175 results about "Volume rendering" patented technology

The invention discloses a sparse view angle indoor reconstruction method based on uncertainty perception depth supervision. The method comprises the following steps: acquiring multi-view angle image data; solving a camera pose based on the SFM; a neural radiation field model based on uncertainty perception is constructed, and modeling is carried out on volume density, color and depth uncertainty parameters of space points; introducing a depth uncertainty synthesis formula into the volume rendering framework; constructing luminosity loss fused with random structure similarity; designing a self-adaptive deep optimization mechanism of uncertainty perception, and optimizing a training process through a progressive uncertainty learning strategy; and applying the trained model to a sparse view angle indoor scene to generate a high-quality three-dimensional reconstruction result and an uncertainty quantization graph. According to the invention, by introducing a fusion method of uncertainty perception and adaptive depth supervision, the problems of poor reconstruction quality and detail missing under a sparse view angle are effectively solved, and the precision and robustness of indoor scene three-dimensional reconstruction are significantly improved.

The invention provides a component freezing recovery method based on a page life cycle interface. In the frame compiling stage, page life cycle event monitoring logic is injected into the component, a freezing and recovering event monitor is automatically registered, a large-data-volume rendering component tree is recognized, a freezing identifier is added, and butt joint of the Vue component tree and a browser freezing event is achieved. When a freezing event is triggered, dependence of a responsive memory for releasing a virtual rolling component in a background label page is ensured by pausing virtual rolling rendering, generating a component state snapshot, creating a rolling position anchor point record, removing non-visual area nodes, retaining placeholder elements and releasing responsive data binding. And after a recovery event is triggered, positioning a rolling position based on anchor point parameters, preferentially reconstructing a visual area component in stages, and recovering the responsive system, so that rolling position zero offset and state zero loss are realized when the page is recovered, and the reconstruction process has no perception delay. According to the method, manual intervention is not needed, and front-end development efficiency and performance optimization are both considered.

The invention discloses a three-dimensional packaging modeling method based on NeRF and a diffusion model, and the method comprises the steps: firstly receiving demand description information, fusing packaging physical constraint conditions for joint coding, and forming a comprehensive condition vector; secondly, a cross-path dynamic information exchange mechanism is established through a geometric generation path and a texture optimization path, an initial three-dimensional packaging model is generated, physical consistency constraints are added, and NeRF is used for conducting volume rendering reconstruction on the initial three-dimensional packaging model; and then introducing an observation distance adaptive dynamic scale perception neural radiation field optimization mechanism, and generating a multi-scale NeRF model after scale perception optimization. And finally, establishing mapping of physical attributes and optical characteristics of the packaging materials, and generating a final three-dimensional packaging model through independent optical calculation and physical synthesis processes of the multi-layer materials. According to the method, the semantic understanding deviation of CLIP coding in vertical field application is solved, and the material reality representation of industrial-grade three-dimensional packaging modeling is realized.

The invention discloses a sparse view angle cultural relic three-dimensional reconstruction method, a storage medium and computer equipment. The method comprises the following steps: acquiring a cultural relic multi-view angle image and camera parameters; constructing a ray and carrying out spatial point sampling on the ray; joint modeling is carried out on the uncertainty of the spatial points, and color, volume density and evidence uncertainty parameters are output in the implicit neural radiation field model at the same time; a progressive sine position coding method based on uncertainty driving is introduced, a low-frequency channel is activated preferentially at the initial stage of training, and high-frequency features are unfrozen dynamically along with uncertainty feedback; after unfreezing is completed, continuous weighted optimization is carried out on the reliable and unreliable areas by utilizing a partition self-supervision consistency mechanism of uncertainty perception, and potential geometric information of the unreliable areas is fully utilized; and generating a high-fidelity three-dimensional reconstruction image and a confidence evaluation result of the cultural relic through volume rendering. According to the method, the cultural relic three-dimensional reconstruction precision and stability under the sparse view angle condition are effectively improved.

