16 results about "Landmark point" patented technology

This application belongs to the field of computer vision and optical measurement technology, specifically disclosing a point cloud stitching method and apparatus based on monocular pose prior marker matching. The method includes: calculating the pose of multiple markers extracted from left and right camera images, determining the normal vector and three-dimensional coordinates of each marker's circular cross-section in its respective camera coordinate system; transforming the three-dimensional coordinates and corresponding normal vectors of each marker in either the left or right camera image to the other camera coordinate system for marker point registration between the left and right camera images; the circular cross-section is a circular cross-section in three-dimensional space with the origin of the camera coordinate system as its vertex, passing through the elliptical curve where the marker is located; and stitching the three-dimensional point cloud of markers based on the marker point registration results of the left and right camera images. This application can significantly improve the accuracy of binocular stereo marker matching and the system's anti-interference capability in complex industrial scenarios.

Systems and methods of landmark detection using deep neural networks with multi-frequency self-attention are provided. The system includes an encoder network that receives an image of an object of interest as an input and generates a multi-frequency feature map as an output. The system also includes an attention layer that receives the generated multi-frequency feature map and refines the generated multi-frequency feature map based on a correlation or association between the received multi-frequency feature maps. The system further includes a decoder network that receives the refined multi-frequency feature map from the attention layer as a second input and generates a landmark detection result based on the second input. The landmark detection result includes a heat map image of the object of interest, and the heat map image indicates locations of landmark points on the object of interest in the image.

The invention relates to an SLAM back-end optimization method based on sparse matrix factorization and a hardware accelerator architecture, and the method comprises the steps: constructing a Hessian matrix through employing a binary sparse coding mechanism according to the actual observation coordinates of feature points on an image plane and the predicted coordinates of road sign points projected to a camera coordinate system; performing Schel elimination processing on the Hessian matrix based on binary sparse coding to obtain a linear equation set only containing a camera pose variable; and carrying out extended Kalman filtering on a linear equation set only containing a camera pose variable to realize incremental updating of a state pose and a road sign point coordinate. According to the method, the overall delay of SLAM rear-end optimization can be remarkably reduced, and the real-time requirement in a high-speed moving scene is met.

The method includes accessing a plurality of reference landmark points associated with a plurality of features in a reference image, accessing a reference measurement point corresponding to an asset in the reference image, receiving a target image including the features and the asset, identifying a plurality of target landmark points of the target image corresponding to the features, generating a mapping from the reference landmark points to the target landmark points, determining, using the mapping, a target measurement point corresponding to the asset in the target image, and extracting, from the target measurement point of the target image, a temperature measurement of the asset.

A system and method for hierarchical spectral landmark graphs in cognitive manifolds implements cognition through discrete landmark structures that provide scalability, interpretability, and auditability. The system maintains a landmark graph on a cognitive manifold with vertices representing landmark points selected based on geometric properties including curvature and cognitive trajectory density. A spectral basis derived from the landmark graph encodes long-term semantic structure. Spectral continuation updates the basis when geometric invariants indicate structural change, while enforcing differential plasticity constraints protecting foundational low-frequency modes. Reversible edges constructed with forward and reverse displacement vectors enable auditable trajectory replay through cryptographic certificates and manifold journals. The system generates probability estimates by fusing geometric priors from landmark paths, empirical evidence from simulations, and historical evidence from archived cases. Landmark-conditioned naturalization produces interpretable explanations mapping geometric structures to domain-specific semantic labels, enabling transparent, auditable reasoning grounded in verifiable landmark-based evidence.

The application provides a real-time acupoint tracking method and system and a guasha robot, and relates to the technical field of computer vision recognition. The tracking method comprises a static positioning stage, a two-stage training strategy for fusing traditional Chinese medicine meridian prior knowledge is used to improve the YOLO-Pose model, bone landmark point detection accuracy is improved through meridian perception approximate training and adaptive fine adjustment, and a standardized coordinate system is established based on the bone degree and inch method to realize accurate acupoint mapping; in the dynamic tracking stage, a two-stage optimization complementary filter is used to smooth the acupoint coordinates in the video sequence, and high-frequency jitter is inhibited through a speed smoothing mechanism and historical trajectory weighted fusion. The application solves the problems of traditional guasha relying on experience and large dynamic tracking jitter, improves image detection accuracy, optimizes the balance relationship between smoothness, response speed and accuracy of the dynamic trajectory generated by the image, and realizes high-precision and high-real-time acupoint positioning and tracking.

