447 results about "3d space" patented technology

The invention discloses an image processing method and device and a storage medium. The method comprises the steps of obtaining a to-be-simulated 3D convolution model and training data; decomposing the 3D convolution model into cascading of a 3D space convolution model and a 3D time convolution model to obtain a pseudo 3D cascading convolution model; training a pseudo 3D cascade convolution modelby using the training data, and obtaining parameters of a 3D spatial convolution model and a 3D time convolution model; converting the 3D space convolution model and the 3D time convolution model intoa 2D space convolution model and a 2D time convolution model; setting a feature rearrangement rule for the 2D spatial convolution model and the 2D time convolution model; mapping model parameters ofthe 3D spatial convolution model and the 3D time convolution model into parameters of a 2D spatial convolution model and a 2D time convolution model to obtain a 2D cascaded convolution model; and performing convolution operation on the image by using the 2D spatial convolution model and the 2D time convolution model. By means of the mode, image processing conducted through 3D convolution operationcan be achieved through the 2D convolution model.

Three-dimensional Gaussian splatting mechanisms that initialize a set of 3D Gaussian distributions, un-project pixels from two-dimensional (2D) planes to 3D space by applying queries to the 3D Gaussians at expected un-projected ray depth positions, and splat the 3D Gaussian distributions on the 2D planes based on the expected un-projected ray depth positions.

Techniques are discussed herein for generating three-dimensional (3D) representations of an environment based on two-dimensional (2D) image data, and using the 3D representations to perform 3D object detection and other 3D analyses of the environment. 2D image data may be received, along with depth estimation data associated with the 2D image data. Using the 2D image data and associated depth data, an image-based object detector may generate 3D representations, including point clouds and / or 3D pixel grids, for the 2D image or particular regions of interest. In some examples, a 3D point cloud may be generated by projecting pixels from the 2D image into 3D space followed by a trained 3D convolutional neural network (CNN) performing object detection. Additionally or alternatively, a top-down view of a 3D pixel grid representation may be used to perform object detection using 2D convolutions.

A method and apparatus generate a 2-dimensional (2D) image. A method for generating a 2-dimensional (2D) image includes obtaining a time index and a view. The method further includes obtaining first coding indices and a first codebook for canonical 3D Gaussians, wherein the canonical 3D Gaussians are 3D Gaussians corresponding to a reference time index, and represent a 3D space corresponding to the reference time index. The method also includes obtaining second encoding indices and a second codebook for a parameter offset, wherein the parameter offset indicates a difference between the canonical 3D Gaussians and 3D Gaussians for the time index. The method further includes reconstructing the canonical 3D Gaussians based on the first coding indices and the first codebook. The method also includes reconstructing parameter offsets of the 3D Gaussians for the time index based on the second coding indices and the second codebook. The method further includes adding the reconstructed canonical 3D Gaussians and the reconstructed parameter offset to reconstruct the 3D Gaussians for the time index. The method also includes generating a second image for the view based on the reconstructed 3D Gaussians.

Methods, apparatus, devices, subsystems, and systems for holographically displaying live scenes are provided. In one aspect, a method includes optically generating an optical hologram of a live scene including one or more three-dimensional (3D) objects; capturing sequential optical holograms of the live scene and generating sequential hologram data associated with the sequential optical holograms of the live scene, each optical hologram being associated with respective hologram data; processing the at least part of the sequential hologram data based on a mathematical function to remove low frequency components to generate digital holograms associated with the live scene; and reconstructing the live scene in a 3D space based on the digital holograms.

The invention relates to the technical field of smart home and artificial intelligence, discloses a smart home equipment automatic arrangement method and system fusing 3D modeling and a neural network, and aims to solve the problems of complex manual operation, high technical threshold and high error rate in the traditional scheme design. The method comprises the following steps: acquiring original three-dimensional point cloud data of a target space through laser scanning equipment, and constructing a centimeter-level precision 3D space model; a deep learning semantic segmentation network is adopted to identify building components and functional areas, and a spatial semantic map is generated; performing multi-modal fusion on the spatial semantic features and user requirements to generate equipment type probability distribution and an optimal position thermodynamic diagram; performing iterative correction on the initial arrangement scheme based on a multi-objective optimization algorithm, and synchronously optimizing wireless signal coverage intensity, equipment energy consumption efficiency and project implementation cost; the optimized 3D arrangement scheme is automatically converted into a two-dimensional vector engineering drawing, and a construction drawing is generated; and extracting an equipment list and material consumption according to the final arrangement scheme, and automatically generating a project quotation and an installation process guidance document. The system comprises a spatial semantic analysis module, an equipment arrangement decision module, an engineering drawing generation module and a multi-objective optimization engine. Through adoption of the technical scheme, full-process automation from original spatial data to constructable drawings can be realized, and the efficiency and specialty of smart home scheme design are remarkably improved.

