Tactical AR helmet real-time environment three-dimensional reconstruction method

Abstract

The invention relates to a tactical AR helmet real-time environment three-dimensional reconstruction method, and the method comprises the steps: constructing a time and direction prediction relation corresponding to an observation direction, generating a pre-view field sector, preferentially distributing reconstruction calculation resources to a pre-view field, and forming a view field dynamic focusing mechanism based on the head micro-motion, neck posture change and electromyographic signals of tactical personnel; on the basis of the spatial correspondence between the sensing data and the historical three-dimensional structure, performing local geometric splitting decomposition type evolution modeling on the region; constructing a local structure time sequence stack for a space region which is not visible at present due to view angle change or shielding outside the pre-view region, so as to carry out view angle recession compensation; monitoring the deformation continuity of each local space rigid block, and carrying out feedback adjustment on an abnormal region through visual prompt marks and local reconstruction operation; and receiving a spatial intelligence metadata packet from a tactical system, converting the spatial intelligence metadata packet into a voxel structure with geometric morphological characteristics, and embedding the voxel structure into the current three-dimensional relationship, so that the tactical information is structurally fused and interactively embedded in a reconstruction space.

Description

TECHNICAL FIELD

[0001] The present application relates to AR helmet real-time environment three-dimensional reconstruction, in particular to a tactical AR helmet real-time environment three-dimensional reconstruction method. BACKGROUND

[0002] Although the prior art such as Chinese patent CN109921818A, an intelligent multi-point real-time three-dimensional interactive helmet display system, has made certain progress in the application of intelligent helmets, there are still deficiencies and drawbacks in some key functions and system efficiency. First of all, the technical solution of this patent realizes the construction of three-dimensional map and real-time position update in the environment limited by GPS signal through the cooperation of wireless communication and infrared camera. However, the defect of this scheme lies in that there is a certain limitation in the acquisition and processing speed of environmental data. Especially in complex indoor environment, although infrared camera can acquire image data, it relies on the conversion relationship between image data to estimate the motion of the camera, which is easily affected by factors such as reflection, shielding and light changes in complex environment, thereby causing the accuracy of three-dimensional map reconstruction to decrease. Therefore, although this scheme can solve some problems by constructing a local three-dimensional map in real time, the overall map reconstruction and accuracy update speed is still difficult to meet the demand of fast response.

[0003] Secondly, the patent scheme relies on infrared camera and SLAM (simultaneous localization and mapping) module to perform real-time update of local three-dimensional map, but the computing resources of this system are concentrated on relatively simple task execution, lacking detailed optimization for different tasks and environments. After the infrared camera acquires image data, the central processing unit performs splicing and fusion of local map, but this method has poor adaptability to large-scale and dynamic environment, especially in complex environments involving rapid movement or complex shielding, the quality of infrared image data is unstable, and the splicing error caused will affect the accuracy of the overall map. In addition, although this patent provides real-time data transmission and cooperative combat functions based on wireless transceiver, the system still has deficiencies in processing multi-user interaction and information sharing. When multiple users perform tasks in complex environment at the same time, the system needs to process and synthesize multiple data from different users, if the information update is not timely or the data transmission lags, it will directly affect the timeliness and accuracy of tactical decision-making. For example, if the system fails to transmit the latest tactical data or real-time position information in time, it may cause coordination errors between tactical personnel, affecting the overall combat effectiveness. Finally, the intelligent helmet in the prior art relies on fixed hardware structure and functional modules, and cannot dynamically adjust the priority of voxels and display content according to task requirements.

[0004] In summary, although the prior art provides a preliminary solution for the application of intelligent helmets and tactical information systems, there are still deficiencies in handling complex dynamic environments, quickly responding to task requirements, improving system collaboration capabilities, and optimizing multi-user interaction. SUMMARY

[0005] The purpose of the present application is to provide a tactical AR helmet real-time environment three-dimensional reconstruction method, thereby solving some of the problems and deficiencies pointed out in the background art.

[0006] The technical scheme adopted by the present application to solve the above-mentioned technical problems is as follows: a tactical AR helmet real-time environment three-dimensional reconstruction method, comprising: based on the head micro-movement, neck posture change and electromyographic signal of the tactical personnel, constructing a time and direction prediction relationship corresponding to the observation direction, for generating a pre-visual field sector of the area to be gazed at, and preferentially allocating reconstruction computing resources to the pre-visual field to form a visual field dynamic focusing mechanism; based on the spatial correspondence relationship between the sensing data in the pre-visual field and the historical three-dimensional structure, extracting a structure residual area, and performing local geometric splitting evolution modeling on the area; For the spatial area that is currently invisible due to the change of viewing angle or occlusion outside the pre-visual field, a local structure time sequence stack is constructed, the historical visible angle of the area is traced back, and the historical structure is projected into the current three-dimensional structure in the form of a spatial mapping patch, so as to compensate for the perspective recession; during the construction of the three-dimensional structure, the deformation continuity of each local spatial rigid block is monitored, and if structure drift or geometric incoherence is detected, a dynamic lag feedback mechanism is triggered to feedback and adjust the abnormal area through visual prompt markers and local reconstruction operations; After the reconstruction and compensation operations are completed, spatial intelligence metadata packets from the tactical system are received, and the data packets are converted into voxel structures with geometric morphological features and embedded into the current three-dimensional relationship, so that the tactical information is structurally fused and interactively embedded in the reconstructed space.

[0007] Further, the time and direction prediction relationship is real-time corrected based on the time series modeling results of the past behavior trajectory of the tactical personnel, for dynamically adjusting the direction weight distribution of the pre-visual field sector; the calculation of the pre-visual field sector depends on the observation direction, and includes visibility projection estimation of the current terrain occlusion structure.

[0008] Further, the geometric difference value in the structure residual area is obtained by comparing the structure integrity evolution rate, and the splitting modeling is started when the local geometric evolution exceeds the dynamic threshold; the space unit after the splitting modeling contains a local reconstruction trust marker, and the trust marker is used for subsequent lag feedback mechanism to determine whether the modeling segment needs to be rolled back.

[0009] Furthermore, the spatial mapping patch generates a transparent time axis attribute field after projection. The attribute field is used to indicate the timeliness of the structure and performs grayscale attenuation processing based on the time since the last visibility. The management of the local structure time sequence stack adopts a cyclic truncation strategy to retain structure frames that appear more frequently than a threshold in the visible area.

[0010] Furthermore, a rotational inertial residual kernel is introduced into the drift detection of the rigid block to independently capture the micro-change region where pose drift occurs; a regional hysteresis buffer counter is introduced into the feedback adjustment mechanism to trigger visual alert when drift occurs continuously for more than three cycles.

