56 results about "Viewing cone" patented technology

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 provides a camera external parameter calibration method and device based on image and point cloud matching, and the technical scheme of the invention is that a point cloud rendering view corresponding to an original point cloud is rotated, so that the point cloud rendering view and a camera image are superposed visually, and initial space association is established; and directly learning and solving accurate camera external parameters from the local relevance between the original point cloud intensity information and the image RGB information in an end-to-end manner by using a deep learning model. According to the scheme, the strict camera-radar orientation consistency requirement in a traditional method is not needed through rough matching, manual participation in the whole process of traditional calibration is replaced, transition from visual alignment to geometric alignment is achieved, and the calibration efficiency and scene adaptability are remarkably improved. Meanwhile, the scheme of the invention initiates a process of visual field cone cutting-perspective projection rasterization-intensity-RGB end-to-end matching, and compared with a traditional feature point method, the calculation complexity is greatly reduced, and the memory occupation is reduced by 80%.

The invention discloses a digital twinning-based three-dimensional visual management system and a digital twinning-based three-dimensional visual management method for an anode assembly workshop. The system comprises a triple binding index module, a unified time axis alignment module, a view cone inverse solution positioning module, a safety linkage arrangement module and a resume playback module, and is provided with a process template library and a network isolation / main / standby module. Three-dimensional positioning and PTZ linkage of an AI / threshold event are realized by establishing mapping of a three-dimensional object, a camera and a PLC point location; the batch / bracket / station events, the PLC time sequence and the video slices are indexed in a unified mode; shadow verification, interlocking check, instruction issuing, read-back and video consistency confirmation and trace leaving are executed according to the process, and a safe closed loop from warning to disposal is formed. The method comprises the steps of access and calibration, triple binding, time axis alignment, event positioning and stream taking, linkage processing and resume filing. The positioning and handling efficiency can be improved, the misoperation risk is reduced, and cross-production-line reuse and tracing evidence obtaining are supported.

The invention discloses a three-dimensional target detection method based on view cone point cloud clustering and 2D detection, and the method comprises the steps: generating a plurality of target bounding boxes with categories according to a 2D image, converting the target bounding boxes to an image coordinate system, and carrying out the screening of 3D point clouds, so as to obtain view cone point clouds; performing a point cloud clustering operation on the view cone point cloud, and calculating a result of the point cloud clustering operation to obtain a 3D frame meeting a preset requirement; and performing IOU calculation of different visual angles on the 3D frames to merge the 3D frames of the same object under different visual angles so as to obtain a 3D bounding box. According to the method, the problem of geometric information loss can be effectively avoided, and the accuracy of the finally obtained 3D bounding box is ensured; the problem that global search is needed due to the fact that 3D point cloud is directly adopted is avoided, the complexity of the calculation process is reduced, and the overall calculation efficiency is effectively improved; the preset frame requirement can be set according to the requirement of a use scene, so that different requirements are met, and the wide use range is ensured.

The invention discloses a mechanical arm fine positioning and closed-loop calibration method and system based on heterogeneous visual fusion, and the method comprises the steps: heterogeneous data synchronous collection and cross-modal joint calibration, semantic segmentation based on prompt learning and high-precision mask generation. Point cloud refining and edge noise elimination in a view cone space; and differential compensation and closed-loop feedback control based on a golden model. A heterogeneous visual fusion framework is constructed, and 3D point cloud noise is constrained by using a sub-pixel edge of a high-resolution 2D image, so that the positioning precision breaks through a sub-millimeter level; a segmentation large model with zero sample generalization ability and a prompt learning mechanism are introduced, so that the system can adapt to various workpieces under a complex background without repeated training; refining point cloud and a standard three-dimensional gold model are registered and fed back to the mechanical arm in a closed loop mode, so that the system can compensate and calibrate drift and accumulative errors in real time, and long-term stable operation is achieved.

