959results about "Three-dimensional object recognition" patented technology

This disclosure is directed to an item-identifying, mobile cart that may be utilized by a user in a materials handling facility to automatically identify a user operating the cart and items that the user places into a basket of the cart. In addition, the cart may update a virtual shopping cart of the identified user to include items taken by the user. The mobile cart may include multiple imaging devices and oriented such that their respective optical axes are directed towards an interior of a perimeter of the top of the basket, and above the top of the basket. The mobile cart may also include an imaging device oriented away from the basket such that a user operating the mobile cart may scan a user identifier using this imaging device to enable recognition of the user.

The application discloses a kind of optical three-dimensional chain measurement system precision prediction method based on AMC theory, belong to optical three-dimensional measurement equipment application field.The measurement error of front-stage measurement equipment is first condensed into position error and attitude error of tool coordinate system by the observation coordinate value of a group of spatial distribution measurement point set;The obtained data is mapped into the coordinate system of lower-stage measurement equipment through the space pose transformation process with the rear-stage measurement equipment synchronization, and the size and nature of the error do not change in the mapping stage, only the value and direction change;Data and the measurement error of the rear-stage measurement device itself are associated, and finally show the composite error value of the coordinate point in the measurement space of the rear-stage measurement device, and the error ellipsoid distribution is geometrically represented, so as to complete the whole error transmission process between adjacent measurement devices.Can provide accurate evaluation and prediction tool for the forward design, performance optimization and reliable application of optical chain measurement system, and promote the chain measurement system to play a greater value in high-end manufacturing and precision measurement field.

The application discloses a three-dimensional target detection method and device fusing coordinate attention and multi-scale features, and relates to the technical field of three-dimensional target detection.The method comprises the following steps: acquiring original three-dimensional point cloud data, performing feature extraction, and generating a two-dimensional bird's eye view feature; performing one-dimensional feature aggregation of the two-dimensional bird's eye view feature in a quadrature direction, extracting direction perception features, determining spatial attention weights, reweighting the two-dimensional bird's eye view feature, and generating enhanced multi-scale features; performing weighted aggregation on the enhanced multi-scale features, and generating a shared feature map; performing attribute prediction processing based on the shared feature map, obtaining a center point heat map and three-dimensional bounding box attributes respectively, performing geometric collaborative optimization on the three-dimensional bounding box attributes, and outputting a detection result.The application decouples global pooling into one-dimensional pooling in horizontal and vertical directions, retains the coordinate prior of each spatial position, and generates bidirectional attention weights on this basis, so that the feature response of the position of a target can be adaptively strengthened.

The application discloses a driver distraction monitoring method, aiming at solving the problems of false alarm or missed alarm of the existing driver monitoring system caused by individual differences of drivers, environmental changes, and single sensor failure. The application obtains the face image collected by the vehicle-mounted DMS camera and the eye movement data collected by the AR glasses worn by the driver in parallel, respectively calculates the first state parameter and its confidence, the second state parameter and its confidence. Based on the confidence evaluation result, the fusion strategy is dynamically selected: when both channels have high confidence, the or logical judgment is adopted; when the judgment results conflict, the high confidence channel result is used as the criterion; when the single channel confidence is lower than the threshold, the channel judgment is suspended and the other channel is completely relied on. In addition, the change of the pupil diameter based on the AR glasses can distinguish visual distraction and cognitive distraction. The application realizes intelligent collaboration of multi-source heterogeneous sensors, and significantly improves the accuracy and robustness of distraction monitoring in complex scenes.

Embodiments of the present application provide a training data acquisition method and a model training method and device, which can be related to the fields of artificial intelligence, three-dimensional generation, etc. The method comprises: acquiring a plurality of first images; for each first image, generating a multi-view corresponding to the first image by a multi-view generation model; inputting the multi-view into a quality evaluation model to evaluate the quality of an object in the multi-view and obtain a quality evaluation result of the multi-view; and if it is determined according to the quality evaluation result of the multi-view that the quality of the object in the multi-view is qualified, regarding the multi-view and the first image corresponding to the multi-view as a set of training data of a three-dimensional generation model. The method generates multi-views corresponding to a plurality of first images, and evaluates the quality of the generated multi-views, thereby screening out high-quality multi-view data for training a three-dimensional generation model, enriching the training data, and further enabling a model with better performance to be obtained based on the training data.

