477 results about "Octree" patented technology

The invention discloses a laser SLAM (Simultaneous Localization and Mapping) method in a dynamic environment of a well mine, which relates to data processing and comprises the following steps of: performing mapping by utilizing a Cartograph SLAM algorithm; segmenting the PCD point cloud map by adopting an octree-based segmentation method according to a preset movement or deformation frequency threshold value of an object in each region in the well mining environment; adjusting the map resolution of each area in the PCD point cloud map according to the segmentation result; a change detection algorithm is used to carry out change detection on an area with object movement or deformation frequency exceeding a threshold value in the segmented PCD point cloud map, and new point cloud data of the change area is updated to the original PCD point cloud map in an increment mode; carrying out registration and fusion on the PCD point cloud map after local updating and an original PCD point cloud map; the global consistency is improved by aiming at the poor global consistency existing in the laser SLAM mapping of the well industry and mining in the prior art.

The invention discloses an interactive large-screen visualization system and method based on multi-dimensional data real-time rendering, relates to the technical field of image rendering, and aims to solve the bottleneck problems of a traditional system in the aspects of data throughput efficiency, rendering delay and interactive response. The method comprises the following steps: dividing space blocks by adopting a non-uniform octree, dynamically adjusting the size of the blocks according to data density, and combining a ZSTD + Hoffman hybrid compression algorithm; a CPU-GPU three-level asynchronous processing assembly line is constructed, and task priorities are dynamically adjusted through a fixation point weight algorithm; the master node dynamically allocates LOD parameters, and the slave nodes achieve special-shaped large-screen sub-pixel-level splicing based on a spherical coordinate system model; a Q-learning algorithm is adopted to predict a user operation sequence, and dual-channel bandwidth adaptive allocation is realized in combination with an ASTP protocol. The system comprises a dynamic partitioning module, a pipeline scheduling module, a multi-node cooperation module, an intelligent preloading module and an exception handling module. Real-time interaction visualization of a large amount of data can be realized, the rendering efficiency is improved, and hardware resource occupation is reduced.

The invention discloses a BIM model automatic generation method and system based on point cloud data, and belongs to the field of building information modelling, and the transmission method comprises the steps: obtaining original point cloud data, and employing a filtering algorithm based on point cloud density adaptive adjustment to carry out the preprocessing of the point cloud data; constructing a voxel octree structure for the preprocessed point cloud data, and performing semantic classification on the point cloud; geometric modeling is carried out based on the segmented point cloud subsets, and a fitting algorithm is adopted to carry out shape completion on a point cloud area; semantic annotation is carried out on the components subjected to geometric reconstruction, a corresponding relation between component types and spatial attributes is constructed, fusion features based on a point feature histogram and a local curvature are adopted, and classification is carried out; a standard BIM component family is converted, and a three-dimensional BIM model is constructed through the mapping relation. According to the method, the voxel octree data structure and the deep semantic segmentation neural network model are combined, division and semantic recognition are performed on the point cloud data, the intelligent degree of the model is improved, and manual intervention is reduced.

The invention relates to the technical field of data processing, in particular to an earthwork balancing method based on discrete elevation points and a free-form surface creation technology, and the method comprises the steps: generating topographic feature partition tags through an improved DBSCAN density clustering algorithm, and constructing a regularized elevation field in combination with octree and R * tree mixed indexes; generating an elevation standard deviation thermodynamic diagram based on Gaussian process regression, and outputting a multi-target digging and filling scheme set in combination with an FEM-DEM coupling model; a multi-layer graph model fused with soil bearing capacity is constructed, a collision-free construction path is optimized through an improved D * Lite algorithm, a positioning track and a planning deviation are fused by using Kalman filtering, and an online sequence extreme learning machine is triggered to realize dynamic updating of the model. According to the method, the complex terrain modeling efficiency and the earthwork volume calculation precision are improved, the resource allocation path is optimized, the construction cost is reduced, and the problem of excavation and filling imbalance caused by data discreteness and modeling staticization in a traditional method is solved.

