An ar adaptive visualization method and system based on user intention and spatial relation learning

Abstract

The application belongs to the technical field of AR human-computer interaction, and specifically discloses an AR adaptive visualization method and system based on user intention and spatial relationship learning, which collects multi-modal data of a user, calculates the Euclidean distance of the user and a virtual element, an angle of sight, and divides different regions; three types of task intentions are identified through gazing and the like to generate an intention weight vector; a three-dimensional fusion interaction correlation degree model of "intention-space-semantic" is constructed, distance, sight, and semantic weight are fused, and a virtual information display level is dynamically controlled; a distance difference value is used to calculate a transition time length, linear interpolation is used to realize LOD smooth switching, and double threshold anti-shake, shelter relocation, and local detail enhancement of a gazing area are matched. The application solves problems such as visual shelter, information overload, excessive information at a long distance, insufficient details at a short distance, neglect of user intention, and frequent flickering and jumping of a picture, and is suitable for scenes such as intelligent factory production line simulation review, equipment state inspection, and industrial operation and maintenance auxiliary decision-making.

Description

Technical Field

[0001] This invention belongs to the field of augmented reality human-computer interaction technology, specifically involving an AR adaptive visualization method and system based on user intent and spatial relational science, which is suitable for the adaptive presentation of augmented reality information in complex industrial scenarios such as intelligent factory production line simulation review, equipment status inspection and industrial operation and maintenance auxiliary decision-making. Background Technology

[0002] Augmented Reality (AR) technology overlays computer-generated virtual information onto a user's view of the real environment, creating a seamless interactive experience. In recent years, with the increasing maturity of head-mounted AR devices (such as HoloLens and Magic Leap), AR technology has been increasingly widely applied in smart manufacturing fields such as production line simulation reviews in smart factories, equipment status inspections, and industrial assembly guidance. By anchoring virtual content such as equipment status parameters, process guidance information, and 3D models to corresponding locations on physical equipment or in space, AR technology provides on-site personnel with an intuitive way to obtain information, thereby significantly improving operational efficiency and decision-making quality.

[0003] However, existing AR information visualization methods still have the following technical shortcomings in practical applications facing complex industrial scenarios:

[0004] First, fixed coordinate anchoring leads to visual occlusion and information overload. Existing AR systems typically fix virtual information panels to the relative spatial coordinates of the physical device. When users move or change their perspective in complex environments such as workshops, the large number of stacked virtual information panels can easily cause severe visual conflicts with real objects, resulting in the occlusion of key information. At the same time, the disordered stacking of information in space further increases the user's visual burden, causing information overload and making it difficult for users to quickly filter out the relevant information needed for the current task from a large amount of fragmented information.

[0005] Secondly, the level of detail is rigid, lacking semantic-level adaptation based on spatial relationships. Existing AR systems generally use Level of Detail (LOD) technology to optimize rendering performance, but its application mainly focuses on reducing the number of geometric faces of the model based on spatial distance to save rendering power. It fails to dynamically adjust the semantic density and content hierarchy of information presentation according to the spatial relationship between the user and the device. Whether the user is performing a macro overview at 3 meters away or a detailed inspection at 0.3 meters away, the system often presents the same information panel, resulting in overly dense visual information at long distances and insufficient detail at close range. In addition, existing LOD switching schemes based on fixed thresholds lack a smooth transition mechanism. When the user moves near the boundary of the area, it is easy to cause frequent screen flickering and jumps, which seriously affects the user's visual experience and task execution efficiency.

[0006] Finally, neglecting user intent leads to passive and rigid information presentation. Existing AR systems struggle to effectively perceive a user's current task intent through implicit behaviors (such as gaze patterns and head movements), such as whether the user is performing a global overview, detailed inspection, or searching for a specific device. The lack of proactive task-oriented information presentation results in the system presenting information in the same way regardless of the operational scenario, failing to proactively adjust the display priority and presentation method based on the user's current task needs. This severely impacts the efficiency of virtual review and on-site operations.

[0007] In summary, how to simultaneously address visual occlusion and information overload in augmented reality environments, achieve semantic-level adaptation based on spatial relationships, proactively perceive user intent and dynamically adjust information presentation accordingly, and ensure a smooth and seamless display transition process has become a pressing technical challenge in this field.

[0008] In view of this, this invention is hereby proposed. Summary of the Invention

[0009] The purpose of this invention is to overcome the shortcomings of the prior art and provide an AR adaptive visualization method and system based on user intent and spatial relational science. It is mainly used to solve the technical problems existing in the current augmented reality information visualization methods, such as visual occlusion and information overload, rigid detail levels and lack of semantic-level adaptation based on spatial relations, passive rigidity of information presentation due to ignoring user intent, and frequent screen flickering and jumping when switching display levels.

[0010] The objective of this invention is achieved through the following technical solution:

[0011] On the one hand, this invention provides an AR adaptive visualization method based on user intent and spatial relational science, comprising the following steps:

[0012] Step 1, Multimodal Data Acquisition: Real-time acquisition of the user's spatial state data and behavioral data through the sensors of the augmented reality device; the spatial state data includes at least the user's head center position, head posture, and gaze direction; the behavioral data includes at least gaze duration, head angular velocity, voice navigation commands, and environmental alarm signals;

[0013] Step 2, Spatial Relationship Calculation and Region Division: Calculate the Euclidean distance between the user's head center position and the spatial position of each virtual element in the scene. The unit vector for determining the user's head gaze direction is determined based on the head posture and the gaze direction. And determine the unit vector pointing from the center of the user's head to the center of the target virtual element based on the spatial position of the user's head center and the spatial position of the virtual element. And then according to and Calculate the angle of view Based on the Euclidean distance Virtual elements are divided into interactive areas, attention areas, and irrelevant areas, based on the aforementioned viewing angle. The virtual element is divided into a focus area, a surrounding area, and an ignored area to obtain the spatial relationship status of the virtual element relative to the user.

[0014] Step 3, User Intent Recognition: Based on the gaze duration and the gaze angle... The system uses head angular velocity, voice navigation commands, and alarm trigger events to identify one or more user intent states, such as detailed analysis, overview browsing, or navigation addressing, and generates a user intent weight vector. The user intent weight vector It includes weighted components corresponding to detailed analysis intent, overview browsing intent, and navigation addressing intent, respectively, and the sum of each component is 1;

[0015] Step 4: Calculate the interaction relevance: Obtain the preset semantic importance weight for each virtual element. According to the user intent weight vector Determine the first weighting coefficient corresponding to the distance weight. The second weighting coefficient corresponding to the angle weight The third weighting coefficient corresponding to the semantic importance weight Based on the Euclidean distance Determine distance weights Based on the aforementioned angle of view Determine angle weights The distance weight With the Euclidean distance The angle weight increases and decreases. With the angle of the line of sight Increases and decreases; the interaction correlation degree of each virtual element at the current moment is calculated according to the following formula. :

[0016]

[0017] Step 5, Adaptive Display Strategy Decision: Based on the aforementioned interaction correlation. The comparison results with preset thresholds determine the display detail level and information panel expansion degree of each virtual element; simultaneously, based on the Euclidean distance... and the degree of interaction Adjust the number of model faces and information display density of virtual elements;

[0018] Step 6, Smooth Rendering and Occlusion Relocation: When the display layer of virtual elements changes, the transition time is calculated based on the difference in Euclidean distance before and after the user moves. and during the transition time The number of model facets and the degree of expansion of the information panel are smoothly and gradually changed. At the same time, it detects whether the virtual elements and their information panels are obscured by real entities. When the preset trigger conditions are met, the panel coordinates are recalculated and the control panel is moved to the visible area through the easing function to eliminate screen jumps.