The disclosure provides a point cloud data processing method and device, an intelligent chip, a storage medium and a vehicle, relates to the field of artificial intelligence, and in particular to the technical field of intelligent transportation, computer vision and automatic driving. The point cloud data processing method is executed by the intelligent chip, and the specific implementation scheme is: according to the width value and the height value in the point cloud data, the point cloud data is grid processed to obtain a target grid unit corresponding to the point cloud data in a predetermined grid; according to the depth value and the reflection intensity value of the point cloud data, the point cloud statistical data corresponding to the target grid unit is updated; and in response to completing the update of the point cloud statistical data corresponding to the grid unit in the predetermined grid according to a frame of point cloud data, a target task of computer vision is executed according to the point cloud statistical data corresponding to the grid unit in the predetermined grid.

Some embodiments of an example method disclosed herein may include receiving point cloud data representing one or more three-dimensional objects; receiving a viewpoint of the point cloud data; selecting a selected object from the one or more three-dimensional objects using the viewpoint; retrieving a neural network model for the selected object; generating a level of detail data for the selected object using the neural network model; and replacing, within the point cloud data, points corresponding to the selected object with the level of detail data.

The present invention relates to a method of verifying that (or determining whether) a biometric feature of a live human is present. More particularly, the present invention relates to a method of verifying that (or determining whether) a biometric feature of a live human is present for use as part of a biometric recognition system and / or method. In particular, the invention relates to a method (100) of determining whether a biometric feature of a live human is present, comprising: using a camera (106), capturing (202) visual data of a presented biometric feature; transmitting (204) a signal towards the biometric feature; using a sensor (104), capturing (206) data related to a reflected signal from the presented biometric feature; and determining (208) whether the visual data and the reflected signal data relate to a biometric feature having realistic dimensions thereby to determine (210) that a live biometric feature is present.

The application discloses a complex reverberation suppression method based on Bayesian matrix decomposition, and relates to the technical field of sonar detection. The method comprises the following steps: vectorizing each three-dimensional sonar image in an original three-dimensional sonar image sequence in time sequence, and then splicing the three-dimensional sonar image sequence into a two-dimensional data matrix; using a Bayesian matrix decomposition method to extract an indicator matrix of low-rank components and non-low-rank components from the two-dimensional data matrix; calculating each non-low-rank component through the low-rank components and the indicator matrix, and selecting a non-low-rank component with the maximum contrast as a sparse target component; reversely vectorizing the two-dimensional sparse target component into a three-dimensional sparse target image sequence; using nonlinear superposition on the three-dimensional sparse target image sequence to suppress fluctuation reverberation and enhance the target, and obtaining a moving target trajectory. Compared with other reverberation suppression algorithms, the application avoids the selection of an optimization algorithm regularization parameter, and can stably implement complex reverberation suppression to detect a moving small target.

The invention relates to a calibration device and a calibration method for a mining vehicle. According to an example aspect, there is provided a method comprising: performing a calibration action of a calibration procedure for a mining vehicle, receiving scan data of the calibration action based on an environmental scan of the calibration action performed by a scanner, wherein the scan data is indicative of at least one position of a work implement part of a work implement relative to a body part of a body, receiving calibration verification reference data indicative of at least one of: a calibration target position of the work implement part relative to the body part for the calibration action, or a target movement range of the work implement part relative to the body part for the calibration action, processing the scan data and the calibration verification reference data to determine at least one deviation of: a current position of the work implement part from the calibration target position, or a current movement range of the work implement part from the target movement range; and verifying the calibration procedure based on the determined at least one deviation.

