53results about How to "The registration result is accurate" patented technology

The present invention discloses a dentition model generation method based on oral cavity scanning data and CBCT (Cone Beam Computed Tomography) data. The method comprises: reading a three-dimensional multi-source data model of an oral cavity; setting and registering a fixed model and a floating model, wherein the fixed model is a triangular mesh model created by means of CBCT, and the floating model is a triangular mesh model of intraoral scanning; manually picking up feature point pairs of the models on the fixed model and the floating model, correcting the feature point pairs according to point cloud curvature of the models, moving feature points to a point with the largest curvature in a surrounding neighborhood, and performing preliminary registration on the feature point pairs by using a point-to-point ICP (Iterative Closest Point) algorithm; calculating a point cloud error of the models after preliminary registration, and by combination with medical parameter requirements, setting a parameter of precise registration; performing precise registration by using an optimized point-to-surface ICP algorithm; and fusing the CBCT data and intraoral scanning data after registration, and outputting tooth root and crown models after registration and fusion. Rapid and precise registration and fusion of data of different devices or data of different postures of patients are implemented.

The invention discloses a target object three-dimensional color point cloud generation method based on Kinect. Firstly a set of RGBD images are photographed around a target object so that the set of RGBD images are enabled to include complete information of the target object; then as for each RGBD image, Ostu segmentation is performed on a depth image and then a foreground mark is acquired and acts as the input of a Grab Cut algorithm, the RGB image is segmented again and then the accurate area of the target object is acquired, and background information is removed; registration is performed on adjacent point cloud fragments by using an improved ICP algorithm so that a transformation relation matrix between the point cloud fragments is acquired; and finally point clouds are spliced by using the transformation relation matrix between the point cloud fragments, and down-sampling is performed to reduce redundancy so that the complete three-dimensional color point cloud data of the target object can be acquired.

The invention discloses a method and a device for registering point cloud data, and belongs to the technical field of computers. The method comprises steps as follows: the point cloud data of every two adjacent frames are registered to obtain target attitude transformation parameters between the point cloud data of every two adjacent frames; the point cloud data of each frame are accumulated according to each target attitude transformation parameter to obtain initial registered point cloud data; registration reference data are extracted from the initial registered point cloud data and have the high registration degree; the point cloud data of each frame are registered again by referring to the registration reference data so as to obtain target registered point cloud data. According to the method and the device for registering the point cloud data, the point cloud data of each frame are accumulated according to each target attitude transformation parameter, after the initial registered point cloud data are obtained, the registration reference data with the higher registration degree are further extracted from the initial registered point cloud data, the point cloud data of each frame are registered again by referring to the registration reference data, and accumulated errors can be reduced, so that a more accurate registration result can be obtained.

The invention discloses an image registration method based on improved mutual information and Harris corner point features. The method mainly solves the problems that through a traditional method, registration consumes much time and registration accuracy is low. The method includes the implementation steps that firstly, an image F to be registered and a reference image R are input; secondly, rectangular frames are placed in any position of the image F to be registered and any position of the reference image R, and most similar rectangular frames are obtained; thirdly, corner point features of the most similar rectangular frames are extracted, and matching points in the similar rectangular frames are obtained; fourthly, the similar rectangular frames on which the matching points are extracted are registered after mismatching points are deleted, and registered transformation parameters are recorded; fifthly, affine transformation is conducted on the image F to be registered through the transformation parameters; sixthly, the image obtained after affine transformation is fused with the reference image R, and then a registered image is obtained. The image registration method has the advantages of being short in consumed registration time and high in registration accuracy, and can be used for registration of medical images, natural images and synthetic aperture radar images.

The invention discloses a visible light-infrared image registration method based on salient region features and edge degree, and is to solve the technical problem of poor registration precision due to low similarity of image features extracted through an existing visible light-infrared image registration method. The technical scheme is characterized by extracting a salient region from an image pair to be registered through a significant region detection method based on adaptive local contrast; describing shape features of the salient region by utilizing a Zernike rotation invariant moment to ensure translation, gray level, scale and rotation invariance of the utilized features; through RANSAC, obtaining a homonymy point pair and carrying out estimation to obtain initial transform parameters; and finally, carrying out optimization on the initial transform parameters by utilizing a three-dimensional joint histogram constructed based on edge degree to obtain a final result. The method improves image registration precision through salient region and edge degree features, the similarity of which are higher, between heterogeneous source images, and invariance of the Zernike moment.