[0019] Furthermore, in step 2, the angle of view... The calculation formula is:

[0020]

[0021] In the formula, This is a unit vector representing the direction of the user's head gaze. The vector pointing from the center of the user's head to the center of the target virtual element.

[0022] Furthermore, in step 2, based on the Euclidean distance... The specific method for dividing virtual elements into interactive regions, attention regions, and irrelevant regions is as follows: when the Euclidean distance... When the distance is less than the first distance threshold, it is determined to be an interactive area; when the Euclidean distance is less than the first distance threshold, it is determined to be an interactive area. When the distance is greater than or equal to the first distance threshold and less than the second distance threshold, it is determined to be a region of concern. If the distance is greater than or equal to the second distance threshold, it is determined to be an irrelevant region;

[0023] Based on the aforementioned line of sight angle The specific method for dividing virtual elements into a focused area, a surrounding area, and an ignored area is as follows: when the angle of view... When the angle is less than the first angle threshold, it is determined to be a focal area; when the angle of view is... When the angle is greater than or equal to the first angle threshold and less than the second angle threshold, it is determined to be the surrounding area. If the angle is greater than or equal to the second angle threshold, it is determined to be an ignored region.

[0024] Furthermore, in step 3, the method for identifying the user's intent is as follows: when the user's gaze is fixed on a virtual element and remains there for more than a first time threshold, it is determined to be an intent for detailed analysis and given a high analysis weight; when the user's head angular velocity is higher than a preset angular velocity threshold and there is no gaze behavior exceeding the first time threshold, it is determined to be an intent for overview browsing and given a high browsing weight; when the system receives a navigation voice command or triggers an emergency alarm event, it is determined to be an intent for navigation addressing and given a high navigation weight.

[0025] Furthermore, in step 4, the first weighting coefficient Second weighting coefficient and the third weighting coefficient The user intent weight vector The specific driving method is as follows: when a detailed analysis intent is identified, the first weighting coefficient is increased. and the second weighting coefficient The value of the third weighting coefficient is increased when an overview browsing intent or navigation addressing intent is identified. The value of .

[0026] Furthermore, in step 5, when the angle of view... When the gaze is in the focal area and the gaze duration exceeds a set time threshold, a high-precision sub-mesh is divided within a set radius around the gaze point, and the texture resolution of the high-precision sub-mesh is multiplied by an enhancement factor. At the same time, edge highlighting rendering is performed to enhance local details.

[0027] Furthermore, in step 6, the transition time The calculation formula is:

[0028]

[0029] In the formula, The preset time coefficient, and These represent the Euclidean distances before and after the user's movement.

[0030] Further, in step 6, during the transition time Within this model, the number of facets is smoothly varied using the following linear interpolation formula:

[0031]

[0032] In the formula, For the number of facets in the model, and These represent the number of target facets under two adjacent LOD standards. For the transition start time, The transition end time, This is the current time.

[0033] Furthermore, in step 6, the preset triggering condition includes a location threshold. and time update threshold When the absolute value of the user's head movement exceeds the position threshold. Or the user's head remains still for more than the time update threshold. Furthermore, if no new instructions are input, the panel coordinates will be recalculated.

[0034] On the other hand, the present invention also provides an AR adaptive visualization system based on user intent and spatial relational science, for performing the above-described method, comprising:

[0035] The data acquisition module is used to collect users' spatial status data and behavioral data in real time;

[0036] The spatial relationship calculation module, connected to the data acquisition module, is used to calculate the Euclidean distance between the center position of the user's head and the spatial position of each virtual element in the scene. Angle with line of sight And based on the Euclidean distance Virtual elements are divided into interactive areas, attention areas, and irrelevant areas, based on the aforementioned viewing angle. The virtual elements are divided into a focus area, a surrounding area, and an ignored area;

[0037] The user intent recognition module, connected to the data acquisition module, is used to identify the user intent state based on the gaze duration, head angular velocity, voice navigation commands, and alarm trigger events, and generate a user intent weight vector. ;

[0038] The interaction correlation decision module, connected to both the spatial relationship calculation module and the user intent recognition module, is used to obtain the preset semantic importance weight for each virtual element. According to the user intent weight vector Determine the first weighting coefficient corresponding to the distance weight. The second weighting coefficient corresponding to the angle weight The third weighting coefficient corresponding to the semantic importance weight According to the Euclidean distance Determine distance weights Based on the aforementioned angle of view Determine angle weights and the distance weight The angle weights and the semantic importance weight A weighted sum is performed to calculate the interaction correlation degree of each virtual element at the current moment. The interaction correlation decision module is also used to determine the relationship based on the following: The comparison result with the preset threshold is displayed as a strategy decision signal;

[0039] An adaptive visualization rendering module, connected to the interaction correlation decision module, is used to adjust the display detail level, model facet number, texture resolution, and information panel expansion degree of the virtual elements according to the decision signal.

[0040] The occlusion detection and layout repositioning module is connected to the data acquisition module and the adaptive visualization rendering module. It is used to detect whether the virtual elements and their information panels are occluded by real entities, and to recalculate the display position of the information panels when occlusion occurs.

[0041] The smooth transition control module, connected to the interaction correlation decision module, the occlusion detection and layout repositioning module, and the adaptive visualization rendering module, is used to calculate the transition time based on the Euclidean distance difference before and after the user moves when switching display levels. It also performs a smooth gradient and an easing function to smoothly move the panel when it is repositioned.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] 1. This invention constructs a three-dimensional fusion interaction correlation model of "intent-space-semantics" to achieve multi-factor collaborative decision-making through interaction correlation, so that information presentation can dynamically adapt to user intent and spatial attention. Compared with the existing technology that only adjusts LOD based on a single factor, it achieves a leap from geometric level to semantic level. By using multimodal implicit behavior fusion to determine user intent and drive the dynamic adjustment of weighting coefficients, information presentation is transformed from passive fixed type to task-oriented active adaptation type. It effectively solves the technical problems of excessive information density at long distances and insufficient details at close distances, and the passive and rigid information presentation caused by ignoring user intent.