The invention discloses a roadside multi-mode fusion sensing method. According to the method, an initial 3D view cone tensor with depth information is generated through a depth estimation network, features of pixel points are projected to a coordinate system in combination with internal and external parameter matrixes of a camera, then a three-dimensional space point coordinate set is generated according to projected feature position information, a Gaussian function is adopted to distribute weights of neighborhood grids, and a three-dimensional space point coordinate set is obtained. And finally, inputting the image features and the features into a fusion unit, respectively carrying out uncertainty modeling on the two types of features, automatically generating a fusion weight, and finally, carrying out normalization weighted aggregation on the features in the neighborhood so as to reduce the position approximation error and obtain the calibrated image features, and finally, inputting the image features and the features into the fusion unit, and respectively carrying out uncertainty modeling on the two types of features and automatically generating a fusion weight. And dynamic optimization is carried out according to a self-supervised loss function in the fusion process, so that the stability of multi-modal feature fusion is ensured, the detection accuracy is high, the performance is relatively good, and the problem of global feature dislocation caused by external parameter deviation and depth error can be avoided.

The present application relates to a kind of WebGL-based pipe network digital twin model light weight rendering method, method includes: resource preconfiguration: the grid data of different three-dimensional model is generated corresponding instance template;Through instance template generation instance is stored in instance pool;In scene graph, create include call data and dynamic data lightweight placeholder, call data is used to extract instance, and dynamic data is used to adjust the dynamic display of instance;Lightweight rendering: set the visibility screening strategy of placeholder, whether current placeholder is visible is judged in turn through the screening condition of level, distance and perspective, view cone intersection;Set instance management strategy, for the placeholder screened as visible, corresponding instance is called from instance pool and is bound, and the binding relationship is recorded, for the placeholder screened as invisible, unbind, and the instance after unbinding is marked as idle;The placeholder of bound instance is rendered and output.Effectively reduce the memory occupation when rendering.

The present application relates to a kind of infrared scene quantitative radiance calculation methods based on light ray tracking module, comprising: obtaining target mask information image, the radiance image of infrared scene, depth information image and position information image, sensor position information, sensor field of view and sensor observation direction;Infrared scene is divided into several view cones according to the resolution of target radiance area image, wherein several view cones and the object surface in infrared scene intersect to form several face elements;The radiance of each face element at the sensor is calculated by the theorem of solid angle projection;The radiance of several face elements at the sensor is accumulated, and the radiance of infrared scene to the sensor is obtained.The method supports full scene, single target, multiple targets as radiation energy source, and can be applied to high complex infrared scene simulation work;Without additional data support, the radiance at the sensor can be calculated without performing redundant work.

The application discloses a multi-dimensional demonstration method and device based on time domain technology and a storage medium, and relates to the technical field of data interaction. The application first acquires a time domain file containing multiple perspectives and / or multiple time points, and encapsulates multi-dimensional index information containing a space index and a time index. Then, a demonstration narrative stream including a key frame sequence of the demonstration and a camera motion path of the key frame is constructed based on the multi-dimensional index information. In the process of playing and rendering, the current camera state is acquired in real time, the prediction information of the camera motion path is combined, and the target display area in the current view cone is calculated. Finally, the corresponding target pyramid tile data is called from the multi-dimensional index information, and progressive decoding and rendering are performed in the order from low to high resolution, so as to realize the non-susceptible loading of the demonstration scene from the overall overview to the local micro details. In this way, the loading efficiency of the multi-dimensional demonstration scene is improved, and the resource consumption is reduced.

The invention relates to the technical field of computer graphics and geological exploration visualization, and discloses a geologic body space penetration processing system based on a domestic digital twin engine technology, and the system comprises a heterogeneous data mapping module which is used for carrying out the normalization interpolation of discrete geological data, and generating a continuous three-dimensional body texture; the view cone depth weighted statistical analysis module is used for carrying out view cone sampling on the body texture and generating a weighted histogram in combination with a depth weight; the time domain damping adaptive threshold value control module is used for solving an instantaneous threshold value based on the histogram and generating a smooth final application threshold value in combination with a historical threshold value; and the multi-dimensional gradient mapping and penetration rendering module is used for calculating comprehensive visual potential energy in combination with the attribute value and the surface orientation factor, and executing fragment elimination or output based on a final threshold value. According to the method, a time domain damping self-adaptive threshold control mechanism is constructed, the time sequence stability of a rendered picture in the dynamic roaming process is ensured, and visual jump is eliminated.