The invention discloses a video sequence registering method based on combination of motion information and background information. The method mainly solves the problem that in the prior art, video sequences in different view points cannot be accurately registered. The method includes the steps that (1) separation of background images and motion targets is carried out on two input video sequences; (2) feature point matching pairs of the background images are obtained, and basis matrixes between the background images are calculated; (3) four sets of matching point pairs corresponding to four spatial points not on the same plane in a three-dimensional space are selected from the background images; (4) intersection points of projection lines, in another video sequence, of points in motion tracks and bipolar lines are obtained; (5) motion track points are matched and a point pair set corresponding to candidate time is obtained; (6) a time line is fitted and time transformation parameters are recovered. The time relation between video sequences can be accurately recovered, registering accuracy is improved and the method can be used for registering video sequences with static backgrounds.

The invention discloses a contour registration method used in train guide rail contour measurement based on machine vision. Contour measurement comprises the first step of collecting steel rail images, the second step of obtaining a single-pixel curved line of a steel rail profile through computer processing, the third step of carrying out calibration restoring and the fourth step of carrying out contour registration. The contour registration method comprises the steps of enabling an actual contour line and a standard contour line to be unified in a coordinate system and utilizing asymptotic lines of the two contour lines to obtain an included angle between the actual contour line and the standard contour line, rotating the actual contour line and enabling the asymptotic lines of the two contour lines to be paralleled, and horizontally moving the actual contour line and enabling ends of the actual contour line to be coincident with ends of the standard contour line. According to the contour registration method, the actual contour line and the standard contour line are unified into the coordinate system and the actual contour line is processed, the asymptotic lines of the actual contour line and the standard contour line are paralleled and coincide at ends, and therefore the registration result is accurate, observation and calculation are carried out conveniently and reliable abrasion loss data can be achieved.

The invention provides a CBCT and laser scanning point cloud data tooth registration method based on supervoxels. The method comprises the following steps: firstly, extracting a tooth model from oralcavity CBCT scanning data according to a region growing method; secondly, processing the tooth model by using a laser scanner, and separating a dental crown in the tooth model from abutment teeth according to the characteristics of the teeth; converting the tooth models from different sources into tooth point cloud data, adding color information to the tooth point cloud data to assist initial geometric alignment of the tooth point cloud, and realizing down-sampling of the point cloud based on a super voxel method; and finally, constructing an alignment metric, and proposing a mutual corresponding matching condition based on the mixed characteristics; two pieces of tooth point cloud data with color information are registered from different angles; according to the method, tooth registrationof different resolutions can be achieved, tooth color information is manually added, the registration precision is improved, the registration time is saved, and a complete tooth model capable of being used for implantation and orthognathic surgery is constructed.

The invention provides a point cloud registration method based a neighbourhood rotary volume and belongs to the technical field of three-dimensional images. The point cloud registration method based the neighbourhood rotary volume aims at solving the technical problems of low registration speed and poor anti-noise performance exist in an existing point cloud registration method. The method comprises the steps of loading and separately calculating a source key point set and a target key point set of source point cloud and target point cloud of which visual angles are adjacent; for each key point, using n times of surface resolution as a neighbourhood, calculating out the rotary volume of neighborhood inner points and using the rotary volume as a description vector, and obtaining a source key point description vector set and a target key point description vector set; according to a calculation result of the last step, calculating a rotary translation matrix; using the rotary translation matrix to rotate and translate the source key point description vector set, and obtaining a middle key point description vector set; according to a final rotary translation matrix, converting the source point cloud to a coordinate system in which the target cloud is located, and obtaining middle point cloud; using an ICP algorithm to precisely registrate the middle point cloud and the target point cloud.