[0011] Furthermore, the tactical intelligence metadata is used... Parsed as a set of multi-level geometric structures with task semantics Based on task urgency, task type, and regional environmental complexity, the system dynamically selects appropriate geometric representations (such as cones, spheres, or bounding voxels) for 3D integration; and records the interaction behavior after the integrated voxel interacts with the user. Based on the action time, operation type, response delay, and voxel influence area information, a behavior response function is generated. The data is uploaded to the tactical system as behavioral feedback logs. A behavior-driven voxel feedback adjustment formula is used to comprehensively evaluate the task saliency, user behavior sensitivity, and regional policy weights of voxels, in order to adjust the voxel embedding priority. in: The cumulative behavioral response intensity function is used to evaluate the performance of a specific intelligence voxel over time. The importance of the response within; The micro-time variables and integral parameters in response evolution; The rate function of user interaction actions per unit time (e.g., frequency of gestures, eye movements, and voice triggers). gaze path offset function during user interaction; The spatial strategy weight of the current voxel represents its importance level in the current combat zone; These represent the moderating factors of interaction rate, attention change rate, and task weight on the overall feedback, respectively. The response saturation control coefficient is used to adjust the voxel's decay sensitivity to long-term interactions; Estimated average attention duration of users for this type of voxel; The process of a user interacting with an intelligence voxel is essentially a dynamic event that is continuous in time, spatially related, and coupled with semantic variables. This event lasts for a period of time. Internal manifestations include various quantifiable behavioral characteristics; the behavioral response intensity function Set as an integral function about time, whose integral core is the instantaneous intensity estimation model of interactive behavior, the overall expression is: Wherein represents the instantaneous behavior response intensity at any time ; Will Further modeling as a nonlinear combination function composed of three parts: the first part represents the user behavior activity in unit time, such as eye movement, gesture, speech interaction frequency; Set the rate function as , because the behavior intensity increases nonlinearly, so use square enhancement: , wherein is the adjustment coefficient; The second part considers the speed of user attention change, that is, the degree of shift of gaze path over time, reflecting the task attraction and user interest jump; Set the path function as , the derivative represents the gaze switching rate, with the weight factor ; The third part describes the tactical weight value carried by the voxel in space, that is, the importance of the voxel changes over time; Let the value be , and saturate it through the Sigmoid function to get: Wherein controls the steepness of the response rise, is the average response time of the user estimated by the system, that is, the central turning point; After combining the above three parts, the complete instantaneous response intensity function is constructed, which is substituted into the integral form to get the final behavior response evaluation function . This function not only has time domain accumulation ability, but also introduces spatial attention derivative term and voxel semantic saturation modulation term, which can analyze the user and voxel interaction process in multiple dimensions, especially suitable for real-time voxel scheduling, information credibility evaluation, and interactive priority sorting key function modules in tactical AR.

[0012] Further, the cone, sphere and frame voxels correspond to attack area indication, target unit identification and restrictive warning area respectively, and the task urgency is linked through three-dimensional geometric color coding; The selection of the geometric structure is based on the dynamic adjustment of the current combat phase of the tactical system, and the frame voxel is embedded in the pre-combat stage, and the sphere or cone voxel is embedded in the combat stage.

[0013] Further, the information voxel triggers a behavior response classification module after the interaction is completed, which classifies the behavior as attention, avoidance or confirmation according to the interaction duration and operation type; the behavior feedback log contains the gaze path change trajectory before and after user interaction, which is used to evaluate the guiding effect of tactical information on spatial cognition.

[0014] Further, if the user performs similar operations on the same category of voxel for three times in succession, the system infers the user's preference for the task state and adjusts the future voxel embedding strategy; conflict detection is performed on the recorded user interaction behavior, and if multiple users in the same battlefield area mark the same information voxel as ignored, the system triggers information effectiveness review.

[0015] The present application has the following advantages: by analyzing the behavior data of tactical personnel in real time and combining the time and direction prediction relationship, the system can dynamically adjust the direction weight distribution of the preview field sector, and preferentially allocate computing resources to the area where the tactical personnel will gaze. It can perform real-time high-precision three-dimensional reconstruction in a complex dynamic environment, ensuring that the key areas can be quickly and accurately presented, and improving the real-time reaction ability of tactical personnel. Through the linkage display of three-dimensional geometric color coding and task urgency, tactical personnel can more intuitively and clearly identify each important area in the tactical environment, such as attack area, target unit and restrictive warning area, etc. Based on the interaction behavior of users and information voxels, the behavior response classification module dynamically identifies the task preferences of tactical personnel, and adjusts the embedding priority of voxels according to these preferences.

[0016] Through the behavior feedback log, conflict detection and information effectiveness review mechanism, the accuracy and timeliness of tactical information are ensured. By detecting the feedback of multiple users on the same information voxel, the system can automatically start the review program when the information conflicts, re-evaluate the effectiveness of the information, and thus avoid the impact of information errors on tactical decision-making, improving the intelligence and reliability of the system. The system's preview field sector calculation not only depends on the user's observation direction, but also combines the visibility projection estimation of the real-time terrain occlusion structure. In the battlefield environment, as the tactical personnel move and the viewing angle changes, the system can dynamically adjust the range of the field of view and the allocation of computing resources according to the real-time environment, ensuring the real-time reconstruction and accurate presentation of key areas. Through the interaction and conflict detection mechanism of multiple users, coordinated combat between multiple tactical personnel can be achieved. When the tactical information of the same area is marked as "ignored" by multiple users, the system can trigger effectiveness review to ensure the uniformity and accuracy of the information, thereby improving the information sharing and coordinated combat capability in team combat. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 Brief process of three-dimensional reconstruction and information fusion of the tactical AR helmet of the present application.

[0018] Figure 2 The key function relationship diagram of the tactical AR three-dimensional reconstruction system of the present application.

[0019] Figure 3 The main flow of intelligence voxel embedding and interactive feedback of the present application.

[0020] Figure 4 The three-dimensional reconstruction flow of the urban tactical AR helmet of the present application in embodiment 1.

