30results about How to "Fast 3D reconstruction" patented technology

The invention discloses a method and a system of three-dimensional reconstruction of a shape structure of trees in a leaved state. The method comprises acquiring three-dimensional point cloud data of trees and carrying out preprocessing and simplifying processing to obtain simplified three-dimensional cloud data set of the trees, carrying out organ segmentation to the simplified three-dimensional point cloud data set of the trees to obtain the three-dimensional point cloud data of each organ, wherein the three-dimensional point cloud data of each organ comprise the three-dimensional point cloud data of a leaf organ, the three-dimensional point cloud data set of a main trunk, the three-dimensional point cloud data set of a main branch and the three-dimensional point cloud data set of a sprout, extracting the frameworks of limbs of the trees according to the three-dimensional point cloud data set of the main trunk, the three-dimensional point cloud data set of the main branch and the three-dimensional point cloud data set of the sprout and optimizing the frameworks of the limbs of the trees, carrying out meshing to the frameworks of the limbs of the trees to generate a three-dimensional mesh model of the limbs of the trees, classifying each limb of the frameworks of the limbs of the trees, and carrying out canopy three-dimensional reconstruction. The fast three-dimensional reconstruction of the shape structure of the trees in the leaved state is achieved, and therefore a method and a high-precision basic model support are provided for the structural analysis of the canopies of the trees, the evaluation of the physiological and ecological characteristics and the like.

The invention discloses a light field microscopic three-dimensional reconstruction method and device based on a deep learning algorithm, and the method comprises the steps: building a light field microscopic imaging system; simulating the light field microscopic imaging system to obtain a point spread function of the imaging system; generating a simulation sample body distribution data set; generating a simulation light field data set through a point spread function and corresponding sample data, and correcting possibly generated noise and a background model; constructing a deep convolutionalneural network, taking the simulation light field data as the input of the network, taking the simulation sample body distribution data as the output of the network, and training the network accordingto a microscopic sample design loss function; and after the training of the model is completed, inputting the light field microscopic data to be reconstructed into the model for testing, and obtaining a prediction value of the distribution data of the corresponding sample body. According to the method, rapid, high-resolution and less-artifact light field data three-dimensional reconstruction canbe realized while the advantage of rapid acquisition of three-dimensional information by the light field is maintained.

This invention relates to computer graphics and mode identification technique field and in detail to use point recreation for super large scale of medical three-dimensional image visualization and comprises the following steps: first to divide the cared part from the medical two-dimensional slice; second to recreate based on point to extract the cared organic three-dimensional surface and to store it in form of point; third to draw with three-dimension to display the extracted surface mold with the drawing method based on points.

The invention relates to a method for the three-dimensional reconstruction of a straight-line optical flow field based on an intelligent optimization algorithm, which is characterized by comprising the following steps of: shooting at least three continuous frames of images of an object in a scene by a fixed camera to form an image sequence, extracting the straight-line profile of the object in the scene and taking a straight line as a characteristic; under a perspective projection model, eliciting a relationship between the motion parameter of the straight-line profile of a moving rigid object in a 3D space and a two-dimensional straight-line optical field on a projection image of the moving rigid object, expressing the relationship by a straight-line optical flow equation group and constructing a physical model and a mathematical model of a three-dimensional reconstruction algorithm. The invention has the advantages of realizing the three-dimensional reconstruction and the moving target division of an object with a straight-line profile in a scene quickly, solving the problems of robustness of the three-dimensional reconstruction algorithm and reconstruction of a shielding boundary of the scene effectively and enhancing the accuracy of three-dimensional reconstruction effectively.

The present invention belongs to the field of image processing and mode recognizing, and relates to a quick three-dimensional reconstructing method applied for detecting fabric defect. The method comprises a BCCSL (Binocular Camera Color SstuctureLight) method which is used for obtaining the three-dimensional coordinate of the fabric ''framework'' and is provided by the present invention and an SFTS (Shape From Two Shading) that is used for the ''detail'' information of the fabric and is provided by the invention. The fabric defect three-dimensional reconstructing method designed by the invention can effectively settle the precision and speed of three-dimensional reconstructing, satisfy the requirement of fabric on-line detection, make preparation for the subsequent fabric defect three-dimensional recognition and increase the quality and grade of Chinese textile.