The invention discloses a contour registration method used in train guide rail contour measurement based on machine vision. Contour measurement comprises the first step of collecting steel rail images, the second step of obtaining a single-pixel curved line of a steel rail profile through computer processing, the third step of carrying out calibration restoring and the fourth step of carrying out contour registration. The contour registration method comprises the steps of enabling an actual contour line and a standard contour line to be unified in a coordinate system and utilizing asymptotic lines of the two contour lines to obtain an included angle between the actual contour line and the standard contour line, rotating the actual contour line and enabling the asymptotic lines of the two contour lines to be paralleled, and horizontally moving the actual contour line and enabling ends of the actual contour line to be coincident with ends of the standard contour line. According to the contour registration method, the actual contour line and the standard contour line are unified into the coordinate system and the actual contour line is processed, the asymptotic lines of the actual contour line and the standard contour line are paralleled and coincide at ends, and therefore the registration result is accurate, observation and calculation are carried out conveniently and reliable abrasion loss data can be achieved.

The invention provides a multi-modal medical image registration fusion method and device and electronic equipment. The method comprises the following steps: acquiring two-dimensional medical images of at least two modals of a patient; respectively inputting the two-dimensional medical images of the at least two modalities into a pre-trained corresponding image segmentation network model so as to respectively obtain the output of the two-dimensional medical image of each modal body position area; and performing three-dimensional reconstruction on the two-dimensional medical image of each modal body position area, and then performing point cloud registration fusion to obtain a multi-modal fused three-dimensional medical image. The method is high in image registration precision, low in time cost, capable of processing complex multi-modal fusion conditions, capable of being applied to non-rigid registration conditions, accurate in registration result and capable of providing accurate treatment reference basis for medical staff.

The invention discloses a remote sensing image registration method based on multiple feature points, and mainly solves the disadvantages of low accuracy of conventional multisource and multispectral remote sensing image registration. The realization steps are that 1. two remote sensing images are inputted; 2. the anisotropic scale space of the input images is constructed; 3. Harris and Hessian operators are respectively used in the anisotropic scale space of the input images to perform feature point detection; 4. the two feature point detection results of the input images are combined so as to generate feature vectors; and 5. the feature vectors of the input images are matched, the mistakenly matched feature point pairs are deleted and registration results are outputted. The remote sensing image registration method based on the multiple feature points has advantages of high registration accuracy and can be used for registration of multisource remote sensing images and multispectral remote sensing images.

The invention discloses a self-adaptation splicing method of X-ray images. The method comprises the following steps that 1. horizontal image bands of two continuous X-ray images are obtained; 2. the variances of the two horizontal image bands are computed respectively, and a template image and a moving image are determined; 3. the template image and the moving image are registered; and 4. an overlap region nonlinear fusion method is used for carrying out fusion on images, and a new image after splicing is obtained. According to the method, the variances of the gray values of the horizontal image bands selected from the overlap region part of the two X-ray images are compared respectively, a clear horizontal image band with more information is selected in a self-adaptation mode as a template to be registered with another image, then the overlap region nonlinear fusion transition method is used for carrying out fusion on input images, and full-scale splicing is achieved.

The invention discloses a rectification skin repair method based on wing body actual measurement data virtual butt joint, which comprises skin registration and repair, and the registration comprises the steps of denoising wing body point cloud data, and carrying out data recombination on a data set obtained after filtering to obtain a key point set; calculating a normal direction feature descriptor of any key point in the key point set and a normal direction feature descriptor of any point in the skin point cloud data; calculating the Euclidean distance between the feature descriptors of the two points by using a K-nearest neighbor tree structure search algorithm, and adding the points with high similarity into the set M; and calling a clustering and Hough voting algorithm for the set M toobtain a local point cloud set matched with the skin point cloud data in the key point set; matching the skin point cloud data with the local point cloud set through a point cloud iterative nearest point algorithm. Skin registration and repair can be completed through computer programs, manpower can be greatly saved, and production efficiency is improved. Meanwhile, the influence of human subjective factors is avoided, and the registration result is more accurate.

