165 results about "Geometric mapping" patented technology

The invention discloses an intelligent control method and system for AR equipment, and the method specifically comprises the steps: determining a pupil center position coordinate and a cornea reflection point position coordinate of a user according to an eye image of the user, and inputting the pupil center position coordinate and the cornea reflection point position coordinate into a geometric mapping model for matching, obtaining user sight direction information; acquiring image brightness value distribution of an area near a user fixation point, determining a plurality of candidate saliency areas according to the image brightness value distribution, and comparing the candidate saliency areas with a preset saliency area template to determine a user attention focus; and acquiring head posture information of the user in real time by adopting a head posture tracking algorithm, and dynamically controlling the shooting direction of a camera of the AR equipment according to the head posture information, the sight direction information of the user and the attention focus of the user. According to the method and the device, accurate alignment of the AR content and the visual focus of the user is realized, and the immersion and interaction experience of the AR application are greatly improved.

The invention provides an unmanned aerial vehicle optimal route planning method and system fusing image recognition and M-RRT and APF algorithms. The point cloud and visual data of a fan are obtained in real time through a multi-mode sensor, key parts of blades are recognized based on deep learning, a geometric mapping model is constructed, the blade tip points are corrected in a clustering mode in combination with DBSCAN, and the shutdown posture is predicted. A global path adopts an M-RRT algorithm improved by a direction heuristic factor to guide and search a key area of a blade; a local track is optimized through a dynamic weight APF algorithm, and a repulsive force field is adjusted in real time to cope with attitude changes. An online re-planning mechanism is introduced, NSGA-III multi-target optimization is triggered when the environment suddenly changes or the tracking error exceeds a threshold value, and the optimal track is generated by integrating energy consumption, time and safety. The system integrates high-precision sensing, self-adaptive planning and dynamic optimization, and the inspection coverage rate and the track safety under the complex shutdown attitude are remarkably improved.

The invention discloses a building engineering progress automatic identification early warning system based on computer vision, which relates to the field of building engineering informatization and comprises a synchronous calibration module, a joint calibration module, a pose generation module, a mapping construction module, a resampling module, a mapping registration module and a comparison early warning module. Clock synchronization and rolling readout calibration are carried out on a camera and an inertial measurement unit, a continuous time pose is established in a frame, sampling is carried out according to rows, and an equivalent global shutter frame is generated in combination with plane and pixel-level geometric mapping; outputting a camera pose track and a local measurement map by adopting back-end optimization containing a rolling mechanism, and registering with the building information model; component detection, instance segmentation and state discrimination are completed on an equivalent global shutter frame, pixel domain measurement is unified into engineering quantity under pose and registration constraints, the engineering quantity is mapped to a work decomposition structure and a progress plan, deviation is calculated, and graded early warning is output according to a multi-threshold rule. Long-term stable and reliable operation and evidence traceability of the system are guaranteed through whole-course quality monitoring and threshold grading.

The invention discloses a three-dimensional geologic model subdivision method based on an earth space grid structure, which adopts a plurality of types of earth space grid structures including longitude and latitude grids, octahedral projection grids and Hilbert curve filling grids to carry out three-dimensional division on the earth surface, and combines a layered recursive grid division technology to obtain a three-dimensional geologic model. Multi-level grid unit generation and geometric mapping from coarse resolution to high resolution are achieved, grid parameters are dynamically adjusted according to geologic features, large-size grids are adopted in a simple area to improve efficiency, small-size grids are adopted in a complex area to capture details, redundant units are eliminated in combination with a grid optimization algorithm, and the efficiency is improved. The accuracy and the calculation efficiency of the model are improved; a global unique identifier is generated through a GeoSOT-3D hierarchical recursion technology, union set, intersection and difference set operations of grid units are completed in combination with efficient Boolean operation, and the integrity and topological consistency of geologic model subdivision are ensured.

The invention relates to the technical field of industrial robot welding control, and provides an off-line programming-based elevator car bottom welding path field debugging method, which comprises the following steps of: constructing a three-dimensional process simulation model of a welding robot and an elevator car bottom, and establishing a geometric mapping relationship between the model and a physical scene; virtual welding scene parameters and robot body parameters are configured in the offline programming system; generating an initial welding track based on the 45-degree inclined welding surface, and executing multi-dimensional collision detection and kinematics verification; the optimized welding track is coded into an industrial robot instruction in an ANSI / UTF-8 format, and the industrial robot instruction is deployed to a physical welding system; collecting multi-source sensing data in the welding process in real time, wherein the multi-source sensing data comprise six-axis attitude parameters, process parameters and environmental parameters; and a dynamic comparison mechanism of the actual welding data and the simulation prediction data is established, and when the deviation exceeds a preset threshold value, an iterative optimization closed loop is triggered. The method has the beneficial effects that the efficiency is improved, the precision is guaranteed, the process stability is enhanced, and the compatibility is expanded.