The invention belongs to the field of image processing and pattern recognition, and relates to an RGBPSP (red green blue phase shift profilometry) three-dimensional color reconstruction method applied to online detection for fabric defects. In the RGBPSP three-dimensional color reconstruction method, a color camera and a color light source projecting apparatus are adopted, the color light source projecting apparatus does not need to project a series of light information to a measured object but always projects fixed light information to the measured fabric, accurate phase information of each measurement point of a measured scene is obtained by the aid of a phase correcting method, and accordingly three-dimensional coordinate information of the measured scene is obtained. The RGBPSP three-dimensional color reconstruction method has the advantages that the proper three-dimensional reconstruction precision and the proper three-dimensional reconstruction speed can be effectively realized, requirements on online detection for the fabric are met, follow-up three-dimensional fabric defect identification is prepared, and the quality and the grade of Chinese fabrics are improved.

The invention provides an electromagnetic emission point positioning method based on an urban environmental physical model. The method comprises the following steps: 1) establishing an urban digital three-dimensional model; 2) based on the principle of a mirror image method, establishing a model of the urban electromagnetic environment by utilizing a ray tracing kernel function in a toolbox of Matlab; 3) setting an electromagnetic emission point in the established urban electromagnetic environment model, collecting corresponding DC data, and training a pre-constructed machine learning model byusing the DC data as a sample to obtain a trained model; and 4) measuring real DC data at a set position in a real city environment, and inputting the real DC data into the trained model to identifythe position of the electromagnetic emission point. By applying the embodiment of the invention, the requirement of rapid modeling of the urban electromagnetic environment can be well met.

The invention relates to a monocular polarization three-dimensional reconstruction method, comprising the following steps: (a) obtaining light intensity images of an object and polarization images ofseveral objects with different polarization angles; (b) obtaining a first object surface normal according to that object polarization image; (c) obtaining a second object surface normal line accordingto that light intensity image of the object by use an SFS algorithm; (d) correct that first object surface normal using the second object surface normal to obtain a third object surface normal; (e) realizing that three-dimensional reconstruction according to the surface normal of the third object. The surface normal of the object obtained by the SFS method is used for correcting the surface normal obtained by the polarization information, thereby realizing the monocular three-dimensional reconstruction, reducing the complexity of the system structure and improving the operation speed.

The invention discloses a teleoperation method for an unstructured environment based on 5G + AR. The method comprises the following steps: firstly, carrying out rapid three-dimensional reconstructionon an unstructured environment where a robot is located, and mapping a reconstructed scene into a virtual reality space; and then, capturing a gesture action of a user, establishing a virtual model ofthe hand of the user, and mapping the gesture action into the virtual reality space, thereby realizing interaction between the user and the virtual object; finally, introducing an auxiliary technology to help the user finish teleoperation work more accurately and quickly.

The invention relates to a GPU (Graphics Processing Unit) frame based three-dimensional reconstruction method of a dotted line optical flow field, comprising the following steps of: firstly building a three-dimensional motion and structure uniform model based on pixel point and profile straight-line optical flow field reconstruction; computing point optical flows at a low resolution ratio, and computing straight-line optical flows at a high resolution ratio; setting the straight-line expression dimensions of an image to be consistent with the dimensions of pixel points so as to unify the pixel point and profile straight-line optical flow field reconstruction to one three-dimensional reconstruction computation model; and finally adopting a VC++language programming implementation algorithm. The invention has the technical effects of being capable of fast completing the three-dimensional reconstruction, accurately processing the details of an object and integrally realizing the three-dimensional reconstruction of the object.