The invention discloses a point cloud registration method based on a differential evolution algorithm and a TrimmedICP algorithm. The TrimmedICP algorithm can solve the problem of low overlapping rate in point cloud registration, but needs the setting of a proper initial value of a rotation matrix and a proper initial value of a translation matrix at first, otherwise it will fall into the local optimum; the differential evolution algorithm can generate an initial population through a random means and makes the population distributed in all ranges through using mutation, crossover and selection operation; and the differential evolution algorithm and the TrimmedICP algorithm are combined, and therefore, the problem of difficulty in selecting the initial values of the rotation matrix and the translation matrix can be solved, a globe optimal solution can be obtained, and thus, an accurate registration result can be obtained.

The invention discloses a contour registration method based on machine vision train guide rail contour measurement. The contour measurement comprises the following steps: S1, acquiring a steel rail image; S2, processing the acquired steel rail image by a computer to obtain a single-pixel curve of the cross section of the steel rail; S3, calibrating and restoring; S4, carrying out the contour registration, wherein the contour registration comprises the followings sub steps: S41, integrating a real contour line and a standard contour line into a coordinate system, and obtaining an angle between the real contour line and the standard contour line by utilizing an asymptotic line of the two contour lines; S42, rotating the real contour line, and enabling the asymptotic lines of the two contour lines to be parallel to each other; S53, translating the real contour line, and enabling the end point of the real contour line to be overlapped with the end point of the standard contour line. According to the registration method, the real contour line and the standard contour line are integrated into the coordinate system, the real contour line is processed, the asymptotic lines of the real contour line and the standard contour line are parallel to each other, and the end points of the two contour lines are overlapped, so that the registration result is accurate, convenience in observation and calculation can be realized, and reliable abrasion data can be obtained.

The invention discloses a 6DOF object attitude estimation method based on deep learning point cloud matching, and relates to the field of machine vision. The method comprises the following steps: 1, carrying out the key point detection of a to-be-recognized image according to an object key point detection network model based on deep learning, obtaining pixel coordinates of key points of an objectin the to-be-recognized image, and converting the pixel coordinates into image coordinates for a subsequent PNP algorithm; 2, solving the PNP algorithm according to the 3D coordinates and the image coordinates of the key points to obtain an initial attitude of the object in the to-be-recognized image, and using the initial attitude for an initial value of subsequent ICP algorithm registration; and3, performing object segmentation according to the key points of the object in the to-be-recognized image to obtain a local point cloud containing attitude information of the object, and registeringthe local point cloud with a template point cloud by using the ICP algorithm to obtain a pose of the object in the to-be-recognized image.

The invention belongs to the field of medical instruments, and particularly relates to a surgical navigation space registering method based on a probe. A point cloud coplanar four-point registering method is adopted in the method, registration is conducted on patient space point cloud obtained by the probe and image space point cloud, and a primary registration result of the two groups of point cloud is obtained; the point cloud iterative registration method is used according to the primary registration result, and the registration of the two groups of point cloud is completed. By means of themethod, it is shown through the result that the method is easy to implement, high in registration efficiency and stable in precision and can be applied in a surgical navigation system to improve theregistration precision of the surgical navigation system.

The invention provides a multi-view stereoscopic vision body surface positioning tracking method. The multi-view stereoscopic vision body surface positioning tracking method comprises the following steps that S1, calibrating cameras in each group of vision assemblies; S2, performing simulation geometric reconstruction on the CT data: during reconstruction, retaining point cloud data and target point information of a body surface in the CT data, and simulating a spatial position of the CT point cloud data in a left camera coordinate system; S3, collecting body surface point cloud data through visual components on the two sides of the treatment bed; S4, projecting grating stripes to the surface of the human body by each group of structured light projectors, and modulating the phases of the grating stripes; Carrying out phase splitting by a computer; Obtaining three-dimensional information of the body surface; S5, fusing the point cloud data collected on the two sides of the body surface;And S6, performing three-dimensional accurate registration on the collected and fused body surface point cloud data and CT reconstructed body surface point cloud data. The multi-view stereoscopic vision body surface positioning tracking method has no radiation damage to human bodies and is low in equipment cost.