The invention discloses a geophysical three-dimensional visualization and geologic model construction system, which belongs to the technical field of geophysical data processing and three-dimensional geologic modeling and comprises a multi-source data acquisition module, a volume rendering module, an interactive segmentation module, an uncertainty expression module, a multi-scale fusion module, an attribute interpolation module and a model construction module. Multi-source geophysical data are converted into a visual three-dimensional visual model through a volume rendering technology, an interactive segmentation algorithm is adopted to accurately extract a geologic body boundary, an uncertainty expression mechanism is innovatively introduced to quantify the reliability of the model, and a multi-scale fusion technology is integrated to realize seamless connection from a regional structure to an ore deposit fine structure. The Kriging interpolation functional product physical property parameter three-dimensional distribution field is provided, the spatial cognitive ability and the decision-making efficiency under the complex geological condition are remarkably improved, and a powerful visual analysis platform is provided for geological exploration and resource evaluation.

The invention belongs to the technical field of high-power microwave effect tests in laboratories, and relates to a lossless reinjection data synchronization method and device for a multi-mode test body, and the method comprises the steps: S1, data collection and transmission: collecting electric field data in the multi-mode test body, and transmitting the electric field data out of a shielding cabin of the multi-mode test body; s2, an edge processing step: sending the electric field data transmitted to the outside of the shielding cabin into an edge GPU node, and carrying out voxelization processing and WebGL body rendering; s3, a data synchronization step: synchronizing the three-dimensional rendering data stream to a plurality of user side browsers in real time through WebRTC and HTTPS (Hyper Text Transfer Protocol over Secure Socket) protocols; s4, time synchronization is carried out, a unified high-precision time reference is provided through a Whit-Rabbit clock, and sub-nanosecond alignment of multi-channel data is achieved; and S5, a data storage step: generating a hash abstract for the synchronous data and writing the hash abstract into the block chain. By means of the method, the safety, the real-time performance and the reliability of data transmission can be effectively improved.

The invention discloses an automatic driving scene three-dimensional reconstruction method based on multiple view angles and continuous depth modeling. The method comprises the following steps: realizing high-precision and real-time three-dimensional scene modeling by designing an LMV-NeRF model; according to the method, multi-view image data and a continuous depth representation technology are fused, camera parameters are utilized to align feature maps, and global consistency is enhanced through a cross-view attention mechanism; generating a group of depth values by adopting a hierarchical sampling strategy, predicting color and volume density in combination with a volume rendering formula, and finally generating a target image and a depth map under a new view angle; and obtaining a data set of the multi-view image in the automatic driving scene to train the model, and selecting an optimal model to obtain the target image in the new view. By fusing multi-view image data and a continuous depth representation technology, high-precision and real-time three-dimensional reconstruction of the automatic driving scene is realized, and the method is particularly suitable for dynamic environment perception and decision support in the automatic driving scene.

The invention relates to a three-dimensional scene cross-domain migration method based on neural radiation field representation, computer equipment, a readable storage medium and a program product. The three-dimensional scene cross-domain migration method comprises the steps that a real image is acquired for three-dimensional scene reconstruction; inputting the three-dimensional scene information into a geometric module to extract content features; inputting the three-dimensional scene information into a source domain coloring module to extract source domain style features; performing volume rendering to obtain a first rendered image, training and optimizing the geometric module and the source domain coloring module until the expectation is met, and freezing the geometric module; inputting the three-dimensional scene information into a target domain coloring module to extract target domain style features; performing volume rendering to obtain a second rendered image; performing style migration on the real image to obtain a stylized image; comparing loss training to optimize a target domain coloring module, and freezing the target domain coloring module until expectation is met; and receiving a specified view angle direction, respectively outputting features by using a geometric module and a target domain coloring module, and obtaining a target domain image through volume rendering.