[0021] Figure 5 The three-dimensional reconstruction and intelligence dynamic feedback flow of the urban tactical AR helmet of the present application in embodiment 2. DETAILED DESCRIPTION

[0022] The specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0023] The specific embodiments of the present application will be described in detail below with reference to the accompanying drawings. Figure 1The application provides a tactical AR helmet real-time environment three-dimensional reconstruction method. The method mainly realizes real-time collection and analysis of head micro-movement, neck posture change and electromyographic signals of a tactical personnel, constructs a time and direction prediction relationship corresponding to an observation direction, generates a pre-visual field sector of an area to be gazed at, and preferentially allocates computing resources to the area, so as to form a visual field dynamic focusing mechanism. First, the head micro-movement and neck posture change of the tactical personnel are monitored in real time through a plurality of sensors (such as IMU, gyroscope, accelerometer, electromyographic sensor, etc.) built in the helmet. Through analysis of these signals, the observation direction change trend of the tactical personnel and the area to be gazed at are calculated. This process can accurately predict the next observation direction of the tactical personnel by constructing a time and direction prediction relationship model. In combination with the needs of a tactical scene, a pre-visual field sector of the area to be gazed at can be generated. The size and accuracy of the pre-visual field are closely related to the stability of the current observation direction, the complexity of the surrounding environment, and the current action of the personnel. Then, in the three-dimensional reconstruction process, the computing resources are preferentially allocated to the pre-visual field area, and the dynamic focusing mechanism is used to ensure efficient reconstruction of important areas, so that the system can complete three-dimensional reconstruction of the most critical area in the shortest time and achieve the optimal real-time response effect. Secondly, based on the spatial correspondence relationship between the collected sensor data in the pre-visual field and the historical three-dimensional structure, a structure residual area is further extracted. In this process, by comparing the current reconstructed scene with the historical structure information, the system can identify the area that is different from the historical structure, that is, the structure residual area. These areas are usually changes or dynamic areas in the scene. Subsequently, for these structure residual areas, a local geometric splitting evolutionary modeling method is used for processing. In this method, the local structure information is more accurate by performing fine-grained geometric splitting on each structure residual area. Each splitting area is modeled by a geometric evolution algorithm, which can gradually adjust the reconstruction strategy according to the shape and position of the area to optimize the reconstruction accuracy and efficiency, thereby realizing fine three-dimensional reconstruction.

[0024] In the process of three-dimensional reconstruction, for the space region that is currently invisible due to the change of viewing angle or occlusion outside the preview domain, a local structure time stack is constructed for processing. The system records the historical reconstruction information of each local space region in time sequence, and establishes a local structure time stack combined with the historical viewing angle data, to realize the backtracking of the historical visible angle of the region. When the observation direction of the tactical personnel changes, the system can backtrack the visible angle of the region in the historical data, compare the difference between the historical viewing angle and the current observation direction, calculate the historical reconstruction state of the region, generate historical structure patches, and project these patches into the current three-dimensional structure in a spatial mapping manner, ensuring that even if a region is invisible in the current viewing angle, the historical structure of the region can be integrated into the current three-dimensional model through the compensation mechanism, and the perspective retroactive compensation is realized. Through this method, the system can use historical information to repair the region that is invisible in the current viewing angle in a dynamic environment, and maintain the consistency and continuity of the overall three-dimensional reconstruction. Further, in the process of constructing the three-dimensional structure, the system monitors the deformation continuity of each local rigid block. The system analyzes the spatial stability and geometric continuity of each local rigid block in real time to determine whether it has drifted or is geometrically incoherent. If any form of structural drift or deformation anomaly is detected, the system will trigger a dynamic lag feedback mechanism. Under the action of this mechanism, the system indicates the problem area through visual prompt markers, displays the marker information of the abnormal area on the AR helmet display screen, and performs compensatory repair on the area through local reconstruction operation. After detecting the abnormal area, the system will preferentially perform fine modeling on the structure of the area, use historical information, sensing data and real-time updated environmental data for weighted repair, so as to restore the accurate morphology of the area, and ensure the accuracy and stability of the three-dimensional reconstruction result.

[0025] The system will receive a spatial intelligence metadata package from the tactical system after completing the three-dimensional structure reconstruction and perspective compensation operation. This data package contains dynamic information related to the tactical mission, such as enemy positions, friendly positions, important marked areas, target tracking information, etc. After receiving the spatial intelligence metadata package, the system will first analyze it and interface the tactical information with the current three-dimensional structure. The system will convert the spatial intelligence metadata package into a voxel structure with geometric shape characteristics based on the coordinate system information, intelligence category, and task priority in the data package. Voxel structure is the basic unit representing information elements in three-dimensional space, each voxel has a specific coordinate and volume in space, and can carry various information related to the tactical mission, such as target type, position, threat level, etc. By converting tactical information into voxel structure, the system can embed the information into the current three-dimensional space model. The embedding process is through voxelization of tactical information, which is convenient for presentation in AR display devices and seamless integration with existing environment three-dimensional reconstruction data. In this process, the embedding of voxels is not just a simple spatial superposition, but a dynamic embedding process based on spatial coordinates, information classification and task requirements, ensuring that each voxel is properly prioritized in the reconstructed space according to its importance and real-time nature, so that tactical information can be accurately positioned and updated in real time in the three-dimensional reconstruction space. The process also includes interactive operation support for embedded voxels, users can interact with embedded tactical information through the interactive interface of the AR helmet, such as selecting, hiding, marking or adjusting specific voxels through gestures, voice or gaze operations. During the interaction process, the system will dynamically update the display state and spatial position of the relevant voxels, so that users can obtain the most intuitive environmental feedback according to the current tactical requirements.

[0026] The accompanying drawings are incorporated in and constitute a part of this specification. Figure 2In terms of pre-visual sector calculation and dynamic adjustment, the time and direction prediction relationship is corrected in real time based on the time series modeling results of the tactical personnel's past behavior trajectory. In this process, the system analyzes the tactical personnel's historical behavior trajectory data, such as head movement, line of sight change, gait, etc., and uses time series modeling methods to construct a prediction model that can predict the direction of the tactical personnel's gaze in real-time tactical environment. This prediction model not only considers the short-term behavior patterns of the tactical personnel, but also combines long-time sequence data to accurately predict the tactical personnel's line of sight direction and target area. The system dynamically corrects the prediction results by comparing the tactical personnel's current behavior data with the historical behavior trajectory in real time, thereby improving the accuracy of the visual angle prediction. When the tactical personnel's behavior or task changes, the system can adjust the direction weight distribution of the pre-visual sector based on the real-time corrected prediction relationship. This adjustment process is based on the tactical personnel's current behavior patterns and environmental dynamic changes, so that the direction weight of the pre-visual sector more accurately reflects the tactical personnel's real-time line of sight changes and environmental focus points. Next, the calculation of the pre-visual sector not only depends on the tactical personnel's current observation direction, but also considers the visibility projection estimation of the terrain occlusion structure in the current environment. When calculating the pre-visual sector, the system first makes a prediction based on the tactical personnel's actual visual angle direction, and then performs environmental modeling through real-time perception data such as surrounding obstacles, buildings, natural terrain, etc., to estimate which areas in the current scene are visible and which areas are not visible due to occlusion or visibility restrictions. Through these real-time estimations, the system can generate accurate visible area projections and adjust the sector size and location of the pre-visual sector based on the environmental occlusion, ensuring that the reconstruction system prioritizes processing important areas within the tactical personnel's current line of sight. In addition, if there is an occlusion or obstacle change in the environment, the system will also update the direction and size of the pre-visual sector in real time, ensuring that the tactical AR system always focuses computing resources on important areas in the user's field of view, achieving rapid response to dynamic environments.