The invention discloses a method for optimizing micro-crack segmentation based on deep learning and super-resolution reconstruction, and the method comprises the steps: cutting an image in real time, obtaining an image block which takes a component as a target main body, and synchronously recording a homography matrix for geometric mapping; selecting an amplification strategy to improve the resolution, and recording a scale mapping relation; inputting the enhanced image block into a double-flow network; adaptive fusion and reconstruction are carried out on the two branch features, and a high-resolution texture image is output; generating a geometrically corrected ortho-image, fusing the geometrically corrected ortho-image with original illumination information, and outputting a corrected image with a known pixel size; identifying cracks, spalling and honeycomb diseases in parallel; generating a unified defect confidence map; calculating real geometric parameters of the BIM in a BIM global coordinate system through coordinate back projection; and generating quantitative defect reports and maintenance suggestions. The method has the advantage that seamless connection between the detection result and the BIM global coordinates is realized.

The invention provides a coal face mining auxiliary method based on visual and point cloud data fusion modeling, and the method comprises the steps: collecting the visual image data of a coal face, collecting the three-dimensional point cloud data through a point cloud sensor, and obtaining the motion state information through an inertial sensor; based on internal and external parameters and IMU data of joint calibration, mapping point cloud data to an image plane, establishing a geometric mapping relationship between three-dimensional points and pixels, generating an attribute point cloud model containing geometry, color and semantics, and generating a high-precision global three-dimensional digital twinning model; and identifying and tracking equipment instances in the digital twinborn model, and resolving motion parameters of key components based on a URDF kinematics model and a principal component analysis algorithm to support remote control decisions. According to the invention, high-precision real-time sensing and three-dimensional visualization of the working space size, the equipment connection relation and the movement position of the coal face can be realized, and the remote control precision and the response efficiency of unmanned mining are remarkably improved.

The invention discloses a method for constructing a three-dimensional geographic entity and a three-dimensional electronic map, which comprises the following steps of: firstly, classifying acquired basic surveying and mapping data based on a constructed multi-level classification framework, and determining an entity category to which each element in the basic surveying and mapping data belongs; then, according to entity category design mapping data corresponding to the elements, converting the mapping data into a geometric mapping relation and an attribute mapping relation of a two-dimensional geographic entity, and through the geometric mapping relation and the attribute mapping relation, converting the mapping data and the attribute information corresponding to the elements into geographic entity geometry and geographic entity attributes; semantized geographic two-dimensional entities are generated to effectively inherit and structurally reuse stock data, and massive waste and repeated construction of data resources are avoided. And then, selecting a corresponding three-dimensional conversion method according to the type of the two-dimensional geographic entity, and converting the two-dimensional geographic entity into a three-dimensional geographic entity. And finally, constructing a corresponding three-dimensional electronic map according to the three-dimensional geographic entities corresponding to all the elements.

The invention discloses a cross-view multi-target collaborative perception and global tracking method. The method comprises the following steps: receiving original video streams of different views in parallel; carrying out feature extraction and content perception feature recombination processing; establishing a homography transformation matrix representing the mapping relation between the visual angles; performing forward geometric mapping by using the homography transformation matrix, and performing reverse mapping, secondary confirmation and state updating; a joint cost model based on space-time-appearance fusion is constructed, the potential association degree of unassociated residual targets is quantified, and missing targets are recalled; performing duplicate removal and optimization on the targets in the association set; and training and optimizing the network. According to the embodiment of the invention, the method achieves the capturing of a pixel-level target, and improves the texture and edge expression capability of a tiny target in a deep feature map. Efficient fusion of multi-dimensional spatio-temporal information among visual angles is realized; higher robustness and tracking continuity are achieved, and the integrity and fineness of a final trajectory result are improved; and the recall is ensured, and the generation of redundant tracks is avoided at the same time.

