16 results about "Point set registration" patented technology

The present application relates generally to identifying regions of interest in images, including but not limited to whole slide image region of interest identification, prioritization, de-duplication, and normalization via interpretable rules, nuclear region counting, point set registration, and histogram specification color normalization. This disclosure describes systems and methods for analyzing and extracting regions of interest from images, for example biomedical images depicting a tissue sample from biopsy or ectomy. Techniques directed to quality control estimation, granular classification, and coarse classification of regions of biomedical images are described herein. Using the described techniques, patches of images corresponding to regions of interest can be extracted and analyzed individually or in parallel to determine pixels correspond to features of interest and pixels that do not. Patches that do not include features of interest, or include disqualifying features, can be disqualified from further analysis. Relevant patches can analyzed and stored with various feature parameters.

The present application relates generally to identifying regions of interest in images, including but not limited to whole slide image region of interest identification, prioritization, de-duplication, and normalization via interpretable rules, nuclear region counting, point set registration, and histogram specification color normalization. This disclosure describes systems and methods for analyzing and extracting regions of interest from images, for example biomedical images depicting a tissue sample from biopsy or ectomy. Techniques directed to quality control estimation, granular classification, and coarse classification of regions of biomedical images are described herein. Using the described techniques, patches of images corresponding to regions of interest can be extracted and analyzed individually or in parallel to determine pixels correspond to features of interest and pixels that do not. Patches that do not include features of interest, or include disqualifying features, can be disqualified from further analysis. Relevant patches can analyzed and stored with various feature parameters.

The application discloses a sub-pixel precision robust PIN needle position accuracy detection method and device and a storage medium. The method comprises the following steps: acquiring an image of a PIN needle of a product to be detected, wherein the image can present contour information of a needle body / needle tip; positioning a PIN needle region to obtain a region of interest of each PIN needle; performing sub-pixel segmentation on the PIN needle in the region of interest of each PIN needle to obtain a sub-pixel level precision needle body / needle tip segmentation result; obtaining needle body / needle tip center coordinates according to the result; if the number of obtained needle body / needle tip centers is not equal to the number of template centers, the product is determined as an unqualified product; if the number is the same, a coordinate system O is established according to the obtained needle body / needle tip center coordinates; point set registration is performed on the coordinate system O and a template center coordinate system, and whether the position accuracy meets the detection requirement is determined according to a point set registration result. Compared with an existing position accuracy detection method, the sub-pixel segmentation can guarantee high segmentation precision and reduce the hardware cost of a high-precision imaging scheme.

The present invention discloses a point set registration method based on continuously differentiable least trimmed squares estimation. The present invention proposes a continuously differentiable least trimmed squares criterion based on the least squares criterion and the least trimmed squares criterion. The cost function of this criterion is continuous and differentiable, so it can be more easily applied to gradient-based learning strategies, solving the defect that the least trimmed squares criterion is not differentiable. The present invention applies the continuously differentiable least trimmed squares criterion to the point set registration problem, which can better handle the point set registration situation in the presence of outliers, making the registration process as little affected by outliers as possible, so as to obtain a relatively accurate registration result. By using the continuously differentiable least trimmed squares criterion, the present invention can effectively avoid the problem that the robustness of the algorithm is sensitive due to improper selection of the trimming parameter. The present invention avoids the need for an additional sorting step in the traditional LTS criterion-based registration method, thus simplifying the algorithm flow and making the application more convenient.

The invention discloses a circuit board marking and positioning method and system for general material identification and point set registration, and the method comprises the steps: obtaining a teaching image of a teaching circuit board, calling a general material identification model to carry out the identification of the teaching image, and obtaining a teaching material point set; receiving a specified marking point in the teaching image, and recording a relative displacement relationship of the marking point relative to the associated material point; obtaining a to-be-marked image of a to-be-marked circuit board, and calling the general material recognition model to recognize the to-be-marked image to obtain a to-be-marked material point set; point set registration is carried out on the teaching material point set and the to-be-marked material point set, and transformation parameters from a teaching image coordinate system to a to-be-marked image coordinate system are obtained; mapping the associated material points to a to-be-marked image coordinate system based on the transformation parameters, and determining marking point pixel coordinates of the to-be-marked circuit board according to the relative displacement relationship; and converting the marking point pixel coordinates into marking point coordinates under a marking equipment coordinate system. The invention aims to improve the stability of marking positioning.