The invention relates to a robot vision automatic scanning path planning method based on a three-dimensional grid model, and belongs to the related field of machining and manufacturing. According to the method, a scanning view cone model of a scanner is obtained according to a series of constraints of a surface structured light scanner, and through Gaussian mapping and clustering analysis, through segmentation and combination of triangular patches, a three-dimensional grid model is obtained; the method comprises the steps of obtaining a plurality of areas convenient to scan of a workpiece, generating candidate viewpoints according to a scanning view cone model, introducing viewpoint measurement quality parameters, selecting optimal viewpoints, and carrying out path planning on all the obtained optimal viewpoints according to a greedy algorithm, so that automatic viewpoint arrangement and automatic path generation are realized; the problems of repeated measurement, measurement omission, low efficiency and the like of manual teaching viewpoint workpiece measurement are solved, and the method has the advantages of high precision, high efficiency, high integrity, automation and the like.

The application relates to the technical field of target detection, in particular to a space passable area detection method and device, a detection equipment and a storage medium, wherein the method comprises the following steps: obtaining three-dimensional point cloud data and an environment image of a target area; performing plane projection and view cone projection on the three-dimensional point cloud data respectively to obtain multiple surface features and view cone projection features; fusing the multiple surface features and the view cone projection features to obtain a space feature map; fusing the space feature map and the environment image under the same time and space to obtain a fused feature map; inputting the fused feature map into a trained detection model; and outputting a space passable area of the target area. Thus, the related art is generally used for detecting and identifying grounded obstacles, pedestrians, vehicles and the like, and the resolution and generalization are not strong, the spatial position of the obstacle cannot be accurately determined, and the problems of avoiding collision and causing accidents and the like are solved.

The application discloses a fuselage skin viewpoint generation and detection method and device based on an unmanned vehicle, and the method comprises the following steps: acquiring a fuselage point cloud model of the airplane; performing cell decomposition on the contour of the fuselage point cloud model according to the fuselage point cloud model, a camera coverage range and a shooting overlap rate requirement; acquiring a normal vector and a center point of an effective cell; calculating initial pose information of a camera according to orthoview and equidistance shooting constraint conditions; generating an optimized target viewpoint according to minimum safety distance constraint conditions and lifting device height constraint conditions; and detecting the viewpoint coverage rate of the target viewpoint according to a view cone elimination strategy and an occlusion elimination strategy. The fuselage surface coverage problem can be well solved by using the unmanned vehicle system to automatically collect data, perform cell decomposition and generate and optimize the viewpoint, so that the efficiency and safety are improved, the coverage degree is ensured, and the data collection quality is ensured.

The invention discloses a high-fidelity scene reconstruction method based on laser radar-vision-inertial navigation fusion. The method comprises the following steps: carrying out online scanning and segmentation on a laser radar point cloud according to a camera sampling interval; acquiring an initial camera attitude by a laser radar-inertial navigation odometer, and jointly optimizing a re-projection error and a luminosity error through error state iteration Kalman filtering to obtain a refined camera attitude; an octree is adopted to remove map points outside a view cone, survival map points, an image and a refined camera attitude are combined for Gaussian micro rendering, meanwhile, a Gaussian ellipsoid cloning process is improved, the influence of motion blur is eliminated, overlapped fog shadow and back propagation loss are avoided, Gaussian ellipsoid parameters are iteratively optimized, and high-fidelity scene reconstruction is achieved. According to the scene reconstruction method provided by the invention, the problem of insufficient robustness of the traditional SLAM technology in the outdoor environment is improved, and meanwhile, the performance of the 3DGS in the outdoor environment is optimized.

The application discloses a three-dimensional target detection method based on RGBD data and a view cone, and comprises the following steps: acquiring a feature image of a scene RGB image according to the scene RGB image; inputting the feature image of the scene RGB image into a region generation network to acquire a bounding box of a to-be-detected object and a category vector of the to-be-detected object; acquiring a 3D view cone point cloud according to the bounding box of the to-be-detected object and a depth image of the scene; inputting the position coordinates of points in the 3D view cone point cloud and the category vector of the to-be-detected object into a PointNet network with an attention mechanism to acquire the center point coordinates of the to-be-detected object; and further acquiring the parameters of a 3D bounding box of the to-be-detected object to detect the three-dimensional to-be-detected object. The application simultaneously considers the plane information and the depth information of a scene, improves the detection precision, and makes the detection result more accurate while ensuring the efficiency.