The present application relates to a kind of road traffic sign based on vehicle-mounted laser point cloud fast extraction and modeling method, comprising the following steps: S1, the extracted vehicle-mounted laser point cloud is segmented into block according to road geometric shape using point cloud segmentation mode and discard the point above track in block, after translation rotation, projection to plane generates bird's eye view reference graph;S2, construct road sign directed target detection data set under bird's eye view;S3, joint national road sign shape specification requirement constructs road traffic sign template library;S4, train improved directed target detection model;S5, it is projected to point cloud coordinate system, obtains point cloud coordinate system;S6, road traffic sign is modeled to space 6D vector;S7, complete road traffic sign in high-precision map modeling work.The present application can quickly complete projection conversion and extract road traffic sign to complete high-precision map modeling task, construct road sign directed target detection data set, with good precision and efficiency.

The present application relates to the technical field of breast modeling, and particularly relates to a breast modeling method. A plurality of target optical images of different perspectives corresponding to a breast to be modeled and target image data and target ultrasound data corresponding to the breast to be modeled are acquired; surface geometry prior data, internal structure prior data and dynamic mechanics prior data corresponding to the breast to be modeled are constructed based on the target optical images, the target image data and the target ultrasound data; an enhanced feature vector is generated based on the surface geometry prior data, the internal structure prior data and the dynamic mechanics prior data; and a target breast model corresponding to the breast to be modeled is generated based on the enhanced feature vector. The accuracy requirement of surgical planning and prosthesis design is met, and soft tissue deformation characteristics such as breast droop and asymmetry are accurately restored; dependence on special high-precision scanning equipment is eliminated, and high-quality reconstruction can be realized based on conventional multi-source clinical data, which is convenient for popularization and application in a conventional clinical environment.

A system for modeling a three-dimensional structure utilizing two-dimensional segments comprising a memory and a processor in communication with the memory. The processor extracts a plurality of two-dimensional segments corresponding to the three-dimensional structure from a plurality of images indicative of different views of the three-dimensional structure. The processor determines a plurality of three-dimensional candidate segments based on the extracted plurality of two-dimensional segments and adds the plurality of three-dimensional candidate segments to a three-dimensional segment cloud. The processor transforms the three-dimensional segment cloud into a wireframe indicative of the three-dimensional structure by performing a wireframe extraction process on the three-dimensional segment cloud.

This invention discloses a method and system for feature extraction of multi-layer stacked cardboard boxes based on 3D vision. Addressing the problem that existing technologies cannot accurately quantify surface anomalies and multi-dimensional stacking features of cardboard boxes in complex stacking scenarios, this invention acquires the original point cloud and RGB images of the stacking scene; after preprocessing, a height distribution histogram is constructed and clustered to obtain initial point cloud clusters; warped anomalies are eliminated through curvature analysis and normal vector consistency to obtain accurate upper surface point clouds; principal component analysis is used to obtain the length, width, and normal vector direction of the cardboard box; basic geometric features, damage features, occlusion relationships, overhang / overlap features, tilt features, and multi-layer interlacing features are extracted by combining RGB images; finally, the feature vector of each cardboard box is output in a structured manner. This invention achieves refined point cloud processing, effectively eliminates warped points, systematically quantifies complex stacking features, and integrates multimodal data, significantly improving the ability to identify and extract features from cardboard boxes in complex logistics scenarios.

This invention provides a real-time 3D environmental reconstruction system based on semantic interaction. It integrates semantic prior information into the mesh optimization process, enabling clear differentiation between adjacent but semantically different objects, thus generating a mesh model that is more topologically accurate and geometrically precise. By employing an energy function that includes a geometric term, a semantic consistency term, and a semantically modulated smoothness term to optimize the vertex positions of the initial semantic mesh, it can actively reduce smoothness constraints at the boundaries of different semantic objects while maintaining surface smoothness, effectively preventing the blurring of object edges and generating object representations with clear outlines and sharp boundaries. Therefore, this invention transforms the traditional passive mapping process of "perception-processing-mapping" into an intention-driven active mapping process of "query-perception-processing-highlighting on the map," generating high-fidelity, textured 3D mesh models with semantic interaction capabilities in real time.