The invention discloses a building BIM modeling method based on three-dimensional point clouds, and the method comprises the following steps: S1, carrying out the collection of densely matched building point clouds and laser building point clouds, and carrying out the denoising and downsampling of the building point clouds; s2, registering the cross-source building point cloud data based on a feature point four-point set registration algorithm to obtain complete building point cloud data; s3, plane segmentation is carried out on the building point cloud, a building boundary is extracted, and components are extracted based on nearest neighbor search; and S4, carrying out fusion, regularization and classification on component boundaries, extracting component parameters, carrying out secondary development based on Revit, reading the component parameters, and realizing BIM automatic modeling. According to the method, creation of the BIM model can be accurately completed, manual participation in the process of generating the BIM model through the point cloud is greatly reduced, the efficiency and accuracy of generating the BIM model through the point cloud can be improved, and certain method support is provided for construction of a smart city.

The invention provides a multi-stage skyline positioning method, which belongs to the technical field of navigation, and specifically comprises the following steps: acquiring panorama image data of a mountain range image in a mountainous area, generating a mountainous area landscape image, and extracting a skyline corresponding to the mountainous area landscape image; extracting a local aggregation descriptor feature, an SVD feature and a height frequency vector from the landscape image, splicing the features into a unified retrieval vector, screening an initial skyline set by taking an Euclidean distance as a measure, and taking a first number of skylines ranked from small to large as a candidate skyline set; using a ResNet network to extract depth features and calculate Euclidean distances between the depth features and the candidate skyline set, and selecting a second number of skylines before the Euclidean distances are ranked from small to large as a target skyline set; and point set registration is carried out, final sorting is carried out according to the registered Euclidean distance, and the position information corresponding to the sample with the minimum distance is selected as the positioning result of the mountain landscape image. Through the scheme of the invention, the positioning accuracy and adaptability are improved.

The application discloses a bimodal developing navigation marker system, which comprises a marker body, a visible characteristic module, a positioning and registration module, a fusion display module and a deformation compensation module. The marker is made of metal titanium and coated with a polymer coating containing micro-bubbles, and can be developed in CT and ultrasonic images simultaneously. The system realizes automatic alignment of the CT model and the B ultrasonic image through a point set registration algorithm, and realizes real-time prediction of the non-rigid deformation of soft tissue by using an optical flow algorithm and correction of the position of the three-dimensional model, so that intraoperative dynamic navigation and accurate puncture are realized. The system has the advantages of strong bimodal visibility, fast registration speed, high puncture precision and real-time respiratory compensation, and is suitable for kidney puncture, biopsy and interventional therapy navigation.

The present invention relates to a method encompassing acquiring data as to the geometric structure / topology of a predefined surface section as well as to an additional physical property assigned to the surface section, determining variability of the geometric structure / topology as well as of the additional physical property over the surface section, and determining from these at variabilities an expected stability of co-registering datasets describing the predefined surface section. The present invention further relates to a corresponding computer program and a corresponding system for carrying out this method.

The invention relates to the technical field of intelligent auxiliary driving, and provides an electronic outside rear-view mirror follow-up visual field adjusting method and electronic equipment so as to track a target, set a marker line, delimit a visual field tracking range according to the marker line, ensure that a tracking target clock is located in the center area of a display picture and further ensure the driving visual field. Meanwhile, during use collection, an image feature point set of a collected image frame is synchronously obtained, the movement distribution change of a tracking target in a current image collection area is further determined, then the marker line is adaptively adjusted, and the view tracking area of the current collected image frame is moved to execute follow-up view adjustment. Stable tracking of the view tracking area is achieved through point set registration, display is smooth, the method is suitable for matching combination of different tractors and trailers, universality is good, and the safety of the vehicle is effectively improved.

The present application belongs to the technical field of underwater multi-sensor fusion matching positioning, and particularly relates to a seabed terrain and gravity-magnetic fusion matching method based on Bayesian filtering, a program, an equipment and a storage medium. The present application converts the selection of corresponding points of seabed terrain into a graph isomorphism problem, improving the accuracy and robustness of the selection of corresponding points. The present application uses an isogram nearest point iteration-based seabed gravity-magnetic information corresponding point selection algorithm, improving the accuracy of the selection of corresponding points. The present application uses a filter-based point set registration algorithm, converting the solving of rotation and translation parameters into a filtering problem, making the rotation and translation parameters explore in a dynamic space, using a nonlinear Bayesian filter to optimize the problem, making the algorithm converge to a global optimal solution. The present application uses Bayesian filtering to fuse the gravity-magnetic matching result, the terrain matching result and the inertial navigation result, obtaining an optimal estimation of the pose.