[0044] 2. This invention achieves a refined description of the spatial relationship between the user and virtual elements through dual-threshold spatial region division (distance dimension: interactive area / attention area / irrelevant area; gaze dimension: focused area / peripheral area / ignored area), providing accurate spatial weight input for interaction correlation calculation. On this basis, when the gaze angle is in the focused area and the gaze duration exceeds a set time threshold, local detail enhancement is triggered, that is, high-precision sub-mesh is divided around the gaze point and the texture resolution is improved. Combined with edge highlight rendering, it simulates the high-precision visual perception characteristics of the human eye's fovea, so that high-precision rendering resources are focused only on the core area currently being gazed at by the user, effectively reducing the overall rendering overhead while ensuring visual experience.

[0045] 3. This invention achieves continuous and smooth changes in the number of model faces and the extent of information panel expansion during LOD level switching by calculating the transition time based on the difference in Euclidean distance before and after user movement and using linear interpolation for smooth transition. This eliminates the frequent flickering and jumps in the screen caused by traditional fixed threshold switching. Combined with a dual-threshold anti-shake mechanism of position threshold and time update threshold, it effectively filters out invalid triggers caused by slight head movements of the user. The easing function makes the panel repositioning process smooth and continuous, significantly improving visual comfort and interaction coherence. At the same time, the occlusion detection and panel repositioning mechanism ensure that virtual information is always in a readable position in complex physical environments, fundamentally solving the problems of visual occlusion and information overload.

[0046] 4. This invention constructs a complete closed-loop data processing link at the system level, from data acquisition, spatial relationship calculation, intent recognition, interaction correlation decision-making to adaptive rendering output. The connection relationship between each module is clear and the data flow is well-defined, realizing integrated collaboration of "perception-computation-decision-rendering". It can be deployed in various industrial scenarios such as intelligent factory production line simulation review, equipment status inspection and industrial operation and maintenance auxiliary decision-making. At the same time, through the synergistic cooperation of semantic-level LOD and local detail enhancement, the overall rendering overhead is effectively reduced while ensuring visual experience, and the 3D rendering performance and frame rate stability are improved. Attached Figure Description

[0047] The accompanying drawings are incorporated in and form part of this specification, and together with the description serve to explain the principles of the invention.

[0048] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 Flowchart of the AR adaptive visualization method based on user intent and spatial relationship science provided by the present invention;

[0050] Figure 2 A module architecture diagram of the AR adaptive visualization system based on user intent and spatial relationship science provided by this invention;

[0051] Figure 3 A schematic diagram of a spatial region division geometric model based on Euclidean distance and the angle between lines of sight provided by the present invention;

[0052] Figure 4 This diagram illustrates a comparison between the smooth interpolation process of model facet number and information display density during the transition period when switching LOD levels, as provided by this invention, and the traditional LOD hard switching. Detailed Implementation

[0053] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples consistent with some aspects of the invention as detailed in the appended claims.

[0054] To enable those skilled in the art to better understand the technical solutions of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0055] Example 1

[0056] This embodiment provides an AR adaptive visualization method based on user intent and spatial relational science. This method can be deployed in head-mounted augmented reality devices (such as HoloLens 2) for scenarios such as intelligent factory production line simulation review, equipment status inspection, and industrial operation and maintenance auxiliary decision-making.

[0057] As an optional implementation, a pre-implementation step of scene registration and virtual element initialization is included before performing the following steps. Specifically, virtual elements such as the digital twin model of the production line to be inspected or reviewed, the 3D model of the equipment, the status data panel, and navigation signs are registered in the physical space coordinate system. During the registration process, it is necessary to configure the ontological attributes (including equipment type, name, unique identifier, etc.) and semantic importance weight for each virtual element. (Based on the pre-defined importance levels of the equipment, for example, assigning higher weight to core equipment, medium weight to auxiliary equipment, and lower weight to background environmental elements), a multi-level LOD model (including 3D model resources with different numbers of facets and precision), and a set of displayable information fields (including information content at different levels such as equipment name, status color blocks, core operating parameters, complete operating parameters, and historical trend charts). The above configuration information is stored in the system for use in the interaction correlation calculation in step 4. Through the above pre-configuration, it is ensured that subsequent method steps can accurately obtain the spatial location, semantic importance weight, and display related content of each virtual element.

[0058] Step 1: Multimodal data acquisition

[0059] This step involves using sensors in the augmented reality device to collect real-time spatial and behavioral data from the user. The spatial data includes the user's head center position, head posture, and gaze direction; the behavioral data includes gaze duration, head angular velocity, voice navigation commands, and environmental alarm signals.

[0060] Specifically, the user's head center position in the spatial state data is obtained through joint localization by the inertial measurement unit (IMU) and depth camera built into the head-mounted AR device, and is represented as coordinates in three-dimensional world space. Preferably, a SLAM (Simultaneous Localization and Mapping) positioning system provided by the device is used, with a sampling frequency of no less than 30Hz to ensure real-time tracking accuracy of the user's head movements; head posture is obtained by fusing data from a three-axis gyroscope and a three-axis accelerometer in the IMU to determine the user's head orientation; the gaze direction is obtained in real time by detecting the position of the user's pupils and corneal reflection points in the eye-tracking sensor (such as an infrared camera array) built into the head-mounted AR device. Preferably, the sampling frequency of the eye-tracking sensor is no less than 60Hz, and the gaze direction measurement accuracy is better than 2°.

[0061] Furthermore, the gaze duration in the behavioral data is obtained by ray intersection detection based on the spatial position of the gaze direction and the virtual elements in the scene. When the gaze direction intersects with the bounding box of a virtual element, the duration of gaze at that element is accumulated. The head angular velocity is obtained by differential calculation of head posture data and is used to determine whether the user is in a rapid scanning state. Voice navigation commands are collected by the microphone array built into the head-mounted AR device and converted into navigation intentions by the voice recognition module. Environmental alarm signals are received in real time through the industrial IoT interface or the device's own status monitoring system, including but not limited to temperature alarms, abnormal vibration alarms, pressure over-limit alarms, and equipment shutdown / fault alarms.

[0062] In addition, this step also acquires scene depth maps and spatial mesh data using a depth camera for occlusion detection in subsequent steps. The scene depth map records the distance information between each point in the environment in front of the user and the sensor in the form of a depth image; the spatial mesh data is constructed in real time by the SLAM system based on the depth map and sensor pose, describing the geometry of the physical environment around the user in the form of a triangular mesh.

[0063] In a preferred embodiment, the acquisition of the above-mentioned multimodal data adopts a multi-threaded parallel processing architecture. The data streams of each sensor are fused after the timestamps are aligned to ensure that the data used for subsequent spatial relationship calculation, intent recognition and occlusion detection have a consistent time reference.

[0064] Step 2: Spatial Relationship Calculation and Region Division

[0065] This step first calculates the spatial relationship between the user's head center position collected in step 1 and the spatial position of each pre-registered virtual element in the scene.