The invention provides a neural radiation field compression rendering method based on decomposition expression, and relates to the technical field of computer graphics, and the method comprises the steps: inputting a multi-view image and camera internal and external parameters, carrying out the ray sampling, generating a space sampling point, carrying out the mixed feature coding of the space sampling point, and obtaining a multi-view image; obtaining three-dimensional voxel features, aligned and fused three-plane features and position codes, and splicing the three features to form fused features; inputting the fusion features into a factorization neural BRDF rendering network to predict volume density, geometric latent features, material parameters and reflection features, and improving the view angle color under the compression condition through a BRDF modulator; a ray weight is predicted through a plane-ray combined modeling module, and a compressed neural radiation field model is obtained through combined optimization of miniaturized body rendering, weighted reconstruction loss and compression constraint and is used for target view angle image rendering; according to the method, the storage overhead of the neural radiation field model is reduced, and meanwhile, the synthesis quality and rendering consistency of the new view angle under different compression ratios are improved.

The invention relates to the technical field of occupancy network training, and discloses a self-supervised occupancy network training method and system based on 3D Gaussian splash rendering, and the method comprises the steps: fusing the 3D voxel features of a vehicle-mounted image with embedded features containing initial Gaussian attributes, obtaining the fused voxel features, and carrying out the decoding based on a Gaussian attribute decoder, obtaining a 3D Gaussian ellipsoid of each voxel grid; projecting the 3D Gaussian ellipsoid to a 2D image space by using a Gaussian splash rendering technology to obtain a 2D depth map; and generating a target view through linear sampling based on the 2D depth map and the vehicle-mounted image at the adjacent moment, and calculating luminosity reprojection loss based on the vehicle-mounted image and the target view to train the occupancy network. According to the method, the volume rendering calculation complexity and the reasoning time proportion are reduced, and the operation efficiency of the self-supervised occupation network is improved; and meanwhile, rendering can be carried out on the premise of not reducing the resolution of the depth map and supervising the quality, and the training effect and performance of the self-supervised occupancy network are improved.

There is provided a system and method for determining a three-dimensional model of an object using a projector and a light sensor arranged in a stereo configuration. The method including: receiving pixel intensity values captured by the light sensor for a plurality of images of a scene containing the object, wherein the images of the scene each capture different projection patterns emitted by the projector onto the scene; determining, using the pixel intensity values from a set of the plurality of images, correspondence between a projector plane and a camera plane to form a three-dimensional model of the object, the correspondence determined using a trained artificial neural network model that uses a combination of volume rendering and geometric representation; and outputting the three-dimensional model of the object.

The invention discloses an extensible reconstruction method for a super-large scene based on 3D (three-dimensional) Gaussian Splitting. The balance of the number of sub-block images is ensured through image perception self-adaptive blocking; a low-contribution view is removed through SSIM pruning; a priori depth is obtained based on virtual stereo rendering, a confidence mask is generated through multi-view verification, boundary consistency loss is constructed, and Gaussian densification is driven; and finally, carrying out TSDF fusion to obtain a seamless grid. Experiments show that reconstruction is completed in a 2.7 km scene within 6 h, MatrixCity-Aerial F1 reaches 0.784 and is improved by 41% compared with City GaussianV2, precision, efficiency and boundary consistency are considered, and the method is suitable for being applied to oversized scenes such as digital twin cities and automatic driving.

The invention provides a Bayesian nerve radiation field modeling method and system based on uncertainty perception and dynamic importance sampling. The method comprises the following steps: replacing a full connection layer in a multi-layer perceptron with a Bayesian linear layer to obtain a Bayesian neural radiation field BN-NeRF model; acquiring a data set containing house source photos of different viewing angles and corresponding camera positions, and training a BN-NeRF model by using the data set; performing preliminary coarse sampling on each light passing through the house source scene to obtain a coarse sampling point set, and performing uncertainty evaluation on each sampling point in the coarse sampling point set by adopting a trained BN-NeRF model; according to the uncertainty evaluation result corresponding to the preliminary coarse sampling, performing secondary sampling on each light passing through the housing resource scene to obtain a fine sampling point set; and integrating the coarse sampling point set and the fine sampling point set to generate a final sampling point set, calculating the color and volume density of each sampling point in the final sampling point set by adopting a trained BN-NeRF model so as to carry out volume rendering, and generating a final house viewing picture.