The application relates to a data fusion method and system based on visual deep learning and laser point cloud, and a medium, the method comprising F1: during vehicle driving, target point cloud data is acquired based on a vehicle-mounted laser radar and is time-stamped, and target image data is acquired based on a vehicle-mounted camera and is time-stamped; F2: the target point cloud data and the target image data are time-synchronized, and synchronized target point cloud data and synchronized target image data are output; F3: all point cloud data in a view cone corresponding to an image detection frame is extracted and clustered, and point cloud clustering results in the view cone are output. The application not only solves the correlation failure problem of current perception system visual and laser post-fusion, but also solves the frequent sudden braking problem caused by the fact that, when an automatic driving vehicle drives on uneven road, a rut mark or a road protrusion not affecting driving is clustered as an obstacle, and the driving fluency of the automatic driving vehicle is improved.

The invention relates to the field of image processing, in particular to an industrial gas leakage detection method based on infrared thermal imaging, which comprises the following steps of: acquiring video streams by using a plurality of infrared thermal imagers, and constructing a gas basic geometric voxel grid through a background difference and view cone intersection method; thirdly, the sight line penetration depth and a statistical variable coefficient are calculated, and adaptive normalized radiation feature mapping based on statistical moments is constructed so as to solve errors caused by traditional homogeneous hypothesis; then, measuring and calculating a three-dimensional form compactness factor, generating a heterogeneous space density field weight in combination with the relative depth, and fitting a real diffusion gradient; and finally, calibrating a physical mapping coefficient by using single-point sensor data, and performing integral calculation on the real-time mass flow in combination with a three-dimensional optical flow method. According to the invention, the complex density gradient in the gas can be effectively reduced, and high-precision three-dimensional visual display and quantitative flow monitoring of gas leakage are realized.

The weak constraint 3D WebGIS video fusion method belongs to the technical field of virtual video fusion, and is characterized in that: including the following steps: S1, three-dimensional model construction; S2, three-dimensional model optimization adjustment; S3, 3D WebGIS platform construction; S4, access to video stream data; S5, model view matrix and projection matrix calculation; S6, view cone construction and occlusion detection; S7, rendering based on WebGL fragment shader; S8, video image dynamic texture mapping; S9, video projection edge feathering; S10, dynamic parameter adjustment. The weak constraint 3D WebGIS video fusion method solves the problem of video projection occlusion to a certain extent, realizes video fusion under weak constraint, improves the video fusion effect, and has high practicability, applicability and adaptability.

The application discloses a reconstruction method and device of a three-dimensional scene, and relates to the field of computer graphics. The method simulates the process of light rays from the camera optical center to the three-dimensional scene through a light ray travel module to determine the spatial coordinates of a plurality of sampling points in the three-dimensional scene; decomposes a color feature space and a density feature space to determine a matrix and a vector pair corresponding to the spatial coordinates of each sampling point; interpolates each matrix and vector pair to determine the color feature vector and the density of each sampling point; determines the spherical harmonic feature vector of each sampling point according to the color feature vector of each sampling point and the spherical harmonic coding of surface modeling; determines the anti-aliasing feature of each sampling point according to the color feature vector of each sampling point and the view cone coding; splices the spherical harmonic feature vector and the anti-aliasing feature to input the predicted color of each sampling point output by a target multi-layer perception; and performs volume rendering on the predicted color and the density to determine the reconstructed three-dimensional scene, thereby efficiently and high-quality reconstructing the three-dimensional scene.

The present disclosure provides a 3D point cloud segmentation method, device, electronic equipment and storage medium, the method comprising: obtaining 3D point cloud data and image data corresponding to the same sampling time and the same scene; converting the image data into a set of 2D masks, each 2D mask identifying the contour area of an object contained in the image data; converting the pixel coordinates of each 2D mask into a 3D point cloud coordinate system, thereby determining the corresponding view cone space of each 2D mask in the 3D point cloud coordinate system; for each 2D mask, segmenting the 3D point cloud data located in the view cone space corresponding to the mask in the 3D point cloud data to determine the point cloud block where the object is located. Accordingly, it is realized that the 3D point cloud data does not need to be manually labeled, and the mask generated by the image can guide the point cloud segmentation, thereby reducing the cost of manual labeling; it does not depend on specific sensor parameters, adapts to different hardware configurations, reduces the adaptation cost of different platforms, and enhances the generalization ability.