The invention discloses a real-time dynamic registration method suitable for RGBD-SLAM and a machine readable storage medium. On the basis of the SLAM of the depth camera, edges are extracted by applying phase consistency according to each frame of infrared image and color image provided by the TOF and the camera, then normalization cross-correlation is calculated by adopting a blocking strategy,an optimal matching pair is obtained, and the offset is calculated, so that accurate registration of the infrared image and the color image is realized. According to the invention, real-time registration can be realized, the registration result is accurate, the operation of the SLAM can be guided more correctly, better AR experience can be brought to a user, and the life of people is more convenient.

Disclosed is a three-dimensional laser point cloud rapid relocation method, on the basis of a priori map, the data operand is reduced based on two-dimensional rasterization, the real-time performance is improved, meanwhile, a preliminary candidate scene set is obtained based on a Jaccard coefficient. On this basis, screening is carried out on the preliminary candidate scene set based on a data main direction and a Pearson correlation coefficient to obtain a candidate scene set, so that the real-time performance is improved, three-dimensional clustering is carried out based on an Euclidean clustering method, and then a bipartite graph is constructed for each historical frame scene in the candidate scene set; three-dimensional similarity measurement is completed based on cosine similarity after a maximum matching relation is found based on a Hungarian matching algorithm, and finally a unique candidate scene is obtained, so that the obtained corresponding relation is more reliable in solution, higher in robustness, more accurate in registration result, high in overall real-time performance of the method and high in accuracy. And finally, a 3D-NDT algorithm is utilized to obtain an attitude transformation relation matrix between two frames to complete repositioning. According to the method, unique candidate scene screening is carried out based on cosine similarity.

The invention discloses a dynamic contrast enhancement magnetic resonance image detection method, by which non-rigid registration measured based on complexity and similarity of a residual is applied to detection of a magnetic resonance image. The dynamic contrast enhancement magnetic resonance image detection method comprises the following steps: preprocessing an image, performing image interpolation by a B-spline function method, calculating the measuring similarity, setting circulating iteration of a threshold, and working out a difference image as a relatively precise image area display image. The invention as MRI contrast dynamic study, and a conventional pixel intensity values based on a similar approach as compared to the non-rigid registration, under the same experimental conditions, with more accurate registration of the resulting image. In the dynamic contrast enhancement magnetic resonance image detection method, a dynamic contrast magnetic resonance image is taken as a study object; and compared with the conventional non-rigid registration method based on similar intensity values of pixels, the dynamic contrast enhancement magnetic resonance image detection method has the advantage that a image registration result is more accurate under the same experimental condition.

The embodiment of the invention provides a method for realizing pulmonary nodule adaptive matching based on CT image skeleton registration. The method comprises the following steps of data preparation, lung and skeleton point cloud data extraction, three-dimensional point cloud data registration and pulmonary nodule adaptive matching. . Firstly, lung image data and pulmonary nodule data are subjected to three-dimensional point cloud rigid transformation registration based on the characteristic of small human skeleton change characteristics, so that alignment of the lung and pulmonary nodule data before and after follow-up visit is realized. Secondly, an FGR algorithm is adopted, so that the method is obviously superior to ICP and other local refinement algorithms in the aspects of operation speed and registration accuracy. Thirdly, RMSE is adopted as a lung point cloud registration error, adaptive matching of pulmonary nodules is achieved, manual intervention of pulmonary nodule registration is little, the automation degree is high, and the registration result is accurate; and fourthly, normalization processing is carried out on the CT image data, so that the robustness of the algorithm can be improved, and the method can be widely applied to different types of CT equipment and DICOM data with different pixel spacing values.

The invention relates to a tissue and organ registration method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring a first processing image of a first image and a second processing image of a second image, and performing rigid registration on a target organ in the first processing image and a target organ in the second processing image to obtain a rigid registration result, and performing elastic registration on the target organ in the rigid registration result and the target organ in the second processing image to obtain an elastic registration result, and processing the first image according to the rigid registration result and the elastic registration result to obtain a registered image. According to the method, rich information of the target organ contained in the first image can be registered into the second image, registration is carried out based on the conversion relation of the feature points in the target organ, and compared with a traditional registration method that registration is carried out only based on the surface feature points of the target organ, the obtained registration result is more accurate, and the robustness is high.