The invention belongs to the field of computer graphics and computer vision, and provides an explicit volume rendering primitive densification method based on detail perception gradient, which is suitable for a three-dimensional reconstruction task under a sparse view angle or sparse point cloud condition. According to the method, a detail perception gradient index is introduced, the fitting capability of a current model to a fuzzy region in an image is dynamically evaluated, and whether point refinement operation is executed or not is judged according to the fitting capability. Compared with a traditional densification strategy depending on position gradient, the index provided by the invention can more accurately identify a detail missing region, and the rationality of density distribution is improved. Furthermore, according to the method, visual angle correlation and gradient normalization information are combined, a threshold function is set to split and reconstruct Gaussian primitives, and low-overhead and controllably-distributed point set enhancement is achieved. While the compactness of the whole model is kept, the reconstruction quality of a complex structure and an edge region is effectively improved, and the method can be widely applied to various application scenes such as real-time rendering, underwater perception and sparse reconstruction.

The invention relates to the technical field of rice phenotype extraction, and discloses a rice fine phenotype extraction method based on three-dimensional vision, a terminal and a storage medium. The method comprises the following steps: acquiring a multi-view image of a rice plant by using a depth camera, and performing camera attitude estimation on the multi-view enhanced image to obtain camera internal and external parameters corresponding to each view; constructing a neural radiation field network, training the neural radiation field network by taking the enhanced image, the contour mask and internal and external parameters of the camera as input, and learning color and density distribution of a scene by utilizing a volume rendering technology; extracting surface points based on a mixed threshold strategy by using the trained neural radiation field network, and generating a rice three-dimensional point cloud; inputting the rice three-dimensional point cloud into a point cloud instance segmentation network, and outputting an instance mask containing each organ of the rice and a segmentation result of a three-dimensional bounding box; and counting phenotypic parameters of the rice according to a segmentation result. According to the method, the fine instance segmentation of the high-density point cloud is realized while the reconstruction precision is ensured, and the phenotypic analysis of the spike fraction is supported.

A system associated with an immersive experience framework may include an immersive virtual location data store containing information about a plurality of three-dimensional scenes (with each scene being associated with an immersive virtual location). An immersive virtual location tool may receive, from a creator, an immersive virtual location request (e.g., including an environment description). A request prompt is created based on the environment description and transmitted to a text-to-video generative AI model. A video of the virtual location is received from the generative AI model and converted into a three-dimensional scene using a volume rendering technique. Information about the scene is stored in the immersive virtual location data store and a user can interact with the scene using a substantially real-time experience interaction engine. In some embodiments, a JSON file describing the scene is directly generated using a LLM without creating the video.

The invention discloses a light field image coding method, which comprises the following steps of: constructing a multi-domain information coupled anchor point feature optimization module: taking a light field image and self-adaptive learning anchor point features as input, fusing space, angle and EPI information of the light field image with the self-adaptive learning anchor point features, and outputting light field information embedded anchor point features; constructing an inter-anchor-point context transformation module: taking the center anchor point feature embedded by the light field information and the features corresponding to the nearest anchor points as input, performing nonlinear transformation on the features corresponding to the anchor points, and outputting compact center anchor point features; carrying out compressed encoding on the center anchor point features by adopting an entropy model; predicting the attributes of Gaussian primitives based on the decoded anchor point features, and reconstructing a light field multi-view image in combination with a volume rendering technology; and constructing and training a 3DGS-based light field image coding network: inputting an original light field multi-view image, and outputting a reconstructed light field multi-view image based on a compact center anchor point feature.

The invention discloses a dynamic nerve radiation field real-time three-dimensional reconstruction method and device based on window attention and gradient equilibrium, and the method comprises the steps: carrying out the stratified sampling of rays, and obtaining sampling points; projecting the sampling points to six groups of two-dimensional feature planes; querying corresponding feature vectors from the six groups of feature planes through bilinear interpolation, and generating spatio-temporal features through Hadamard product fusion; inputting the fused spatial-temporal characteristics into a decoding network of a Swin Transform, and outputting the color and the volume density of a corresponding sampling point; and carrying out volume rendering integration along the ray of the camera to generate a composite image, and optimizing and updating the model according to a multi-target loss function between the composite image and a real image. The invention aims to solve the problems of large memory consumption, poor rendering quality, long training time and poor geometric consistency in the prior art, and realizes real-time three-dimensional reconstruction of a dynamic scene by using six-plane grid representation, Swin Transform feature generation and a gradient equilibrium mechanism.