The invention belongs to the field of computer graphics, and discloses a fast parallel multi-scale smooth symbol distance curved surface reconstruction method and system, and the method comprises the steps: carrying out the local fitting of point cloud data through employing a moving least square method, and calculating the Gaussian curvature of a point cloud; constructing a curvature adaptive octree based on Gaussian curvature, wherein the octree is used for dynamically adjusting local resolution and capturing geometric details to improve curved surface reconstruction precision; global fitting is carried out through an implicit function, and a zero contour surface is obtained through an octree contour surface extraction algorithm; and multiple threads are introduced in the process of iteratively solving the linear system to carry out parallel computing. The adaptive octree based on Gaussian curvature can effectively capture complex geometric details, and the algorithm realizes high-precision curved surface reconstruction under different scales. In addition, the local and overall calculation efficiency is remarkably improved through parallel calculation, and the high efficiency of the algorithm in large-scale data processing is ensured. The method provided by the invention provides an efficient and accurate solution for complex curved surface reconstruction.

The invention provides a part surface efficient three-dimensional reconstruction method based on adaptive octree sampling, and belongs to the field of surface reconstruction. The part surface efficient three-dimensional reconstruction method based on self-adaptive octree sampling comprises the following steps of performing multi-stage filtering and denoising on collected point cloud data; octree nodes are adaptively divided according to local point cloud density and curvature features, the curvature of each node is calculated, feature region marking is carried out, voxel downsampling processing is carried out on non-feature regions, curvature weighted random sampling processing is carried out on feature regions (such as cracks and recesses), and high resolution of the regions is reserved; a multi-resolution space grid is generated based on octree hierarchy, a hierarchical symbol distance field is constructed, and the Poisson surface reconstruction process is accelerated. The method can be used in the fields of surface topography analysis, process optimization, quality control and the like, is particularly suitable for scenes requiring efficient and accurate surface topography detection, and has a wide application prospect.

The invention discloses a method for rendering a large-scale point cloud at a web end, and belongs to the technical field of large-scale point cloud rendering, and the method comprises the steps: S1, carrying out the data preprocessing of point cloud data, and carrying out the octree division of the whole point cloud; s2, on the basis of octree division, performing Lod hierarchical data division on leaf nodes of the whole octree; s3, data which needs to be rendered currently are screened out from the octree through view cone cutting; s4, high-level data and low-level data are separated from the screened data needing to be rendered, the low-level data are used for real-time rendering, and the high-level data are used for non-real-time rendering; and S5, combining the real-time rendering result and the non-real-time rendering result, and outputting all rendering results. According to the method, a similar double-buffer structure is used, and large-scale point clouds can be rendered in a real-time and non-real-time mixed mode at a web end.

The invention discloses a self-adaptive octree three-dimensional magnetic method inversion method fused with abnormal region recognition, and belongs to the technical field of three-dimensional magnetic method data inversion processing. Comprising the steps that initial inversion is rapidly completed on a coarse grid, the overall contour of an underground structure is obtained, a fuzzy c-means clustering algorithm is introduced, and a target area with significant magnetic anomaly is automatically extracted from a coarse solution; and constructing a multi-level octree grid in the extracted abnormal region to realize fine subdivision of the complex geological boundary, and further executing high-precision inversion by using a fine model as a new starting point. The adaptive octree grid has the characteristics of high efficiency, small grid number and small data fitting difference. The process of coarse inversion, intelligent identification, local refinement and fine inversion can be circularly executed as required, and computing resources are dynamically allocated. Compared with a traditional global fine grid scheme, the inversion process has the advantages that the inversion precision is maintained and even improved, and meanwhile, the number of grid units is remarkably reduced.

The invention provides a multi-modal data fusion method and device based on dynamic Gaussian modeling and a vehicle. According to one embodiment of the invention, the multi-modal data fusion method based on dynamic Gaussian modeling comprises the following steps: obtaining continuous multi-frame multi-modal data, obtaining 3D voxel features and multi-view image features through the multi-modal data, constructing a dynamic Gaussian parameterized model based on the 3D voxel features, and fusing the dynamic Gaussian parameterized model and the multi-view image features through the dynamic Gaussian parameterized model. Generating a multi-view-angle Gaussian feature by using a dynamic Gaussian parameterized model and the multi-view-angle image feature, and projecting the multi-view-angle Gaussian feature to a dynamic resolution BEV space to obtain a BEV feature, and obtaining a 3D voxel feature by adopting dynamic voxelization, and the dynamic voxelization is based on octree sparse voxel coding, and the resolution is adaptively adjusted according to the scene complexity. According to the invention, the resource utilization rate can be optimized and calculated while the sensing precision is ensured.