A computer-implemented method for 3D reconstruction of a target is provided, comprising obtaining an initial global reconstruction of the target in a 3D space, inferred by a global machine learning model; providing, to a user, an initial visualisation of the target based on the reconstruction; receiving, from the user, at least one indication of at least one point of interest in the visualisation; resampling at least one first subsection of the target based on the at least one point of interest to obtain local data, wherein the local data is associated with the subsection based on spatial information that associates the local data with a point in 3D space; inputting the resampled local data and spatial information into a local feature machine learning model to obtain at least one 3D reconstruction of the target, wherein the local feature machine learning model has been trained to output a target reconstruction from local data of resampled subsections, and wherein the 3D coordinate system of the local 3D reconstruction aligns with the global 3D reconstruction; and merging the global 3D reconstruction with the local 3D reconstruction. A corresponding computer system and computer readable medium may also be provided.

This invention relates to a method for visually recognising and tracking objects and events in 3D space with high accuracy, reduced computational load and fast setup, using virtual sensors that are designed to ensure high reliability and reduce costs by maximising their potential for accurate object recognition for given visual input data. The main task of a virtual sensor is to provide cost-effective and reliable data and replace the use of humans and / or hardware sensors to detect and recognise objects or events that are important to a given entity, such as an industrial enterprise and various business entities. The solution is well suited for industrial applications, especially for tracking larger objects of known shape or appearance, such as in large industrial halls, warehouses without fixed racking systems, container docks, train docks, car parks, etc. It is particularly suitable for tracking coils, metal pieces and larger building components. The method provides a reliable source of i data for obtaining a digital twin of the monitored objects, which represents real-time information about each stock keeping unit (SKU) in the warehouse (WH), bringing significant benefits to WH managers and thus saving human labour spent on searching for materials, improving management flow, reducing overall equipment effectiveness (OEE) of vehicles, improving throughput, quality, etc.

Provided herein are systems for determining a location of a gross target volume of a patient. In some examples, systems can include one or more processors that are configured to obtain image data associated with a plurality of images of a lesion of a patient. For each image, the one or more processors can be configured to backproject points representing the lesion into the 3D space to determine a plurality of distribution confidence values for a subset of voxels within the three-dimensional space. The one or more processors can be configured to determine a three-dimensional confidence distribution based on confidence values from the plurality of distribution confidence values corresponding to each voxel of the 3D space and determine a position of the lesion within the 3D space based on the 3D confidence distribution.

The invention provides a logistics track spatial data acquisition method based on a three-dimensional model, and relates to the technical field of data processing, and the method comprises the steps: 1, collecting the dynamic position data of a logistics object in a three-dimensional space in real time through a multi-source sensor, the three-dimensional space coordinate data comprising horizontal direction coordinates and vertical direction coordinates; 2, inputting the dynamic position data into a pre-constructed three-dimensional environment model, and generating a preliminary three-dimensional track through a space registration algorithm; and step 3, performing real-time updating processing on the preliminary three-dimensional trajectory, dynamically correcting trajectory data through a time sequence analysis algorithm, fusing obstacle information in the three-dimensional environment model, optimizing the precision of the preliminary three-dimensional trajectory, and obtaining optimized three-dimensional trajectory data. According to the invention, high-precision acquisition, dynamic optimization and precise correction of the logistics track data are realized, the accuracy and environmental adaptability of the logistics track data are improved, reliable data support is provided for intelligent logistics management, and the logistics efficiency is improved.

The invention provides an occupied grid prediction method and device, electronic equipment and a readable storage medium. The method comprises the following steps: acquiring continuously acquired multiple frames of visual images and laser point cloud data; converting the visual image to a point cloud scene to obtain image point cloud data; fusing the image point cloud data and the laser point cloud data to obtain fused voxel features; performing motion perception prediction on the dynamic voxels in the fused voxel features through the occupancy flow to obtain motion perception features corresponding to the dynamic voxels; and performing state prediction based on the motion perception features to obtain an occupation prediction result. Visual images are converted into image point cloud data, and the image point cloud data and laser point cloud data are fused in a 3D space to obtain voxel features, so that information loss caused by multi-mode mutual projection can be avoided, and meanwhile, motion perception is directly performed on fused voxels through an occupation stream, so that complete space information can be effectively captured and utilized, and the accuracy of the motion perception is improved. Therefore, more accurate and comprehensive occupied grid prediction is realized.