The invention discloses a multi-view three-dimensional point set recovery method based on a monocular camera. The method specifically comprises the following steps: estimating internal parameters of amonocular camera, calculating external parameters of the camera, collecting and preprocessing multiple viewpoints of an image, extracting feature points of the processed image, matching the feature points, and solving an initial three-dimensional point cloud according to a matched feature point set; and calculating, denoising and segmenting the initial point cloud to obtain processed three-dimensional point cloud data of each independent object, and performing fitting processing on the obtained point cloud data to realize reconstruction of each object in multiple views acquired by the monocular camera. According to the multi-view three-dimensional point set recovery method based on the monocular camera, recovery of the three-dimensional point set of the object in the scene can be completed by using a small number of images.

A facet-based complex scene three-dimensional reconstruction method is characterized by including: firstly, obtaining intrinsic and extrinsic parameters and distortion coefficients in a depth camera and a color camera by three-dimensional calibration; secondly, calculating three-dimensional information of a complex scene by the depth camera according to the principle of triangulation; thirdly, defining a point model as a facet, with each value as an attribute of the point model including the attributes such as normal vector and radius; fourthly, subjecting each facet to surface 'internal, external and intersection' relation test with another facet; fifthly, obtaining confidence neighborhood of the facets which are judged intersected, calculating boundaries of the most adjacent facets on the point model, and resampling the facets in the confidence neighborhood; sixthly, three-dimensionally drawing the complex scene according to results of Boolean operation in the step 5. The facet-based complex scene three-dimensional reconstruction method is highly timely and available for effectively extracting three-dimensional point cloud data of the complex scene.

The invention relates to the technology of structured light measurement and specifically relates to a three-dimensional visual sense reconstruction method, comprising following steps: establishing a three-dimensional measurement system; projecting a sine projection pattern with phase shift by a projector and capturing modulated sine pattern on a measured object surface with a camera in real time by use of a multi-frequency phase measuring profilometry structured light scan method, and calculating measured object surface phase information according to the multi-frequency phase measuring profilometry phase developing method; calculating the relative depth of the measured object by use of torsion resistance which changes with the surface depth of the measured object for the phase pattern in space, and realizing three-dimensional reconstruction according to the relative depth. The method can be applied in situations where complicated calibration cannot be realized due to condition limiting, calibration is difficult to carried out, or only low precision three-dimensional coordinate image reconstruction is required to quickly realize three-dimensional reconstruction; the method gives a solution to the problem that the present three-dimensional reconstruction method can not be carried out with insufficient calibrating data; the method has wide applications in the field of archaeology and three-dimensional data visualization.

The invention relates to a video image processing method and device, electronic equipment and a storage medium, and the method comprises the steps: inputting a first frame image in a target video into a three-dimensional reconstruction network and a video frame coding network, and obtaining a three-dimensional reconstruction result of a target object in the first frame image and the image features of the first frame image; inputting the image feature of the i-th frame of image in the target video and the three-dimensional reconstruction result corresponding to the i-th frame of image into a time sequence feature extraction network to obtain the time sequence feature of the i-th frame of image; inputting the (i + 1)-th frame of image in the target video into the video frame coding network to obtain image features of the (i + 1)-th frame of image; generating a three-dimensional reconstruction result corresponding to the (i + 1)-th frame of image based on the image feature of the (i + 1)-th frame of image and the time sequence feature of the i-th frame of image; and updating the value of i to i + 1, and repeatedly executing the step of inputting the time sequence feature extraction network to generate the three-dimensional reconstruction result corresponding to the (i + 1)-th frame of image until i = N. According to the invention, the precision and efficiency of object three-dimensional reconstruction can be improved.

The invention relates to a device for extracting three-dimensional phenotypic information of small plants and seedling plants. The device comprises an upper computer control center, a precise motion controller and a visible light imaging module. The imaging module contains an image acquisition module, a loading rotary table and a visual light source. The precise motion controller has a single axismotion control function and can be communicated with the upper computer. The upper computer is mainly used for communicating with a lower computer and extracting three-dimensional information of theplant images. The device utilizes a machine vision technology and mechanical automation technology, realizes fast and non-destructive extraction of three-dimensional phenotypic information of small plants and provides a fast and efficient scientific and technological means for researching phenotypic characters of small plants or seedling plants for researchers in the related fields.