The invention relates to the technical field of three-dimensional model reconstruction, in particular to an existing building three-dimensional model reconstruction system based on point cloud data, and the system comprises a point cloud data processing module which carries out the octree recursive space subdivision of input point cloud data, builds a hierarchical index structure, and generates a multi-view fusion point set. In the method, octree recursive space subdivision is performed on the input point cloud data to establish a hierarchical index structure, so that the key area of the building surface can be positioned more accurately in subsequent data processing, data redundancy is reduced, and the precision and efficiency of three-dimensional model construction are improved; then voxel depth field data is obtained by calculating a truncation symbol distance function value of voxels, the description accuracy of a building surface structure is improved, and an initial surface grid is formed through contour surface extraction; meanwhile, plane slicing is conducted on the multi-view fusion point set in the preset building main axis direction, and contour details of all building structure sections are captured.

The invention discloses a multi-thread and LOD-based point cloud rendering method, and relates to the field of video rendering, and the method comprises the steps: dividing point cloud data into n data blocks; calculating an initial position of each data block in the point cloud file, and creating n threads; extracting m pieces of LOD point cloud data with different levels of detail from the octree of each data block; calculating the proportion # imgabs0 # of the size of the current point cloud in a rendering viewport and the distance # imgabs1 # from a viewpoint to a camera, and calculating the detail level # imgabs4 # of the current to-be-rendered LOD point cloud according to a preset viewport proportion threshold value # imgabs2 # and a preset sight distance threshold value # imgabs3 #; according to # imgabs5 #, selecting point cloud data of a corresponding detail level from the m pieces of LOD point cloud data; for each interactive operation event, parallel coordinate transformation and illumination calculation are carried out on the LOD point cloud data to be rendered currently in different threads; submitting a calculation result to a renderer for rendering; aiming at low rendering efficiency in the prior art, the method and the device can efficiently render large-scale point cloud data.

The invention discloses an urban district regulation construction management operation and maintenance method and system based on three-dimensional digital twinning, and the method comprises the steps: collecting multi-source data, including geographic raster data, BIM model data and mobile phone signaling data, of an urban district, and achieving the coordinate alignment of the multi-source data through an LM algorithm; the multi-source data generates feature data composed of coding tables defining urban features through feature engineering, and an octree is constructed based on the feature data; defining leaf nodes of the octree as voxels, and applying multi-scale convolution to each voxel to obtain super voxels with space-time labels; constructing a constraint network based on the super voxels and a construction, management or operation and maintenance specification constraint library; and judging the risk area of the city district by calculating the space-time constraint intensity of the constraint edge, and outputting a corresponding risk area list. According to the invention, the problems of multi-source data integration, real-time dynamic monitoring and risk prediction in the planning, construction, management, operation and maintenance processes of the urban district are solved.

The invention relates to the technical field of geological analysis, and discloses a method for fusing a three-dimensional geological model and a WebGIS (Web Geographic Information System), aiming at solving the problems of poor data intercommunity, limited spatial analysis capability, difficulty in real-time updating and low model loading efficiency of the existing method, and the scheme mainly comprises the following steps: extracting geological model data, and carrying out standardized format processing; lightweight processing is carried out, and geological attributes and model nodes are bound one by one; the format is converted into a binary format embedded with a metadata label, and data integrity verification and automatic error correction are carried out in the conversion process; constructing a non-uniform adaptive octree spatial index, a multi-level attribute data architecture and a real-time updating mechanism of attribute data; after the geological model and the terrain are precisely registered and fused, a dynamic loading mechanism, a high-performance rendering mechanism and a geological model real-time updating mechanism are constructed, and a three-dimensional geological analysis tool is integrated. According to the method, efficient loading, real-time rendering, dynamic updating and spatial analysis of the three-dimensional geologic model in the WebGIS are realized.

The invention relates to an OpenGL (Open Graphics Library)-based large-scale three-dimensional model object quick pickup method. The method comprises the following steps: converting an interactive operation of a user into a three-dimensional space structure; a CPU constructs octree space acceleration structures based on a bounding box of an object in a large-scale three-dimensional model, and leaf nodes, intersecting with a pickup ray or a pickup view pyramid, of an axis alignment bounding box in each octree space acceleration structure are screened out to serve as candidate leaf nodes; collecting triangles contained in all candidate leaf nodes, and constructing a candidate triangle set; dynamically allocating thread groups and threads of the GPU based on the number of candidate triangles in the candidate triangle set; and performing highlight display on the candidate triangles intersected with the pickup ray or the pickup view cone, and feeding back IDs of the corresponding candidate triangles. According to the method, the data transmission quantity and the video memory access pressure are remarkably reduced, the pickup response time is greatly shortened in a large-scale three-dimensional model scene, and the requirements of real-time interaction and visual operation can be met.

