Indoor AR navigation system and method based on sparse spatial map and dynamic path fusion

Abstract

The present invention discloses an indoor AR navigation system and method that integrates a sparse spatial map with a dynamic path, and relates to the field of augmented reality navigation technology. The system comprises a sparse map construction module, a human-computer interaction adaptation module, a landmark marker management module, a dynamic path fusion module, and an AR navigation execution module. The sparse map construction module is used to realize three-dimensional reconstruction and persistent storage of indoor scenes. The human-computer interaction adaptation module is used to solve ground height parameters in real time and dynamically correct the spatial coordinate system. The landmark marker management module is used to generate editable marker objects containing spatial coordinates and attribute data. The dynamic path fusion module is used to topologically associate user-defined paths with algorithm-planned paths. The AR navigation execution module is used to realize path visualization, dynamic path optimization, and AR navigation through a path matching algorithm. The present invention can overcome the shortcomings of existing indoor navigation technologies, such as low positioning accuracy, high mapping cost, and poor dynamic adaptability.

Description

Technical Field

[0001] The present invention relates to the field of augmented reality navigation technology, and in particular to an indoor AR navigation system and method that integrates a sparse spatial map with a dynamic path. Background Art

[0002] AR indoor navigation is one of the key applications of indoor positioning technology. Indoor navigation has long faced three core challenges: low positioning accuracy, high mapping costs, and poor dynamic adaptability. Traditional satellite navigation (such as GPS) is completely ineffective indoors due to signal attenuation. Positioning solutions based on Bluetooth beacons and Wi-Fi fingerprints require the deployment of extensive hardware, resulting in high costs and complex maintenance.

[0003] In recent years, indoor augmented reality navigation technology has rapidly developed. Its core relies on the integration of environmental perception, positioning algorithms, and virtual-reality fusion capabilities. Traditional solutions often use high-precision dense 3D maps (based on LiDAR or photogrammetry) as the navigation foundation. However, dense 3D reconstruction consumes significant computing resources and is difficult to run in real time on mobile devices. While visual SLAM technology can achieve autonomous positioning through cameras and IMUs, it is prone to feature point loss in textureless areas (such as solid-color walls), leading to positioning drift. Summary of the Invention

[0004] In order to solve the above technical problems, the present invention provides an indoor AR navigation system that integrates sparse spatial maps with dynamic paths, including

[0005] The sparse map construction module, based on visual inertial odometry cameras and multi-sensor fusion technology, uses the SLAM algorithm to simultaneously construct 3D sparse maps of indoor scenes and locate mobile devices. It also uses the RANSAC algorithm to remove abnormal feature points and build a map database containing environmental semantic information.

[0006] The human-computer interaction adaptation module, based on the ray detection method and the dynamic calculation technology of the ground height, corrects the three-dimensional space coordinate system in real time. The ground height parameter is updated based on the location clicked by the user.

[0007] The landmark marker management module generates editable marker objects by matching feature points with the coordinates of the sparse spatial map in the cloud. The marker objects contain spatial coordinates, attribute data, and semantic labels.

[0008] The dynamic path fusion module uses a topological association mechanism to fuse user-defined paths with algorithm-planned paths. The path arbitration unit then dynamically optimizes the paths by calculating multiple weighted indicators, including path length, traffic status, and user preference. The path arbitration unit compares and prioritizes paths by calculating basic weights, including path length, traffic obstruction rate, and user preference. The user preference weight increases as the number of times a user selects a path increases.

[0009] The AR navigation execution module integrates the map location detection unit and the path matching algorithm to perform path visualization and dynamic path optimization, thereby performing AR navigation.

[0010] The technical solution further defined in the present invention is:

[0011] Furthermore, in the sparse map construction module, multi-sensor fusion includes timestamp alignment visual error, IMU and depth sensor data fusion; the specific method of using the RANSAC algorithm to remove abnormal feature points is: by randomly selecting the minimum sample point to fit the model, calculating the error between the data point and the model, and setting a basic threshold, the points with an error less than the basic threshold are divided into inliers, and the rest are outliers; repeat the above process, and finally select the model with the most inliers as the optimal model, thereby removing abnormal feature points.