The application discloses a kind of based on multi-modal vision's livestock and poultry meat tissue recognition and intelligent segmentation method and system, belong to image data processing technical field.For solving the problem that artificial segmentation is low in standardization degree, it is difficult to accurately avoid bone and muscle fiber trend, the multispectral image and three-dimensional point cloud data of the same livestock and poultry fresh meat carcass are obtained, and multi-modal feature map is generated after space-time registration and feature fusion;Bone structure, fat distribution and muscle fiber trend vector field are identified from the feature map using a deep learning model;Based on the identification result, a multi-objective optimization function is constructed, and a three-dimensional segmentation path is obtained from the preset starting point to the end point, avoiding bone, following the muscle fiber trend and satisfying the optimal meat yield. The application is used for livestock and poultry meat intelligent segmentation, and can improve the meat yield and cutting standardization level.

The application discloses a model similarity judgment method and device, equipment and a storage medium, and relates to the technical field of computer vision, which comprises the following steps: determining each distribution direction vector according to three-dimensional point cloud data of a to-be-judged model; generating two-dimensional point cloud data according to each distribution direction vector and the three-dimensional point cloud data; generating a triangle set of the to-be-judged model according to each two-dimensional point cloud data; and if a triangle similar to each triangle in the triangle set of the to-be-judged model is searched in a triangle set of a target model, it is determined that the to-be-judged model is similar to the target model. In the foregoing manner, the three-dimensional model local coordinate system constructed with each distribution direction vector as the coordinate axis direction is used to perform dimension reduction processing on the three-dimensional point cloud data, so that the data calculation amount is reduced and the calculation efficiency is improved, and then the similarity between the models is judged through the similarity of the triangles, so that the calculation cost of judging the similarity of the models can be effectively reduced, and the efficiency of judging the similarity of the models is improved.

The application discloses a kind of full-process supervision systems based on visual offline identification, and the application relates to component production technical field, it is solved that existing technology mostly stays in single size comparison level, lack of the problem of full-process integrated supervision mechanism from model generation, standard matching, automatic splicing to assembly verification, the unified check feature of the present application is constructed by built-in midpoint, plane midpoint and feature vector, corresponding three-dimensional standard model can be matched from model database quickly and accurately, the identification logic is stable, and the anti-interference is strong, effectively avoid the model mismatching problem caused by similar appearance;Through the vector comparison of reference feature and component feature, length difference and included angle calculation, the assembly accuracy can be quantitatively judged, and the assembly standard, correctable and substandard state can be distinguished, the visualization and digitization of assembly quality are realized, the on-site quick error correction and traceability are facilitated, and the assembly reliability and overall production quality of flow line are effectively improved.

The present disclosure generally relates to performance recreation, and in particular, the recreation of observed human performance using reinforcement learning. In this regard, a first object is identified from a plurality of objects. The manipulation of the first object is tracked from a first position to a second position. A characterization of the manipulation is generated. A policy that controls a mechanical gripper to recreate the manipulation is generated based on an iteratively increasing cumulative award. The mechanical gripper iteratively recreates the manipulation to increase a cumulative award with each recreation.

The application belongs to the technical field of target detection, and particularly relates to a 3D target detection method and device and a medium. The method comprises the following steps: S1, obtaining 3D point cloud data of a to-be-detected region; and S2, inputting the 3D point cloud data into a pre-trained 3D target detection model to obtain class information and a bounding box of a detection target. The 3D target detection model comprises a 3D backbone network, a neck network and a head network. The 3D backbone network is a residual network with an InceptionNeXt module added after each standard 3D residual block. The InceptionNeXt module is used to average divide the output features of the standard 3D residual block into four components in the depth channel direction, and input the four components into four branches respectively. The first branch extracts local detailed features, the second branch extracts long-range dependence in the horizontal direction, the third branch extracts long-range dependence in the vertical direction, and the fourth branch is an identity mapping. The application solves the technical problem of slow operation speed of a large kernel convolution model in the prior art.