The application provides a large rotary equipment blade measurement system calibration method based on a topological structure landmark point coding, and belongs to the field of blade appearance three-dimensional measurement. Firstly, the three-dimensional landmark points are projected onto a two-dimensional plane by using a principal component analysis method. Then, a local geometric descriptor is generated for each point on the projection plane. The cross-frame matching and trajectory tracking of the landmark points are realized through the descriptor matching. Finally, the center point and normal vector of all trajectories are extracted, the average value of the center point coordinates and the average value of the normal vector are calculated, and the rotary shaft axis linear solution is taken as the rotary shaft pose calibration. The application takes a disc with circular landmark points as a calibration module, designs a position tracking descriptor based on the topological relationship between the landmark points in the disc plane, realizes the landmark point position tracking, has important significance for the rotary shaft pose calibration, and provides a strong technical foundation for the subsequent blade splicing measurement based on the axis calibration.

The application provides a face recognition and permission control method and system based on multi-algorithm fusion, and belongs to the technical field of biological feature recognition. The application obtains an access path sequence of a target object and a spatial coordinate sequence of facial landmark points, encodes the curvature of the coordinate neighborhood to obtain a morphology vector, and maps the transition probability of the region in the path sequence to obtain a behavior vector. The maximum cross-correlation coefficient of the two is calculated to obtain a projection matrix pair, the morphology vector and the behavior vector are mapped to a unified sub-space superposition to obtain a correlation vector and add the vector to a vector set. A directed graph is constructed with the vectors in the set as nodes and the kernel function similarity as weights, and a manifold matrix is obtained through graph Laplacian difference and characteristic decomposition. The low-dimensional components corresponding to the correlation vector are mapped to the feature space of the preset permission level clustering cluster to determine the belonging permission level cluster and the centricity, and the identity identification and the access permission instruction are output. The application realizes accurate identity recognition and access permission control of the target object.

According to one aspect of the disclosure, a computer-implemented method of dynamic head generation is provided. The method can include computing a linear blend skinning (LBS) binding system associated with a virtual avatar head based on a set of mesh deformations for a plurality of poses of the virtual avatar head. The LBS binding system can include a set of skin joints and a set of joint weights. Each joint weight of the set of joint weights can be assigned to a vertex of a mesh of the virtual avatar head in a neutral pose. The method can include computing an adjusted head control cage of the virtual avatar head based on a correspondence between the mesh of the virtual avatar head in the neutral pose and a template head control cage using a landmark point prediction model or a regression model. The method can include animating the virtual avatar head based on the LBS binding system and the adjusted head control cage.

A system and method for adaptive geometric diffusion projection enables mapping of heterogeneous high-dimensional representations onto a shared low-dimensional manifold without neural network training. The system maintains landmark points in source spaces and computes their spectral coordinates through graph Laplacian eigen decomposition based on semantic similarities. New input points are projected via harmonic extension, computing weighted interpolations of nearby landmark spectral coordinates. A geometric optimization process refines positions while maintaining manifold constraints through tangent space projections. The system continuously monitors geometric invariants including principal angles, spectral gaps, and curvature distributions. When invariants exceed thresholds, targeted adaptations occur: spectral basis updates using warm-started iterations, landmark set augmentation in high-residual regions, or parameter adjustments. The approach supports logarithmic computational scaling, enables streaming operation on continuous data, and handles multimodal inputs through reliability-weighted consensus. The system maintains projection quality indefinitely through continuous geometric monitoring and local adaptations.