The invention relates to the technical field of signal processing of intelligent voice equipment, and discloses an equalization method and system for a Chirp modulation signal of audio equipment, and the equalization method for the Chirp modulation signal of the audio equipment comprises the steps: analyzing the non-stationary characteristic of the Chirp modulation signal through an information geometric mapping algorithm, and generating non-stationary complexity mapping; processing the Chirp modulation signal by using a multi-scale entropy analysis algorithm to generate a multi-scale complexity spectrum; extracting long-range correlation characteristics of the Chirp modulation signals by adopting a fractal analysis algorithm, and outputting Hurst index characteristic mapping; constructing a complexity-perceived adaptive equalization system; a time-frequency joint optimization algorithm is constructed, and global optimization and adaptive adjustment of equalization parameters are realized; signal complexity characteristics are quantized by using a multi-scale entropy analysis algorithm, and the adaptability of the equalization system to signals with different complexities is enhanced.

The invention relates to the field of ship design and manufacturing, in particular to a ship outfitting piece three-dimensional reconstruction method based on a single RGB image. The method comprises the following steps: acquiring a multi-view data set of an outfitting piece, and performing geometry and texture optimization on an input single RGB image through a diffusion model to generate a multi-view image consistent with the features of the outfitting piece; performing feature extraction on the multi-view image by using a convolutional network to generate a geometric feature sequence and a texture feature sequence; reconstructing a three-dimensional model of the fitting-out part through a geometric mapping network in combination with three-dimensional geometric prior; and carrying out real-time geometric verification and detail optimization on the generated three-dimensional model. According to the method, through combination of deep learning and the geometric mapping network, the three-dimensional model of the ship outfitting piece is rapidly reconstructed from a single RGB image with high precision, the modeling efficiency and precision are remarkably improved, manual intervention is reduced, the precision requirement of ship design and manufacturing is met, and the practical value is high.

The invention relates to the technical field of airborne wind power measurement, and discloses an airborne wind measurement laser radar method and system. The method comprises the following steps: generating a scanning time sequence identifier and a path correction mark structure through scanning beat synchronization and stray light shielding processing; executing time-frequency demodulation and quality gating based on the back scattering signal to generate an observation wind speed initial value sequence; performing time alignment and motion compensation in combination with inertial navigation data, and generating a wind speed vector grid fragment through spatial interpolation and scanning geometric mapping; and finally, generating a configuration updating instruction structure through coordinate alignment and grid splicing. According to the invention, the precision and the anti-interference capability of wind field measurement in an airborne environment are effectively improved, and the data reliability and the system stability under a complex meteorological condition are remarkably enhanced through multi-source data fusion and a self-adaptive compensation mechanism.

The invention provides a target detection and matching method, system and device based on a ship visible light infrared image and a medium, and belongs to the technical field of ship target detection. A ship target in the image is detected and recognized, and NHA features are extracted; calculating the NHA feature similarity of the ship target in the image, and determining a matching target; calculating a homography matrix according to the key feature points, and constructing a homography matrix library based on the homography matrix; realizing coverage of geometric mapping relations under different focal lengths in the homography matrix library; and detecting potential targets around the mapping position in the infrared image, supplementing missed ship targets, and fusing the complemented and detected targets into a final detection result. The method not only ensures the high-precision corresponding relation of the target between the visible light image and the infrared image, but also can supplement the missed target through geometric mapping when the deep learning algorithm is missed, thereby improving the integrity of target detection.

The invention discloses a tunnel cross-sectional image analysis method based on image processing, and aims to solve the problems that a surface image is not clear in correspondence with a transient seismic wave method, a geological radar and a resistivity imaging section space, and anomalies at a certain distance in front are difficult to map to a tunnel face. According to the method, anisotropic reforming is carried out by adopting a Fourier neural operator, cross-modal Transform registration of a micro attention mask based on sector geometry, curve mileage and section polar coordinate position coding is combined, analytic geometry mapping and uncertainty propagation are matched, tunnel face structure traces and water seepage texture evidences are fused, and the tunnel face structure traces and the water seepage texture evidences are combined. The technical effects of accurate positioning of abnormity on the tunnel face, position confidence range estimation, risk grading early warning, stripe artifact suppression, abnormal boundary reservation and output of structured results of mileage stake numbers, azimuth angles, distance intervals and the like are achieved.