The application discloses a light field image encoding method, comprising: constructing a multi-domain information coupled anchor point feature optimization module; taking a light field image and adaptively learned anchor point features as inputs, fusing spatial, angle and EPI information of the light field image with the adaptively learned anchor point features, and outputting anchor point features embedded with light field information; constructing an anchor point inter-context transformation module; taking center anchor point features embedded with light field information and features corresponding to the nearest neighbor anchor points as inputs, performing nonlinear transformation on the features corresponding to the anchor points, and outputting compact center anchor point features; and using an entropy model to compress and encode the center anchor point features; predicting the attributes of a Gaussian cell based on decoded anchor point features, and reconstructing a light field multi-view image in combination with volume rendering technology; constructing and training a 3DGS-based light field image encoding network; inputting an original light field multi-view image, and outputting a reconstructed light field multi-view image based on compact center anchor point features.

The application provides a method, device, equipment and medium for reducing image shooting requirements, and relates to the field of computer vision. The method comprises: collecting multi-view images of a scene to be rendered; sampling on a camera ray corresponding to a given pixel point to obtain a sampling point coordinate and a view direction; calculating a hour angle h, an declination δ' and a solar elevation angle θ s , a solar azimuth angle Φ s , obtaining a global light direction vector, and encoding; encoding the shooting time of the image; inputting the sampling point coordinate, the view direction, the light encoding and the time encoding into a neural network to obtain the color and the volume density of the sampling point under the corresponding view; obtaining the final predicted pixel color by using volume rendering, calculating the color loss; and updating the neural network parameters by back propagation. The application can weaken or even eliminate the influence of light changes on outdoor shooting, and reduce the shooting requirements of neural radiance field images.

The patent application provides a complex environment three-dimensional scene reconstruction method based on a multi-view neural network, and the method mainly comprises two steps: scene information capturing: obtaining scene data from a plurality of views through a high-resolution camera, and ensuring that all details of a complex scene are covered; the captured image is standardized, including color correction and noise removal, to ensure image quality and consistency. And three-dimensional model reconstruction: performing three-dimensional reconstruction on the acquired scene data by using a neural network. Firstly, light sampling is performed on a scene through camera parameters and pose information, and an occupancy grid is created to optimize the sampling efficiency. Thirdly, decomposing the sampling data into feature vectors and inputting the feature vectors into a neural network which is divided into a density network and a color network and is used for generating volume density and color information of the scene; and finally, obtaining a three-dimensional scene model through a volume rendering technology. The method has the advantage that a high-precision three-dimensional model can be quickly generated in a complex environment through an efficient sampling and data processing technology.

A method and server for volume rendering of 3D scenes are provided. The method comprises training a given machine-learning algorithm (MLA) of a plurality of MLAs to identify a boundary between a plurality of interpenetrated objects to be rendered in a given 3D scene, by applying a signed distance function (SDF) loss function configured to penalize a respective predicted SDF value, generated by the given MLA during a given training iteration, for a given point of a training 3D scene, in response to the respective predicted SDF value generated by the given MLA being equal to the respective predicted SDF value generated by an other MLA of the plurality of MLAs.

The invention provides an unstructured grid volume rendering method and device based on a neural point field, and relates to the technical field of data processing, and the method comprises the steps: extracting an outer surface grid of a physical field domain from original unstructured grid data; adaptive sampling is carried out on the physical field space gradient modulus length, and a training data set containing sampling point coordinates and corresponding physical field values is obtained; training a pre-built neural point field model by adopting a strategy containing multi-scale Fourier feature coding; generating a projection light ray for each pixel of the image plane, determining a sampling range of the projection light ray in a physical field domain by using the outer surface grid, and generating a space sampling point in the sampling range; and inputting the spatial sampling points into the neural point field model to generate a predicted field value, and generating a corresponding pixel color. The invention provides an efficient solution for real-time and high-precision visual analysis of a large-scale unstructured grid physical field.