The invention discloses a seal identification control method, which comprises the following steps: collecting template seals: collecting images of seals to be identified into a database through a CCD or a scanner, storing them in a picture format as template seals; Text extraction: use the four operation steps of binarization, skeleton extraction, frame extraction and seal extraction to extract the seal to be recognized; seal registration: the seal registration operation is divided into two steps, rough registration and fine registration , the rough registration first adjusts the print to be recognized and the template to approximately the same position and orientation, and the fine registration further adjusts the two prints to almost the same position and orientation; seal identification: multi-level recognition is adopted The strategy and the multi-feature classification fusion decision-making method are used to identify the printed text to be recognized and the printed text of the template. The control method overcomes the shortcomings in the prior art, and can greatly improve the seal recognition reliability and recognition efficiency.

The invention discloses a 2D-3D image registration algorithm. The algorithm comprises the following steps: (1) scanning an affected part of a patient before an operation to obtain a preoperative 3D image; carrying out positive and lateral perspective scanning on the corresponding affected parts of the patient in an operation to obtain positive and lateral 2D perspective images; (2) constructing aspace model of intraoperative shooting positive and lateral perspective imaging equipment, and calculating and obtaining an initial pose of the preoperative 3D image in the space model according to the step (1); (3) generating a positive side position DRR image of the preoperative 3D image by adopting the space model constructed in the step (2), and carrying out template matching on the positive side position DRR image and the corresponding intraoperative 2D perspective image to obtain a rough matching transformation pose of the preoperative 3D image; and (4) performing accurate optimization matching by using a gradient descent optimization algorithm to obtain a final transformation pose of the preoperative 3D image. According to the method, rough matching is carried out by adopting template matching, and then the optimal result is obtained by using the gradient descent optimization algorithm, so that a very accurate registration result can be obtained while the registration speed is ensured.

The embodiments of the present invention provide an image registration method, an image registration device, an image splicing method and an image splicing device. The image registration method includes the following steps that: a correlation coefficient image of a reference image and a floating image are calculated; the gradient image of the correlation coefficient image is calculated; the pixelsof the extreme value of a correlation coefficient in the correlation coefficient image is determined according to the pixels of the extreme value of a gradient in the gradient image; and the reference image and the floating image are registered according to the pixels of the extreme value of the correlation coefficient. The image registration method and device are high in universality, small in calculation amount, are more accurate in registration result, and can ensure real-time operation. The image splicing method and the image splicing device are realized on the basis of the image registration method. The speed and accuracy of the image splicing method and the image splicing device are improved.

The invention discloses a point cloud speed measurement method based on an improved ICP, and the method comprises the steps of point cloud pre-processing: removing the noise points in the point cloud data, and removing the dynamic objects in the point cloud data through clustering and detection means; point cloud matching: obtaining the pose of a carrier when the current frame of point cloud is collected, and further optimizing the pose of the carrier corresponding to the current frame of point cloud by using an improved ICP method and on the basis of the pose; local graph optimization: carrying out local graph optimization on all poses in a given time interval including the latest calculated pose; and speed calculation: representing the pose change experienced by the carrier by using an angular axis method, and then calculating the current speed of the carrier. According to a three-dimensional point cloud pre-processing method, the outliers and burr points in the initial point cloud data can be removed, and the influence of noise points on the subsequent matching step is greatly reduced; a small-area plane is projected to a large-area plane, so that the projection effect is greatly close to the real situation, and a registration result is more accurate.

The invention relates to a change detection method based on a plurality of controllable pyramids, which is used for carrying out change detection on the difference between two images (Ipre and Icur) of the same scene shot at different moments, and comprises the following steps: decomposing the two images (Ipre and Icur) into sub-bands with different direction scales through the plurality of controllable pyramids, wherein each sub-band comprises two component amplitudes A and a phase phi; performing image registration by adjusting the phase difference between different sub-bands of the two images; reconstructing the different sub-bands after phase adjustment to obtain a final registered image; analyzing the change of the phase to obtain the real change of the scene.