[0027] In terms of geometric discrepancy detection and fragmentation modeling in structural residual areas, the geometric discrepancy value of the structural residual area is obtained by comparing the evolution rate of the structural integrity curve. Specifically, when performing three-dimensional reconstruction, the system constructs a structural integrity curve based on historical data and current reconstruction data, which reflects the integrity of the three-dimensional structure in the current environment over time. By comparing the evolution trend of the current structure data with the historical data, the system can calculate the geometric discrepancy value, which is specifically represented as the difference between the current structure and the expected structure. This difference value can reflect the areas of structural changes in the environment in real time, i.e., those structural residual areas caused by dynamic factors such as occlusion, object movement, etc. To effectively handle these residual areas, when the geometric discrepancy of the structural residual exceeds the predetermined dynamic threshold, the system will start the fragmentation modeling process. Fragmentation modeling is a local geometric fragmentation operation on the difference area, which subdivides the area into multiple smaller spatial units. The geometry and position of these small units can more accurately reflect the true structure of the residual area, ensuring the accuracy of the reconstruction. Each spatial unit after fragmentation modeling contains a local reconstruction trust label. The trust label is used to measure the reliability of the spatial unit in the reconstruction process, and its value represents the reconstruction accuracy and reliability of the area. The trust label will be dynamically adjusted according to the geometric structure changes and relative importance in the area, and when the label value is low, it means that the reconstruction of the area has a large uncertainty and needs to be further verified or corrected. The trust label plays a crucial role in the subsequent lag feedback mechanism. When the system detects structural drift or geometric inconsistency, the lag feedback mechanism will determine whether to roll back the modeled fragments based on the value of the trust label. If the reconstruction trust label of a certain area is too low, the system will roll back the reconstruction result of that area and re-model it to ensure that the final three-dimensional reconstruction structure maintains high precision and high consistency.

[0028] In terms of transparent timeline attribute management of spatial mapping patches and dynamic update of local structure temporal stack, the spatial mapping patches generate a transparent timeline attribute field after projection, which is used to indicate the timeliness of the structure. Specifically, when the tactical AR system performs spatial compensation based on historical data, the historical structure data of certain areas needs to be projected into the current three-dimensional model through mapping patches to complete the invisible areas caused by occlusion or changes in viewing angle. In order to ensure the timeliness of these patches, the system adds a transparent timeline attribute field in each projected patch. According to the time difference between the generation time of the patch and the latest visible data, the attribute field indicates the timeliness of the patch, that is, whether the information in the patch is still valid or needs to be updated. Over time, the transparent timeline attribute in the patch is dynamically updated, reflecting the aging degree and availability of the structure. In application, the transparent timeline attribute field controls the transparency of the patch through the relationship with time, so that the patch far from the current time gradually becomes transparent, indicating that the historical information of the area gradually loses its importance over time, thereby realizing the gray attenuation processing. In this way, the tactical AR system will preferentially display the current valid structure information when presenting the real-time three-dimensional reconstruction result, and gradually weaken the visual effect of the outdated or inefficient historical data, thereby improving the real-time response and data effectiveness of the system. Secondly, the management of the local structure temporal stack adopts a circular truncation strategy to optimize the use of system memory and computing resources. Under this strategy, the local structure temporal stack only retains the structure frames that appear in the visible area more than a predetermined threshold. Specifically, the system records all the structure frames generated in the three-dimensional reconstruction process of each local area, and stores these frames in the temporal stack in chronological order. In order to avoid excessive memory occupation or waste of computing resources, the system sets a threshold to filter out the structure frames that frequently appear in the current visible area. When the appearance frequency of a certain structure frame exceeds the threshold, the system retains it in the temporal stack to ensure that the frame data is effectively processed and updated. Conversely, for structure frames that appear less frequently, the system removes them from the temporal stack to release storage space and maintain system efficiency.

[0029] In terms of rigid block drift detection and feedback adjustment mechanism, the drift detection of rigid blocks is achieved by introducing a rotational inertia residual kernel to independently capture the micro-variation regions where pose drift occurs. During the three-dimensional reconstruction of the tactical AR system, the structures in space are divided into multiple rigid blocks, each representing a local region that maintains relatively stable shape and position in space. During real-time reconstruction, the system obtains the motion information of the helmet through an inertial measurement unit (IMU) and analyzes the attitude changes of these rigid blocks using the rotational inertia residual kernel. The rotational inertia residual kernel is a calculation method based on rotation matrix and inertial sensor data, which accurately detects and quantifies the small pose drift of each rigid block relative to its ideal position. When a rigid block experiences pose drift, the system can detect the micro-variation in that region through the rotational inertia residual kernel, and then determine whether there is a drift phenomenon. This method can effectively identify structural drift caused by rapid head movement, environmental changes or other factors, ensuring that the system can respond in time and make necessary adjustments. Next, the feedback adjustment mechanism introduces a regional lag buffer counter to effectively manage and respond to drift phenomena. When the system detects pose drift in a rigid block, the regional lag buffer counter counts the drift and continuously tracks its changes. The counter records the time and frequency of drift occurrence to determine whether the drift is a persistent error. If the drift of a rigid block occurs more than three times in a cycle, the system will trigger a visual prompt mechanism to provide visual markers through the AR helmet display interface, prompting the user to the presence of geometric discontinuity or drift error regions in the current field of view. Visual prompts can be in the form of flashing borders, color changes or transparency changes to attract the user's attention.

[0030] In conjunction with the attached Figure 3 In terms of geometric fitting and interactive feedback mechanism of information voxels, the tactical information metadata is parsed into a multi-level geometric structure set with task semantics , and according to the task urgency, task type and regional environment complexity, the appropriate geometric expression form (such as cone, sphere or edge box voxel) is dynamically selected for three-dimensional fitting. The tactical system determines which regions need high-precision reconstruction according to the priority of the task and environmental changes, and adjusts the geometric form for fitting accordingly to ensure accurate and rapid reconstruction of the most critical task areas. After the fitting voxels interact with the user, the system records the interaction behavior and generates a behavior response function based on the action time, operation type, response delay and voxel affected area of the behavior, which is uploaded to the tactical system as a behavior feedback log for subsequent analysis and decision-making reference.