The application discloses a three-dimensional medical image segmentation method based on CircularUMamba, and belongs to the technical field of medical image analysis and deep learning. The method comprises the following steps: acquiring 3D medical image data as a training set, and performing pretreatment and preliminary feature extraction on the image data; constructing a backbone segmentation network based on CircularUMamba, processing the features output by S1 by using the network, and finally outputting a multi-level voxel-level segmentation prediction probability map; training the network constructed by S2 by using a combined loss function, optimizing the network parameters by minimizing the total training loss, and obtaining a trained image segmentation model; and using the trained model to obtain a voxel-level segmentation prediction probability map as a segmentation result for an image to be segmented. The application can improve the recognition accuracy of micro lesions and fuzzy tissue boundaries.

A campsite patrol route planning method and system based on a UAV, in the method, by analyzing the three-dimensional point cloud data around the fence of the campsite, first, the surface of the mesh fence is identified, and the low-texture background area is screened out by using the normal vector variance feature. For each monitoring point on the fence, search for an unobstructed area center in the low-texture background area as a background target anchor point, and establish a virtual line-of-sight vector from the background anchor point to the monitoring point. Based on this line-of-sight vector, the reverse optical axis is generated by extending outside the fence, and the optimal shooting flight point is determined in combination with the preset imaging depth. The application is used to improve the accuracy of automatic defect recognition based on a UAV.

This application relates to the technical field of 3D display and discloses a lightweight 3D display method based on multi-source data fusion, comprising: acquiring spatial observation data and target display parameters from different data sources; dividing a 3D spatial region into different spatial blocks based on the target display parameters, and identifying potential units of interest in the spatial blocks; calculating the spatial structure contour provided by each type of spatial observation data and the stability contribution of each type of spatial observation data in each potential unit of interest; selecting a fusion data source for each potential unit of interest based on the spatial structure contour and stability contribution; and performing lightweight 3D display on each potential unit of interest based on the corresponding fusion data source. This application can reduce the computational load and improve the real-time performance of 3D display while ensuring the stability of 3D structure representation.

The application relates to the field of image segmentation technology in computer vision, in particular to a zero-shot image segmentation method and system based on a three-dimensional model of an object. The method comprises the following steps: for any untrained object, a three-dimensional model of the object is rendered into a two-dimensional RGB reference image under a plurality of preset discrete viewing angles by using a graphics rendering engine, the reference image is input into a visual basic model DINOv3, a global feature vector representing semantic information of the object is extracted, and a reference feature library is constructed; a target image to be segmented is input into a visual basic model SAM2, and a plurality of candidate object masks are generated; for each candidate mask, an object image corresponding to the candidate mask is cropped from the target image, and the object image is input into the DINOv3 to extract a semantic feature vector of the candidate object; by calculating the cosine similarity between the candidate object feature vector and each template feature vector in the reference feature library, the top similar degrees with the highest values are selected, and an arithmetic mean value of the similar degrees is calculated, and the mean value is taken as the classification confidence of the candidate object mask; the class label of each candidate object mask is determined according to the confidence, and the class label of the target object and a corresponding pixel-level segmentation mask are output.

This application provides an obstacle detection method, apparatus, and system, and a computing device, and relates to the image processing field. The method includes: obtaining an image; then constructing a depth map of the image based on structure constraint information; and after completing depth map construction, processing the depth map to obtain a region identifier map including a plurality of regions, where any one of the plurality of regions is a traveling region or a non-traveling region. The structure constraint information includes a semantic type of each sample in the image, and the non-traveling region is considered as an obstacle. In this way, depth map construction is guided based on the semantic types of the samples, so that depth distribution of the samples in the depth map better complies with a depth distribution rule corresponding to the semantic types to which the samples belong. This improves accuracy of the depth map, and helps accurately recognize the traveling region like the ground and the non-traveling region like the obstacle, thereby improving obstacle recognition accuracy.