[0012] As mentioned above, in the indoor AR navigation system that integrates sparse spatial maps with dynamic paths, in the human-computer interaction adaptation module, the ray detection method is specifically as follows: the system emits a virtual ray from the starting point of the camera along the specified direction, detects the intersection with the collision body on the ray path, and returns the detection value if the ray intersects with the collision body; the ground height dynamic solution technology is used to solve the height of the mobile device from the ground. Through the ground height solution algorithm within the system, the parameter value obtained by ray detection is combined with the three-dimensional space coordinates to finally solve the relative height of the mobile device and the ground.

[0013] As mentioned above, in the indoor AR navigation system that integrates sparse spatial maps with dynamic paths, in the landmark marker management module, feature point matching is performed by combining ORB with FAST key point detection and BRIEF descriptor. FAST is used to detect corner points in the image by detecting the brightness changes of pixels in a circle around the pixel to determine the corner point. ORB estimates the direction of the key points detected by FAST; BRIEF generates descriptors by comparing the intensities of pixel pairs in the image; ORB improves BRIEF, and the generated descriptor is a binary vector. In AR navigation, feature point matching is performed by matching the ORB feature points in the real map identified by the camera with the pre-stored virtual scene feature points.

[0014] As mentioned above, in the indoor AR navigation system that integrates sparse spatial maps with dynamic paths, in the dynamic path fusion module, the topological association mechanism integrates the user-defined path with the algorithm-planned path in the following way: by topologically associating the nodes in the map with the nodes of the customized path to form a topological graph, the optimal path from the starting point to the end point is found in the topological graph through a search algorithm.

[0015] As mentioned above, in the indoor AR navigation system that integrates sparse spatial maps and dynamic paths, in the AR navigation execution module, the map location detection unit performs AR navigation. After the map matching is completed, the system compares the feature points of the real map with the feature points of the saved sparse spatial map in the spatial coordinate system, and calculates the relative direction and position of the location in the real map based on the relative position.

[0016] The present invention also provides an indoor AR navigation method that integrates a sparse spatial map with a dynamic path, which is applied to the indoor AR navigation system that integrates a sparse spatial map with a dynamic path as described above, and includes the following steps:

[0017] S1. Create a map and collect environmental data synchronously through multiple sensors. Use the SLAM algorithm to build a 3D sparse map of the indoor scene and store feature points and semantic labels in the cloud database.

[0018] S2. Calculate the ground height in real time based on the ray detection method. When the user interaction triggers the coordinate system correction, update the spatial coordinate axis offset and reproject the virtual object.

[0019] S3. Match the current mobile device position in the cloud map and generate an AR location information marker that supports dynamic data binding and modification to set the location required for navigation in the map;

[0020] S4. Between the set navigation locations, the user-defined path and the algorithm-planned path are integrated to generate a multi-objective optimized path through dynamic path fusion and path comparison and optimization algorithm. Based on the starting and ending points selected by the user, the optimal path is dynamically planned by combining the predefined path data with the path data obtained by algorithm analysis.

[0021] S5. Execute AR indoor navigation. The user selects a target location in the current map. The system will generate a visual multi-target optimized path to guide the user to the target location.

[0022] Furthermore, in step S1, feature points in the environment are captured by a camera, including corner points and edges, and the pose, i.e., position and orientation, of the mobile device is estimated in real time in combination with a visual inertial odometry or a pure visual odometry, while a sparse three-dimensional spatial map is constructed.

[0023] The motion model of a mobile device is used to describe the change of posture over time and is expressed as:

[0024] p t =f(p t-1 ,u t )+w t

[0025] Among them, f(p t-1 ,u t ) is the motion function, describing the movement of the mobile device from p t-1 to p t Transformation of u t is the control input, including speed and angular velocity; w t is motion noise;

[0026] The observation model of a mobile device is used to describe the observation of feature points in the environment by the mobile device and is expressed as:

[0027] z t =h(p t ,m)+v t

[0028] Among them, h(p t ,m) is the observation function, describing the position p of the mobile device t The process of observing feature point m; z t is the observed value, i.e. the coordinate of the feature point; v t is the observation noise.

[0029] In the indoor AR navigation method based on the fusion of sparse spatial maps and dynamic paths, in step S3, the feature points of the real map obtained by the camera are compared with the feature points of the cloud map to achieve pose matching. Feature point matching is achieved by minimizing the reprojection error:

[0030]

[0031] Among them, z ti is the time t for feature point m i The observed value, p t is the pose of the mobile device, h(p t ,m i ) is the feature point m i Predicted values ​​projected onto the image plane.