[0066] For any virtual element in the scene Let the coordinates of its center position in three-dimensional world space be... The center position of the user's head is Then the Euclidean distance between the user and the virtual element is... The calculation formula is:

[0067]

[0068] In a preferred embodiment, the center position of a virtual element can be the center point of its geometric bounding box, or it can be a semantic anchor point (such as the location of a key component of a device) that is pre-defined based on the semantic attributes of the virtual element.

[0069] Based on the head posture and gaze direction collected in step 1, determine the unit vector of the user's head gaze direction. And determine the unit vector pointing from the center of the user's head to the center of the target virtual element based on the spatial position of the user's head center and the virtual element's position. And then according to and Calculate the angle of view The calculation formula is:

[0070]

[0071] Angle of sight This reflects the degree to which virtual elements are deviated from the center of the user's gaze, i.e. The smaller the value, the closer the virtual element is to the direction the user is looking.

[0072] like Figure 3 As shown, this embodiment is based on Euclidean distance. The virtual elements are divided into three distance zones: when Less than the first distance threshold When it is determined to be an interactive area, Greater than or equal to And less than the second distance threshold When it is determined to be an area of ​​attention, Greater than or equal to The area is then determined to be irrelevant. Preferably, The value ranges from 0.3m to 1.0m, more preferably, It is 0.5m; The value ranges from 1.5m to 3.0m, more preferably, It is 2.0m.

[0073] Furthermore, based on the angle of view The virtual elements are divided into three viewing areas: when Less than the first angle threshold When it is determined to be the focus area, Greater than or equal to And less than the second angle threshold When it is determined to be the surrounding area, Greater than or equal to This is determined to be an ignored region. Preferably, The value ranges from 20° to 40°, more preferably, It is 30°; The value ranges from 45° to 55°, more preferably, It is 50°.

[0074] Through the aforementioned two-dimensional spatial region division, each virtual element is assigned a distance region label and a line-of-sight region label (i.e., the spatial relationship state of each virtual element = distance region label + line-of-sight region label), which together constitute the complete spatial relationship state of the virtual element relative to the user. For example, a virtual element may be simultaneously located in the "interaction region" and the "focus region," indicating that the user is very close to the element and is looking at it; another virtual element may be located in the "irrelevant region" and the "ignored region," indicating that the user is far away and is not paying attention to it.

[0075] Step 3: User Intent Recognition

[0076] This step identifies the user's current task intent based on the multimodal behavioral data collected in step 1, including detailed analysis intent, overview browsing intent, and navigation / addressing intent. Figure 3 Types.

[0077] In this embodiment, the detailed method for identifying intent is as follows: when the user's gaze is fixed on a virtual element and remains there for more than a first time threshold. At this point, the system determines that the user intends to perform detailed analysis. Specifically, the system continuously monitors the intersection of the user's gaze direction and virtual elements; if the gaze continuously points at the same virtual element for more than [a certain amount of time], [the system will detect the intersection]. Detailed analysis and intent determination are triggered at specific times. Preferably, The value ranges from 0.8s to 3.0s, more preferably, The time is 1.5 seconds. Once the intent is determined to be for detailed analysis, the intent weight vector... Assigning analysis weights The dominant weight, i.e., the analytical weight. Significantly higher than browsing weight and navigation weight .

[0078] In this embodiment, the method for recognizing the user's browsing intent is as follows: when the user's head angular velocity is higher than a preset angular velocity threshold... And no more than When the user maintains a continuous gaze, it is determined that the user intends to browse the overview. Preferably, The value ranges from 100° / s to 300° / s, more preferably, The speed is 200° / s. Once the intent is determined to be an overview browsing intent, the intent weight vector... Assigning browsing weight Based on dominant weight, i.e., browsing weight Significantly higher than the analysis weight and navigation weight .

[0079] In this embodiment, the navigation addressing intent is identified as follows: when the system receives a voice navigation command (such as the user saying "Take me to see machine tool number 3") or triggers an emergency alarm event, it determines that the user is in a navigation addressing intent. Once a navigation addressing intent is determined, it is then added to the intent weight vector. Assigning navigation weights With dominant weight, i.e., navigation weight Significantly higher than the analysis weight and browsing weight .

[0080] The three intents mentioned above may coexist. For example, after an alarm is triggered, a user may quickly scan the screen while asking a question via voice. In this case, the system performs weighted fusion based on the confidence level of each signal to generate the final intent weight vector. And satisfy In a preferred embodiment, the dominant weight means that the value of one component in the intention weight vector is not less than 0.5, and the other two components are distributed proportionally within the remaining range to ensure that the sum of the three is 1.

[0081] Step 4: Calculation of Interaction Relevance

[0082] This step calculates the quantified interaction correlation for each virtual element in the scene. This serves as the core basis for subsequent display strategy decisions.

[0083] First, obtain the preset semantic importance weight for each virtual element. The weighting criteria include, but are not limited to, equipment type (higher weight for core production equipment, medium weight for auxiliary equipment, and lower weight for background environmental factors), importance level (higher weight for critical equipment than for general equipment), and alarm level (semantic importance weight when an emergency alarm is triggered). It is set to the maximum value of 1.0.

[0084] Based on Euclidean distance Determine distance weights Its value varies The distance weight increases and then decreases until it reaches zero. In a preferred embodiment, the distance weight... The calculation method is as follows: 1.0 is taken in the interaction region, linearly decreasing from 1.0 to 0 in the attention region, and 0 in the irrelevant region. In another preferred embodiment, an exponential decay function can also be used. Calculation, where This is the preset attenuation coefficient.

[0085] Based on the angle of view Determine angle weights Its value varies It increases and decreases, taking 1.0 in the focused region, linearly decreasing from 1.0 to 0 in the surrounding region, and taking 0 in the ignored region.

[0086] Based on the user intent weight vector generated in step 3 Determine the first weighting coefficient corresponding to the distance weight. The second weighting coefficient corresponding to the angle weight The third weighting coefficient corresponding to the semantic importance weight The three weighting coefficients were determined using a combination of offline calibration and online adaptive methods. First, the distance, angle, and semantic importance metrics were normalized, and then, based on validation samples containing actual user selections, [the following steps were taken]. and Under the constraint of [unclear], with the goal of improving the accuracy of candidate object ranking, the optimal weight combination for different intent types is determined through grid search. During actual interaction, the weight combinations are weighted and fused based on the probabilities of various intents output by the intent recognition model, thus obtaining the weighting coefficient used to calculate the current interaction relevance. In a preferred embodiment, when a component in the intent weight vector is dominant (value greater than 0.6), the corresponding weighting coefficient is not less than 0.5.

[0087] Finally, the interaction degree of each virtual element at the current moment is calculated using the following formula. :

[0088]

[0089] This formula will use distance weights Angle weight and semantic importance weight By fusing the data according to the weighted coefficients driven by intent, the quantitative unification of three-dimensional information of "intent-space-semantics" is achieved.