A cross-regional and cross-view learning (C2RV) framework is provided for sparse-view reconstruction in cone-beam computed tomography (CBCT) by advantageously leveraging cross-region and cross-view feature learning to enhance representation of a point in 3D space before estimating an attenuation coefficient of the point. Specifically, multi-scale 3D volumetric representations (MS-3DV) are first introduced, where features are obtained by back-projecting multi-view features at different scales to the 3D space. Explicit MS-3DV enable cross-regional learning in the 3D space, providing richer information that helps better identify different internal anatomy structures. Hence, features of the point can be queried in a hybrid way, i.e. multi-scale voxel-aligned features from MS-3DV and multi-view pixel-aligned features from projections. Instead of considering queried features equally, scale-view cross-attention (SVC-Att) is used to adaptively learn aggregation weights by self-attention and cross-attention. Finally, multi-scale and multi-view features are aggregated to estimate the attenuation coefficient.

Embodiments are disclosed for determining a spin rate and axis of a ball using deep learning. In some embodiments, a method comprises: training a deep learning network on training images of spinning balls, each spinning ball having at least one feature point in a time series that forms a two-dimensional (2D) ellipse image in a 2D plane; capturing, with at least one camera, a series of images of a ball; predicting, with the trained deep learning network, spin measurements associated with the ball based on the series of images; determining a spin rate of the ball based on the spin measurements; determining coefficients of a 2D ellipse model based on the spin measurements and the spin rate; and determining, with the at least one processor, a spin axis of the ball in 3D space based on the 2D ellipse model and the spin rate.

The invention discloses a virtual projection 3D oral and maxillofacial key point detection method based on deep learning, and aims to solve the problem of insufficient 3D image key point detection precision and efficiency in the prior art. The method comprises the following steps: projecting 3D image data into a plurality of 2D projection images by using a virtual projection technology; then key point detection and marking are carried out on the 2D image by using a deep learning network model, and 2D mark point coordinates are obtained; and finally, performing 3D reconstruction by adopting a virtual back projection method, and back-projecting the 2D mark point information to a 3D space to obtain 3D mark point coordinates. According to the method, 2D key point detection and 3D projection and reconstruction technologies are combined, rapid, high-precision and automatic detection of 3D image mark key points is realized, and the method can be widely applied to the fields of oral medical image analysis and the like.

The invention provides a computer implementation method and device for determining space occupation information, equipment and a medium, multiple groups of binocular images collected by binocular cameras arranged in different directions of a moving body can be acquired, then image features and depth features of each group of binocular images are determined based on the multiple groups of binocular images, and then the space occupation information is determined. And projecting the image features and the depth features of each group of binocular images to an overlook coordinate system, determining target aerial view features in the overlook coordinate system, and finally, determining space occupation information in the look-around space of the moving body according to the target aerial view features. The target aerial view feature reflects the 3D space information around the moving body only through the 2D format, the storage space occupied by the 3D space information in the calculation process and the calculation complexity are reduced, and the calculation precision and the calculation efficiency of determining the space occupation information based on the target aerial view are further improved.

The invention relates to the technical field of industrial fault diagnosis, in particular to a multi-sensor and cross-working-condition industrial fault diagnosis method. The method comprises the following steps: S1, acquiring equipment operation data of a plurality of sensors under different working conditions through a data acquisition system, and dividing the equipment operation data into a training set, a verification set and a test set; s2, constructing the original data into a 3D space-time collaborative tensor as model input; s3, CBT, LMSCB, ESRM and KAN-TD are embedded into a network architecture, LightM-ConvKNet intelligent fault diagnosis modeling is completed, then pre-training of the model is completed by using a source domain sample, and fine tuning is performed on the model based on a target domain sample; and S4, inputting the target domain test set into the fine-tuned model to generate a fault diagnosis result. By adopting the method, the limitation of the CNN convolution kernel size is broken through, the local correlation among sensor data can be fully acquired, the parameter quantity is reduced, the calculation efficiency is improved, and the calculation complexity is reduced.

The invention relates to the technical field of lane line detection, in particular to a 3D-BEV lane line sensing method based on polar coordinates, and the method comprises the steps: carrying out the 3D lane line detection through employing the polar coordinates to replace a conventional method based on a rectangular coordinate system, employing a sliding long-strip-shaped window convolution to adapt to a polar coordinate system, and carrying out the 3D lane line detection through employing the sliding long-strip-shaped window convolution. According to the method, the lane line in the picture can be accurately identified in the polar coordinate system, accurate positioning is carried out in the 3D space, the false detection rate is greatly reduced, the radial distribution characteristic of perspective projection can be more naturally matched, and the modeling precision of the geometric structure of the lane line is remarkably improved.