[0032] As described above, in the indoor AR navigation method that integrates sparse spatial maps with dynamic paths, in step S3, based on the matched sparse map, the target location for navigation is set in the space. The system visualizes the location information through the information visualization placement algorithm, generates an AR location information marker in the map scene that supports dynamic binding and modification of attribute data, and sets its position and name.

[0033] The beneficial effects of the present invention are:

[0034] (1) In the present invention, a 3D reconstruction method for indoor scenes is realized by sparse spatial map construction technology, which effectively solves the current high-efficiency and real-time problems of SLAM in complex dynamic environments; for areas with low texture, high dynamics, and large areas without feature points;

[0035] (2) In the present invention, existing AR navigation technologies generally have problems such as low positioning accuracy, low path planning efficiency, and even large deviations. The present invention constructs a visual landmark marking system to generate augmented reality location information markers that support dynamic binding and modification of attribute data. The relative coordinates of areas without significant features are recorded in the system and a virtual landmark and dynamic path fusion mechanism is set for them. The user-defined path is topologically associated with the algorithm-planned path, thereby effectively solving the above problems.

[0036] (3) In the present invention, through the dual data fusion architecture of sparse maps and dynamic paths, combined with the path correction mechanism of human-machine collaboration, the coordinated improvement of navigation accuracy and path flexibility is achieved, which is particularly suitable for multi-story and cross-regional indoor continuous navigation scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 Schematic diagram of the structure of the AR navigation system in Example 1 of the present invention;

[0038] Figure 2 Schematic diagram of the AR navigation method in the second embodiment of the present invention;

[0039] Figure 3 This is a schematic diagram of the construction of a spatial map in the third embodiment of the present invention;

[0040] Figure 4 This is a schematic diagram of the ground height and location setting of the sample area in the third embodiment of the present invention;

[0041] Figure 5 This is a schematic diagram of indoor AR navigation in a sample area in Example 3 of the present invention. DETAILED DESCRIPTION Example 1:

[0042] This embodiment provides an indoor AR navigation system that integrates sparse spatial maps and dynamic paths. Figure 1 As shown, it includes a sparse map construction module, a human-computer interaction adaptation module, a landmark marking management module, a dynamic path fusion module and an AR navigation execution module.

[0043] The sparse map construction module, based on Visual-Inertial Odometry (VIO) cameras and multi-sensor fusion technology, uses the SLAM (Simultaneous Localization and Mapping) algorithm to simultaneously achieve three-dimensional sparse map construction of indoor scenes and mobile device (mobile phone) positioning, and uses the RANSAC (Random Sample Consensus) algorithm to eliminate abnormal feature points and build a persistent map database containing environmental semantic information.

[0044] Multi-sensor fusion refers to improving the robustness of SLAM mapping by aligning visual errors, IMU (Inertial Measurement Unit) and depth sensor data through timestamps.

[0045] RANSAC abnormal feature point removal refers to fitting the model by randomly selecting the minimum sample point, calculating the error between the data point and the model, and setting a basic threshold. Points with errors less than the basic threshold are classified as inliers, and the rest are outliers; repeating the above process, and finally selecting the model with the most inliers as the optimal model, thereby eliminating abnormal feature points and achieving robust fitting of the data.

[0046] The sparse map construction module uses the VIO camera and multi-sensor fusion to solve the device posture. According to the movement of the device, it synchronously scans the feature points of the map in the real environment, records them in the spatial map, eliminates errors, and finally uploads them to the cloud database.

[0047] The human-computer interaction adaptation module, based on the ray detection method and the dynamic calculation technology of the ground height, corrects the three-dimensional space coordinate system in real time. The ground height parameter is updated by the location clicked by the user.

[0048] The ray detection method refers to the system emitting a virtual ray from the starting point of the camera in a specified direction, detecting the intersection of the ray path with the collision body, and returning the detection value if the ray intersects the collision body; the ground height solution technology is used to solve the height of the mobile device from the ground. Through the ground height solution algorithm within the system, the parameter value obtained by ray detection is combined with the three-dimensional space coordinates to finally solve the relative height of the mobile device and the ground.

[0049] The human-computer interaction adaptation module helps users control the system settings through UI interface buttons. When the user clicks the button related to editing the ground height in the system menu, the system will request to open the device camera, and then the system will scan the environment map. After the scan is completed, the system will emit a virtual ray to detect the intersection of the ray path with the collision body to set the height of the device from the ground to determine the initial height value.