[0031] Next, in order to feedback adjust the voxel, a behavior-driven voxel feedback adjustment formula is introduced, which is used to comprehensively evaluate the task significance of the voxel, the user behavior sensitivity and the regional strategy weight, so as to dynamically adjust the priority of the voxel embedding. The formula expression is as follows: Wherein: is the cumulative behavior response intensity function, which is used to evaluate the response importance of a certain intelligence voxel in time ; is the micro-time variable in response evolution, the integral parameter; is the user interaction action rate function (such as gesture, eye movement, voice trigger frequency) per unit time; is the gaze path offset function when interacting with the user; is the spatial strategy weight of the current voxel, representing its importance level in the current combat area; respectively represent the adjustment factor of interaction rate, attention change rate and task weight on the overall feedback; is the response saturation control coefficient, which is used to adjust the decay sensitivity of the voxel to long-time interaction; is the average attention duration estimate of the user to this type of voxel.

[0032] The derivation of the above formula is based on time series analysis and user behavior pattern, the main purpose is to quantify the response intensity of interaction behavior through integral function, and to consider the interaction frequency, attention shift and the importance of tactical area through polynomial and nonlinear function weighting.

[0033] The interaction process between the user and the intelligence voxel is essentially a dynamic event of time continuity, space correlation and semantic variable coupling, which shows a variety of quantifiable behavior characteristics in time . In order to more accurately capture these characteristics, the behavior response intensity function is set as an integral function with respect to time, and its integral core is the instantaneous intensity estimation model of interaction behavior, which is expressed as: Wherein, represents the instantaneous behavior response intensity at any time . In order to further refine this model, The nonlinear combination function is decomposed into three parts: the first part represents the user behavior activity in unit time, such as eye movement, gesture, speech, and other interaction frequencies. Let the rate function be Since the behavior intensity increases nonlinearly, the square enhancement is used: where is the adjustment coefficient. The second part considers the speed of user attention change, that is, the degree of shift of the gaze path over time, reflecting the task attraction and user interest jump. Let the path function be The derivative represents the gaze switching rate, with a weight factor The third part depicts the tactical weight value carried by the voxel in space, that is, the importance of the voxel changes over time. Let the value be and saturate it through the Sigmoid function to obtain: where, controls the steepness of the response rise, is the average response time of the user estimated by the system, that is, the central turning point.

[0034] After combining the above three parts, the complete instantaneous response intensity function is constructed, and the integral form is substituted to obtain the final behavior response evaluation function The function not only has time-domain accumulation ability, but also introduces spatial attention derivative and voxel semantic saturation modulation terms, which can analyze the user and voxel interaction process in multiple dimensions, especially suitable for real-time voxel scheduling, information credibility evaluation, interaction priority sorting, and other key function modules in tactical AR.

[0035] In terms of geometric construction mapping of task information and combat phase adaptability adjustment, cone, sphere and edge box are used to represent different tactical information, specifically, cone voxel is used to represent the indication of attack area, sphere voxel is used to identify target units, and edge box voxel is used to represent restrictive alert area. Each kind of geometric voxel corresponds to different tactical tasks in space, and is linked with task urgency through three-dimensional geometric color coding, so as to intuitively show the priority and urgency of the task. The system adopts different color mapping schemes to represent the task urgency of each voxel according to the task characteristics and priority of each area. For example, the cone voxel of the attack area adopts red color, indicating high priority and urgent attack task; the sphere voxel of the target unit adopts yellow color, indicating the task of alert or tracking; and the edge box voxel of the restrictive alert area adopts blue color, indicating the monitoring area with medium priority. The combination of these color coding and geometric construction can provide users with an intuitive and clear tactical view, helping tactical personnel quickly understand the task layout and priority in the current environment. The selection of geometric construction is based on the dynamic adjustment of the current combat phase of the tactical system. In the pre-combat phase, the system usually selects the embedded edge box voxel to construct and clarify the boundary of the tactical area, such as defense position, blockade line, etc., and the accurate definition of these areas is crucial for formulating tactical plans and resource allocation; while in the combat phase, the system dynamically switches to embedded sphere voxel or cone voxel, which are suitable for representing more dynamic tactical elements, such as enemy targets, attack directions, etc. In this case, sphere voxel can accurately represent the position and shape of target units, while cone voxel can be used to indicate the attack path and the range of threat area. Through this dynamic geometric voxel selection mechanism, the tactical AR system can intelligently adjust the shape of voxel in three-dimensional reconstruction according to different combat phases and task types, so as to more accurately reflect the tactical requirements.

[0036] In terms of interactive feedback and behavioral response classification of intelligence voxels, after the interaction with the user is completed, the intelligence voxel triggers the behavioral response classification module. The system classifies the behavior according to the interaction duration and operation type through this module, and marks the behavior as "attention", "avoidance" or "confirmation". The behavioral response classification module evaluates the user's behavior in real time according to the user's interaction mode with the intelligence voxel (such as gestures, eye movements, voice commands, etc.) and the interaction time. If the user stays on a particular intelligence voxel for a long time and repeatedly touches or gazes, the system will mark the behavior as "attention", indicating that the user has a high interest and attention to the information of the voxel; if the user shows avoidance behavior during interaction with the voxel, such as quickly moving away from the line of sight or repeatedly canceling the selection, the system will mark it as "avoidance", indicating that the information has a lower priority for the user's current task; and when the user's interaction with the voxel shows explicit confirmation behavior, such as long gaze, voice confirmation or gesture confirmation, the system will mark the behavior as "confirmation", indicating that the user has accepted or verified the tactical information. Subsequently, the behavior feedback log will contain the gaze path change trajectory before and after the user's interaction, and be used to evaluate the guiding effect of tactical information on spatial cognition. Specifically, the system records the movement trajectory of the user's line of sight during the interaction, including the change in the line of sight from the initial gaze position to the target voxel after the end of the interaction. This process can provide in-depth analysis of how users perceive and process tactical information. By tracking the changes in the gaze path, the system can evaluate the spatial cognition guiding effect of tactical information and determine whether the user can effectively obtain key information from the environment, whether there is information blocking, understanding deviation or cognitive bottleneck. The behavior feedback log not only provides detailed records of user behavior for the tactical system, but also provides data support for subsequent tactical decision optimization, which can help the system adjust the priority and method of information display, optimize the presentation of tactical data and the decision-making process of users.

