66 results about "Orientation (computer vision)" patented technology

The invention relates to the technical field of computer vision, and discloses a three-dimensional scene construction method and device, equipment and a storage medium, and the method comprises the steps: collecting the data of a real three-dimensional scene, and obtaining image data and point cloud data; performing target detection according to the image data and the point cloud data, and determining the object category and the three-dimensional bounding box of each object in the real three-dimensional scene; for each object, determining target image data including the object from the image data; generating a first three-dimensional model of the object according to the object category of each object and the target image data; adjusting the size of the first three-dimensional model of the object according to the size information carried by the three-dimensional bounding box of each object to obtain a second three-dimensional model of the object; and constructing a virtual three-dimensional scene corresponding to the real three-dimensional scene according to the position information and the orientation angle information carried by the three-dimensional bounding box of each object and the second three-dimensional model of the object. According to the invention, the three-dimensional scene can be constructed quickly and accurately.

The invention provides a three-dimensional model view generation method and device, a storage medium and electronic equipment, and is applied to the technical field of computer vision. The method comprises the steps of obtaining a target three-dimensional model; adjusting the orientation of the target three-dimensional model by using a pre-trained orientation discrimination model to obtain the target three-dimensional model with the adjusted orientation; based on the target three-dimensional model with the adjusted orientation, multi-view rendering is conducted on the target three-dimensional model according to preset parameters, a view rendering result is obtained, and the view rendering result comprises multiple views of the target three-dimensional model. According to the method, orientation adjustment is carried out on the target three-dimensional model through an automation technology, and preset parameter rendering is carried out based on the orientation adjustment, so that multiple views are quickly and efficiently generated, large-scale three-dimensional model processing is supported, the generation speed is remarkably improved, the rendering time is shortened, and the requirement for digitization of a large number of cultural relics is met.

This invention belongs to the field of image processing and computer vision technology, and discloses a lightweight single-image super-resolution reconstruction method based on a directional stripe hybrid network. It captures directional structural information of the image from both the spatial and frequency domains through horizontal stripes, vertical stripes, local window attention of the directional stripe self-attention module, and direction-specific convolution of the directional high-frequency enhancement module. The directional stripe self-attention module effectively captures long-distance dependencies extending along specific directions; the directional high-frequency enhancement module extracts high-frequency components through frequency domain decomposition, combines them with direction-specific convolution, and introduces a dual-guided adaptive enhancement mechanism. It identifies texture-rich regions through local variance maps and structural boundaries through edge maps, generating adaptive weights to guide feature enhancement, accurately enhancing high-frequency edges and textures with different orientations while effectively suppressing noise. The two work together to achieve spatial-frequency domain dual-level directional modeling, significantly improving the reconstruction quality of images rich in directional textures, such as architectural scenes and cartoon images, while maintaining a lightweight design.

The invention relates to the technical field of computer vision and image processing, in particular to a space target three-dimensional reconstruction method based on epipolar line mapping constraint dense search. The processing can reduce the dependence on transformation matrix projection parameters in the weak texture region matching process. A kernel line coordinate system is established, orientation processing is carried out on the relative direction of a matched feature image pair, a new inter-pixel mapping relation is obtained, the inter-pixel mapping relation is used as dense search constraint in a triangulation network, and dense search has a better search effect in a weak texture area. And integrating the kernel line mapping search result and the existing epipolar constraint relationship, and finally performing multi-view reconstruction to obtain a target physical model. According to the method, dense point searching between matched image pairs can be effectively carried out, pixel point level searching precision can be achieved, and three-dimensional reconstruction of a space target is achieved.