The invention discloses an automatic texture color mapping method for large-scale point cloud data. The automatic texture color mapping method comprises the following steps: firstly, realizing geometric alignment of geographic texture data and point cloud data; then realizing release of intra-core buffer by means of a mass point cloud space segmentation and retrieval management technology, and segmenting the mass point cloud into a plurality of regional point cloud blocks; carrying out filtering processing on the regional point cloud blocks to filter out discrete flying spots; secondly, considering correction constraints of different texture source information to realize seamless generation of geographic texture data; secondly, efficient point cloud preliminary color assignment is achieved by means of the spatial relation between geometric mapping and local aggregation of regional point cloud blocks and geographic texture data; secondly, eliminating the texture information (color) of the wrong coloring point cloud by using the self-shielding judgment result of the point cloud; and finally, realizing spatial merging of regional point cloud block data sets by using an efficient out-of-core file merging technology. According to the method, the dependence on hardware equipment (such as a large memory and a high graphics card) is avoided, and the block processing of massive point clouds is efficiently realized.

The invention belongs to the field of image processing, and provides a projection image automatic correction method and system based on deep learning, and the method comprises the steps: controlling a projector to output a preset test image; acquiring a first image of a user at a viewing angle by the mobile terminal; acquiring a second image at a projection angle by a shooting device of the projector; extracting features and predicting adjustment parameters for projection correction based on a projection image correction model of deep learning; the adjustment parameters are issued to the projector through a network; and performing pre-distortion processing on an image signal to be projected based on the geometric mapping parameter, and performing pixel-level adjustment of brightness and color based on the compensation parameter to generate a corrected projection output image. According to the method, the usability and accuracy of projection image correction can be improved.

The invention discloses a high-definition map processing method and device based on mobile laser radar data, and the method comprises the steps: obtaining sensor data collected by a plurality of sensors of an autonomous vehicle, carrying out the probability data processing, obtaining a structured 3D point cloud, carrying out the recognition of the 3D point cloud, distributing a semantic tag, and creating a geometric mapping function; the method comprises the following steps: constructing a 3D point cloud image, converting local Euclidean coordinates by adopting a geometric mapping function to obtain a semantic segmentation result, classifying and correcting the semantic segmentation result by adopting a confusion network to obtain object attributes, constructing a semantic graph of the 3D point cloud image based on the object attributes and the geometric mapping function, and outputting a high-definition map corresponding to the semantic graph. According to the method, the high-definition map is evaluated and optimized to obtain a path navigation result of the autonomous vehicle, multi-target optimization is applied to processing the mobile laser radar point cloud to create the high-definition map, a confusion matrix is used to merge uncertainty simulation and probability semantic mapping, and the reliability and accuracy of map generation are improved.

The invention provides a three-dimensional model feature recognition method based on a heterogeneous graph neural network, and belongs to the technical field of computer-aided manufacturing. Comprising the following steps: selecting a three-dimensional model database with feature tags; respectively and sequentially obtaining a topological set, a geometric information set and a normalized geometric information set of each three-dimensional model, and respectively carrying out geometric feature extraction on each piece of normalized geometric information to obtain a corresponding geometric mapping feature set; constructing an initial heterogeneous graph neural network, dividing data of all three-dimensional models in the three-dimensional model database into a training set, a verification set and a test set by using all geometric mapping feature sets, and training, verifying and testing the initial heterogeneous graph neural network in sequence to obtain a final heterogeneous graph neural network; and performing feature recognition on the three-dimensional model by using the final heterogeneous graph neural network. According to the invention, the feature recognition capability and recognition accuracy of the complex three-dimensional model can be effectively improved.

The invention discloses a team state evaluation method based on an index dimension system. The method comprises the following steps: firstly, constructing a comprehensive and customized evaluation system through a preset evaluation dimension and an evaluation index corresponding to the evaluation dimension; then obtaining target evaluation index scores of the team based on the evaluation indexes, converting an evaluation system into computable data, accumulating the target evaluation index scores based on the evaluation dimensions to obtain evaluation dimension scores corresponding to the evaluation dimensions of the current team, and comprehensively reflecting the overall performance of the team in the evaluation dimensions; obtaining an evaluation dimension mapping score through equal ratio mapping processing based on the evaluation dimension score in different years, and eliminating annual difference; the evaluation dimension mapping score is visually processed and displayed through a radar map, and improved strategies such as team states, auxiliary decision making and resource allocation are visually presented.