[0090] Step 5: Adaptive Display Strategy Decision

[0091] This step is based on the interaction relationships calculated in step 4. The comparison results with preset thresholds determine the display detail level and information panel expansion degree of each virtual element.

[0092] In a preferred embodiment, two thresholds are preset. and ( > ).when Higher than At this time, the virtual elements are displayed with a high level of detail and the complete information panel expands, showing the device name, status color blocks, complete operating parameters, and historical trend charts; when Between and In between, control the virtual elements to display at a low to medium level of detail and collapse or simplify the information panel, displaying the device name, status color blocks, and core operating parameters; when Below At this time, virtual elements are automatically hidden or reduced to the lowest transparency, displaying only the device name and status color block to avoid information overload. Simultaneously, based on Euclidean distance... and interaction correlation Adjust the number of model faces and information display density of virtual elements to optimize rendering performance while maintaining visual quality.

[0093] In a preferred embodiment, when the angle of view When the gaze is in the focal area and the gaze duration exceeds a set time threshold (preferably, the set time threshold ranges from 0.5s to 1.0s), a high-precision sub-mesh is divided within a set radius range (preferably, the radius ranges from 3cm to 10cm) around the gaze point, and the texture resolution of the local mesh is multiplied by an enhancement factor (preferably, the enhancement factor is 1.5 times to 4 times). At the same time, edge highlighting rendering is performed to achieve local detail enhancement, simulating the high-precision visual perception characteristics of the human eye's fovea.

[0094] Step 6: Smooth Rendering and Occlusion Repositioning

[0095] In traditional LOD (Level of Detail) switching schemes, when a user moves near the region boundary, the number of model faces changes instantaneously between adjacent levels, easily causing frequent screen flickering and jumps, severely impacting the user's visual experience. To solve this problem, this invention calculates the transition time based on the Euclidean distance difference before and after the user's movement when the display level of virtual elements changes. The calculation formula is:

[0096]

[0097] In the formula, The preset time coefficient is expressed in seconds per minute (s / m). and These represent the Euclidean distances before and after the user's movement. In a preferred embodiment, The value of α ranges from 0.1 s / m to 0.5 s / m, and more preferably, α is 0.2 s / m.

[0098] During the transition time Inside, the number of model facets The following linear interpolation formula is used for smooth evolution:

[0099]

[0100] In the formula, and These represent the number of target facets under two adjacent LOD standards. For the transition start time, The transition end time, The current time is displayed. The expansion level of the information panel also uses the same linear interpolation method for a smooth transition, which will not be elaborated further.

[0101] Figure 4 This illustrates the smooth interpolation process of model facet count and information display density during the transition period in the aforementioned LOD level switching process, where the horizontal axis represents time or relative distance, and the vertical axis represents model facet count or information display density. This contrasts with traditional LOD's direct hard switching at region boundaries (corresponding to...). Figure 4 middle The abrupt change curve at time (is different), and this invention achieves the transition from... through a linear interpolation formula. arrive The continuous and smooth change corresponds to Figure 4 middle to The gradual curve over time effectively eliminates abrupt changes in the image.

[0102] Simultaneously, this invention also utilizes the scene depth map and spatial grid data acquired in step 1 to map the spatial position of virtual elements onto the depth map or spatial grid for comparison, detecting whether virtual elements and their information panels are occluded by real entities. When performing anti-occlusion panel repositioning or information following movement, a position threshold is set. and time update threshold Only when the absolute value of the user's head movement exceeds the position threshold. (Preferred, The value ranges from 0.5cm to 2.0cm, or the user's head remains stationary for more than the time update threshold. (Preferred, When the value ranges from 0.5s to 2.0s and there is no new instruction input, the panel coordinates are recalculated. After recalculation, the interface is smoothly transitioned by an easing function (preferably an ease-in-out easing function), so that the panel accelerates at the beginning of the movement and decelerates when approaching the target position, thereby eliminating screen jumps.

[0103] Through the complete process of steps 1 to 6 above, adaptive control of the presentation of virtual information in the augmented reality environment is achieved. Specifically, step 1 uses multimodal sensors to collect real-time user spatial state data, behavioral data, and scene geometric data, providing a data foundation for subsequent spatial relationship calculation and intent recognition; step 2 quantifies the spatial relationship between the user and each virtual element by calculating the Euclidean distance and the gaze angle and performing two-dimensional spatial region division; step 3 achieves accurate identification and quantitative expression of the user's intent for detailed analysis, overview browsing, or navigation by fusing multimodal behavioral information such as gaze duration, head angular velocity, voice navigation commands, and environmental alarm signals; step 4 constructs a three-dimensional fusion interaction correlation model of "intent-space-semantics," fusing distance weight, angle weight, and semantic importance weight according to the weighted coefficients driven by intent, and calculates the interaction correlation of each virtual element. Step 5 according to The comparison results with the preset threshold are used to dynamically determine the display detail level of each virtual element and the degree of information panel expansion. Step 6 uses the transition time calculation based on the distance difference and linear interpolation to smooth the transition, combined with the dual threshold anti-shake mechanism and occlusion repositioning, to achieve a smooth transition without jumps during LOD level switching and panel movement.

[0104] The above six steps constitute a complete closed-loop data processing chain from "perception (step 1) → calculation (steps 2 and 3) → decision (steps 4 and 5) → rendering (step 6)". The steps are connected by the following data flow: the output of step 1 serves as the input of step 2 and step 3, the outputs of step 2 and step 3 are input to step 4, the output of step 4 serves as the input of step 5, and the decision result of step 5 is executed and output in step 6. Together, they realize the dynamic adaptation of information presentation according to user intent and spatial attention, and complete the transformation from passive fixed presentation to task-oriented active intelligent presentation. This solves the technical problems existing in AR information visualization methods, such as visual occlusion and information overload, rigid detail levels and lack of semantic-level adaptation based on spatial relationships, passive rigidity of information presentation due to ignoring user intent, and frequent screen flickering and jumping when switching display levels.

[0105] Example 2

[0106] This embodiment provides an AR adaptive visualization system based on user intent and spatial relational science, used to execute the method described in Embodiment 1 above. Figure 2 The system's modular architecture diagram is shown below. Figure 2 The connections between the modules and their respective functions are explained in detail.

[0107] Please see Figure 2The system comprises a data acquisition module, a spatial relationship calculation module, a user intent recognition module, an interaction correlation decision module, an adaptive visualization rendering module, an occlusion detection and layout relocation module, and a smooth transition control module. The output of the data acquisition module is connected to the inputs of the spatial relationship calculation module, the user intent recognition module, and the occlusion detection and layout relocation module. The outputs of the spatial relationship calculation module and the user intent recognition module are connected to the input of the interaction correlation decision module. The output of the interaction correlation decision module is connected to the inputs of the adaptive visualization rendering module and the smooth transition control module. The input of the occlusion detection and layout relocation module is also connected to the output of the adaptive visualization rendering module, whose output is connected to the input of the smooth transition control module. The output of the smooth transition control module is connected to the input of the adaptive visualization rendering module. These modules constitute a closed-loop data processing chain from data perception, spatial relationship calculation, intent recognition, correlation decision to adaptive rendering output. The system can be deployed in a head-mounted augmented reality device, with each module implementing its function through software programs and running on the device's processor.