The invention is applicable to the technical field of optical three-dimensional digital imaging, and provides a three-dimensional digital imaging method and device for a folded phase based on phase mapping. The three-dimensional digital imaging method comprises the steps of acquiring an image containing information of a measured object, and calculating a folded phase value of each pixel point on imaging planes of two imaging devices according to the image; calculating coordinates of n candidate space three-dimensional points corresponding to each pixel point via phase mapping by using the folded phase value of each pixel point on the imaging plane of the first imaging device and preset first calibration data; re-projecting the n candidate space three-dimensional points to the imaging plane of the second imaging device to obtain n candidate projection points, and interpolating folded phase values of the candidate projection points; and screening out correct space three-dimensional points from the candidate space three-dimensional points by taking the folded phase value of each pixel point on the imaging plane of the first imaging device and the folded phase values of the n candidate projection points corresponding to the pixel point as a basis so as to realize three-dimensional digital reconstruction. The three-dimensional reconstruction time is shortened according to the method provided by the invention.

The invention relates to a rapid three-dimensional modeling method based on voxels and a system thereof. The N RGBD cameras with the number larger than 1 are sequentially placed around a storage platform at intervals to shoot a target object, and then an image processing unit, a voxel cutting unit, a voxel fusion unit, a voxel filling unit, a grid extraction unit and a texture generation unit are arranged. A vivid three-dimensional reconstruction model of the target object is obtained. The problem that information is insufficient when three-dimensional reconstruction is carried out on the target object under a limited view angle is solved, and meanwhile rapid three-dimensional reconstruction is carried out on the target object. The method is rapid and accurate.

The invention discloses a fire scene positioning method and system based on an unmanned aerial vehicle. The method comprises the following steps: acquiring data of physical object markers or special buildings arranged in multiple places around the fire scene to serve as markers, and establishing a self-determining coordinate system; manually pointing out two of the markers from pictures shot by the unmanned aerial vehicle, combining the click positions and an attitude angle to calculate the position of the unmanned aerial vehicle under the self-determining coordinate system so as to perform unmanned aerial vehicle positioning; manually pointing out the position of the same key point according to the two pictures shot by the unmanned aerial vehicle, combining the click position, the attitude angle and two shooting positions of the key point, calculating the position coordinate of the key point in the self-determining coordinate system, and completing positioning of the fire scene relative to the key point, wherein the position of the unmanned aerial vehicle is the shooting position. According to the method disclosed by the invention, the fire scene is accurately and rapidly positioned, the key point and buildings are marked, intuitive site conditions are provided for fire scene commanders, the fire scene commanders are assisted in making disaster relief tactic decisions, and theefficiency is greatly improved.

The invention discloses a method and a system of three-dimensional reconstruction of a shape structure of trees in a leaved state. The method comprises acquiring three-dimensional point cloud data of trees and carrying out preprocessing and simplifying processing to obtain simplified three-dimensional cloud data set of the trees, carrying out organ segmentation to the simplified three-dimensional point cloud data set of the trees to obtain the three-dimensional point cloud data of each organ, wherein the three-dimensional point cloud data of each organ comprise the three-dimensional point cloud data of a leaf organ, the three-dimensional point cloud data set of a main trunk, the three-dimensional point cloud data set of a main branch and the three-dimensional point cloud data set of a sprout, extracting the frameworks of limbs of the trees according to the three-dimensional point cloud data set of the main trunk, the three-dimensional point cloud data set of the main branch and the three-dimensional point cloud data set of the sprout and optimizing the frameworks of the limbs of the trees, carrying out meshing to the frameworks of the limbs of the trees to generate a three-dimensional mesh model of the limbs of the trees, classifying each limb of the frameworks of the limbs of the trees, and carrying out canopy three-dimensional reconstruction. The fast three-dimensional reconstruction of the shape structure of the trees in the leaved state is achieved, and therefore a method and a high-precision basic model support are provided for the structural analysis of the canopies of the trees, the evaluation of the physiological and ecological characteristics and the like.