The invention discloses a parking lot license plate fee evasion prevention visual identification method and system, and relates to the technical field of intelligent parking management. The method comprises the following steps: step 100, synchronously obtaining a lane panoramic image and 3D spatial data of a corresponding object; step 200, determining whether there is a multi-vehicle foldover area in the panoramic image based on double-feature constraints of license plate semantic features and geometric features; step 300, based on the constraint of a standard license plate template library, carrying out pixel splitting and image completion on the multi-vehicle foldover area to obtain an independent and complete single license plate image; step 400, performing character recognition on the single license plate image, and verifying the authenticity of the license plate in combination with the 3D space data; and step 500, realizing multi-license-plate independent charging and vehicle-by-vehicle gate opening control based on an identification result, and storing related data for tracing. According to the method, the problems of misrecognition and missed recognition fee evasion caused by multi-vehicle foldover are solved, license plate counterfeiting fee evasion is prevented, and the benefits of the parking lot are guaranteed.

Apparatuses, systems, and techniques of using one or more machine learning processes (e.g., neural network(s)) to predict occupancy using an image input. In at least one embodiment, image data is processed using a neural network to predict occupancy in a 3D voxel space. In at least one embodiment, image data is processed using a neural network to detect objects in a 3D space.

The invention relates to the technical field of aquaculture, in particular to a net cage culture fish ingestion behavior recognition and accurate feeding system and a method thereof.The system comprises a multi-modal data acquisition module arranged around a net cage and used for acquiring video data and biological density information of fish behaviors, a data preprocessing module used for coding and compressing the video data, and a data processing module used for processing the coded and compressed video data; the biological density information is packaged, the 3D space-time attention analysis module receives the processed data, space-time features are extracted through a 3D deep learning model, and fish feeding behaviors are recognized. The multi-modal feature fusion module adaptively fuses the spatial-temporal features of the video data and the biological density information to generate ingestion intensity data, and the feeding control module calculates the vibration frequency according to the ingestion intensity data and controls the opening state of a netting door in a net cage, so that precise feeding is realized. The problem that a single data source is low in recognition accuracy in a complex underwater environment is solved, the recognition accuracy of the system under various water quality conditions is improved, and accurate feeding of fishes is achieved.

Aspects of this technical solution can allocate one or more portions of a mesh in a three-dimensional (3D) space to one or more processing units associated with the one or more circuits, the mesh associated with a surface of an object in the 3D space, transform, by the one or more of the processing units, one or more of the portions of the mesh from the 3D space into corresponding second meshes in a two-dimensional (2D) space, segment, by the one or more of the processing units, according to a determination that a distortion of a portion of the mesh among the one or more of the portions of the mesh is below a threshold of parameterization, the portion of the mesh into two further portions of the mesh, where the further portions are among the one or more portions of the mesh, and generate, according to a determination that the distortion of the portion of the mesh among the one or more of the portions of the mesh is at or above the threshold of parameterization, an output mesh including the one or more portions of the mesh.

A system comprises an intraoral scanning device and a processor. The intraoral scanning device comprises a wand including a probe, one or more light projectors disposed in the probe and configured to project a structured light pattern, wherein the structured light pattern comprises first pattern features of a first type and second pattern features of a second type, and two or more cameras disposed in the probe and configured to acquire one or more sets of images. The processor is configured to solve a correspondence problem within each set of images such that first points in 3D space are determined based on a captured subset of the first pattern features and a corresponding projected subset of the first pattern features and second points in 3D space are determined based on a captured subset of the second pattern features and a corresponding projected subset of the second pattern features.

A head-wearable apparatus improves user interactions with virtual interface elements in augmented reality (AR) environments. The apparatus tracks hand movements in 3D space, calculating velocity vectors and positions of fingers. For each virtual interface element, it determines a UI-to-finger vector and calculates alignment with the finger's velocity vector. Interaction zones are dynamically adjusted based on this alignment and velocity magnitude. The system evaluates consistency between finger movement and element locations to predict intended targets. Interactions are triggered when fingers enter adjusted zones of predicted targets. This approach reduces erroneous selections and improves interaction accuracy, even for closely-spaced elements. The invention applies to various AR scenarios, enhancing user experiences in applications like gaming and education.