[0108] Specifically, the data acquisition module is used to collect users' spatial state data, behavioral data, scene depth maps, and spatial grid data in real time. Specifically, the data acquisition module connects to the inertial measurement unit, depth camera, eye-tracking sensor, and microphone built into the head-mounted AR device. The collected data includes the user's head center position, head posture, gaze direction, gaze duration, head angular velocity, voice commands, scene depth maps, and spatial grid data. After being timestamped and aligned, the collected data is sent to the spatial relationship calculation module, the user intent recognition module, and the occlusion detection and layout relocation module, respectively.

[0109] The spatial relationship calculation module is connected to the data acquisition module and is used to calculate Euclidean distance based on the user's head center position, gaze direction, and the spatial position of virtual elements. Angle with line of sight And based on Euclidean distance Virtual elements are divided into interactive areas, attention areas, and irrelevant areas, based on the angle of view. Virtual elements are divided into a focused area, a surrounding area, and an ignored area. The output of the spatial relationship calculation module is transmitted to the interaction correlation decision module.

[0110] The user intent recognition module is connected to the data acquisition module to determine the user's intent based on gaze duration and gaze angle. The system identifies one or more user intent states—detailed analysis, overview browsing, or navigation addressing—based on head angular velocity, voice navigation commands, and alarm trigger events, and generates a user intent weight vector. Specifically, the user intent recognition module determines the intent to perform detailed analysis based on the user's gaze being fixed on the same virtual element for more than a first time threshold; it determines the intent to browse overview based on the user's head angular velocity being higher than a preset angular velocity threshold but without any gaze behavior exceeding the first time threshold; and it determines the intent to navigate based on voice navigation commands or emergency alarm events. The output of the user intent recognition module is transmitted to the interaction correlation decision module.

[0111] The interaction correlation decision module is connected to the spatial relationship calculation module and the user intent recognition module, respectively, and is used to obtain the preset semantic importance weight of each virtual element from the system. The semantic importance weights The pre-configuration process has been pre-set and stored in the system (the specific setting method is described in the pre-configuration section of this embodiment, i.e., pre-set according to factors such as device type, importance level, and alarm level, for example, assigning higher weight to core devices, medium weight to auxiliary devices, and lower weight to background environmental elements; when an emergency alarm is triggered, the semantic importance weight is...). (Set to the highest value of 1.0). The interaction relevance decision module is based on the user intent weight vector. Determine the first weighting coefficient corresponding to the distance weight. The second weighting coefficient corresponding to the angle weight The third weighting coefficient corresponding to the semantic importance weight According to Euclidean distance Determine distance weights According to the angle of sight Determine angle weights and distance weights Angle weight and semantic importance weight A weighted sum is performed to calculate the interaction correlation of each virtual element at the current moment. The interaction correlation decision module is also used to determine the relationship between the interaction and the decision-making process. The comparison result with the preset threshold outputs a display strategy decision signal, including decision results such as show, hide, downgrade, expand, local enhancement or navigation highlight, and sends the decision signal to the adaptive visualization rendering module and the smooth transition control module.

[0112] The adaptive visualization rendering module is connected to the interaction correlation decision module, and is used to adjust the display detail level of virtual elements, the number of model faces, texture resolution, and the expansion degree of the information panel based on decision signals. Specifically, the adaptive visualization rendering module adjusts the display detail level of virtual elements, the number of model faces, texture resolution, and the expansion degree of the information panel based on interaction correlation. Based on the comparison results with preset thresholds, the virtual elements are displayed in a high level of detail, medium-low level of detail, or hidden / degraded mode, and the information panel is expanded, collapsed, or simplified accordingly, while also adjusting the display based on Euclidean distance. and interaction correlation Adjust the number of model faces and the information display density. In a preferred embodiment, the adaptive visualization rendering module is also used to adjust the viewing angle. When the gaze is in the focal area and the gaze duration exceeds a set time threshold, a high-precision sub-mesh is divided within a set radius around the gaze point, and the texture resolution of the local mesh is multiplied by an enhancement factor. At the same time, edge highlighting rendering is performed to enhance local details.

[0113] The occlusion detection and layout relocation module is connected to the data acquisition module and the adaptive visualization rendering module. It uses scene depth maps and spatial grid data to detect whether virtual elements and their information panels are occluded by real entities, and recalculates the display position of the information panels when occlusion occurs. Specifically, the occlusion detection and layout relocation module compares the spatial position of virtual elements with the depth map or spatial grid to determine whether there is spatial overlap between virtual elements and real entities. When occlusion is detected, it recalculates the display position of the information panels within the visible area and sends the relocation result to the smooth transition control module and the adaptive visualization rendering module.

[0114] The smooth transition control module is connected to the interaction correlation decision module, the occlusion detection and layout repositioning module, and the adaptive visualization rendering module, respectively. It is used to calculate the transition time based on the difference in Euclidean distance before and after the user moves when switching display levels. It performs linear interpolation for smooth gradation and easing functions for smooth movement during panel repositioning. Specifically, the smooth transition control module adjusts the display level according to the formula... The transition time is calculated, and a linear interpolation formula is used to smoothly evolve the number of model patches within the transition time. A position threshold is used during panel repositioning. and time update threshold As a trigger condition, it only occurs when the absolute value of the user's head movement exceeds a position threshold. Or remain stationary for longer than the time update threshold Furthermore, when there are no new instructions, the coordinates are recalculated, and after recalculation, a smooth sliding transition of the interface is achieved through an easing function.

[0115] In a preferred embodiment, the above modules run on the processor of the head-mounted AR device and use a real-time operating system for task scheduling to ensure that the end-to-end latency from data acquisition to rendering output does not exceed 50ms, so as to provide a smooth augmented reality interactive experience.

[0116] Example 3

[0117] To more clearly illustrate the workflow and technical effects of this invention in practical applications, an exemplary description is provided below using a specific industrial field application scenario.

[0118] Scenario Setting: In the machining workshop of a smart manufacturing company, inspection experts wear head-mounted AR devices (such as HoloLens2) to conduct virtual reviews or routine inspections. The system pre-registers the digital twin model of the production line to the physical space, including equipment such as CNC machine tools and air compressors, and configures multi-level LOD models and data panels.

[0119] Phase 1: Long-Distance Overview (Macro-level presentation dominated by intent weighting)

[0120] The expert stood at the workshop entrance, a distance from the core equipment, "Air Compressor A03". At this point, the result is calculated using step 2. This means that the area is divided into irrelevant regions, and the expert's head is rotating to scan the entire area, with the head angular velocity exceeding a preset angular velocity threshold. Step 3: Determine the current intent vector. Browse weight At its highest level, the user is currently browsing in an overview mode.