The invention discloses an intelligent human body modeling method and system based on a three-dimensional dot matrix, and relates to the field of intelligent modeling, and the method comprises the steps: carrying out the structural analysis of the semantic level of an original three-dimensional point cloud, and predicting a space displacement field pointing from the surface of observed clothes to the surface of an internal real human body under the semantic guidance. By applying the displacement field, double transformation of point cloud stripping and standardization is carried out on the point cloud, so that repeated fine tuning and optimization are carried out on the body type and posture parameters of the human body model in a standardized space, and finally a high-fidelity human body three-dimensional model is generated through driving. In this way, the complex clothes interference problem is deconstructed into a staged geometric mapping and parameter optimization process, and therefore the robustness and accuracy of human body three-dimensional reconstruction under the real wearing condition are effectively improved.

The invention discloses a geometric analytic expression three-dimensional reconstruction method, and relates to the technical field of computer vision and three-dimensional environment perception. According to the method, for an application scene with a target area located in a known space attitude plane, an accurate geometric mapping relation among a navigation coordinate system, a camera coordinate system and an imaging plane is constructed by fusing a visible light image, a single-point distance measurement value and attitude information output by an attitude sensor (such as an IMU); and pixel-level depth estimation is realized by adopting an analytic expression calculation strategy, so that high-precision and low-complexity three-dimensional reconstruction of a coplanar region is completed.

The invention relates to the technical field of video collection, in particular to a video synchronous collection method and device based on an optical far image, and provides the following scheme: obtaining a calibration image of to-be-synchronized content and a user posture parameter; calculating a geometric mapping matrix and a point spread function through the calibration image and a preset calibration sequence, and performing geometric predistortion and frequency domain compensation on the to-be-synchronized content to obtain a compensation frame; and combining with the rotation range predicted by the attitude parameters, generating a sub-frame sequence by adopting multi-view rendering of time multiplexing, and outputting the sub-frame sequence by a display end, so as to realize stable far-image watching experience under the condition of small desktop size. According to the scheme, visual fatigue caused by close gaze can be relieved, and the intelligent level of learning interaction is improved.

The invention belongs to the field of structure optimization design, and particularly discloses a multi-model topological optimization method and system for a complex geometric model, and the method comprises the steps: building a corresponding finite element model according to a geometric model of a wing corrugated structure; dividing a design domain for the geometric model, and establishing a topological model corresponding to the finite element model; geometric mapping is carried out on the shared design domain through manufacturing constraints with repeated modes, a mapping relation of a local coordinate system for processing a large deformation area is established, a topology model capable of being synchronously optimized is formed, and an output file of the topology model is exported; and carrying out synchronous loop optimization on the plurality of topology models by utilizing a parallel computing interface, and judging optimization results of the topology models according to a convergence criterion until the topology models reach the convergence criterion, so as to obtain a final optimization model. According to the method, the geometric mapping effect and the multi-working-condition synchronous optimization effect of the complex geometric model can be improved.

The invention discloses a determination method for an injection orifice of an enhanced vapor injection compressor, the opening angle theta open and the closing angle theta close of the injection orifice in an enhanced vapor injection loop in the compressor are determined, accurate determination of the position of the injection orifice is realized by combining a three-circle positioning geometric mapping method, and the injection orifice can be accurately determined through geometric constraints of a base circle, an auxiliary circle and a rotor inner circle trajectory circle. The settable area of the jet orifice on the partition plate is precisely defined, the maximum allowable outline of the jet orifice is determined through the inscribed circle, it is ensured that the jet orifice can be reliably opened and sealed within the calculated ideal angle range in the rotor movement process, gas backflow and mechanical interference are prevented, the timing precision of enhanced vapor injection and the control accuracy of the gas injection amount are improved, and the service life of the jet orifice is prolonged. The adaptability and the operation stability of the compressor under different working conditions are improved, and the injection efficiency is improved.

The invention relates to the technical field of medical auxiliary diagnosis, and discloses a hypospadias penis curvature measuring method based on physical simulation inversion. The invention aims to solve the technical pain point that the penile erection curvature and the internal fibrosis focus distribution are difficult to quantify noninvasively and accurately before the operation in the existing hypospadias diagnosis and treatment. According to the method, a standard white mold containing fine anatomical partitions is constructed, personalized geometric mapping is realized by using a static look-around image, and constraint physical parameter inversion is performed in combination with a multi-attitude dynamic stress image, so that the digital twin with real biomechanical properties is reconstructed in a virtual environment. According to the method, the erection bending form can be accurately predicted in the non-erection state, the mechanical property abnormal area range is positioned, operation planning is assisted, experience medical treatment is spanned to accurate digital medical treatment, and the wound diagnosis and treatment and operation risks are remarkably reduced.