3D objection detection is a computer vision task that generally refers to detecting (e.g. classifying and localizing) an object in 3D space from an image or video that captures the object. This computer vision task has many useful applications, such as autonomous driving applications which rely on the detection of 3D objects in a local environment to make autonomous driving decisions. State-of-the-art 3D object detectors generally rely on machine learning, but current training processes for these detectors do not specifically address false negative detections, or missed objects, which are often caused by occlusions and / or cluttered backgrounds in the given image / video. Reducing false negatives is crucial for many downstream applications, particularly autonomous driving applications which rely on accurate detection of obstacles for making safe driving decisions. The present disclosure provides for a multi-stage training process that reduces false negative detections by 3D object detectors.

The application discloses a 3D printing multi-sector multifunctional metasurface and a design method thereof. The multi-sector multifunctional metasurface is designed based on five polarization multiplexing multifunctional metasurfaces with different phase control means, then the five metasurfaces are respectively faced to five different directions in space and integrated by using a 3D printing technology, and a multi-sector metasurface is formed. Finally, the designed multi-sector metasurface can integrate different electromagnetic functions in five directions in a 3D space. The application designs a polarization multi-sector metasurface based on the 3D printing technology. The polarization multi-sector metasurface can integrate electromagnetic functions of non-common polarization waves on a single device, and can complete wave front control in multiple directions in space, so that the electromagnetic wave control range is expanded from only transmission and reflection to multi-direction control, and the information capacity of the electromagnetic device and the space utilization rate of the wave front control are greatly improved.

The application discloses an intelligent path planning method and system for a mechanical arm spray head of a deicing vehicle, and can realize automatic and accurate deicing on a complex three-dimensional surface of an airplane. The method comprises the following steps: target surface extraction and reconstruction are performed on a deicing area of the surface of the airplane to generate a continuous three-dimensional surface; the three-dimensional surface is mapped to a two-dimensional parameter plane by surface parameterization; a 2D waypoint sequence is generated on the two-dimensional parameter plane in combination with the effective spraying width of the nozzle; 3D space inverse mapping is performed based on the 2D waypoint sequence, the spray head posture of the deicing vehicle is aligned based on the result of the inverse mapping, and the three-dimensional target posture of the nozzle at each waypoint is obtained; the target posture of the nozzle is converted into specific angle instructions of each joint of the mechanical arm of the deicing vehicle; on the basis of the joint angle instructions, joint trajectory smoothing and S-type speed planning are performed, and finally the action path of the spray head of the mechanical arm of the deicing vehicle is generated.

The invention discloses a bulk cargo cabin real-time fusion detection method based on a laser point cloud, and the method comprises the steps: obtaining a real-time three-dimensional scene point cloud of a port scene, and obtaining a deck plane equation and a normal vector; projecting the three-dimensional scene point cloud to a two-dimensional deck plane, and constructing an intensity image taking point cloud intensity as information and a depth image taking projection distance as information; obtaining an intensity mask image from the intensity image; acquiring a depth edge image according to the depth image, and fusing the depth edge image and the intensity mask image to obtain a fused edge image; contour extraction is carried out on the fused edge image, polygon fitting is carried out, a quadrangle in the fused edge image is reserved as a candidate cabin, and a minimum enclosing rectangle of the quadrangle is obtained; and obtaining four vertex coordinates of the minimum bounding rectangle, and back-projecting the four vertex coordinates to the three-dimensional space to obtain three-dimensional space rectangular coordinate points of the final cabin. The method is suitable for a complex environment and a dynamic scene in a port operation area, and the cabin in the dynamic scene can be obtained in real time.

A method includes providing an image to a feature extraction model to generate feature data. The image depicts a portion of a first device and a portion of a second device. The feature data includes coordinates representing key points of each of the first and second devices depicted in the image. The method also includes obtaining position data indicating a position in 3D space of a connector of the first device. The method further includes providing the feature data and the position data to a trained autonomous agent to generate a proposed maneuver to mate the connector with a connector of the second device.

The application provides a large scene graph switching method, system and device and a storage medium, and the method comprises the following steps: S1, constructing a point cloud map and a graph database of a large scene; S2, robot positioning according to the graph map and real-time switching of the graph. The application firstly cuts the high-precision point cloud map of the complete large scene according to the graph, which is used for constructing the graph database of the complete large scene; then dynamically loads the graph around the current position of the robot according to the graph database, which is used for composing the graph map of the local large scene, and is used for real-time calculation of the position of the current robot in the large scene; and according to the current robot position and the predicted action path, it is decided which graphs in the graph database are switched to at present, and the graph map of the local large scene is updated in real time; the robot can realize fast and real-time graph switching in the super large scene without feeling, and accurate 3D space positioning can be realized on the hardware device with limited computing power.