[0050] Ground height update triggering methods include: the user clicks the ground height setting button, the system generates an AR entity, the user clicks it to place it in the scene, and the system detects the height difference between the device and it to set the ground height.

[0051] The landmark marker management module generates an editable marker object by matching feature points with the coordinates of the sparse spatial map in the cloud. The marker object contains spatial coordinates, attribute data and semantic labels.

[0052] Feature point matching is performed by combining Oriented FAST and Rotated BRIEF (ORB) with FAST (Features from Accelerated Segment Test) key point detection and BRIEF (Binary Robust Independent Elementary Features) descriptors. FAST is used to detect corner points in an image by measuring the brightness changes of pixels around the pixel. ORB estimates the orientation of the key points detected by FAST, making them rotation-invariant.

[0053] BRIEF is a method based on binary descriptors, which generates descriptors by comparing the intensities of pixel pairs in an image. ORB improves BRIEF to make it more suitable for rotation invariance. The generated descriptor is a binary vector with fast calculation speed and small storage space. In AR navigation, feature point matching is achieved by matching ORB feature points in the real map identified by the camera with pre-stored feature points of the virtual scene.

[0054] The coordinate alignment methods include: online relocalization technology based on the ORB-SLAM3 framework, which achieves real-time matching of the current device posture with the cloud map through virtual and real feature point matching; and the semantic segmentation algorithm U-Net is used to classify and annotate the attribute data of the marked object.

[0055] The landmark management module enables users to add location landmark information to the cloud map through the interface with the cloud database, integrate feature point information with three-dimensional coordinates, and realize the integration of landmark marking and spatial mapping. Figure 1The core of landmark management is feature point matching, which is achieved by combining ORB with FAST key point detection and BRIEF descriptors to achieve high-precision and high-efficiency feature point matching.

[0056] The dynamic path fusion module uses a topological association mechanism to fuse user-defined paths with algorithm-planned paths, and achieves dynamic optimization by calculating multiple weighted indicators (path length, traffic status, and user preference) through a path arbitration unit. The path arbitration unit compares and optimizes paths by calculating basic weight indicators, including path length weight, traffic obstruction rate weight, and user preference weight. The user preference weight increases as the user selects the path multiple times.

[0057] The specific method of the topological association mechanism to integrate the user-defined path with the algorithm-planned path is as follows: by topologically associating the nodes in the map with the nodes of the custom path to form a topological graph, the optimal path from the starting point to the end point is found in the topological graph through the search algorithm.

[0058] The topological association mechanism includes: users can place virtual markers in the virtual map and customize paths by connecting virtual markers; the nodes of the path map are composed of three-dimensional coordinates, traffic status markers and dynamic weight parameters; user-defined paths are injected into the map through a priority evaluation model, and the priority weight coefficient is higher than that of the algorithm-planned path; customized paths with repeated starting points are judged by comparing weight values.

[0059] Dynamic Path Fusion combines user-defined paths with paths generated by the system algorithm based on weights calculated by the Path Arbitration Unit. Custom Paths, specifically the system's custom path feature, allows you to create a path that meets your needs based on pre-set landmarks within the map.

[0060] The AR navigation execution module integrates the map location detection unit and the path matching algorithm to perform path visualization and dynamic path optimization to realize AR navigation.

[0061] During AR navigation, after the map matching is completed, the map location detection unit compares the feature points of the real map with the feature points of the saved sparse spatial map in the spatial coordinate system, and calculates the relative direction and position of the location in the real map based on the relative position.

[0062] A path matching algorithm specifically refers to a navigation path calculation and display algorithm that combines virtual and real-world data. This algorithm imports optimal path data and uses it to integrate the virtual path with the real world. This algorithm, based on virtual path data, combines the virtual path with reality, displays it on a map via the phone's camera, and performs real-time detection and dynamic adjustments to the path. In this embodiment, a dynamic window approach (DWA) is used to sample the user's current location and speed, generate speed samples, and optimize the path for real-time obstacle avoidance.

[0063] Users start AR navigation by scanning the real map and matching it with the cloud. Navigation ends when the user is about to reach the navigation destination. Example 2:

[0064] This embodiment provides an indoor AR navigation method that integrates a sparse spatial map with a dynamic path, which is applied to the indoor AR navigation system that integrates a sparse spatial map with a dynamic path proposed in the first embodiment. Figure 2 As shown, the following steps are included:

[0065] S1. Create a map and synchronously collect environmental data through multiple sensors. Use the SLAM algorithm to build a three-dimensional sparse map of the indoor scene and store feature points and semantic labels in the cloud database.

