathing animations, allowing the user to more quickly identify the target branch when they actually approach the intersection. Similarly, in an underground parking lot, as the user gradually approaches the entrance to Zone A, the salience of the "A3 Zone" and "Entrance to Zone A" objects can be gradually increased as the remaining path distance decreases, while non-critical objects such as "Exit" and "Charging Parking Area" remain weakly displayed or unchanged.

[0103] Furthermore, in some implementations, constructing the decision corridor solely based on geometric expansion may still introduce some spatial areas that are practically unusable at the current stage. This is particularly true in scenarios with L-shaped corners, corridors obstructed by pillars, shop fence bends, or continuous distribution of obstacles on the outer side of a driveway. Although a sampling location may have a certain lateral expansion space based on its boundary and entrance relationship, this expansion space is not visible from the current location, and the user cannot directly perceive objects in this area at the current stage. Including such areas in the decision corridor could lead to an overly large range for subsequent candidate object extraction, reducing the focus of guidance at the current stage. Therefore, in some implementations, for each sampling location in the forward main guidance zone, a line-of-sight ray pointing from the current location to the corresponding sampling location can be determined. When the line-of-sight ray intersects with a non-transparent barrier boundary, the occlusion truncation component of the corresponding sampling location is determined based on the difference between the visible distance from the current location to the first intersection location and the ray path distance to the corresponding sampling location. Then, the occlusion truncation component is subtracted from the lateral expansion amount corresponding to each sampling location to determine a decision corridor that is closer to the current visible and perceptible range.

[0104] In some implementations, non-transparent barrier boundaries may include wall boundaries, fence boundaries, physical store facade boundaries, impassable obstruction boundaries formed by continuous columns, closed gate boundaries, or other boundaries that block the line of sight and prevent direct passage. A line of sight ray can be understood as a detection line segment emanating from the current position toward the corresponding sampling position. For each sampling position, it can be determined whether the line of sight ray intersects with the aforementioned non-transparent barrier boundary; if an intersection occurs, the length between the current position and the first intersection position is taken as the visible distance, and this visible distance is compared with the ray path distance between the current position and the corresponding sampling position. When the difference is small, it indicates that the area where the sampling position is located is basically within the visible range of the current position, and the occlusion cutoff component can be small or zero; when the difference is large, it indicates that although the sampling position is geometrically expandable, it is actually located behind an obstruction, and the occlusion cutoff component can be increased accordingly. The occlusion cutoff component can be expressed as an absolute subtraction of the original lateral expansion amount, or as a reduction amount determined according to the difference ratio. For a completely obstructed sampling position, the expansion amount in its corresponding direction can even be reduced to zero.

[0105] In this way, the decision corridor will no longer simply cover all geometrically accessible areas, but will instead focus on covering the areas that users can actually observe and use to make judgments at the current stage.

[0106] For example, a user is walking straight along an L-shaped corridor in a shopping mall, with a lateral opening behind a corner. If geometric expansion is based solely on the boundary and entrance relationship, the area near the corner might be prematurely included in the current decision corridor. However, before the user approaches the corner, this opening is obscured by a solid wall corner and is not visible from the current position. In this case, a line-of-sight ray can be emitted from the sampling position of the forward main guide belt near the corner, and the intersection of this ray with the wall boundary can be detected. If it is found that the line of sight intersects with the wall boundary before reaching the sampling position, the occlusion truncation component of the corresponding sampling position can be determined, and the original lateral expansion amount can be deducted, thereby preventing the currently invisible area behind the corner from being prematurely included in the decision corridor. Similarly, in an underground parking lot, if there is a visual obstruction formed by continuous pillars and barriers on one side of the driveway, the system can also reduce the expansion amount on that side in the same way, making the decision corridor more closely match the area in front of the driveway that the user can currently perceive.