The invention provides a full-scene group activity understanding method based on mixed hierarchical representation, and mainly relates to a group activity understanding problem in the field of computer vision. The method comprises the steps of constructing a data set and a neural network framework, training a network, testing the network and the like. Wherein the neural network framework firstly extracts time sequence features, space environment features and the like to form rich individual features, and classifies individual actions; then, taking the individual features as nodes, reasoning information between the nodes at an individual interaction level and an action interaction level by using the proposed graph reasoning strategy, and fusing the information into group features; meanwhile, topological graph structures are constructed according to the position coordinates and the orientation respectively, a graph neural network is introduced for representation, and hierarchical features of a group structure are fused; and finally, splicing the group features and the group structure hierarchical features to identify the group activity type. According to the method, the multi-level representation of the group activity is fully realized, and the simultaneous identification of the individual actions of the group members and the whole activity type is realized.

A method, device, storage medium, and electronic device for generating a three-dimensional model view provided by the present invention are applied to the field of computer vision technology. The present invention obtains a target three-dimensional model; uses a pre-trained orientation discrimination model to adjust the orientation of the target three-dimensional model to obtain a target three-dimensional model with the adjusted orientation; based on the target three-dimensional model with the adjusted orientation, performs multi-view rendering on the target three-dimensional model according to preset parameters to obtain a view rendering result, where the view rendering result includes multiple views of the target three-dimensional model. The present invention adjusts the orientation of the target three-dimensional model through an automated technique and performs rendering with preset parameters based on this, thereby quickly and efficiently generating multiple views, thus supporting the processing of large-scale three-dimensional models, significantly improving the generation speed and shortening the rendering time to meet the needs of a large number of cultural relic digitizations.

The application provides a monocular three-dimensional object detection method, device, equipment and product, and relates to the technical field of computer vision.The method comprises the following steps: acquiring a to-be-detected image; inputting the to-be-detected image into a monocular three-dimensional object detection model to obtain a two-dimensional detection result, object depth, observation angle and size of the to-be-detected object output by the monocular three-dimensional object detection model; generating a three-dimensional center point of the to-be-detected object in a three-dimensional space according to a two-dimensional center point and the object depth; generating an orientation angle of the to-be-detected object according to the three-dimensional center point and the observation angle; and generating a three-dimensional detection result according to the three-dimensional center point, the orientation angle and the size.The application significantly optimizes the prediction performance of a truncated object in monocular three-dimensional detection, is not limited to the type of a detection object, and can perform monocular three-dimensional detection on normal objects and truncated objects.

The application discloses an image generation method based on a diffusion model, and relates to the field of computer vision and remote sensing image processing. Through a direction and size perception attention mechanism, the application realizes the cooperative and accurate control of the target position, direction and size in remote sensing image generation, and the geometric consistency of the generated image and the input layout is significantly improved. Based on a complete layout rebalancing, conditional generation and quality screening process, the application can automatically and high-quality synthesize a data set, and the accuracy of a downstream rotating target detector is obviously improved on multiple benchmark data sets, effectively solving the problems of data scarcity, uneven distribution and high labeling cost in remote sensing detection tasks, and providing reliable data support for remote sensing target detection tasks. The method can be used for generating synthetic remote sensing data sets with consistent layout and accurate geometric properties, and is especially suitable for generating remote sensing target detection training data containing various orientations and sizes.

The invention discloses an image generation method based on a diffusion model, and relates to the field of computer vision and remote sensing image processing. According to the invention, through a direction and size perception attention mechanism, cooperative accurate control of the position, direction and size of the target is realized in remote sensing image generation, and the geometric consistency of the generated image and the input layout is significantly improved; on the basis of complete layout weight balance, conditional generation and quality screening processes, data sets can be automatically synthesized in a high-quality mode, the precision of downstream rotating target detectors on multiple reference data sets is obviously improved, the problems of data scarcity, unbalanced distribution and high labeling cost in remote sensing detection tasks are effectively solved, and the accuracy of the remote sensing detection tasks is improved. And reliable data support is provided for a remote sensing target detection task. The method can be used for generating a synthetic remote sensing data set with consistent layout and accurate geometric attributes, and is particularly suitable for generating remote sensing target detection training data containing various orientations and sizes.

