64results about How to "Reduce registration error" patented technology

An embodiment of the invention provides a panorama image stitching method and device which can perform good processing on a plurality of images when a camera is in non-pure rotation and displacement exists, and can allow the distortion generation rate of stitched images to be reduced. The method includes the following steps of: designing a homographic matrix according to local areas of various feature points in various images; using the designed homographic matrix to detect the feature points and delete the wrong feature points; performing meshing processing on the images; constructing a global energy function based on mesh vertices and various feature points after the wrong feature points are deleted; solving the optimized mesh vertices; performing coordinate transformation on pixel points of various images based on the optimized mesh vertices; and performing stitching processing on various images by using various transformed pixel points.

The invention discloses an image stitching method based on characteristic point plane similarity, comprising first, respectively extracting characteristic points of two images to be stitched; matching the characteristic points to obtain characteristic point matching pairs; screening the characteristic point matching pairs based on plane similarity, the screening method including first randomly selecting a smallest sampling set; calculating the homography matrix of the smallest sampling set; calculating the residual error between each matching pair and a corresponding homography matrix to form a residual vector; and calculating the plane similarity between each two matching pairs according to a residual vector, and furthermore screening the matching pairs; calculating a transformational matrix according to screened matching sets, and transforming two images to be stitched into a same coordinate system; searching for an optimal stitching line; and fusing images according to the optimal stitching line to obtain an image stitching result. The method employs characteristic point matching pairs which are screened based on plane similarity to perform registering and stitching, and can improve image stitching accuracy and robustness.

The invention relates to a three-dimensional registering method for a single image, and the method comprises three steps: camera calibration, interactive modeling and camera registration. The step of camera calibration comprises the substeps: extracting rough-resolution blanking points through employing a grid method, purifying the rough-resolution blanking points through combining with PClines spatial conversion and an alignment linear point detection algorithm, obtaining candidate blanking points, carrying out the optimization and updating of the candidate blanking points to obtain typical blanking points, taking the typical blanking points as the camera calibration features to build a Manhattan world model, carrying out the analysis of the typical blanking points, and obtaining the internal and external parameters of a camera, wherein the internal and external parameters comprises a focal length and a rotation matrix. The step of camera registration is mainly using a method based on the linear feature alignment to carry out the registering of a three-dimensional model and register the three-dimensional model in a unified three-dimensional scene. The method can iron out the defects that the number of accumulative errors of the three-dimensional registration technology of the single image based on the geometric projection information is larger, and obtain a more accurate three-dimensional reconstruction effect.

The invention provides a medical image registration method based on combination of mutual information and gradient information, comprising steps of performing weighting on mutual information similarity and adding difference values of a gradient model into gradient similarity, performing space changing on a floating image to be registered and calculating similarity measure of a reference image and the changed floating image. The invention provides a similarity measure function of weighting mutual information and gradient information. The similarity measure can attain a maximization value through constantly optimizing search and changing space, and the registration is successful. The similarity measure function is employed in the process of the registration, which can effectively solve the problems that the medical image registration algorism is low in accuracy and the robustness is not strong.

The invention provides a high-efficiency registration method aimed at a CT and optical scanning tooth model, and belongs to the field of computer oral cavity recovery. The method comprises the steps of (1) data preparation in which 3D visualization is carried out on CT data of the oral cavity to obtain a CT reconstructed tooth image, visualization is carried out on optical scanning data of the tooth crown and gum part to obtain an optical scanning tooth crown image, and the CT reconstructed tooth image and the optical scanning tooth crown image are arranged in the unified world coordinate system; (2) initial registration in which characteristic points are selected manually to process the CT reconstructed tooth image and the optical scanning tooth crown image and further to obtain an initial registration model; and (3) accurate registration in which characteristic points are selected automatically to process the initial registration model and further to obtain an accurate registration model. Problems in registration data with scale transformation and different resolution are solved, and a registration result is high in convenience speed and low in registration error.