[0066] The process of creating and saving a sparse map is as follows:

[0067] Creating a sparse spatial map mainly relies on SLAM technology, which uses a camera to capture feature points in the environment (such as corners, edges, etc.), and combines visual inertial odometry or pure visual odometry (VO) to estimate the pose (position and orientation) of the mobile device in real time, while building a sparse three-dimensional spatial map.

[0068] The sparse spatial map constructed in this embodiment ensures the accuracy and real-time performance of the map by optimizing feature point matching and pose estimation algorithms, while supporting persistent storage, providing AR applications with stable environmental understanding and positioning capabilities.

[0069] The motion model of a mobile device describes the change of its posture over time and is expressed as:

[0070] p t =f(p t-1 ,u t )+w t

[0071] Among them, f(p t-1 ,u t ) is the motion function, describing the movement of the mobile device from p t-1 to p t Transformation of ut is the control input, including speed and angular velocity; w t is the motion noise (assumed to be Gaussian noise).

[0072] The observation model of a mobile device describes the observation of feature points in the environment by the mobile device and is expressed as:

[0073] z t =h(p t ,m)+v t

[0074] Among them, h(p t ,m) is the observation function, describing the position p of the mobile device t The process of observing feature point m; z t is the observed value (the coordinate of the feature point); v t is the observation noise.

[0075] S2. Calculate the ground height in real time based on the ray detection method. When the user interaction triggers the coordinate system correction, update the spatial coordinate axis offset and reproject the virtual object.

[0076] The process of solving the ground height is as follows:

[0077] Spatial perception technology based on user interaction behavior, combined with the initial value of the ground height stored in the variable, and using ray detection to detect the location where the user clicks, solves the ground height parameters in real time and dynamically corrects the spatial coordinate system.

[0078] S3. Match the current mobile device position in the cloud map and generate an AR location information marker that supports dynamic data binding and modification to set the location required for navigation in the map.

[0079] The process of matching the current device pose is as follows:

[0080] The real-world map feature points obtained by the camera are compared with the cloud map feature points to achieve pose matching. Feature point matching is a key step in SLAM and is achieved by minimizing the reprojection error:

[0081]

[0082] Among them, z ti is the time t for feature point m i The observed value, p t is the pose of the mobile device, h(p t ,m i ) is the feature point m i Predicted values ​​projected onto the image plane.

[0083] The process of generating an augmented reality place information marker that supports dynamic binding and modification of attribute data is as follows:

[0084] Based on the matching sparse map, the desired navigation target location is set within the space. The system uses an information visualization placement algorithm to visualize the location information, generate AR location information markers in the map scene that support dynamic binding and modification of attribute data, and set their positions and names, providing users with intuitive location management functions. Users can set locations for multiple saved maps.

[0085] S4. Between the set navigation locations, the user-defined path and the algorithm-planned path are integrated, and a multi-objective optimized path is generated through dynamic path fusion and path comparison and optimization algorithm.

[0086] The process of generating a multi-objective optimization path is as follows:

[0087] Create custom routes with dynamic route creation and management. This provides users with an intuitive and flexible way to define connections between locations. Innovations lie in dynamic route creation, location-based route definition, data persistence and real-time updates, and modular design. This route creation method is designed to efficiently and conveniently solve navigation problems in large areas without feature points.

[0088] The navigation path is calculated using a dynamic path fusion and path comparison algorithm. The system dynamically plans an optimal path based on the user's selected starting and ending points, combining predefined path data with path data analyzed by the algorithm. During the path planning process, the system avoids duplicate paths and dead loops to ensure path validity.

[0089] S5. Execute AR indoor navigation. The user selects a target location in the current map. The system will generate a visual multi-target optimized path to guide the user to the target location.

[0090] The process of AR navigation is as follows:

[0091] The system renders the navigation path in real time, overlaying the virtual path onto the displayed scene. The path's starting and ending points are dynamically adjusted based on the user's location and the target location. During navigation, the system detects the user's distance from the target location in real time. Based on a path priority evaluation model, a path matching algorithm is used to achieve precise in-scene navigation and dynamically optimize the navigation path. Navigation is complete when the user approaches the target location (the distance to the target point is less than the preset distance). Example 3:

[0092] This example is for AR indoor navigation in a large-scale scenario, in a university library, using the following steps:

[0093] Step 1: Figure 3 As shown, first use the map creation function in the menu to scan the characteristic area in the library, not limited to Figure 3 The area shown.