A calibration method for performing calibration in a computer vision system using a three-dimensional reference object asymmetric as viewed in any direction and having predetermined dimensions includes calculating, based on three-dimensional shape data about the three-dimensional reference object and three-dimensional measurement data obtained through three-dimensional measurement of the three-dimensional reference object, a position and an orientation of the three-dimensional reference object relative to a measurement unit coordinate system defined for the three-dimensional measurement unit, and calculating a reference-measurement unit transformation matrix representing a coordinate transformation between the measurement unit coordinate system and a reference coordinate system defined for the three-dimensional reference object.

This invention provides a starting point annotation correction method based on panoramic image and scene graph matching, relating to the fields of computer vision and building information processing technology. The method includes: generating sets of triangles with the same orientation combination—a scene graph triangle set and an actual scene triangle set—based on the actual scene corner point set and the scene graph feature point set, respectively; matching the triangle pair with the highest similarity between the two sets by calculating the similarity of the triangle's interior angles and side length ratios, establishing a correspondence between local obstacles in the scene graph and the actual scene; mapping the physical starting point position of the actual scene to the scene graph coordinate system based on the affine transformation parameters determined by the triangle pair, and outputting the corrected starting point coordinates. This invention can achieve accurate and efficient restoration of the current layer layout of a construction site.

This application belongs to the field of computer vision and 3D point cloud processing technology. It provides a convolution method for anisotropic point clouds based on local geometric adaptation. The embodiments of this disclosure estimate the principal orientation of the local geometric structure by performing principal component analysis on the neighborhood point cloud and constructing an affine transformation matrix. This matrix is ​​then used to dynamically deform the basic kernel points of the convolution operator, dynamically aligning the shape and orientation of the convolution kernel with the local structure of the point cloud. Integrating this convolution operator into a point cloud registration network allows for feature extraction from 3D point clouds, particularly those from underwater multibeam echo sounders. The convolution operator constructed using this method effectively captures the structural priors of strongly anisotropic point clouds, solving the problems of structural information loss and limited feature extraction capabilities caused by isotropic kernels when processing such data.

The present disclosure provides a method, device, electronic device and medium for driving a virtual image, which relates to the field of artificial intelligence technology, specifically computer vision, augmented reality, virtual reality, deep learning and other technical fields, and can be applied to scenes such as metaverse and digital human. The method may include: obtaining human joint point positions corresponding to multiple joints of a target human body; obtaining human joint rotation data based on the human joint point positions, the human joint rotation data including the respective rotation angles of the multiple joints; determining the orientation of at least one key joint among the multiple joints based on the human joint rotation data; obtaining updated human joint rotation data by updating the rotation angle of at least one other joint among the multiple joints based on the orientation of the at least one key joint; and driving the virtual image of the target human body based on the updated human joint rotation data.

The invention relates to a sit-up counting detection method and system, and belongs to the technical field of computer vision, and the method comprises the steps: obtaining the position of a terminal device, dividing target persons into three rows based on the position of the terminal device, setting a preset number of point locations from each row close to the terminal device, and determining the head orientation of each row of target persons; under the condition that counting detection is started, first key point coordinates of each frame of target personnel in a counting time range are obtained; calculating the position score of each frame of target personnel based on the first key point coordinates, and calculating the number of sit-ups of the first row of target personnel according to the position score of each frame of target personnel, the position change score, the counting score and the action standard score in each getting-up process; and calculating the number of sit-ups of the second row of target persons based on the second key point coordinates and calculating the number of sit-ups of the third row of target persons based on the third key point coordinates. High-precision and high-robustness automatic counting of sit-ups in a complex real scene is realized.

The application provides a full-scene group activity understanding method based on a mixed hierarchical representation, and mainly relates to a group activity understanding problem in the field of computer vision. The method comprises the following steps: constructing a data set and a neural network framework, training a network, testing a network and the like. The neural network framework firstly extracts time sequence features, spatial environment features and the like to form rich individual features, and classifies individual actions; then, taking the individual features as nodes, information between the nodes in the individual interaction level and the action interaction level is inferred by using a proposed graph reasoning strategy, and is fused into group features; meanwhile, a topological graph structure is constructed by using position coordinates and orientations, and a graph neural network is introduced for representation, and is fused into group structure hierarchical features; finally, the group features and the group structure hierarchical features are spliced to identify a group activity type. The application fully represents a multi-level group activity, and simultaneously identifies individual actions of group members and a whole activity type.