The invention relates to the technical field of three-dimensional models, in particular to a drawing method for making a three-dimensional model based on three-dimensional laser point clouds, which comprises the following steps of: sequentially reading point cloud data files of the three-dimensional laser point clouds, inserting coordinate values of each point into an octree structure, and when octree node buffer areas in a memory are full, drawing a three-dimensional laser point cloud into the octree structure; and writing the node data in the buffer area into a hard disk, and traversing the hard disk octree node set from bottom to top. According to the method, the three-dimensional laser point cloud data is inserted step by step, and the hierarchical aggregated octree structure is utilized, so that the point cloud data processing efficiency is effectively improved, and the memory occupation pressure is reduced; representative points in an octree structure are fused with an average normal to construct a continuous function field, a grid model is adaptively generated in a multi-detail-level mode, and the detail retaining capacity and drawing precision of the model are improved. Furthermore, accurate identification and positioning of topological features are realized by adopting calculation of a cell complex sequence and a topological feature noise rank.

The invention discloses a rapid local repair method for a self-intersecting triangular mesh. According to the method, local processing and rational number calculation are combined, aiming at an input low-quality triangular mesh, firstly, the mesh quality is optimized through a preprocessing step, intersection areas are positioned, then the mesh is partitioned based on octree division and connected component analysis, and the intersection areas and non-intersection areas are separated. In each intersection area, a shape-preserving boundary three-dimensional grid optimization method based on rational numbers is adopted to ensure the robustness of intersection calculation, and surface grids are extracted after successful conversion into floating-point number representation. And finally, seamlessly splicing the repaired local grid and the non-intersecting grid, and improving the overall grid quality through an edge re-gridding method based on curvature self-adaption. According to the method, the non-intersecting high-quality grid can be quickly generated, the method is suitable for downstream applications such as finite element analysis, meanwhile, the calculation overhead is remarkably reduced, and the algorithm robustness is improved.

A point cloud data transmission method according to embodiments may comprise the steps of: acquiring point cloud data; encoding, by application of a geometry tree, geometry information including positions of points of the point cloud data; encoding attribute information including attribute values of the points of the point cloud data on the basis of the geometry tree; and transmitting the encoded geometry information, the encoded attribute information, and signaling information, wherein the step of encoding the attribute information comprises: a step of generating level of details (LoDs) on the basis of the geometry tree and a step of compressing the attribute information on the basis of the LoDs.

A three-dimensional data encoding method includes: encoding geometry information of each of three-dimensional points based on one of a first geometry information encoding method of encoding using octree division and a second geometry information encoding method of encoding without using octree division; and generating a bitstream including the geometry information encoded and a geometry information flag indicating whether the encoding was performed based on the first geometry information encoding method or the second geometry information encoding method. In the generating of the bitstream: when the encoding is performed based on the first geometry information encoding method, the bitstream including a parameter set used in octree division is generated; and when the encoding is performed based on the second geometry information encoding method, the bitstream not including the parameter set used for octree division is generated.

A system and method for encoding and decoding the geometry of 3D point clouds using octree-based data structures are disclosed. The method involves encoding and decoding bitstreams containing octree structure information and vertex data, including the presence and position of vertices on cuboid edges corresponding to leaf nodes. The decoding process determines triangles connecting vertices within each cuboid, which are voxelized to reconstruct the 3D point cloud. To enhance voxelization accuracy, triangles may be extended along one or more sides based on a sampling distance parameter (dsampldsampl) or adaptive halo parameters. The encoding process utilizes similar principles to encode the octree structure and vertex information, supporting geometry reconstruction with high fidelity. The system employs the Möller-Trumbore algorithm and barycentric coordinate calculations with constraints based on dsampldsampl for voxelization. Extensions may include fixed or adaptive parameters encoded within the bitstream.