A digital image processing method. The method includes printing a first set of reference marks on one side of a substrate with a first print engine; printing a second set of reference marks on the same side of the substrate as the first set of reference marks with a second print engine; sensing both sets of reference marks on the substrate with an image sensing unit and generating a digital image of the reference marks; performing image analysis on the digital image to obtain an image-to-image distortion map where the image-to-image distortion map is a local measure of difference between the first set of reference marks and the second set of reference marks; and generating a compensated customer image by using the image-to-image distortion map to reduce registration errors when using the first and second print engines.

The invention provides a three-dimensional point cloud registration method adopting a CPD (coherent point drift) algorithm based on an affine transformation model, relates to a spraying robot position registration method based on point cloud registration and aims to solve the problems of high complexity, long program running time, low registration accuracy and poor robustness of the conventional mainstream three-dimensional point cloud registration algorithm CPD. The specific process comprises the following steps: 1, obtaining a group of three-dimensional point cloud data to serve as to-be-registered point cloud; 2, taking the obtained point cloud data as a reference point set; 3, calculating a covariance and initiating B and t; 4, parallelly computing P by a GPU; 5, solving parameters B, t and sigma<2> when an objective function gets maximum; 6, iterating steps 4 and 5 repeatedly until the covariance is smaller than a set threshold, solving B and t as well as a finally registered result point set when the covariance is smaller than the set threshold, and performing spraying on a to-be-sprayed object according to the final registration result. The three-dimensional point cloud registration method applies to the field of spraying by spraying robots.

The invention discloses an automatic three-dimensional point cloud registration method applicable to a ToF (Time of Flight) camera and a system. Diffuse reflection material is laid and stuck on the upper surface of a turntable, round marking points are arranged at intervals on the diffuse reflection material, and the ToF camera and the controllable turntable are both connected with a computer. Data are acquired through the ToF camera, the turntable is recognized in the three-dimensional point cloud, the turning angle of the turntable is combined for carrying out coarse registration on the three-dimensional point cloud, measured object point cloud is acquired through background segmentation, marking point cloud is then acquired through recognizing the marking points on the turntable in a strength image and combining the coarse registration three-dimensional point cloud, registration of each view angle marking point cloud is completed by using a local gradual registration method, and finally, the marking point cloud registration result is applied to the measured object point cloud, and the registration process is completed. Compared with the prior art, the method and the system have the advantages of good accuracy, strong robustness, good usability, low cost, and wide application.

The invention belongs to the field of medical image processing and application and relates to a high-precision automatic neuronavigation space registration method. The method comprises, through a global optimization point set registration method based on Gaussian mixture models, transforming coordinates of a point cloud obtained by a laser scanner and a point cloud obtained through CT (computed tomography) to achieve primary registration; then by means of a disordered point automatic registration method, performing secondary registration on the result of the primary registration and patient space reconstructed by the CT during neuronavigation. Usage results indicate that the high-precision automatic neuronavigation space registration method is easy to implement, reliable in precision, applicable to clinical application and capable of being integrated in an existing navigation system, thereby greatly improving the precision of the navigation system.

The invention provides a laser detection system for three-dimensional imaging of a moving object and an imaging method. The imaging method comprises the following steps that object slice images are obtained through the laser detection system; a reference image is selected from the object slice images; mutual information registration is conducted on the reference image and an image to be registered; whether the registration succeeds or not is judged, the offset amount and rotating amount of the image to be registered with respect to the reference image are output if the registration succeeds, the image to be registered is corrected according to the offset amount and the rotating amount to achieve space consistency between the image to be registered and the reference image, and a particle swarm search algorithm is continuously adopted to optimize the offset amount and rotating amount and transform the image to be registered if the registration does not succeed; time delay of each pixel point in the area where the object is located is calculated; corresponding distance information is calculated by adopting a mutual correlation algorithm; the distance information corresponding to each pixel is directly converted into a three-dimensional image of the object. For the moving object, the slice images different in gating distance are utilized, the imaging difficulty of the moving object is greatly reduced, and the application range for a high-speed object is wider.