A layout of labels for annotating a rendered image is optimized. A rendered image is obtained, and locations of a plurality of regions of interest in the rendered image are determined. Semantic information associated with the plurality of regions of interest is obtained. Based on the semantic information and the locations of the plurality of regions of interest, and, taking into account a visibility of the labels and a further visibility of the regions of interest in the rendered image, the layout of the labels for annotating the plurality of regions of interest in the rendered image is determined.

The application discloses a digital human reconstruction method with high-fidelity triangular mesh and material texture mapping, and belongs to the technical field of computer graphics and digital human reconstruction. S1: performing space point sampling on each frame of picture corresponding to monocular video based on ray tracing, and deforming the sampling points to distribution under a standard posture; S2: acquiring geometric information and color information of global space points; S3: performing integration on the sampling points on each light ray to obtain volume rendering results, and completing first-stage optimization; S4: selecting a target frame, initializing a three-dimensional mesh, and generating a human body geometric surface; S5: acquiring material texture properties of the corrected human body geometric surface through a material network; S6: realizing differentiable rendering on the corrected human body geometric surface; S7: introducing an information fusion strategy to generate dense body rendering results under a virtual perspective, and supervising second-stage optimization; and S8: finally generating a digital human with a high-quality triangular mesh surface and material texture properties.

The invention relates to an AI narrative design generation method and system for package design, and the method comprises the steps: constructing a knowledge graph according to a preset theme, and obtaining a preset theme knowledge graph; according to the preset subject knowledge graph, generating packaging content through an AIGC engine; performing style unification and content coherence processing on the visual elements of the packaging content through AI-driven narrative logic in combination with preset theme information in the preset theme knowledge graph to obtain preset theme packaging content; and performing three-dimensional rendering on the preset theme packaging content to obtain a final packaging design. By adopting the technical scheme of the invention, the visual element narrative coherence can be realized, and the immersive interaction of consumers is achieved by real-time rendering and multi-dimensional interaction.

The application discloses a sparse-view indoor reconstruction method based on uncertainty-aware depth supervision, which comprises the following steps: acquiring multi-view image data; solving camera pose based on SFM; constructing a neural radiance field model based on uncertainty awareness, modeling the volume density, color and depth uncertainty parameters of a space point; introducing a depth uncertainty synthesis formula in a volume rendering framework; constructing a photometric loss fused with random structure similarity; designing an adaptive depth optimization mechanism based on uncertainty awareness, optimizing the training process through a progressive uncertainty learning strategy; and applying the trained model to a sparse-view indoor scene to generate high-quality three-dimensional reconstruction results and uncertainty quantification graphs. The application effectively solves the problems of poor reconstruction quality and missing details under sparse views by introducing a fusion method of uncertainty awareness and adaptive depth supervision, and significantly improves the accuracy and robustness of indoor scene three-dimensional reconstruction.

The application provides a kind of audio-driven three-dimensional digital human generation method and system based on neural radiance field, first, a pose preserving prior generation model is constructed, by fusing audio features and facial pose features, the cross-modal interaction relationship between the two modal features is learned using self-attention mechanism and cross-attention mechanism, to generate facial pose prior information synchronized with input audio. Then, a neural radiance field rendering model is constructed, three-plane hash coding technology is used to efficiently sample the three-dimensional space, and the facial pose prior and audio features are dynamically fused through the attention mechanism. Finally, the generated pose prior is input into the neural radiance field rendering model for volume rendering, and the light cumulative transmittance calculation is introduced to generate high-fidelity three-dimensional digital humans. The application improves the realism and expressiveness of digital humans, reduces the rendering computation complexity, and realizes real-time digital human generation with dynamic coherence and lip-synchronized.