[0107] Furthermore, in some implementations, while subtracting the lateral expansion amount solely based on the occlusion truncation component can prevent excessive expansion of the decision corridor, it may also lead to another problem in turning scenarios near the next scene switching node: certain pre-guidance areas that will immediately enter the field of view after passing the next scene switching node and play a crucial role in confirming the next direction may be completely clipped because they are currently behind occlusion, resulting in insufficient forward-looking prompts when the system approaches corners, intersections, or entrances. Therefore, in some implementations, the directional transition amount between the first path direction before the next scene switching node and the second path direction after that node can be further determined. Based on this directional transition amount, the turning contour of the passage boundary in the neighborhood of the next scene switching node, and the positional relationship of the corresponding sampling position relative to the next scene switching node, the turning pre-display component of the corresponding sampling position is determined. Subsequently, the turning pre-display component can be superimposed on the lateral expansion amount corresponding to each sampling position after subtracting the occlusion truncation component, thereby obtaining a decision corridor that balances current visibility and the needs of near-turning pre-guidance.

[0108] In some implementations, the first path direction can be understood as the main forward direction from the current position to the next scene switching node, and the second path direction can be understood as the main extension direction when entering the target passage branch after passing the node. The difference between the two can be characterized by a directional transition amount. This directional transition amount can be recorded in the form of an angle or in the form of a preset directional level, such as a slight turn, a significant turn, and a large-angle turn. The passage boundary turning profile can be understood as the boundary turning shape formed by the wall boundary, guardrail boundary, fence boundary, or lane boundary in the neighborhood of the next scene switching node, such as a corridor corner, a lobby opening, or the outer boundary of a lane fork. By combining the directional transition amount and the boundary turning profile, it is possible to determine which side of the area, which is not yet fully in the field of view but will soon appear after continuing forward, is more worthy of being included in the decision corridor in advance when approaching the next scene switching node.

[0109] In some implementations, the positional relationship of a sampling location relative to the next scene switching node can be characterized by the remaining distance along the path, the path segment number, or the relative order of appearance. Generally, the closer a sampling location is to the next scene switching node, the larger its corresponding turn preview component can be; while sampling locations that are farther away are still mainly constrained by the current visible range.

[0110] For example, when the sampling position is less than 3 meters, 5 meters, or the length of a path segment from the next scene switching node, a certain amount of pre-visualization extension can be added to the target turning side based on the directional transition amount and boundary turning contour; when the sampling position is far away, no pre-visualization extension is added or only a small amount is added. This pre-visualization extension can be represented as a lateral additional width on the target turning side, or as a local fan-shaped, strip-shaped, or boundary-fitting area extending towards the target branch direction after the node. As long as it can reflect a pre-guidance area that "is not fully visible at present, but will immediately enter the field of view after passing the node and will be helpful for the next step of judgment," it can be regarded as the turning pre-visualization component of this application.

[0111] For example, a user is walking straight along the outer corridor of the atrium on the third floor of a shopping mall. The next scene transition node ahead is a connection node that requires a right turn to enter the east wing dining area. If only the occlusion truncation component is used for correction, the area near the inside of the right-turn entrance, currently partially occluded by the corner, may be completely removed. This would make it difficult for the system to present key pre-guidance information such as "east wing" and "dining area" in advance when approaching the node. In this case, the directional transition amount between the first path direction from the current position to the node and the second path direction of the right-turn branch after the node can be analyzed. Combined with the guardrail turning contour and the right entrance opening boundary in the node's neighborhood, the right-turn pre-display component can be determined. Subsequently, after subtracting the occlusion truncation component from the original lateral expansion amount, the directional pre-display component is superimposed at the sampling position near the node, so that the decision corridor gradually expands towards the right entrance neighborhood as the user approaches the node. For example, in an underground parking lot, when a user is about to turn from the main lane into the branch road of area A, the system can also increase the pre-display extension towards the target branch road at the sampling position near the fork based on the transition amount between the main lane direction and the target branch direction, as well as the turning shape of the entrance side boundary. This ensures that the decision corridor does not over-include invalid occlusion areas, and can cover the target entrance neighborhood that will soon enter the field of view when approaching the branch.