The invention relates to the field of medical equipment, and relates to a volume data visualization method for surgical navigation based on a PPC (Panel Personal Computer). A navigation system adopted in the method mainly consists of the PPC, a navigator, an auxiliary infrared position indicator, a reference frame, an adaptor and the like, wherein point-to-point real-time communication can be realized between the PPC and the navigator through a wireless communication technology, and navigation information is transmitted and displayed on the PPC. A main navigation model realized by the method is that the image information of a cross section, a vertical plane and a coronal plane corresponding to a navigation probe point can be displayed on the PPC in real time; furthermore, a section image obtained by intersecting the extending surface of the PPC and the head of a patient is displayed in real time through the PPC. According to the method, standard section images can be displayed, any section image can be displayed through the position of the PPC, tumours and lesions can be auxiliary positioned by doctors, and convenience and help can be brought for surgical operation.

The invention relates to a multi-resolution non-rigid head medical image registration method on the basis of image edges, which pertains to the field of information fusion. The method utilizes searches gradually becoming accurate to transform parameters during the registration process. At first, wavelet transformation is used for detecting the edge of an image; a edge pyramid image is formed on the basis of an edge image; then relatively wide search is carried out in the lowest scale level (the roughest level) of the edge pyramid image; the best transformation for registering the two images is found; each subsequent level takes the search result of the former level as a center; and search scope is reduced to continually search until the search in the highest scale level is carried out. Compared with the registration directly using the original image, the method can reduce calculated amount, improve accuracy and can register images of different spatial resolution.

The invention provides a registration target, a registration method, a registration device, an electronic device and a storage medium. The registration target comprises a fixed part and an acquisitiondevice connected with the fixed part, wherein the acquisition device is used for simultaneously acquiring data of a plurality of points on a to-be-registered object; thus, the registration target provided by the invention can acquire the data of the plurality of points on the cartilage surface and / or the hard bone surface of a to-be-registered bone at one time; and the registration time of the to-be-registered bone is greatly shortened. Meanwhile, point cloud data of the cartilage surface and / or the hard bone surface of the to-be-registered bone can be collected at one time through the registration target provided by the invention, so sampling points are not needed to be selected manually; operation is more convenient; and registration errors are effectively reduced. The registration method provided by the invention can obtain the information of bone surface soft tissues, so effective information is provided for bone ligament balance analysis.

The invention relates to a method for elastic registration of stereo NMR brain images based on machine learning. The machine learning method is used to obtain the best dimension of the computation attribute vector on each point in the reference image, from which the obtained best attribute vector keeps discrepancy of the greatest extent from the attribute vector on each point around, and conformability of the greatest extent with the attribute vector on the corresponding point of the training sample. Based on the significance and consistency condition of the attribute vector on each point of the image, a standard for evaluating a key point is defined. The key point is selected automatically and hierarchically in each registration stage via the machine learning method, thus preventing the registration process from trapping in a local minimum value point. Finally, the machine learning based frame is combined with the existing registration arithmetic to complete the elastic registration of stereo NMR brain images. The invention can enhance precision and robustness of registration of both real MR images and simulated MR images, thereby establishing a foundation for the feasibility and accuracy of subsequent clinical applications.

The invention provides a feature registration method based on branch points. The method includes: firstly obtaining a green channel of an eye fundus image, and extracting a blood vessel center line; extracting a vascular tree in the eye fundus image by employing a blood vessel center line extracting technology based on gradient and divergence; then extracting feature points of the eye fundus vascular tree by employing an eight-neighborhood table look-up method, and deleting erroneous feature points to obtain Harris angle point coordinates corresponding to correct branch points; performing feature extraction on the obtained branch points by employing an improved feature description algorithm, obtaining corresponding branch angles and vascular lengths, and forming feature vectors; and finally obtaining feature point pairs used for final matching by employing Euclidean distances between the vectors, obtaining a transformation matrix of a reference image relative to an original image by employing a random sample consensus algorithm, performing basic modification on a registered result image, and obtaining the eye fundus image with accurate registration. According to the registration method, good effect is achieved in a disclosed FIRE eye fundus registration database.