[0037] In terms of preference inference and conflict detection of user interaction behavior, if the user performs similar operations on the same category of voxels for three consecutive times during the interaction process, the system infers that the user has a preference for this type of task state, and adjusts the embedding strategy of future voxels accordingly. The system analyzes the frequency and similarity of the user's operations on specific types of voxels (such as target identification, threat area, path indication, etc.) during the interaction process. When the user performs similar operations (such as long gaze, repeated clicks or selection of the same marker, etc.) on the same category of voxels for three consecutive times, the system infers that the user has a certain preference for this type of task state. For example, if the user selects or confirms the same target or the same area on the displayed tactical interface for three consecutive times, the system will identify that this area or task has a high priority. Based on this preference inference, the system will dynamically adjust the future voxel embedding strategy, giving higher rendering priority and more detailed display to related voxels, to ensure that the user can obtain the most relevant tactical information. Further, the system also performs conflict detection on the recorded user interaction behavior, especially in a multi-user cooperative combat environment. If multiple users in the same battlefield area mark the same intelligence voxel as "ignore", the system will trigger intelligence validity review. Specifically, the system detects whether there are multiple users in the same area who have marked the same intelligence voxel as "ignore". If such behavior occurs, the system believes that the intelligence voxel has a problem, that is, the information is outdated, irrelevant or misleading. In this case, the system will start the intelligence validity review mechanism to further verify the accuracy, timeliness and relevance of the intelligence. During the review process, the system will re-evaluate the validity of the voxel by reviewing other sensor data, historical intelligence or feedback information from different users, to ensure that the tactical information provided to the user always remains efficient, accurate and useful.

[0038] Embodiment 1: In conjunction with the accompanying Figure 4 In this embodiment, in a city tactical environment, tactical personnel A wears a tactical AR helmet to perform a three-dimensional environment reconstruction task. In this task, the main goal of the tactical personnel is to find and mark enemy hiding positions, while adjusting the reconstruction accuracy of the visible area in real time according to the urgency of the task and environmental changes, in order to obtain the most valuable tactical intelligence.

[0039] During this process, the action trajectory and head micro-movements of tactical personnel A are collected in real-time through IMU sensors, gyroscopes, accelerometers, and electromyography sensors. The tactical system continuously acquires behavioral data of the tactical personnel. These data include head movements, eye movement data, and small electromyography signals. Through the analysis of these signals, the system establishes a time and direction prediction relationship related to the observation direction. This relationship is based on the time series modeling results of the historical behavior trajectory of the tactical personnel and is corrected in real-time. For example, the head movement pattern of tactical personnel A in the past few minutes shows a certain regularity, and the tactical system predicts the next observation direction of the tactical personnel based on these regularities.

[0040] The system finds that tactical personnel A has always scanned the left area first and then quickly turned to the right area in the past three movements. Therefore, the system can predict that the tactical personnel will focus on the right front area in the next few seconds. The tactical system corrects the direction prediction relationship in real-time and allocates weight to the right front area where the tactical personnel will soon be looking in the preview area sector. Since the observation direction of tactical personnel A has changed, the system dynamically adjusts the reconstruction resources of the right area and allocates more computing resources to this area that will soon be focused on, while reducing the allocation of computing resources to areas that are no longer focused on, such as the left area.

[0041] After dynamically adjusting the direction weight distribution of the preview area sector, the tactical AR helmet begins to calculate the visibility projection estimation of the right front area in real-time. It is set that there is a building in this area that blocks part of the view. The system estimates the current view through terrain data and historical building structure information. For example, the location, size, and shape of the building are taken into account in the calculation, and the system estimates that the current view of tactical personnel A is blocked by 30%. Based on this estimation, the system further subdivides the right front area into two sub-areas, namely the completely visible area and the partially blocked area. The system adjusts the allocation of computing resources according to the visibility of these areas, and performs high-precision modeling for the completely visible area and low-precision modeling for the partially blocked area, ensuring that the calculation efficiency and precision of the reconstruction process are optimized.

[0042] Through this process, the tactical AR helmet can adjust the preview area sector in real-time and dynamically optimize the three-dimensional reconstruction results according to the observation direction and the influence of the terrain blocking structure. It is set that tactical personnel A will soon enter an enemy hiding area, and the system enhances the reconstruction accuracy of this area based on previous speculation and real-time observation dynamics, ensuring that the tactical personnel can clearly identify the enemy hiding position through the AR helmet and perform corresponding tactical operations.

[0043] The task of tactical personnel A is to explore and mark enemy hideout areas while performing real-time environmental perception. As tactical personnel A gradually enters the complex urban battlefield environment, the tactical AR helmet continuously updates the environmental three-dimensional data, adjusts the field of view area in real time, ensures high-precision reconstruction of key areas, and quickly responds to dynamic environmental changes. During the exploration process of tactical personnel A, several situations that need to be handled appear, including structural residuals, spatial mapping patches, rigid block drift, etc.

[0044] Firstly, tactical personnel A is walking through a narrow alley and gradually approaching a building. At this time, the AR helmet generates a pre-view sector in the direction of tactical personnel A's line of sight according to the previous view angle and position prediction. Due to the partial obstruction of the nearby building during the movement of tactical personnel A, the AR system finds that the geometric difference value of this area is large. By calculating the difference between the current reconstruction data and the historical data, the system detects the structural residual area of this region. In order to handle these structural residuals, the system compares the structural integrity curve of this area and calculates the structural evolution rate of this area in the past few seconds. For example, set the evolution rate of the structural integrity curve to 0.35 in the past 3 seconds, which means that the structure has changed significantly in this time. Since the evolution rate of this area exceeds the system's dynamic threshold of 0.3, the system decides to start crack modeling. The process of crack modeling divides the residual area into multiple small spatial units and performs local geometric reconstruction, thereby improving the reconstruction accuracy of the area. Each spatial unit is assigned a local reconstruction trust marker, and whether rollback modeling is needed is determined according to the strength of the trust marker. If the trust marker is low, the system will perform rollback reconstruction on the area to ensure the accuracy of the final three-dimensional structure.

[0045] Next, during the execution of tactical personnel A's task, the system finds that some areas cannot be fully visible at the current view angle due to obstruction. In order to compensate for these invisible areas, the system uses spatial mapping patches and projects them into the current three-dimensional reconstruction space. Each patch is assigned a transparency time axis attribute field after projection, which indicates the timeliness of the patch. For example, a patch represents an area that tactical personnel A has not been able to directly observe in the past 10 seconds, and the transparency time axis attribute field of the patch will be grayed out according to the time decay of the patch. If the time distance of the patch from the current view angle is too long, the transparency will gradually decrease, indicating that the effectiveness of the area information is declining. Set the distance of the patch from the last visible time to 20 seconds, the system sets the transparency of the patch to 50%, so that the user can identify the patch as historical data and needs to be updated.

[0046] At the same time, the system dynamically manages all historical frames through the management of the local structure temporal stack. Each time the tactical AR helmet updates the field of view, it stores the structure frame of the visible area and updates it according to the frequency of appearance of the area. If an area appears multiple times in the past 5 seconds (for example, more than 3 times at a set threshold), the structure frame of the area will be retained in the temporal stack, ensuring that the three-dimensional reconstruction data obtained by the user is the latest and most relevant. If the frequency of appearance of the area is lower than the threshold, the system will remove these infrequently appearing structure frames through a circular truncation strategy, saving computing resources and avoiding excessive system load.