The invention relates to a system and method for estimating room coordinate axes based on computer vision scanning of a room and its content objects and a coordinate axis majority voting process. The invention provides a method comprising: capturing a plurality of images of an indoor environment having one or more walls and a plurality of objects, each wall and object having a surface oriented along a corresponding plane; local coordinate axis orientations of at least a portion of the wall and the object are determined based on information received from the AR engine. The method includes estimating a room coordinate axis of an indoor environment based on a voting process, wherein the room coordinate axis is determined using a majority-matched local coordinate axis orientation; and outputting an indication of the room coordinate axis.

The present invention discloses a method for reconstructing the internal morphology of an interface. First, the interface morphology and the interface friction force experimental results are obtained through a rough interface friction experiment; then, the direction angle of the surface morphology is divided into different characteristic directions, and the interface morphology is analyzed by using a deep learning technology based on computer vision to obtain the distribution ratio of the interface morphology corresponding to the characteristic direction, and the morphology orientation ratio recognition result is obtained based on this, and then the "friction force-orientation ratio" scaling rate relationship is jointly constructed with the interface friction force experimental result, and the morphology orientation ratio of the target sample is calculated by combining the relationship between the spring mass block model and the friction force; finally, the reconstructed morphology is obtained according to the morphology orientation ratio and the interface internal morphology reconstruction formula based on the characteristic direction. The present invention can realize the morphology reconstruction and prediction at any time during the friction process of the target sample, and the power spectrum density of the reconstructed interface morphology and the actual morphology are well matched.

The invention relates to the technical field of computer vision and intelligent information processing, in particular to a machining drawing information extraction method based on YOLO + OCR, which comprises the following steps: sequentially carrying out graying processing, Gaussian filtering noise reduction and Canny edge detection on a machining drawing, and extracting and screening drawing frames and drawing areas; detecting a text group in the drawing by adopting a YOLO-OBB algorithm; and performing discrete classification prediction on the rotation angle of the detected text group image. According to the method, text pixel loss caused by direct zooming is avoided through drawing preprocessing, the 360-degree text direction problem is solved by utilizing a YOLO-OBB algorithm and a four-classification neural network model, PMI information is aligned and recognized in combination with a special symbol model, global parameters are extracted through natural language processing, and finally errors are monitored and corrected. The problems of text pixel loss, insufficient direction processing, poor symbol and text association and the like in the prior art are solved.

The invention belongs to the technical field of optical measurement and computer vision, and particularly relates to a photometric three-dimensional deformation monitoring method oriented to a natural light source condition. The method comprises the steps that targets are designed, and a target array is arranged according to the imaging range of a camera and the morphology characteristics of an observation object; measuring three-dimensional coordinates of feature points of a target pattern, and calculating a normal vector of a target surface and internal and external orientation elements of an image; calculating multi-period light source directions and light source intensities based on the target three-dimensional coordinates, the normal and the internal and external orientation elements of the image; calculating a surface normal vector by using a luminosity stereoscopic model based on the multi-period light source direction and intensity; and realizing object surface continuous reconstruction based on the surface normal vector and the imaging geometric constraint. Compared with the traditional method, the method can break through the problem that the traditional photometric stereo depends on an artificial light source and is difficult to adapt to the dynamic change of a natural light source so that the method is not applicable, can realize deformation monitoring in an outdoor environment without the artificial light source, and has the advantages of low cost and wide coverage range.