The invention discloses a three-dimensional non-rigid point cloud registration method based on a consistent point shift algorithm, which relates to a spray painting robot position registration method based on point cloud registration and aims at solving defects of low registration precision, poor robustness and long calculation time existing in the present three-dimensional point cloud non-rigid registration algorithm in conditions of large point cloud data amount, complex deformation, noise, outliers and missing point influences. The method particularly comprises steps: 1, an image acquisition device in the spray painting robot is used for scanning an object waiting for spray painting, and a group of three-dimensional point cloud data is obtained as a to-be-registered point cloud; 2, the obtained point cloud data serve as a reference point set; 3, the covariance sigma between the reference point set and the existing template point set is calculated, and related parameters of the consistent point shift algorithm are initialized; 4, a Gauss kernel matrix is constructed; 5, the final registered result point set is obtained, and according to the final registered result point set, spray painting work is carried out on the object waiting for spray painting. The method of the invention is used for spray painting robot position registration.

A registration system includes nip rollers for conveying a sheet in a process direction when the nip rollers are in an engaged position, and first and second sets of omnidirectional rollers arranged perpendicularly to each other, for conveying the sheet in the process and cross-process directions, respectively. The omnidirectional rollers each allow the sheet to slide, relative to the omnidirectional roller, in a direction parallel to an axis of rotation of the omnidirectional roller. The nip-rollers are movable from the engaged position to a disengaged position, in which the first pair of omnidirectional rollers convey the sheet in the process direction. A feedback system includes sensors for determining registration errors of the sheet and a control system controls the omnidirectional rollers to reduce the registration errors in the sheet while the nip-rollers are in the disengaged position.

The invention discloses a hierarchical multi-piece point cloud rigid registration method, including the steps of: (1) building a connection relation map according to overlapping relations of multiple pieces of point clouds; (2) performing rigid registration on each two pieces of point clouds having a connection relation; (3) extracting independent loops in the point cloud connection relation map; (4) calculating the closing degree of each independent loop according to a point cloud rigid registration result; (5) determining a sequence according to the closing degree of each independent loop, and performing closing operation on each independent loop successively; and (6) performing global error diffusion on the connection relation map. The hierarchical multi-piece point cloud rigid registration method provided by the invention can perform rigid registration on multiple pieces of point clouds and ensure closure of rigid transformation between the point clouds, and has the advantages of small registration error, high operation efficiency and low dependence on initial position conditions of the point clouds.

The invention provides a super-resolution reconstruction method based on learning and adaptive trilateral filtering regularization. The super-resolution reconstruction method specifically comprises the following steps: inputting a low-resolution image block sequence; calculating a dictionary D1 of the low-resolution image blocks and a dictionary Dh of the high-resolution image blocks correspondingto the low-resolution image blocks; obtaining a high-resolution initial interpolation image through a non-uniform interpolation method; calculating sparse representation sparse alpha based on the learning dictionary D1; then, based on the learning dictionary Dh, calculating a high-frequency prior image block hi; down-sampling the obtained high-frequency prior image until the size is the same as that of the high-resolution initial interpolation image to obtain a down-sampled high-frequency prior image, and finally superposing the down-sampled high-frequency prior image and the high-resolutioninitial interpolation image to generate a high-resolution iterative initial image; and calculating a regularization parameter lambda and combining with a trilateral filtering method, and calculating and reconstructing a high-resolution image by adopting a loop iteration mode.