[0094] Step 2: Figure 4 As shown, click the location setting button in the menu to match the map saved in the cloud based on the feature points. After the match is successful, first set the ground height within the map, and then set the target location within the map. If necessary, you can customize special paths between locations. The example shown is the second floor of the university library.

[0095] Step 3: Start navigation. Click the Start Navigation button in the menu. The system will match the feature points of the user's area with the feature points of the sparse map in the cloud. After the match is successful, the system will select the target location to start indoor AR navigation. After the navigation starts, the system will determine the user's location based on the proposed function and guide the user to the destination using a visual virtual path. Figure 5 The figure shows the navigation path from Borrowing Area 1 to Borrowing Area 2 on the second floor of the university library. If the user deviates from the navigation path, the system will recalculate the optimal path based on the user's location coordinates and posture.

[0096] The present invention realizes a three-dimensional reconstruction method of indoor scenes through sparse spatial map construction technology, which effectively solves the current high-efficiency and real-time problems of SLAM in complex dynamic environments. For low-texture, high-dynamic, large-area areas without feature points, existing AR navigation technologies generally have problems of low positioning accuracy, low path planning efficiency and even large deviations. The present invention constructs a visual landmark marking system to generate augmented reality location information marks that support dynamic binding and modification of attribute data, records the relative coordinates of areas without significant features in the system and sets a virtual landmark and dynamic path fusion mechanism for them, and topologically associates user-defined paths with algorithm-planned paths. These two methods effectively solve the above problems.

[0097] The present invention achieves a coordinated improvement in navigation accuracy and path flexibility through a dual data fusion architecture of sparse maps and dynamic paths, combined with a human-machine collaborative path correction mechanism. It is particularly suitable for multi-floor, cross-regional indoor continuous navigation scenarios.

[0098] In addition to the above embodiments, the present invention may also have other implementations. Any technical solution formed by equivalent replacement or equivalent transformation falls within the protection scope of the present invention.

Claims

1. An indoor AR navigation system that integrates sparse spatial maps with dynamic paths, characterized by: include The sparse map construction module, based on visual inertial odometry cameras and multi-sensor fusion technology, uses the SLAM algorithm to simultaneously construct 3D sparse maps of indoor scenes and locate mobile devices. It also uses the RANSAC algorithm to remove abnormal feature points and build a map database containing environmental semantic information. The human-computer interaction adaptation module, based on the ray detection method and the dynamic calculation technology of the ground height, corrects the three-dimensional space coordinate system in real time. The ground height parameter is updated based on the location clicked by the user. The landmark marker management module generates editable marker objects by aligning feature point matching with the coordinates of the cloud-based sparse spatial map. The marker objects contain spatial coordinates, attribute data, and semantic labels. Feature point matching is performed by combining ORB with FAST key point detection and BRIEF descriptors. FAST is used to detect corner points in an image by detecting changes in pixel brightness in a circle around the pixel. ORB estimates the direction of the key points detected by FAST. BRIEF generates descriptors by comparing the intensities of pixel pairs in the image. ORB improves BRIEF and generates a binary vector descriptor. In AR navigation, feature point matching is performed by matching the ORB feature points in the real map identified by the camera with the pre-stored feature points of the virtual scene. The dynamic path fusion module uses a topological association mechanism to fuse user-defined paths with algorithm-planned paths. The path arbitration unit then calculates multiple weighted indicators for dynamic optimization, including path length, traffic status, and user preferences. The path arbitration unit compares and prioritizes paths by calculating basic weight indicators, which include path length weight, traffic obstruction rate weight, and user preference weight. The user preference weight increases as the number of times a user selects the path increases. The AR navigation execution module integrates the map location detection unit and the path matching algorithm to perform path visualization and dynamic path optimization, thereby performing AR navigation.

2. The indoor AR navigation system integrating sparse spatial maps and dynamic paths according to claim 1 is characterized in that: In the sparse map construction module, multi-sensor fusion includes timestamp alignment of visual errors, IMU, and depth sensor data fusion. The specific method of using the RANSAC algorithm to remove abnormal feature points is as follows: by randomly selecting the minimum sample point to fit the model, calculating the error between the data point and the model, and setting a basic threshold, points with an error less than the basic threshold are classified as inliers, and the rest are classified as outliers. Repeat the above process and finally select the model with the most inliers as the optimal model, thereby eliminating abnormal feature points.