The invention relates to a point set registration method and device based on a kernel function and feature constraint, and aims to solve the problem that a traditional ICP algorithm is insufficient in robustness in scenes with noise, shielding and large attitude deviation. The method comprises the following steps: firstly, generating a feature enhancement point set through curvature and density combined feature extraction and mixed sampling; the residual scale is dynamically estimated based on the median absolute deviation in the coarse registration stage, and abnormal matching is suppressed in a weighted mode by combining a Cauchy kernel function; and finally, introducing feature constraint to optimize rigid body transformation parameters in a fine registration stage, and improving the overall registration precision and convergence stability. The method provided by the embodiment of the invention can effectively reduce errors, enhance robustness and convergence speed and realize high-precision point set registration in point set registration tasks with shielding, missing, noise interference and large initial value deviation, and has good engineering application value and popularization prospect.

A method for alignment of volumetric and surface scan images, said method comprising the steps of: receiving a volumetric image and surface scan image, wherein the volumetric image is a three-dimensional voxel array of a maxillofacial anatomy of a patient and the surface scan image is a polygonal mesh corresponding to the maxillofacial anatomy of the same patient; segmenting the volumetric image and surface scan image into a set of distinct anatomical structures by assigning each voxel in the volumetric image an identifier by structure and assigning each vertex or face of the mesh from the surface scan image an identifier by structure, wherein at least one of the distinct anatomical structures are in common between the volumetric and the surface scan image; extracting a polygonal mesh from the volumetric image featuring common structures with the polygonal mesh from the surface scan image; converting both meshes from the volumetric image and from the surface scan to a point cloud; and aligning the converted meshes via point clouds using a point set registration.

The present disclosure relates to a point set registration method and device based on kernel functions and feature constraints, aiming to solve the problem of insufficient robustness of traditional ICP algorithms in scenarios with noise, occlusion, and large posture deviation. The method first generates a feature-enhanced point set through joint feature extraction and mixed sampling of curvature and density; then, in the coarse registration stage, the residual scale is dynamically estimated based on the median absolute deviation, combined with the Cauchy kernel function weighted suppression of abnormal matching; finally, in the fine registration stage, feature constraints are introduced to optimize the rigid body transformation parameters to improve the overall registration accuracy and convergence stability. The method provided in the embodiment of the present disclosure can effectively reduce errors, enhance robustness and convergence speed in point set registration tasks with occlusion, missing, noise interference, and large initial value deviation, and achieve high-precision point set registration, which has good engineering application value and promotion prospects.

The invention discloses a registration method for a tooth and oral cavity endoscope two-dimensional image and a CBCT (cone beam computed tomography) three-dimensional model image, relates to the technical field of image registration, and aims at the tooth and oral cavity endoscope two-dimensional image and the CBCT three-dimensional model image. Image registration is carried out by using real tooth edge contour points in the oral cavity endoscope two-dimensional image and three-dimensional model projection contour points as registration feature points; according to the technical key points, an oral cavity internal three-dimensional model is obtained through oral cavity optical scanning, and tooth gingival margin line feature information is extracted; a YOLOv5s target detection algorithm and a Canny edge detection algorithm are combined to carry out target tooth identification and segmentation and edge contour extraction on a first frame image input by an oral cavity endoscope, so that the image registration efficiency is improved; contour point set registration is carried out in a one-to-many mode through a coherent point drift algorithm, and the rotation angle around each coordinate axis and a registration transformation matrix which should be applied when registration of the three-dimensional model is achieved are rapidly obtained; according to the method, the three-dimensional model is subjected to dimensionality reduction projection, so that the dimensionality of the image to be registered is unified, the real tooth edge contour points in the two-dimensional image of the oral cavity endoscope and the projection contour points of the three-dimensional model are taken as the registration feature points, and the efficiency and accuracy of image registration can be improved.

The invention discloses a printed circuit board automatic optical detection method and system based on edge computing power, and the method comprises the steps: obtaining circuit board design data and / or a standard template image of a target product, and constructing a standard template point set containing a material center point coordinate and a material category mark; collecting an image of the circuit board to be detected; performing weighted point set registration on the set; according to the target space position and the position, the size and the orientation of the corresponding detection frame, the offset, the rotation deviation and the scale change of the material position number on the current circuit board and the overlapping degree index between the current circuit board and the template detection frame are calculated; and comparing the detection index of each material position number on the current circuit board with the corresponding dynamic judgment threshold to obtain a qualification or disqualification judgment result of the circuit board to be detected and each material position number. According to the technical scheme, the manual parameter adjustment frequency is reduced in the multi-variety and working condition drifting production environment, and the reliability and consistency of defect judgment are stabilized.