The invention relates to the technical field of electronic maps, and discloses a matching method for a GNSS position and an image in a crowdsourcing map. The method comprises the step of: matching animage acquired by a camera module with a GNSS position corresponding to an image acquired by a GNSS module according to the same time point; the method further comprises at least one of the followingmanners for improving matching accuracy: (1) setting a clock of the camera module with a clock of the GNSS module as a reference; (2) regulating an output timestamp of the GNSS position; (3) framing from multiple images acquired by the camera module through an onPreviewFrame () function; and (4) after matching, regulating a matching error. The method can solve the problem of high-precision matching of the GNSS position and camera framing, and lowers registration errors.

The invention discloses a global measurement data registration method based on multi-vector constraint, belongs to the field of large-size measurement, and relates to a global measurement data registration method based on multi-vector constraint. According to the method, a plurality of non-coplanar target balls are reasonably arranged in a certain space to serve as common points, a common point data set is obtained through multi-station measurement of a laser tracker, and the station with the minimum overall measurement error serves as a global coordinate system. Normal vectors are then solvedbased on the common points and unitized. And finally, a conversion parameter of the local coordinate system is solved based on the constraint of the normal vector of the redundant unit, and the localdata is registered to the global coordinate system. According to the method, the problem that the data registration precision in a large-size space is abnormally sensitive to common point measurementerrors and the number of common points is fully considered; registration errors of global measurement data can be effectively reduced, the method can be quickly applied to field practice to guaranteethe overall measurement accuracy of geometrical characteristics of large aviation parts, and the method has wide application prospects in the field of large-size measurement.

The invention belongs to the field of human face registration, especially the dynamic face registration method based on 3D camera, Aiming at the problem that the existing face registration can not accurately find the invariant features of human face, which leads to the low registration accuracy and quality, the following solutions are proposed: S1, registering the personal information of the reference person, establishing the reference three-dimensional face model of the reference person, and establishing a spatial coordinate system for the reference three-dimensional face model at the same time; S2, connecting the nose tip of the reference three-dimensional face with the forehead to obtain a reference line 1, and connecting the two eyes of the reference three-dimensional face to obtain areference line 2. The matching point 1 and the matching point 2 are respectively coincided with the reference point 1 and the reference point 2 to carry out the preliminary positioning of the three-dimensional face to be registered, Rotating three-dimensional face to be registered with the line between the coincident reference point one and the reference point two as the axis can reduce the registration error, make the registration accuracy higher, reduce the registration step, and improve the registration efficiency.

The invention discloses a hub point cloud registration method and device, electronic equipment and a storage medium. The method comprises the steps of: acquiring a hub sampling point cloud and a preset hub standard point cloud; performing inclination correction on the hub sampling point cloud, so that the main axis direction of the hub sampling point cloud coincides with the main axis direction of the hub standard point cloud; projecting the corrected hub sampling point cloud and the corrected hub standard point cloud to a reference plane to respectively obtain a sampling point cloud projection image and a standard point cloud projection image; performing image matching on the sampling point cloud projection image and the standard point cloud projection image to obtain rotation amount and translation amount for moving the sampling point cloud projection image to be overlapped with the standard point cloud projection image; determining a main shaft translation amount between the hub sampling point cloud and the hub standard point cloud based on the moved sampling point cloud projection image and the standard point cloud projection image; and moving the hub sampling point cloud according to the rotation amount, the translation amount and the main shaft translation amount. According to the scheme provided by the invention, the registration precision of the hub point cloud can be improved, and the registration error is reduced.

A system for reducing tension fluctuations in a web while printing multiple copies of a print job on the web comprises a printing system with a print station disposed opposite a first side of the web and one or more rollers adapted to receive tension control commands, the tension control commands operating on the first rollers to control the amount of tension in the web. A sensor is used to measure tension changes produced during the printing of the first copy of the print job. A processor is used to determine first tension control adjustments based on the measured tension changes and to use the adjustments to adjust the tension control commands to the rollers in the printing system to change the tension in the web when printing a second copy of the print job, thereby reducing tension fluctuations in the web.