[0121] Based on the interaction correlation formula calculated in step 4, the data panels for various devices are controlled to be at the lowest level of detail (LOD 2) in step 5, displaying only the device name and color blocks representing the device status (e.g., air compressor A03 is displayed as a yellow warning color block). During the display process, if some devices are obscured by factory columns, step 6 detects the obstruction using scene depth map and spatial grid data, and automatically moves the minimalist labels to the visible blank area next to the column, ensuring that the information is always readable. At this stage, the expert's field of vision is clean and clear, unaffected by massive parameters, facilitating a quick acquisition of the overall workshop status.

[0122] Phase Two: Mid-Range Proximity and Smooth LOD Transition (Spatial Relationship Driven)

[0123] Experts noticed the yellow warning color block approaching air compressor A03. At this point, the distance reaches the first distance threshold. With the second distance threshold The "attention zone" between LOD 2 and LOD 1. The system triggers the transition from LOD 2 to LOD 1. Step 6 is based on the transition time formula. The transition time is calculated, and a linear interpolation formula is used within that time period. The number of vibration points on the screen evolves smoothly. As the user approaches, the yellow warning icon unfolds like silk into a folded core parameter panel, displaying the current vibration value and temperature, without any abrupt screen flickering or jumps.

[0124] Phase 3: Close-up detailed examination (local enhancement driven by both gaze and intention)

[0125] Experts approached (Interactive area), and the angle of view (Focus on the area), and continuously observe the vibration sensor area of ​​air compressor A03 for 1.8 seconds (exceeding the first time threshold). Step 3: Based on the fixation duration and the angle of gaze, the intention vector is... Switch to detailed analysis intent ( (Dominant). Interaction correlation in step 4 Significantly increased and exceeded the high brightness threshold Step 5's adaptive visualization manifests as follows: the parameter panel fully expands, displaying a detailed maintenance chart with historical trend lines; simultaneously, local detail enhancement is triggered within a 5cm range of the expert's gaze, doubling the texture resolution and adding a bright, flowing edge effect. When the expert's head makes a slight movement, the dual-threshold dynamic compensation mechanism in step 6 (position threshold) is activated. and time update threshold Keep the expanded information panel anchored, only when the displacement is greater than... Only when the easing function is used to smoothly follow, the UI tremor caused by slight eye movement is effectively eliminated.

[0126] Phase Four: Sudden Alarm Interruption and Address Navigation (Semantic Weight Breakthrough)

[0127] During a detailed inspection of air compressor A03 by experts, another CNC machine tool located further away suddenly triggered a temperature alarm. (The semantic importance weight of this machine tool is not specified.) In step 4, the system raises the value to the highest level (when an emergency alarm is triggered). Even if the user is far away from the machine tool, the degree of interaction remains high. It still exceeds the threshold. Step 3 transfers the system intent to "navigation addressing intent" ( (Dominating), in step 5, the detailed panel of the air compressor A03 immediately shrinks to a semi-folded shape, while a blue guide light band pointing towards the distant alarm machine tool is generated in the expert's field of vision. All non-alarm devices along the way are reduced to the lowest transparency to avoid visual interference with emergency addressing, enabling experts to quickly locate and handle alarm devices, ensuring extremely high efficiency of inspection or virtual review.

[0128] As can be seen from the complete examples of the above four stages, the present invention can adaptively adjust the presentation of virtual information under different spatial distances, different user intentions and different task scenarios, realizing the transformation from passive fixed presentation to active intelligent presentation oriented to task, and effectively solving technical problems such as visual occlusion, information overload, rigidity of detail levels and screen jump in the prior art.

[0129] In summary, the AR adaptive visualization method and system based on user intent and spatial relational science provided in this invention constructs a three-dimensional fusion interaction correlation model of "intent-space-semantics" to achieve multi-factor collaborative decision-making through interaction correlation. This allows information presentation to dynamically adapt to user intent and spatial attention, achieving a leap from geometrical Level of Detail (LOD) to semantical adaptation. By using multimodal implicit behavior fusion to determine user intent and drive dynamic adjustment of weighting coefficients, information presentation transforms from a passive, fixed approach to a task-oriented, proactive adaptation. Through distance difference-based transition time calculation and linear interpolation for smooth gradation, combined with a dual-threshold anti-shake mechanism and occlusion relocation, continuous and seamless transitions are achieved during layer switching and panel movement. This system can be widely applied to complex industrial scenarios such as intelligent factory production line simulation review, equipment status inspection, and industrial operation and maintenance auxiliary decision-making. It effectively reduces user cognitive load and visual interference, improves 3D rendering performance and frame rate stability, and makes the AR interactive experience more natural, efficient, and smooth.

[0130] The above description is merely a specific embodiment of the present invention, intended to enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments without creative effort will be readily apparent to those skilled in the art; the basic principles defined herein can be applied to other embodiments without departing from the spirit and scope of the present invention.

[0131] It should be understood that this invention is not limited to the specific content described above, and various equivalent modifications or substitutions can be made without departing from its scope of protection. The scope of protection of this invention is defined only by the appended claims.

Claims

1. An AR adaptive visualization method based on user intent and spatial relational science, characterized in that, Includes the following steps: Step 1, Multimodal Data Acquisition: Real-time acquisition of the user's spatial state and behavioral data through the sensors of the augmented reality device; The spatial state data includes at least the user's head center position, head posture, and gaze direction; the behavioral data includes at least gaze duration, head angular velocity, voice navigation commands, and environmental alarm signals. Step 2, Spatial Relationship Calculation and Region Division: Calculate the Euclidean distance between the user's head center position and the spatial position of each virtual element in the scene. The unit vector for determining the user's head gaze direction is determined based on the head posture and the gaze direction. And determine the unit vector pointing from the center of the user's head to the center of the target virtual element based on the spatial position of the user's head center and the spatial position of the virtual element. And then according to and Calculate the angle of view Based on the Euclidean distance Virtual elements are divided into interactive areas, attention areas, and irrelevant areas, based on the aforementioned viewing angle. The virtual element is divided into a focus area, a surrounding area, and an ignored area to obtain the spatial relationship status of the virtual element relative to the user. Step 3, User Intent Recognition: Based on the gaze duration and the angle of gaze... The system uses head angular velocity, voice navigation commands, and alarm trigger events to identify one or more user intent states, such as detailed analysis, overview browsing, or navigation addressing, and generates a user intent weight vector. The user intent weight vector It includes weighted components corresponding to detailed analysis intent, overview browsing intent, and navigation addressing intent, respectively, and the sum of each component is 1; Step 4: Calculate the interaction relevance: Obtain the preset semantic importance weight for each virtual element. According to the user intent weight vector Determine the first weighting coefficient corresponding to the distance weight. The second weighting coefficient corresponding to the angle weight The third weighting coefficient corresponding to the semantic importance weight Based on the Euclidean distance Determine distance weights Based on the angle of view Determine angle weights The distance weight With the Euclidean distance The angle weight increases and decreases. With the angle of the line of sight Increases and decreases; the interaction correlation degree of each virtual element at the current moment is calculated according to the following formula. : Step 5, Adaptive Display Strategy Decision: Based on the aforementioned interaction correlation. The comparison results with preset thresholds determine the display detail level and information panel expansion degree of each virtual element; simultaneously, based on the Euclidean distance... and the degree of interaction Adjust the number of model faces and information display density of virtual elements; Step 6, Smooth Rendering and Occlusion Relocation: When the display layer of virtual elements changes, the transition time is calculated based on the difference in Euclidean distance before and after the user moves. and during the transition time The number of model facets and the degree of expansion of the information panel are smoothly and gradually changed. At the same time, it detects whether the virtual elements and their information panels are obscured by real entities. When the preset trigger conditions are met, the panel coordinates are recalculated and the control panel is moved to the visible area through the easing function to eliminate screen jumps.