The invention provides a starting point marking correction method based on panoramic image and scene graph matching, and relates to the technical field of computer vision and building information processing, and the method comprises the steps: respectively generating a triangle set, a scene graph triangle set and an actual scene triangle set which are combined in the same direction according to an actual scene angular point set and a scene graph feature point set; by calculating the similarity of the inner angles and the side length proportions of the triangles, matching the triangle pair with the highest similarity in the two sets, and establishing a local obstacle corresponding relation between the scene graph and the actual scene; and mapping the physical starting point position of the actual scene to the scene graph coordinate system according to the affine transformation parameter determined by the triangle pair, and outputting the corrected starting point coordinate. According to the method, accurate and efficient recovery of the current layer layout of the construction site can be realized.

The invention discloses a tray identification method and device based on structural feature point cloud, equipment and a storage medium, and relates to the technical field of computer vision and detection. Acquiring tray RGB information and tray depth information through a binocular camera; generating a point cloud of the tray based on the depth information, aligning the RGB information with the depth information, and extracting an ROI region of the tray in the point cloud; converting the point cloud data into a two-dimensional plane in a camera coordinate system through the projection matrix; the point cloud data in the two-dimensional plane is rotationally scanned, and the main beam orientation of the tray is calculated; cutting the ROI region in the orientation range of the main beam to obtain point cloud data of the measured surface; projecting the point cloud data of the measured surface to a plane orthogonal to the rotating shaft, and calculating the spatial positioning of the tray through a straight line fitting algorithm; the scheme is high in identification precision, accurate in size measurement, high in processing efficiency, adaptive to a high-beat production line and high in automation degree; the method has strong environmental adaptability and anti-interference capability, and is strong in adaptability and good in robustness.

The invention relates to an azimuth information measurement and transmission system based on computer vision, in particular to an azimuth reference calibration device and method for computer vision measurement equipment, and aims to solve the problem that a method and a device for calibrating the azimuth reference of the vision measurement equipment in the system are lacked at present. And the measurement result of the vision measurement equipment cannot be directly applied to the azimuth transmission system. The calibration device comprises an orientation physical datum piece, a calibration table, a reference datum piece and angle measurement equipment, the orientation physical datum piece is arranged outside the vision measurement equipment to represent the orientation datum of the vision measurement equipment, and the calibration table is used for being erected on the side, along the optical axis of the vision measurement equipment, of the vision measurement equipment to represent the orientation datum of the vision measurement equipment. Comprising a base as well as a collimation measurement cooperative target and an imaging measurement cooperative target which are arranged on the base, and the reference piece is used for measuring the azimuth included angle of the azimuth physical datum piece relative to the collimation measurement cooperative target in cooperation with the angle measurement equipment.

This invention belongs to the fields of cultural relic protection and computer vision technology. Addressing the problems of low efficiency, poor accuracy, and lack of spatial positioning in traditional methods for detecting defects in ancient city walls, this invention discloses an intelligent detection method and system for surface defects of ancient city walls based on 3D reconstruction. The method includes: processing the texture information of a loaded 3D reconstructed model of the ancient city wall using a multi-strategy texture restoration method; analyzing the model's geometric structure to determine the wall's orientation and dimensional parameters; generating virtual camera positions based on geometric features and a meshed serpentine scanning strategy; rendering RGB and depth images at each virtual camera position; performing defect detection using a YOLO model; mapping the 2D detection results to 3D space to generate 3D defect marker color blocks; and fusing multi-viewpoint results to output the detection results of surface defects on the ancient city wall. This invention achieves efficient, accurate, and automated detection of defects in ancient city walls, providing effective technical support for cultural relic protection.

Apparatus and methods relate to dynamic hazard exposure dosimetry (HED). In an illustrative example, a HED computation engine (HEDGE) may automatically determine a relative three- dimensional (3D) position of a person to a hazard source by analyzing an image feed, such as by a computer vision engine. The 3D position of the person may, for example, be determined by evaluating the size and orientation in the image feed of one or more visual indicia worn by the person and / or disposed on the hazard source. A dosage prediction engine may, for example, generate a hazard dosage based on the relative position over time, such as by using predetermined hazard dosage distributions correlated with the 3D position of the hazard source. Various embodiments may advantageously provide a reduced complexity, easily installed and / or maintained real-time monitoring of hazard exposure (e.g., radiation exposure) in controlled environments (e.g., medical operating rooms).