[0047] During the movement of tactical personnel A, the system also found a problem: due to the complexity of the tactical environment, the rigid blocks in the system had a slight drift. In order to detect these drifts, the system introduces a rotational inertia residual kernel to independently capture and identify the micro-variable areas that have pose drift by analyzing the rotational inertia data of the system. For example, the system detects that the head of tactical personnel A has deviated by 1.5° in the past 2 seconds, causing the viewing angle of the helmet to slightly drift. Through the rotational inertia residual kernel, the system can independently judge this slight drift and determine its impact range. To deal with this drift, the system introduces a lag buffer counter, which triggers a visual prompt when the drift occurs and lasts more than three cycle periods. When the system detects that the pose drift has lasted for more than 3 cycles (i.e. 6 seconds), it triggers a visual prompt to remind the tactical personnel that there is a geometric inconsistency in the area. At the same time, the system starts a rollback mechanism to ensure that the reconstruction data of the area is corrected, avoiding three-dimensional structure errors caused by drift.

[0048] Embodiment 2: In conjunction with the attached Figure 5 , assume that tactical personnel A is performing a complex urban counter-terrorism task, in which he wears a tactical AR helmet and uses the AR system to perform real-time environmental three-dimensional reconstruction, while obtaining and analyzing tactical intelligence information through the helmet. During the task, tactical personnel A must constantly adjust his action strategy to efficiently detect and mark enemy hiding positions. To achieve this goal, the system needs to dynamically adjust according to the behavior of tactical personnel A and accurately display information according to the priority of the task.

[0049] At the beginning of the mission, tactical personnel A observes a potential enemy position based on the real-time environmental reconstruction function of the AR helmet and starts to mark the position in detail through the three-dimensional spatial reconstruction data displayed by the helmet. The system corrects in real time based on the past behavior trajectory of tactical personnel A through the time and direction prediction relationship, infers the line of sight direction of tactical personnel A, and dynamically adjusts the direction weight distribution of the pre-viewing area sector according to the prediction results. For example, the behavior trajectory of tactical personnel A in the past 10 minutes shows that tactical personnel A tends to first observe the top area of the building and then turn to the window of the building. According to these historical data, the system adjusts the weight of the pre-viewing area sector in real time, allocating more computing resources to the window area of the building while reducing the allocation of computing resources to other areas (such as the ground or areas far from the target).

[0050] As the mission progresses, structural residual areas appear in the field of view of tactical personnel A, which are caused by the geometric differences due to the obstruction of the building. The system calculates the geometric difference value of the area by comparing the current reconstruction data with the historical data. Assuming that the system finds that the structural integrity evolution rate of the area is 0.4 (which exceeds the set threshold of 0.3), the system therefore starts the cracking modeling mechanism. At this time, the system divides the residual area into multiple smaller spatial units and models these units according to the local geometric evolution. In order to ensure the accuracy of modeling, a local reconstruction trust label is attached to each spatial unit, indicating the reliability of the reconstruction of the area. If the trust label is low, it means that there is a high degree of uncertainty in the area, and the system will roll back the area for reconstruction in subsequent operations.

[0051] While processing the spatial residual, tactical personnel A continues to perform the task and interacts frequently according to the intelligence requirements. For example, tactical personnel A selects and confirms a target position in the AR interface for three consecutive times and gazes at the area for a long time. After detecting this behavior pattern, the system infers that tactical personnel A has a strong attention preference for the task area, based on which the system will adjust the embedding strategy of the area voxel and provide higher display priority for the area. In this process, the system generates a behavior response function and quantifies the interaction behavior of tactical personnel A. The system calculates according to the following formula: In this formula, is the user interaction action rate per unit time, and tactical personnel A frequently triggers 5 interaction operations through gestures and eye movements in the past 3 seconds, therefore . Setting the adjustment factor set by the system, the calculation result of the first term is: Next, set the gaze path of tactical personnel A to have a 2-degree deviation, and the gaze switching rate , and set , the calculation result of the second term is: In addition, set the spatial strategy weight of the current voxel , and the saturation control parameter used by the system , and At this time, the calculation result of the Sigmoid function is: Substitute these values into the formula to get: Set the interaction duration seconds, and the final behavior response strength is: This value indicates that tactical personnel A pays high attention to the region voxel, and the system will increase the embedding priority of the region voxel according to this result to ensure that tactical personnel A can obtain detailed information of the region in time.

[0052] At the same time, the tactical AR system will continue to monitor the behavior of multiple users in the region. If multiple users mark the same intelligence voxel as "ignore", for example, tactical personnel B and tactical personnel C both mark a certain enemy position voxel as "ignore" in the same battlefield region, the system will trigger intelligence validity review. The system first verifies the voxel in real time, checks the accuracy and timeliness of the information, and decides whether to update or re-evaluate the validity of the voxel according to the latest sensor data and tactical system feedback.

[0053] Tactical personnel A enters a complex urban battlefield environment, and as he advances, the tactical system displays a three-dimensional reconstructed battlefield environment in real-time through the helmet. Personnel A's mission is to search for and mark enemy hiding positions while ensuring safety. At this point, the system makes appropriate adjustments according to the current phase of the operation. In the pre-battle phase, the system prioritizes and intensifies the display of the boundaries of the tactical area, using edge voxels to identify important defensive or alert areas. The display of these edge voxels is mainly yellow, and through three-dimensional geometric color coding, it links to the task urgency, for example, if the task urgency of the area is high, the color will change to red, indicating that special attention should be paid to this area. When tactical personnel A enters the enemy's attack range, the system adjusts the geometric voxels in the reconstructed view to cone voxels to indicate the enemy's attack area or potential threats. These cone voxels are color-coded with red and orange, clearly showing the direction of attack and threat intensity. At the same time, the system projects sphere voxels in the scene, which are used to identify the location of enemy targets and provide accurate identification information. The color of these sphere voxels is closely related to their importance, with targets marked in green or blue, indicating that the target is in a priority attack or tracking state.

[0054] As the mission progresses, tactical personnel A stays in a certain area and interacts with the intelligence voxels in the AR helmet. In the tactical system, the system records the interaction process of tactical personnel A with the voxels. Set tactical personnel A to perform a continuous three times confirmation operation on a certain enemy position target displayed on the AR interface: the first click confirms the target position, the second time confirms the accurate coordinates through eye movement, and the third time confirms the attack priority of the target through voice command. The system classifies this behavior through the behavior response classification module, and marks the behavior as confirmation according to the duration of the interaction and the type of operation. According to this behavior classification, the system infers that tactical personnel A pays high attention to the target, and adjusts the voxel embedding priority of the area where the target is located, increases the visibility and accuracy of the target voxel, to ensure that tactical personnel A can clearly see the target and make correct decisions.