[0112] In this way, the formation of the decision corridor is no longer just a simple geometric expansion or a mere clipping of the visible area, but rather achieves a more reasonable balance between current visual constraints and the pre-guidance requirements of nearby nodes. This allows the subsequent selection and display of candidate labels around the decision corridor to better align with the user's actual navigation perception process in complex indoor topological environments. The determination and superposition of the aforementioned occlusion truncation components and steering pre-display components can be completed locally in real-time on the navigation terminal, or the server can send them to the terminal for execution after generating the phased navigation context. As long as the lateral expansion amount can be corrected based on the visual occlusion relationship and the transition relationship between the front and rear directions of nodes, the technical concept of this further implementation method can be realized.

[0113] Based on the same inventive concept, this application also provides a navigation-optimized spatial information visualization annotation system corresponding to a navigation-optimized spatial information visualization annotation method. Since the principle of the system in this application is similar to the navigation-optimized spatial information visualization annotation method described above, the implementation of the system can refer to the implementation of the method, and the repeated parts will not be described again.

[0114] Reference Figure 4 The diagram shown is a schematic of a spatial information visualization and annotation system optimized for navigation provided in an embodiment of this application. The system includes: The data acquisition module 10 is used to acquire indoor space map data and the current and target locations of the navigation terminal; The first processing module 20 is used to determine a navigation path based on the current position and the target position, and identify the next scene switching node along the direction of travel of the navigation path; and to construct a decision corridor corresponding to the current position based on the current position and the next scene switching node. The second processing module 30 is used to extract candidate annotation objects that intersect with the decision corridor or are semantically associated with the next scene switching node from the indoor space map data; and for each candidate annotation object, based on its disambiguation effect on the passage decision at the next scene switching node, determine the decision contribution degree corresponding to each candidate annotation object. The annotation module 40 is used to map each of the candidate annotation objects to the display area corresponding to the current view, determine the target annotation object and the display mode of the target annotation object based on the decision contribution degree and the projection conflict relationship after mapping, and output the visualization annotation result of the target annotation object.

[0115] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A spatial information visualization and annotation method optimized for navigation, characterized in that, include: Acquire indoor space map data, as well as the current location and target location of the navigation terminal; A navigation path is determined based on the current location and the target location, and the next scene switching node is identified along the direction of travel of the navigation path; Based on the current location and the next scene switching node, construct a decision corridor corresponding to the current location; Candidate annotation objects that intersect with the decision corridor or are semantically associated with the next scene switching node are extracted from the indoor space map data; for each candidate annotation object, the decision contribution degree corresponding to each candidate annotation object is determined based on its disambiguation effect on the passage decision at the next scene switching node; Each candidate labeled object is mapped to the display area corresponding to the current view. Based on the decision contribution and the projection conflict relationship after mapping, the target labeled object and the display method of the target labeled object are determined. Output the visual annotation results of the target annotation object.

2. The method according to claim 1, characterized in that, The construction of the decision corridor corresponding to the current location includes: A forward main guide strip is generated along the path centerline between the current position and the next scene switching node; At each sampling position of the forward main guide strip, the forward main guide strip is laterally expanded according to the passage boundary and the entrance of the accessible branch road at the corresponding position; The extended continuous spatial region is defined as the decision corridor.

3. The method according to claim 1, characterized in that, Determining the decision contribution of each of the candidate labeled objects includes: Determine the first topological distance from each candidate labeled object to the target access branch, and the second topological distance to at least one other accessible access branch; Based on the difference between the first topological distance and the second topological distance, the degree of disambiguation of each candidate labeled object for the passage decision is determined; Based on the degree of disambiguation, the decision contribution of each candidate labeled object is determined.