Systems and methods are provided for estimating an axis of a room based on a computer vision scan of the room and its contents, and a process of axis majority voting. A method is provided that includes capturing a plurality of images of the interior environment having one or more walls and a plurality of objects, each wall and object having a surface oriented along a corresponding plane; determining, based on information received from an AR engine, local axis orientations for at least a portion of the walls and objects. The method includes estimating a room axis of the interior environment based on a voting process, wherein a majority of matching local axis orientations are utilized to determine the room axis; and outputting an indication of the room axis.

The invention discloses a point pricking control method for a PTZ camera image set, a medium and electronic equipment, and the method comprises the steps: extracting and describing feature points from the image set, and carrying out the matching of the extracted feature points; based on a matching result, estimating motion between the images by using an essential matrix, and performing rotation averaging on all the images to obtain an image relative orientation result; performing bundle adjustment to optimize a relative orientation result; querying a relative orientation result, and obtaining an internal parameter matrix and an external parameter between the images; calculating a conversion relation of corresponding points between the images by using the internal reference matrix and the external reference between the images; and according to the conversion relation, calculating and marking the predicted position of the control point on the non-pricking point image. Through the computer vision technology and the SFM technology, automatic control point pricking of the PTZ camera image set is achieved, the working efficiency and precision are improved, and the method is suitable for various monitoring and measurement application scenes.

Workstation for performing a task on a workpiece using a tool. Some implementations of the invention allow an adjustable shelf to be used to position a workpiece on the workstation. Some implementations of the invention include a support bar for partially supporting a tool as the task is performed on a workpiece coupled to the workstation. Some implementations of the invention allow a workpiece to be positioned at a reference location or orientation relative to the workstation. Some implementations of the invention allow a tool to perform an action based on determining the position of the tool relative to the workstation using computer vision techniques to analyze a feature of the workstation.

The invention relates to the technical field of computer vision, and discloses a book checking method and system based on computer vision. The method comprises the following steps: acquiring multi-source perception data of the book shelf through three-dimensional scanning equipment, and processing the multi-source perception data to generate standardized point cloud data; the multi-modal fusion network is utilized to jointly extract book surface features and identifier information, and an identification result containing a unique identifier is output; the book orientation angle and the coordinate offset are solved by combining a bookshelf topology model, and accurate pose data are obtained; based on the pose data, a robot motion track is generated by adopting a probability route map algorithm, and a robust inventory path is obtained through online correction; and tracking book loss or misplacement events in real time according to the path, and associating a time sequence log to generate an incremental inventory report. According to the invention, high-precision book identification and state diagnosis are realized, and the inventory automation level is improved.

The invention discloses a main direction rotating target detection method for deformable quadrilateral attention, and belongs to the technical field of computer vision. The method comprises the following steps: 1) loading a rotating target data set, preprocessing an image, and converting the image into a model training format; 2) through a detection head of the deformable quadrilateral attention module, mapping the fixed window into a quadrilateral adaptive to the geometrical shape of the target through a dynamic prediction transformation matrix; 3) a principal component analysis module is integrated on the detection head, and a principal direction and a rotation bounding box are calculated based on vertex coordinates of the target quadrangle; and 4) inputting the test set into the trained model, and outputting an angle-containing bounding box, category and confidence of the rotating target. According to the method, the mAPs on DOTA-v1.0 and HRSC2016 data sets are 79.38% and 98.90% respectively, compared with an existing method, the precision of the method is improved by 3.2%-5.6%, particularly, the detection robustness of irregular and multi-attitude rotating targets (such as ships and airplanes) is remarkably enhanced, and the method is suitable for remote sensing image and aerial photographing target detection scenes.