2. The AR adaptive visualization method based on user intent and spatial relational science according to claim 1, characterized in that, In step 2, the angle of view The calculation formula is: In the formula, This is a unit vector representing the direction of the user's head gaze. The unit vector pointing from the center of the user's head to the center of the target virtual element.

3. The AR adaptive visualization method based on user intent and spatial relational science according to claim 1, characterized in that, In step 2, based on the Euclidean distance The specific method for dividing virtual elements into interactive regions, attention regions, and irrelevant regions is as follows: when the Euclidean distance... When the distance is less than the first distance threshold, it is determined to be an interactive area; when the Euclidean distance is less than the first distance threshold, it is determined to be an interactive area. When the distance is greater than or equal to the first distance threshold and less than the second distance threshold, it is determined to be a region of concern. If the distance is greater than or equal to the second distance threshold, it is determined to be an irrelevant region; Based on the aforementioned line of sight angle The specific method for dividing virtual elements into a focused area, a surrounding area, and an ignored area is as follows: when the angle of view... When the angle is less than the first angle threshold, it is determined to be a focal area; when the angle of view is... When the angle is greater than or equal to the first angle threshold and less than the second angle threshold, it is determined to be the surrounding area. If the angle is greater than or equal to the second angle threshold, it is determined to be an ignored region.

4. The AR adaptive visualization method based on user intent and spatial relational science according to claim 1, characterized in that, In step 3, the user intent is identified as follows: when the user's gaze is fixed on a virtual element and stays there for more than a first time threshold, it is determined to be a detailed analysis intent and given a high analysis weight; when the user's head angular velocity is higher than a preset angular velocity threshold and there is no gaze behavior exceeding the first time threshold, it is determined to be a general browsing intent and given a high browsing weight. When the system receives a navigation voice command or triggers an emergency alarm event, it determines that it is a navigation addressing intention and assigns it a high navigation weight.

5. The AR adaptive visualization method based on user intent and spatial relational science according to claim 1, characterized in that, In step 4, the first weighting coefficient Second weighting coefficient and the third weighting coefficient The user intent weight vector The specific driving method is as follows: when a detailed analysis intent is identified, the first weighting coefficient is increased. and the second weighting coefficient The value of the third weighting coefficient is increased when an overview browsing intent or navigation addressing intent is identified. The value of .

6. The AR adaptive visualization method based on user intent and spatial relational science according to claim 1, characterized in that, In step 5, when the angle of view is... When the gaze is in the focal area and the gaze duration exceeds a set time threshold, a high-precision sub-mesh is divided within a set radius around the gaze point, and the texture resolution of the high-precision sub-mesh is multiplied by an enhancement factor. At the same time, edge highlighting rendering is performed to enhance local details.

7. The AR adaptive visualization method based on user intent and spatial relational science according to claim 1, characterized in that, In step 6, the transition time The calculation formula is: In the formula, The preset time coefficient, and These represent the Euclidean distances before and after the user's movement.

8. The AR adaptive visualization method based on user intent and spatial relational science according to claim 7, characterized in that, In step 6, during the transition time Within this model, the number of facets is smoothly varied using the following linear interpolation formula: In the formula, For the number of facets in the model, and These represent the number of target facets under two adjacent LOD standards. For the transition start time, The transition end time, This is the current time.

9. The AR adaptive visualization method based on user intent and spatial relational science according to claim 1, characterized in that, In step 6, the preset triggering condition includes a location threshold. and time update threshold ; When the absolute value of the user's head movement exceeds the position threshold Or the user's head remains still for more than the time update threshold. Furthermore, if no new instructions are input, the panel coordinates will be recalculated.

10. An AR adaptive visualization system based on user intent and spatial relational science, used to perform the method according to any one of claims 1 to 9, characterized in that, include: The data acquisition module is used to collect users' spatial status data and behavioral data in real time; The spatial relationship calculation module, connected to the data acquisition module, is used to calculate the Euclidean distance between the center position of the user's head and the spatial position of each virtual element in the scene. Angle with line of sight And based on the Euclidean distance Virtual elements are divided into interactive areas, attention areas, and irrelevant areas, based on the aforementioned viewing angle. The virtual elements are divided into a focused area, a surrounding area, and an ignored area; The user intent recognition module, connected to the data acquisition module, is used to identify the user intent state based on the gaze duration, head angular velocity, voice navigation commands, and alarm trigger events, and generate a user intent weight vector. ; The interaction correlation decision module, connected to both the spatial relationship calculation module and the user intent recognition module, is used to obtain the preset semantic importance weight for each virtual element. According to the user intent weight vector Determine the first weighting coefficient corresponding to the distance weight. The second weighting coefficient corresponding to the angle weight The third weighting coefficient corresponding to the semantic importance weight According to the Euclidean distance Determine distance weights Based on the angle of view Determine angle weights and the distance weight The angle weights and the semantic importance weight A weighted summation is performed to calculate the interaction correlation degree of each virtual element at the current moment. The interaction correlation decision module is also used to determine the relationship based on the... The comparison result with the preset threshold is displayed as a strategy decision signal. An adaptive visualization rendering module, connected to the interaction correlation decision module, is used to adjust the display detail level, model facet number, texture resolution, and information panel expansion degree of the virtual elements according to the decision signal. The occlusion detection and layout repositioning module is connected to the data acquisition module and the adaptive visualization rendering module. It is used to detect whether the virtual elements and their information panels are occluded by real entities, and to recalculate the display position of the information panels when occlusion occurs. The smooth transition control module, connected to the interaction correlation decision module, the occlusion detection and layout repositioning module, and the adaptive visualization rendering module, is used to calculate the transition time based on the Euclidean distance difference before and after the user moves when switching display levels. It also performs a smooth gradient and an easing function to smoothly move the panel when it is repositioned.

Images