The application relates to a detection site tracking method and system, a terminal and a storage medium, and relates to the technical field of medical images. The method comprises the following steps: controlling a preset image acquisition component to acquire a first detection torso image of a preset patient; analyzing the first detection torso image to generate a bone landmark point position; acquiring an abnormal detection position of the patient based on the bone landmark point position; acquiring a reexamination torso image; analyzing the abnormal detection position of the patient, the reexamination torso image and the bone landmark point position to generate a reexamination position; and controlling a preset detection starting point prompt component to prompt a doctor to reexamine the patient according to the reexamination position. The application has the effect of improving the comparability of multiple detection results.

The present application relates to the technical field of image processing, in particular to a machine vision classification method for defect detection of an inductor automatic production line, comprising: acquiring an original gray image and a surface depth distribution map of an inductor to be tested on the automatic production line by using an image acquisition device; then extracting global histogram statistical features and local texture features of the gray image in parallel through a convolutional neural network, and performing normalization processing on the depth distribution map to obtain a depth topographic feature vector; then introducing a self-attention weighting module, taking the local texture features as a query vector, and generating a comprehensive feature map by fusing the depth topography and global statistical features; then inputting the map into a region proposal network, and screening out an inductor main body region of interest and a suspected defect region in combination with a preset aspect ratio anchor point; finally extracting semantic features by using a classification recognizer, and analyzing the spatial distribution of landmark points in combination with a polarity discrimination model to determine an inductor defect class label and a polarity direction angle.

A computer-implemented method acquires X-ray images of human body parts and determines an appropriate orientation of the imaged human body parts for diagnostic purposes; the received X-ray images are rotated according to the determined appropriate orientation. Landmark points of the imaged human body parts can be identified to determine the angular orientation of the initial unrotated X-ray image.

The application provides a landmark point display method and device, electronic equipment and computer readable storage medium. The method comprises: acquiring geometric shape information and color information of a landmark point; the color information indicates the color of different display parts in the landmark point; the geometric shape information of the landmark point is processed by using a vertex shader to obtain a first primitive; and the first primitive is processed by using a geometry shader to obtain a second primitive; for a plurality of fragments obtained by rasterizing the second primitive, a fragment shader determines the display part to which each fragment belongs according to the geometric shape information, and determines the color of each fragment based on the display part to which each fragment belongs and the color information; and the landmark point is displayed according to the color of each fragment in the plurality of fragments. The edge of the displayed landmark point is clear and smooth.

The application discloses a kind of dental arch arrangement irregularity quantitative analysis method, it is related to oral digital analysis and orthodontic diagnosis technical field.The application is first based on the extracted occlusion key anatomical landmark point and carries out overall dental arch curve fitting to dental arch;Again, the same group of landmark points is carried out local dental arch curve fitting;By calculating the difference of overall dental arch curve and local dental arch curve on the corresponding position, and using the normalized root mean square error to the difference is quantified, to obtain the numerical index reflecting the degree of dental arch irregularity.To improve the comparability between different individuals, after the error value is normalized, it is divided into normal, mild, moderate and severe irregular grade according to the preset threshold interval.The application simultaneously reflects the overall trend of dental arch and local arrangement deviation;Without manual intervention, it is suitable for large-scale clinical analysis;And high sensitivity to tooth crowding, rotation and local arrangement disorder, can be used for pre-treatment evaluation, efficacy monitoring and dentition data analysis.

This invention relates to a visual observation construction method based on orthogonal ray projection in tangent space, belonging to the field of computer vision and robot state estimation technology. The invention proposes an orthogonal ray projection error model defined within the tangent space of the observation ray. Specifically, the method involves: transforming two-dimensional observation points in the image into observation rays on a unit sphere, constructing an orthogonal basis in the tangent space of these rays; projecting the three-dimensional landmark points to be estimated onto this tangent space basis, and calculating the orthogonal distance between these landmark points and the observation rays as the residual. This method geometrically eliminates the explicit dependence of residual calculation on scalar depth, making the error function linear with respect to feature point position parameters. The visual measurement device includes a feature extraction module, a tangent space construction module, a residual and derivative calculation module, and a state estimation module. This invention provides an accurate Jacobian matrix, significantly improving the numerical stability, convergence speed, and estimation accuracy of visual SLAM and 3D reconstruction systems in scenarios with distant feature points and weak parallax.