The invention provides an organ three-dimensional reconstruction method and system based on a sparse ultrasonic section and a medium, and the method comprises the steps: obtaining a CT image sequence and an ultrasonic image sequence of an organ, and generating a complete point cloud and a sparse point cloud through processing; the point cloud is converted to a unified coordinate system through a registration technology, and position matching is carried out; inputting the matched sparse point cloud and complete point cloud into a point cloud reconstruction network for training, and learning a geometric mapping relation from the sparse point cloud to the complete point cloud; and obtaining a to-be-reconstructed organ sparse point cloud, inputting the to-be-reconstructed organ sparse point cloud into the trained point cloud reconstruction network to generate a complete point cloud, and drawing a three-dimensional contour surface of the organ. According to the invention, intelligent reconstruction of a complete three-dimensional organ contour can be realized through organ contour information obtained by a limited number of two-dimensional ultrasonic sections; the reconstruction precision and the surface integrity of sparse ultrasonic data are remarkably improved, an efficient and accurate three-dimensional reconstruction solution is provided for medical image analysis, and clinical diagnosis and treatment planning are assisted.

The invention discloses a virtual reality roaming method and system based on a multi-view panoramic image model, and belongs to the technical field of computer vision and virtual reality. Through a multi-viewpoint panoramic image set, real-time generation of a high-quality panoramic image from a known viewpoint to any unknown viewpoint is realized based on a spherical projection viewpoint transformation algorithm. Modeling generation methods for a road scene and a region scene are respectively designed according to scene feature differences, so that scene representation with complete coverage is constructed. According to the system, adjacent viewpoints are quickly positioned through a spatial index mechanism, geometric mapping and multi-viewpoint weighted fusion are performed by using the spherical projection algorithm, and a smooth and continuous free viewpoint roaming picture is rendered in a virtual reality environment. According to the method, on the premise of ensuring high fidelity of the scene, real-time rendering is realized on common consumption-level hardware, and the coherence and the degree of freedom of immersive interactive experience are improved.

The invention provides a deployable flexible GMR eddy current testing method and system for a complex curved surface component, and relates to the technical field of nondestructive testing. The method comprises the following steps: establishing a curved surface coordinate system and an expansion domain coordinate system, constructing a geometric mapping relation of a flexible sensing array, and performing clearance compensation, attitude correction and temperature drift correction on multi-frequency response data in an expansion domain to form a unified response data set; defect response distribution is generated through robust reconstruction and reprojected to the curved surface coordinate system, and space positioning and size and orientation recognition of defects are achieved; and adaptively updating the attitude, the path and the excitation scheme in combination with the fitting state parameters until the detection coverage and the confidence meet the requirements. According to the method, high-precision detection and process traceability of the defects of the complex curved surface component are realized.

The invention discloses a formation control method for an unmanned aerial vehicle cluster, and relates to the technical field of intelligent formation of unmanned aerial vehicles, and the method comprises the following steps: S1, obtaining an initial space state of the unmanned aerial vehicle cluster; s2, performing adjacent constraint construction by using the initial state vector; s3, performing geometric mapping by using the local adjacency relation graph; s4, performing global formation generation by using the relative geometric position matrix; and S5, performing dynamic formation control calculation by using the relative geometric position matrix and the target formation reference model. By setting multi-dimensional measurement calculation based on an initial state vector and combining a combined constraint set of a spatial distance, a speed difference and an angular speed difference, a weighted local adjacency relation graph is formed, and according to the scheme, spatial proximity and dynamic consistency can be considered while the adjacency relation is constructed; the generated graph structure not only reflects the geometric closeness degree between the unmanned aerial vehicles, but also quantifies the compatibility of the unmanned aerial vehicles in motion and attitude changes.

The invention provides a semantic scene reconstruction and interaction method and system based on a visual language model, and relates to the technical field of artificial intelligence and robots. The method comprises the following steps: acquiring image data shot by a camera; estimating a world pose of the camera; inputting the image data to a Visual Language Model (VLM) to generate a structured description containing an object semantic tag and quantitative spatial information thereof relative to the camera; calculating the world pose of the object according to the camera pose and the quantitative space information; a node representative of the object is created or updated in a hierarchical scene graph. The invention further provides a corresponding interaction system. According to the method, a rich and visual three-dimensional semantic map facing natural language interaction is constructed by deeply fusing the semantic understanding ability of the VLM and the geometric mapping ability of the SLAM, and the intelligent level of robot environment perception and man-machine interaction is greatly improved.