The invention discloses a laser radar positioning and three-dimensional mapping method based on an implicit neural field, and belongs to the technical field of positioning and mapping. The method comprises the following steps: for first frame data in radar point cloud data, converting an input sampling point from a sensor coordinate system to a world coordinate system by using a unit matrix or a constant matrix to initialize a map; the hierarchical potential features are optimized in a self-supervision mode, an octree data structure is used for storing the optimized potential features to generate a potential feature set, and features in the potential feature set are decoded into symbol distance values through a shallow multilayer perceptron (MLP); starting from a second frame of data of the radar point cloud data, obtaining a predicted pose through a matching mode from the frame to the implicit map every time a frame is input; converting a new input point into a world coordinate system by using the obtained predicted pose, and determining whether a new potential feature needs to be added or not according to a Morton value of the new input point; and completing the construction of the NSDF by repeating the operations of predicting the pose and expanding the NSDF.

The invention provides a public space dynamic risk network assessment method fusing spatio-temporal characteristics, which relates to the technical field of network assessment, and comprises the steps of acquiring basic data, adaptively dividing grid units by adopting an octree algorithm, establishing a connection relation graph, calculating a risk propagation probability by utilizing a graph neural network, and calculating a risk propagation path by applying an improved A * algorithm. And finally outputting risk level distribution, propagation path risk change and prevention and control deployment suggestions. According to the invention, refined quantitative evaluation of public space risks is realized, and risk early warning accuracy and prevention and control measure pertinence are improved.

The invention relates to a rapid conversion method and system from explicit geometry to implicit geometry, and the method comprises the steps: constructing a linear boundary body hierarchical structure for an input triangular grid, and constructing an octree structure; nodes of the octree structure are used as sampling points, and an adaptive grid unsigned distance field is calculated in parallel for each sampling point; calculating a pseudo normal of the triangular mesh by using an angle weight pseudo normal algorithm, and calculating a symbol of the unsigned distance field of the adaptive mesh by using the pseudo normal of the triangular mesh; and determining data of the sign distance field according to the adaptive grid unsigned distance field and the sign of the adaptive grid unsigned distance field, and completing conversion from implicit geometry to explicit geometry. According to the method, the conversion from implicit geometry to explicit geometry can be quickly realized, the conversion efficiency is greatly improved while high precision is kept, and the performance of graphic processing, physical simulation and application scenes can be improved.

A method for encoding and decoding, an encoder and decoder for a point cloud. The method for encoding a point cloud to generate a bitstream of compressed point cloud data, in which the point cloud's geometry is represented by an octree-based structure with a plurality of nodes having parent-child relationships by recursively splitting a volumetric space containing the point cloud into sub-volumes each associated with a node of the octree-based structure, includes: determining an occupancy pattern for a parent node based on the occupancy of its child nodes; determining a planar context information for at least one of the child nodes; and entropy encoding / decoding the occupancy pattern parent node based on the determined planar context information to produce encoded / decoded data for the bitstream.

The invention discloses an adaptive edge detection iterative denoising point cloud reconstruction method and system. A regularization parameter and a kernel bandwidth are automatically set through point cloud normalization and local covariance noise estimation; an implicit curved surface is fitted based on a parameterized Gaussian field, and initial curved surface construction without a normal direction is realized. Edge detection is carried out by utilizing Voronoi covariance measurement, and the edge detection is continuously mapped into a projection weight, so that feature details of the point cloud are reserved in an acute angle region, and smooth denoising is realized in a flat region. The regularization intensity is dynamically adjusted through residual trend analysis to enhance detail representation, and a triangular mesh is extracted and output in combination with octree discretization and contour surface extraction. The method has unsupervised adaptivity, has outstanding performance in the aspects of edge fidelity, noise robustness and complex structure adaptability, can effectively process point cloud data with high noise, missing normal direction or irregular structure, and is suitable for the fields of engineering survey, reverse modeling, cultural relic digitization and the like.

A method for encoding a three-dimensional (3D) point cloud into a bitstream, includes: obtaining octree information including an octree structure of a volume, wherein the octree structure includes a plurality of cuboids as leaf nodes of the octree structure; providing a threshold range; determining vertex positions Vj on edges of each cuboid as vertex information, wherein the vertex position Vj on the respective edge is determined by points of the point cloud in a distance to the respective edge smaller than the threshold range; and encoding the octree information and the vertex information into a bitstream; wherein the threshold range is dependent on the sizes of the cuboids.