The invention provides a camera visual cone range generation method, a camera positioning method and related equipment. The method comprises the following steps: determining a shooting azimuth angle of a camera based on a road course angle and a shooting direction of the camera; and calculating a view cone range capable of covering the road surface based on the erection height, the latitude and longitude coordinates and the shooting azimuth angle of the camera. Constructing a spatial index based on the view cone range of each camera; based on the target object, searching in the spatial index, and searching a view cone range containing the target object or intersecting with the target object; and determining a camera shooting the target object according to the view cone range in the search result. According to the method, the visual cone range of each camera is accurately generated, and whether the target object is located in the visual field of a certain camera or not can be quickly judged through the constructed spatial index, so that the time for manual search and judgment is greatly shortened, and the video calling efficiency is improved.

The invention relates to a network digital twinborn model lightweight rendering method based on WebGL. The method comprises the steps of resource pre-configuration, wherein grid data of different three-dimensional models are used for generating corresponding instance templates; generating instances through the instance template and storing the instances in an instance pool; a lightweight placeholder containing calling data and dynamic data is created in the scene graph, the calling data is used for extracting the instance, and the dynamic data is used for adjusting dynamic display of the instance; light-weight rendering: setting a visibility screening strategy of placeholders, and judging whether the current placeholders are visible or not according to screening conditions of intersection of levels, distances, viewing angles and visual cones in sequence; setting an instance management strategy, binding placeholders screened to be visible and corresponding instances called from an instance pool, recording a binding relationship, unbinding placeholders screened to be invisible, and marking the unbound instances as idle; and rendering and outputting placeholders of the bound instances. And the memory occupation during rendering is effectively reduced.

The invention discloses a 3D scene loading optimization method and device based on a WeChat applet and a medium, and relates to the technical field of scene loading optimization, and the method comprises the steps: obtaining an original 3D model file, and splitting the original 3D model file to obtain a multi-stage 3D resource file; constructing a scene frame corresponding to the multi-level 3D resource file to determine an interactive 3D scene graph; based on the interactive 3D scene graph, obtaining a resource loading sequence through view cone priority analysis; according to the resource loading sequence, determining a dynamic renderable 3D scene through asynchronous queue loading; and performing rendering process optimization on the dynamic renderable 3D scene to obtain an adaptive 3D scene. By means of the method, the technical problems that in an existing WeChat applet environment, the 3D scene loading speed is low, the rendering performance is poor, and the network flow requirement is high are solved.

The invention discloses a three-dimensional target detection model training method and device, a medium and a program product, and relates to the technical field of computer vision and automatic driving. Local BEV features are uniformly constructed by adopting a distance-independent feature sampling strategy, a virtual view cone is dynamically constructed based on a feature size and a preset depth range, and meanwhile, the depth prediction precision is improved by utilizing a depth diffusion strategy and occupation coding. According to the method, uniform distribution of features in the BEV space is realized from semantic and geometric dimensions, inconsistent information distribution of targets in near and far regions in the BEV space is effectively relieved, and stable and sufficient image features are aggregated for sparse point cloud features. And multi-modal fusion is carried out on an enhanced image BEV feature map formed by splicing the local BEV features and the global image BEV features and the global point cloud feature map for target detection, so that the three-dimensional target detection model obtained according to the invention has higher target detection precision.

The invention discloses an aerial video real-time labeling method and device. The method comprises the steps that an unmanned aerial vehicle fence is obtained according to the position of the unmanned aerial vehicle after the unmanned aerial vehicle takes off, the unmanned aerial vehicle fence is stored in a fence library, the range of the unmanned aerial vehicle fence is a circular area with the longitude and latitude point of the unmanned aerial vehicle as the center, and the radius of the circular area is a real number larger than 0; all point locations in a camera coordinate system of the unmanned aerial vehicle are obtained, the point locations meeting the view cone equation are reserved, and the point locations not meeting the view cone equation are removed; and aggregating the point locations according to the height of the unmanned aerial vehicle. According to the invention, the problems of low efficiency, resource waste and inaccurate scattering point when the unmanned aerial vehicle loads the scattering point information in real time in the prior art are solved.