3. The indoor AR navigation system integrating sparse spatial maps and dynamic paths according to claim 1, characterized in that: In the human-computer interaction adaptation module, the ray detection method is specifically as follows: the system emits a virtual ray from the starting point of the camera along the specified direction, detects the intersection with the collision body on the ray path, and returns the detection value if the ray intersects with the collision body; the ground height dynamic solution technology is used to solve the height of the mobile device from the ground. Through the ground height solution algorithm in the system, the parameter value obtained by ray detection is combined with the three-dimensional space coordinates to finally solve the relative height of the mobile device and the ground.

4. The indoor AR navigation system integrating sparse spatial maps and dynamic paths according to claim 1, characterized in that: In the dynamic path fusion module, the topology association mechanism fuses the user-defined path with the algorithm-planned path in the following way: topologically associating the nodes in the map with the nodes of the customized path to form a topology graph, and then finding the optimal path from the starting point to the end point in the topology graph through a search algorithm.

5. The indoor AR navigation system integrating sparse spatial maps and dynamic paths according to claim 1, characterized in that: In the AR navigation execution module, when the map location detection unit is performing AR navigation, after the map matching is completed, the system compares the feature points of the real map with the feature points of the saved sparse spatial map in the spatial coordinate system, and calculates the relative direction and position of the location in the real map based on the relative position.

6. An indoor AR navigation method integrating a sparse spatial map with a dynamic path, applied to the indoor AR navigation system integrating a sparse spatial map with a dynamic path as claimed in any one of claims 1 to 5, characterized in that: The following steps are involved: S1. Create a map and collect environmental data synchronously through multiple sensors. Use the SLAM algorithm to build a 3D sparse map of the indoor scene and store feature points and semantic labels in the cloud database. S2. Calculate the ground height in real time based on the ray detection method. When the user interaction triggers the coordinate system correction, update the spatial coordinate axis offset and reproject the virtual object. S3. Match the current mobile device's pose to the cloud map and generate AR location information markers that support dynamic data binding and modification, so as to set the required navigation location on the map. Compare the real-world map feature points obtained by the camera with the cloud map feature points to achieve pose matching. Feature point matching is achieved by minimizing the reprojection error. Among them, z ti is the time t for feature point m i The observed value, p t is the pose of the mobile device, h(p t ,m i ) is the feature point m i The predicted value projected onto the image plane; S4. Between the set navigation locations, the user-defined path and the algorithm-planned path are integrated to generate a multi-objective optimized path through dynamic path fusion and path comparison and optimization algorithm. Based on the starting and ending points selected by the user, the optimal path is dynamically planned by combining the predefined path data with the path data obtained by algorithm analysis. S5. Execute AR indoor navigation. The user selects a target location in the current map. The system will generate a visual multi-target optimized path to guide the user to the target location.

7. The indoor AR navigation method integrating sparse spatial maps and dynamic paths according to claim 6, characterized in that: In step S1, feature points in the environment are captured by a camera, including corner points and edges, and the pose, i.e., position and orientation, of the mobile device is estimated in real time in combination with a visual inertial odometry or a pure visual odometry, while a sparse three-dimensional spatial map is constructed. The motion model of a mobile device is used to describe the change of posture over time and is expressed as: p t =f(p t-1 ,u t )+w t Among them, f(p t-1 ,u t ) is the motion function, describing the movement of the mobile device from p t-1 to p t Transformation of u t is the control input, including speed and angular velocity; w t is motion noise; The observation model of a mobile device is used to describe the observation of feature points in the environment by the mobile device and is expressed as: z t =h(p t ,m)+v t Among them, h(p t ,m) is the observation function, describing the position p of the mobile device t The process of observing feature point m; z t is the observed value, i.e. the coordinate of the feature point; v t is the observation noise.

8. The indoor AR navigation method integrating sparse spatial maps and dynamic paths according to claim 6, characterized in that: In step S3, according to the matched sparse map, based on this map, the target location for navigation is set in the space. The system visualizes the location information through the information visualization placement algorithm, generates an AR location information marker in the map scene that supports dynamic binding and modification of attribute data, and sets its position and name.

Images