The invention discloses a rendering method, which comprises the following steps: constructing an octree, and storing a to-be-rendered object in a to-be-rendered three-dimensional scene in a child node of the octree to obtain a space management octree; dividing the view cone into different view cone levels; traversing the space management octree, and determining a first intersection part between the space of the child node and the view cone and a view cone level to which the first intersection part belongs; distributing the first intersection part to different rendering queues according to the view cone hierarchy to which the first intersection part belongs; shielding and removing the first intersection parts in the different rendering queues to obtain current visible objects to be rendered; and rendering the current visible to-be-rendered object. According to the method, the octree is constructed to manage the three-dimensional scene space, the view cone levels are divided, the rendering queues are distributed, the view cone removing and multi-level shielding removing technologies are combined, the visible objects are accurately judged, the rendering load is effectively reduced, and the method has the advantages that the rendering performance is improved, space management is optimized, and the complex scene processing efficiency is improved.

The invention relates to the field of CIMs, and discloses a method and system for automatically degrading a CIM6-level three-dimensional model to CIM5 based on a semantic rule engine, and the method comprises the steps: obtaining CIM6-level three-dimensional model data, and constructing a scene hierarchical tree based on the three-dimensional model data; identifying the category, the attribute, the service label and the importance constant of each object in the scene hierarchical tree; calculating geometric complexity, semantic importance and topological dependency of each object, and calculating a comprehensive weight; matching the interval of the comprehensive weight and executing a corresponding processing operation based on the matched interval; performing grid merging and texture baking operation on the reserved object; performing grid instantiation on the proxy object after proxy replacement; collision holes in the model are detected through an octree in combination with a GJK algorithm to complete topological consistency verification, a lightweight CIM5-level model data packet is generated, and a difference report is output. According to the method, the consistency of the key semantic information and the topology is kept while the data volume of the CIM6 model is reduced.

The invention discloses a hierarchical construction method and system for massive point cloud details, and the method comprises the steps: generating a multi-scale and multi-level pyramid data storage structure on the basis of a spatial octree by using a downsampling mode considering object space features, building an LOD (Level OfDetails) mechanism, and carrying out the hierarchical construction of the massive point cloud details. A detail level access mechanism which gradually changes along with the scale is provided for subsequent rendering of massive point clouds, and the problem that detail transition of the massive point clouds is not smooth in the rendering process is solved; the problem of contradiction between efficient access of massive point clouds and huge cost is solved from the essential mechanism, the massive data structure and mechanism are innovatively designed from multiple aspects of data storage and rendering display, and efficient data response efficiency is provided for real-time point cloud rendering.

The invention relates to an interstellar medium radiation transfer calculation method and device based on ray tracing. The method comprises the following steps: discretizing a three-dimensional space model of an interstellar medium by using a Cartesian grid to obtain a plurality of grid units, and emitting a plurality of uniformly distributed light rays at the center of each grid unit to construct a discretized radiation transfer equation; for each of the light rays, screening out the grid units through which the light rays penetrate by adopting an octree algorithm; calculating the length of a path where each light ray intersects with the grid units when the light ray passes through the grid units by adopting a flat plate method; and solving a discretized radiation transfer equation according to the path length to obtain a parameter equation of the interstellar medium. By adopting the method, the calculation amount of the ray tracing numerical integration in the radiation transfer equation can be reduced.

The invention relates to a zero-sample open vocabulary scene understanding method based on semantic instance generation, and the method comprises the steps: obtaining multi-frame image data including depth information, obtaining 2D mask information through segmentation for each frame of image data, carrying out the semantic feature extraction of a 2D mask, projecting the semantic feature to a 3D space, obtaining a 3D point cloud segment, and carrying out the segmentation of the 3D point cloud segment; combining the 3D point cloud fragments into 3D instances through time sequence grouping combination by adopting a semantic-guided under-segmentation filtering and combination threshold linear attenuation strategy to obtain a 3D instance set; for semantic features of each instance in the 3D instance set, performing feature aggregation processing on the premise of considering intra-class similarity and inter-class difference; and based on the 3D instance set after feature aggregation, constructing a hybrid octree-graph structure to realize zero-sample open vocabulary scene understanding. Compared with the prior art, the problems of inaccurate instance segmentation, inaccurate feature aggregation and low spatial representation efficiency are effectively relieved.