4. The method according to claim 1, characterized in that, The determination of the target annotation object and the display method of the target annotation object includes: Determine the projection area of ​​each candidate annotation object after it is mapped to the current view; In response to the projection regions of at least two candidate annotation objects satisfying a preset conflict condition, the candidate annotation object with higher decision contribution is retained as the target annotation object, and at least one of the following processing methods is performed on the candidate annotation object with lower decision contribution: hiding, shrinking display, and offset display.

5. The method according to claim 1, characterized in that, In response to a change in the floor identifier corresponding to the next scene switching node, the method further includes: Extract the first object corresponding to the entrance of the cross-floor access facility on the current floor and the second object corresponding to the exit of the cross-floor access facility on the target floor from the candidate labeled objects; Establish a cross-layer association display relationship between the first object and the second object; In the visual annotation results, the first object and the second object are highlighted together.

6. The method according to claim 3, characterized in that, The determination of the decision contribution of each of the candidate labeled objects further includes: Obtain the current direction of travel of the navigation terminal; Determine the angle between each candidate annotation object and the current direction of travel, as well as the front-to-back position relationship of each candidate annotation object relative to the current position; In response to a candidate annotation object being located in front of the current position and the angle between the candidate annotation object and the current direction of travel being less than a preset angle threshold, the decision contribution of the candidate annotation object is increased; in response to a candidate annotation object being located behind the current position or the angle between the candidate annotation object and the current direction of travel being greater than a preset angle threshold, the decision contribution of the candidate annotation object is decreased.

7. The method according to claim 4, characterized in that, Also includes: Based on the remaining path distance from the current location to the next scene switching node, adjust the display salience of the target labeled object; As the remaining path distance decreases, the display salience of the target labeled object directly associated with the next scene switching node increases.

8. The method according to claim 2, characterized in that, The step of defining the extended continuous spatial region as the decision corridor includes: For each sampling position of the forward main guide band, determine the line-of-sight ray pointing from the current position to the corresponding sampling position; In response to the intersection of the line-of-sight ray and the non-transparent barrier boundary, the occlusion truncation component of the corresponding sampling position is determined based on the difference between the visible distance from the current position to the first intersection position and the ray path distance to the corresponding sampling position. The decision corridor is determined by subtracting the occlusion truncation component from the lateral expansion amount corresponding to each sampling location.

9. The method according to claim 8, characterized in that, Determining the decision corridor includes: Determine the directional transition amount between the first path direction of the navigation path before the next scene switching node and the second path direction after the next scene switching node; Based on the directional transition amount, the turning contour of the passage boundary in the neighborhood of the next scene switching node, and the positional relationship of the corresponding sampling position relative to the next scene switching node, the turning pre-display component of the corresponding sampling position is determined. The turning pre-display component is used to characterize the pre-guidance area that will be entered into the field of view after passing the next scene switching node. The decision corridor is determined by subtracting the occlusion truncation component from the lateral expansion amount corresponding to each sampling position and superimposing the steering pre-display component.

10. A spatial information visualization and annotation system optimized for navigation, characterized in that, include: The data acquisition module is used to acquire indoor spatial map data as well as the current location and target location of the navigation terminal; The first processing module is used to determine a navigation path based on the current position and the target position, and to identify the next scene switching node along the direction of travel of the navigation path; Based on the current location and the next scene switching node, construct a decision corridor corresponding to the current location; The second processing module is used to extract candidate annotation objects that intersect with the decision corridor or are semantically associated with the next scene switching node from the indoor space map data; and for each candidate annotation object, based on its disambiguation effect on the passage decision at the next scene switching node, determine the decision contribution degree corresponding to each candidate annotation object. The annotation module is used to map each of the candidate annotation objects to the display area corresponding to the current view, and determine the target annotation object and the display mode of the target annotation object based on the decision contribution degree and the projection conflict relationship after mapping; Output the visual annotation results of the target annotation object.

Images