The invention discloses a human skeleton three-dimensional model acquisition method and system based on a two-dimensional medical image, and the method comprises the steps: carrying out the discretization of a three-dimensional model of a skeleton, obtaining a point cloud file describing the surface shape of the skeleton, and enabling the point cloud file and information obtained in a corresponding X-ray plain film to form a training set; all the point cloud files in the training set are converted into the same coordinate system through a registration process, the corresponding relation among all the point cloud file points in the training set is established, and the training set is updated; deformation vectors between the remaining point cloud files and the reference shape corresponding points are calculated, the deformation vectors of all points in the point cloud form a point cloud deformation domain, and an average deformation domain between the point cloud and the reference shape point cloud is calculated; establishing a posterior probability model by using Bayesian inference; and inputting information which is not included in the X-ray plain film in the training set, then outputting a point cloud file of the reconstruction model, and generating a skeleton three-dimensional model. The time and economic cost for acquiring the skeleton three-dimensional model are reduced, and rapid three-dimensional reconstruction of the skeleton is realized.

The invention discloses a rapid three-dimensional reconstruction system and method. The system comprises more than three camera devices and a host, wherein the camera devices surround a reconstructed object and are erected in a non-parallel manner; the camera devices can obtain color images and depth images at the same time, and the host receives the images collected by the camera devices and carries out registration to generate point clouds. The method comprises the steps of camera calibration, image acquisition, image registration, point cloud generation and point cloud registration. According to the invention, three-dimensional reconstruction can be rapidly carried out on the tested object through a non-contact method, the tested object is not damaged, the method is rapid and accurate, and online real-time three-dimensional reconstruction can be realized.

The present invention belongs to the field of image processing and mode recognizing, and relates to a quick three-dimensional reconstructing method applied for detecting fabric defect. The method comprises a BCCSL (Binocular Camera Color SstuctureLight) method which is used for obtaining the three-dimensional coordinate of the fabric ''framework'' and is provided by the present invention and an SFTS (Shape From Two Shading) that is used for the ''detail'' information of the fabric and is provided by the invention. The fabric defect three-dimensional reconstructing method designed by the invention can effectively settle the precision and speed of three-dimensional reconstructing, satisfy the requirement of fabric on-line detection, make preparation for the subsequent fabric defect three-dimensional recognition and increase the quality and grade of Chinese textile.

A monocular panoramic three-dimensional vision sensor focusing on objects comprises a visual detection box and a micro processor, wherein the visual detection box is used for shooting a to-be-reconstructed object, and the micro processor is used for carrying out image processing to the to-be-reconstructed object, reconstructing the object in a three-dimensional mode and outputting reconstruction results. The micro processor comprises an image reading module, a sensor calibration module, an image processing module, a three-dimensional reconstruction module and a result output module, and the image processing module is used for segmenting an image taken directly about the to-be-reconstructed object and a reflected image from an ellipsoidal mirror about the to-be-constructed object from a picture, carrying out perspective projection transformation processing to the reflected image from the ellipsoidal mirror about the to-be-reconstructed object, and obtaining a top view of a horizontal projection plane and four types of basic views, namely a main view, left side view, a right side view and a rear view. The three-dimensional reconstruction module is used for reconstructing a real shape of the to-be-reconstructed object based on corresponding relations, including 'the length in the main view needs to be aligned with the length in the top view, the height in the main view needs to be aligned with the height in the left side view and the height in the left side view, and the width in the top view needs to be the same as the width of the left side view and the width in the right side view', among the basis views. The monocular panoramic three-dimensional vision sensor focusing on the objects is high in practicability, low in calculation resource consumption, simple and reliable in hardware, great in real-time processing ability and convenient to maintain and use.

The invention provides a three-dimensional reconstruction algorithm of the inner wall of the pipeline based on the monocular image sequence. The algorithm is based on the monocular image sequence, determines the coordinates of the center of the pipeline through the monocular image, determines the straight line on this basis, and merges the straight line into a two-dimensional image. Obtain an image of the inner wall of the pipeline. Based on the image of the inner wall of the pipeline, using the principle of triangulation, the elevation of the inner wall of the pipeline is calculated through the difference between the two images of the inner wall of the pipeline, and the three-dimensional reconstruction of the inner wall of the pipeline is realized. The invention has the characteristics of strong adaptability, simple realization and fast rebuilding speed.