[0055] To further improve the accuracy of tactical decision-making, the system also records the gaze path change trajectory before and after user interaction. When tactical personnel A confirms the target, the system records his eye movement trajectory, analyzes the change of his gaze path, and evaluates how tactical information affects his spatial cognition. For example, set tactical personnel A's eye movement trajectory to show that he quickly shifts from initially focusing on the top of the building to the target area around the window and the building. The system evaluates the effectiveness of the guidance of tactical information through the data of the gaze path change trajectory, and uploads these data as a behavior feedback log to the tactical system, providing data support for subsequent tactical decision optimization.

[0056] During the subsequent task execution, Tactical Personnel A continues to interact with the system, and the system infers his preferences for different tactical information based on his behavioral patterns. For example, Tactical Personnel A selects and confirms the same enemy marker in a certain area on the displayed interface for three consecutive times, and the system infers that Tactical Personnel A pays high attention to the enemy target in this area through analysis of the interaction data. According to this behavior, the system adjusts the voxel tessellation strategy for this area, increases the rendering priority of this area, and enhances the detail accuracy of this area to provide more accurate information for Tactical Personnel A.

[0057] At the same time, the tactical AR system also detects conflicts in the recorded user interaction behaviors. Suppose that Tactical Personnel B and Tactical Personnel C are also executing similar tasks in the same battlefield area, and they simultaneously perform the "ignore" operation on the same enemy marker voxel displayed. After detecting this conflict, the system considers that there is a problem with the validity of the enemy intelligence voxel. The system triggers the intelligence validity review mechanism to re-evaluate the validity, accuracy, and timeliness of the voxel. If the enemy target information is outdated or the target no longer exists, the system adjusts the display state of the voxel and updates it according to new intelligence data to ensure that Tactical Personnel can obtain the most reliable and accurate information.

[0058] The tactical AR helmet can adjust the geometry according to different stages of combat, infer tactical preferences based on user interaction behaviors, and optimize the validity of intelligence information through a conflict detection mechanism. Ultimately, the system can provide accurate three-dimensional reconstruction, real-time feedback, and dynamic adjustment throughout the task process of Tactical Personnel A, ensuring that their decisions in complex environments are always based on the most reliable tactical information, thereby improving combat efficiency and the success rate of task completion.

[0059] The above shows and describes the basic principles, main features, and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above examples, and the above examples and descriptions in the specification are only to illustrate the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A tactical AR helmet real-time environment 3D reconstruction method, characterized in that: Based on the head micro-motion, neck posture changes and electromyographic signals of the tactical personnel, a time and direction prediction relationship corresponding to the observation direction is constructed to generate a pre-visual field sector of the area to be gazed at, and the reconstruction computing resources are preferentially allocated to the pre-visual field to form a visual field dynamic focusing mechanism; based on the spatial correspondence between the sensing data in the pre-visual field and the historical 3D structure, a structure residual area is extracted, and local geometric splitting evolution modeling is performed on the area; For the spatial area that is currently invisible due to changes in viewing angle or occlusion outside the pre-visual field, a local structure time stack is constructed, the historical visible angle of the area is traced back, and the historical structure is projected into the current 3D structure in the form of a spatial mapping patch to compensate for the perspective recession; during the construction of the 3D structure, the deformation continuity of each local spatial rigid block is monitored, and if structure drift or geometric inconsistency is detected, a dynamic lag feedback mechanism is triggered to adjust the abnormal area through visual cues and local reconstruction operations; After the reconstruction and compensation operations are completed, spatial intelligence metadata packets from the tactical system are received, and the data packets are converted into voxel structures with geometric shape characteristics and embedded into the current 3D relationship, enabling the structural fusion and interactive embedding of tactical information in the reconstructed space.

2. The tactical AR helmet real-time environment 3D reconstruction method of claim 1, wherein The time and direction prediction relationship is based on the time series modeling results of the tactical personnel's past behavior trajectory for real-time correction, which is used to dynamically adjust the direction weight distribution of the pre-visual field sector; the calculation of the pre-visual field sector depends on the observation direction and the visibility projection estimation of the current terrain occlusion structure.

3. The tactical AR helmet real-time environment 3D reconstruction method of claim 2, wherein The geometric difference value in the structure residual area is obtained by comparing the structure integrity evolution rate, and the splitting modeling is started when the local geometric evolution exceeds the dynamic threshold; the space unit after splitting modeling contains a local reconstruction trust marker, which is used for the subsequent lag feedback mechanism to determine whether the modeling segment needs to be rolled back.

4. The tactical AR helmet real-time environment 3D reconstruction method of claim 3, wherein The spatial mapping patch generates a transparent time axis attribute field after projection, which is used to indicate the timeliness of the structure and perform gray attenuation processing according to the distance from the last visible time; the management of the local structure time stack adopts a circular truncation strategy, retaining structure frames with a frequency higher than the threshold in the visible area.

5. The tactical AR helmet real-time environment 3D reconstruction method of claim 4, wherein The drift detection of the rigid block introduces a rotational inertia residual kernel to independently capture the micro-variation area where the pose drift occurs; the feedback adjustment mechanism introduces a region lag buffer counter, which triggers a visual cue when the drift appears more than three times in a cycle period.

6. The tactical AR helmet real-time environment 3D reconstruction method of claim 5, wherein The voxel structure embedding process supports the parsing of intelligence metadata into multiple levels of geometric construction forms and maps different forms of tasks including cones, spheres, and edge box voxels; the intelligence voxel records the response behavior after the user interacts with it and returns it to the tactical system to form a behavior feedback log.

7. The tactical AR helmet real-time environment 3D reconstruction method of claim 6, wherein The cone, sphere and frame voxels correspond to attack area indication, target unit identification and restrictive warning area respectively, and the task urgency is linked through three-dimensional geometric color coding; the selection of the geometric structure is based on the dynamic adjustment of the current combat phase of the tactical system, and the frame voxel is embedded in the pre-combat phase, and the sphere or cone voxel is embedded in the combat phase.

8. The tactical AR helmet real-time environment 3D reconstruction method of claim 7, wherein The intelligence voxel triggers a behavior response classification module after the interaction is completed, the classification module marks the behavior as attention, avoidance or confirmation according to the interaction duration and operation type; the behavior feedback log contains the gaze path change trajectory before and after the user interaction, which is used to evaluate the guiding effect of tactical information on spatial cognition.

9. The tactical AR helmet real-time environment 3D reconstruction method of claim 8, wherein If the user performs similar operations on the same category of voxels for three times in succession, it is inferred that the user has a preference for the task state of this category, and the future voxel embedding strategy is adjusted; Conflict detection is performed on the recorded user interaction behavior, and if multiple users in the same battlefield area mark the same intelligence voxel as ignored, intelligence effectiveness review will be triggered.

Images