The invention relates to the field of interactive simulation, in particular to a simulation deduction multi-mode interactive data processing method and system, and the method comprises the steps: inputting a deduction instruction through an interactive end after commanders and fighters scenario, subsequently constructing a virtual scene, generating a deduction model in the virtual scene, and determining a deduction path of the deduction model based on the deduction instruction. Synchronously generating a view cone, moving the view cone along a deduction path in the virtual scene, fixing the cone end of the view cone at a plurality of moving points for surrounding, determining visual mutation rendering interference characteristics of the view cone tail end projection area passing through each model, setting labels for a plurality of sub-virtual scenes, and subsequently performing rendering adjustment on the sub-virtual scenes based on the labels. According to the interactive deduction simulation system and method, when the battlefield environment is simulated, the situation that the precision of the model jumps or flashes and suddenly changes during view angle moving and switching is reduced, the situation that the model rendering is not smooth and abrupt is reduced, and the interactive deduction simulation effect is improved.

The application provides a columnar height segmented supervision and local spline decoding three-dimensional semantic occupancy prediction method and system, the method comprises the following steps: step 1, input multi-view image, laser radar point cloud and vehicle pose information; step 2, construct bird's eye view BEV column grid, and generate columnar height segmented true value, voxel visibility mask and occlusion soft label; step 3, extract multi-view image features, and perform view cone space conversion to generate bird's eye view BEV features; step 4, predict ordered height segmented parameters based on bird's eye view BEV column, and recover three-dimensional semantic occupancy results by using local spline decoding.The application unifies true value and predicted object into a limited height segment in the same bird's eye view BEV column, so that the supervision target and network output are one-to-one corresponding in expression level, and learning deviation caused by inconsistent expression is reduced.

The invention relates to the technical field of computer vision and three-dimensional modeling, and discloses a random data enhancement method and device based on three-dimensional space structure modeling and a storage medium, and the method comprises the steps: carrying out the three-dimensional geometry and appearance modeling of a target object, building a scene space model, and defining an internal reference matrix of a virtual camera; generating a camera position and a camera orientation meeting a view cone constraint condition; projecting the target object and the scene space model to a two-dimensional image plane based on the internal reference matrix of the virtual camera and the constructed external reference matrix to generate a two-dimensional basic image; and on the basis of the two-dimensional basic image, generating a final training sample. According to the method, three-dimensional geometry and appearance modeling are performed on the target object, the target object is projected to the two-dimensional image plane based on the internal parameter matrix and the external parameter matrix, and the two-dimensional basic image is generated by combining illumination model rendering, so that the problem that a traditional two-dimensional data enhancement method cannot simulate view angle change and illumination interaction is solved.

The invention discloses a multi-view-angle remote mixed reality image enhancement method and system, and relates to the technical field of remote communication, and the method comprises the steps: obtaining user view angle perception data, calculating the relative pose of a user view angle and a main view angle, and dynamically adjusting the rendering range of the main view angle; receiving a main view angle data stream based on side nodes, constructing a user real-time pose matrix, carrying out pose-driven perspective transformation and view cone dynamic cutting, fusing and dynamically adjusting a main view angle rendering range, and generating a high-fidelity rendering view covering the main view angle range; on the basis of generation of a high-fidelity rendering view covering a main view angle range, implementing a deep layered enhancement strategy, highlighting details of a close shot by adopting a local contrast enhancement technology, and performing defogging and color restoration on a long shot to realize adaptive visual enhancement; and establishing a user interaction real-time feedback closed loop, transmitting to a user terminal through a low-delay network, driving the main view to dynamically update, and realizing a multi-view remote mixed reality image enhancement scheme. According to the invention, the image rendering efficiency is improved.

Disclosed are a graphics processing device, a system and a method for graphics processing. A graphics processing device includes a computing resource; and a neural rendering pipeline that invokes the computing resource to perform the following steps: obtaining a plurality of view parameters, wherein the plurality of view parameters are respectively associated with a plurality of view cameras, the plurality of view cameras are configured to take a same scene at different view angles; based on the plurality of view parameters, calculating an equivalent monocular view parameter associated with an equivalent monocular camera; filtering out a set of neural features within a view cone space associated with the equivalent monocular view parameter; decoding the set of neural features into a set of neural primitives; and rendering the set of neural primitives into a plurality of images based on the plurality of view parameters, respectively.