The invention discloses an object distance detection method of a robot, and the method comprises the following steps: acquiring a color image acquired by a color camera; identifying the image size ofa target object included in the color image; determining whether the distance of the object meets a preset linear fitting requirement or not according to the image size of the target object; when thedistance of the target object meets the preset linear fitting requirement, acquiring the distance between the target object and the robot according to the preset corresponding relation between the image size of the target object and the distance. The distance between the robot and the target object can be quickly determined only through simple judgment and search; the requirement for hardware computing resources can be lowered, the robot cost is reduced, the distance is directly calculated according to the size of the image, registration errors caused by the fact that the image of the target object is small can be reduced, and the accuracy of the calculated distance can be improved.

The invention belongs to the field of medical equipment, and particularly relates to a handheld three-dimensional scanner-based spatial registration method for surgical navigation. The method adopts facial recognition feature point matching registration to perform preliminary registration on space point clouds of a to-be-tested person obtained by the handheld three-dimensional scanner and image space point clouds obtained by a medical image; a preliminary registration result uses a point cloud iterative registration method to complete the registration of the space of the to-be-tested person ofsurgical navigation and the image space; implementation results show that the method has fast registration speed and stable registration precision, and can meet the requirements of a surgical navigation system and improve the registration accuracy of the surgical navigation system.

The invention belongs to the technical field of image processing, and discloses an infrared image homologous registration method based on improved angular point main direction distribution. The methodcomprises the following steps: respectively calculating a plurality of angular points in two infrared images to be registered; an angular point main direction distribution method is improved; allocating a main direction to each corner point; establishing a matching relationship between the descriptor and the main direction; therefore, the rotation invariance of the image is realized; and finally,on the basis of extracting a plurality of PIIFD descriptors from an infrared image to be registered and a reference infrared image, registration is carried out by adopting a nearest neighbor priorityBBF and combining a bilateral matching algorithm. The whole algorithm is simple and efficient in calculation and high in accuracy, and the problems of high registration difficulty and low registration precision caused by low infrared image resolution and large texture information loss are effectively solved.

The invention relates to an electric test structure for measuring epitaxial graphic offset, comprising a buried layer formed on the surface of a semiconductor substrate, an epitaxial layer formed on the surface of the semiconductor substrate, a plug region equidistantly which is formed in the epitaxial layer on the surface of the semiconductor substrate and has a preset offset different with the buried layer along the same direction, a contact hole formed on the surface of the plug region as well as a first conductive connection, a second conductive connection and a third conductive connection which are sequentially formed at the contact hole on the adjacent plug region. The method for measuring the epitaxial graphic offset comprises the step of respectively carrying out an electric test on the electric test structure with different preset offsets. In the invention, the electric test structure is utilized to measure the epitaxial graphic offset, not only is the measuring accuracy higher, but also the registration error of a graph after extension to the graph before extension can be effective reduced.

The invention discloses a space registration method and device for a robot and image equipment, and electronic equipment. The method comprises the steps: building an intermediate coordinate system based on a marker group, processing a perspective image to generate a perspective image coordinate system, and determining the relative pose relation between the perspective image coordinate system and the intermediate coordinate system; determining the relative pose relationship between a robot coordinate system and the intermediate coordinate system, and further determining the relative pose relationship between the robot coordinate system and the perspective image coordinate system; establishing an image equipment coordinate system based on the circle center of a perspective scanning area, anddetermining the relative pose relationship between the perspective image coordinate system and the image equipment coordinate system; and determining the relative pose relationship between the robotcoordinate system and the image equipment coordinate system to realize space registration of the robot and the image equipment. According to the space registration method provided by the invention, registration errors are reduced, and a doctor can realize rapid and high-precision space registration of the surgical robot and the medical imaging equipment by performing basic medical imaging equipment operations and image analysis.