CN112837409A - A method for reconstructing a three-dimensional human body using mirrors - Google Patents
A method for reconstructing a three-dimensional human body using mirrors Download PDFInfo
- Publication number
- CN112837409A CN112837409A CN202110146923.8A CN202110146923A CN112837409A CN 112837409 A CN112837409 A CN 112837409A CN 202110146923 A CN202110146923 A CN 202110146923A CN 112837409 A CN112837409 A CN 112837409A
- Authority
- CN
- China
- Prior art keywords
- dimensional
- human body
- mirror
- parameters
- reconstructing
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/80—Analysis of captured images to determine intrinsic or extrinsic camera parameters, i.e. camera calibration
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
- G06T2207/10012—Stereo images
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30196—Human being; Person
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computer Graphics (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method for reconstructing a three-dimensional human body by using a mirror, which inputs a picture containing the mirror, the human body and a human body mirror image, takes the human body mirror image as an image shot by a virtual camera, and reconstructs the respective orientation, position and three-dimensional attitude parameter of the human body and the mirror image by using the constraints of sharing, mirror symmetry, reprojection and the like of the three-dimensional attitude parameter of the human body, thereby effectively solving the problems of depth ambiguity and shielding of the human body attitude at a single visual angle. In addition, the invention also provides a method for estimating the normal direction of the mirror surface and the camera internal reference by using the human semantic key points, which can explicitly restrict the human orientation and further improve the reconstruction precision.
Description
Technical Field
The invention belongs to the technical field of computers, and particularly relates to a method for reconstructing a three-dimensional human body by using a mirror.
Background
The single-view three-dimensional human body reconstruction means reconstructing a three-dimensional posture of a human body from a picture obtained by shooting with a single camera. The most obvious challenge of this problem is the depth ambiguity of a single view angle, such as whether the arm is projected forward or backward to the image plane, from the perspective of a two-dimensional keypoint. The existing methods mostly utilize the prior information of the human body to reduce the ambiguity, such as punishing abnormal limb bending angles or using a discriminator to judge whether a certain posture of the human body is reasonable or not. These methods are effective for some simple body poses, but cannot handle complex situations. In addition, the current method cannot accurately recover the global position of the person in the camera system, and the consistency of the relative positions of the person and the person is lacking in a multi-person scene.
Disclosure of Invention
The invention aims to provide a method for reconstructing a three-dimensional human body by using a mirror, and the depth ambiguity of single-view-angle human body reconstruction can be effectively eliminated by using human semantic key points to estimate the normal direction of the mirror surface and camera parameters aiming at the defects of the prior art.
The purpose of the invention is realized by the following technical scheme:
according to a first aspect of the present invention, there is provided a method of reconstructing a three-dimensional human body using mirrors, comprising the steps of:
1. estimating the two-dimensional human body posture: inputting a picture containing a mirror, a human body and a human body mirror image, obtaining a two-dimensional enclosure frame by using a human body detector, taking the human body and the human body mirror image as two human bodies, and respectively detecting two-dimensional key points of the human body and the human body mirror image by using an existing two-dimensional human body posture estimator such as HRNet.
2. Sharing constraint of three-dimensional attitude parameters: considering that the person in the mirror and the real body have the same motion (after left-right interchange), the real body and the mirror image can be constrained to share the same set of three-dimensional attitude parameters. The three-dimensional posture parameters can be three-dimensional key point relative coordinates, rotation angles (such as shaft angles) between joints of the skeleton model, posture parameters of a human body model (such as SMPL) and the like. The parameters can be mapped to the human body through a predefined mapping relation, such as positive kinematics of a skeleton model or the mapping of the parameters of the human body model to the human body, so as to obtain the three-dimensional posture of the whole human body.
3. Optimizing: and constructing an objective function based on the three-dimensional attitude parameter sharing constraint, and optimizing to obtain a group of three-dimensional attitude parameters shared by the human body and the mirror image.
Further, the objective function also includes mirror symmetry constraints, represented by a parallel term and a perpendicular term, respectively.
The parallel items are specifically as follows: by utilizing mirror symmetry information, key points corresponding to the person and the mirror image (such as the right wrist corresponding to the mirror image of the left wrist of the person) are connected into a straight line, and any two connected straight lines are constrained to be parallel to each other, namely the cross product in the direction of the two straight lines is small enough.
The vertical terms are specifically as follows: furthermore, theoretically these straight lines should be parallel to the mirror normal, but considering that the mirror normal is unknown, while the midpoints of the straight lines are theoretically located on the mirror, the line connecting any two straight lines perpendicular to their midpoints is more loosely constrained, i.e. the dot product of the straight line direction and the line connecting the midpoints of the two straight lines should be sufficiently small.
Further, the method also comprises the steps of re-projection and prior constraint: both of these are common constraints. The reprojection constraint enables the projection of the three-dimensional key points to be consistent with the two-dimensional key points, and the prior constraint plays a role in regularization, so that the optimized human body does not deviate from the initial estimation too far, and the situation that the posture is too distorted is avoided. In the optimization process in the step 3, the three-dimensional attitude parameter sharing constraint, the mirror symmetry constraint, the reprojection and the prior constraint can be combined to construct an objective function.
Further, in the optimization process in the step 3, an optimization algorithm such as L-BFGS is adopted for optimization, and the final result is obtained through convergence.
According to a second aspect of the present invention, a method for estimating specular normal and camera parameters by using human semantic key points is provided, which comprises the following steps:
1. estimation of the mirror normal: different from the traditional method, the method faces great difficulty in finding the corresponding relation between the internal point and the external point of the mirror surface, the invention fully utilizes the semantic information of the human body, connects the previously estimated human body and the two-dimensional key points corresponding to the mirror image into a straight line, and calculates the intersection point among a plurality of straight lines formed by a plurality of groups of key points, wherein the intersection point is the vanishing point vertical to the direction of the mirror surface. Because a plurality of semantic key points exist on a human body, sufficient corresponding relation can be provided, and compared with common key points in a scene, vanishing points can be calculated more robustly. In the case of the known camera parameters, the normal to the mirror surface can be directly calculated from the vanishing point and the internal parameters. Once the specular normal is determined, an explicit specular normal constraint can be added to the previous optimization process, i.e., the connecting lines of all corresponding keypoints should be parallel to the estimated specular normal.
2. Estimation of camera internal parameters: in some cases, camera parameters are unknown, such as network pictures. Consider here the estimation of camera parameters with multiple orthogonal vanishing points. Assuming that the pixels of the camera are square, without distortion, and the principal point of the camera is at the center of the image, the two orthogonal vanishing points can recover the camera parameters. Furthermore, if the mirror surface is rectangular (referring to a large mirror, such as a dancing mirror, etc.), the intersection point of the mirror projected on the image screen can be calculated by two parallel edges of the mirror edge, so as to obtain one or two (mutually orthogonal) vanishing points parallel to the direction of the mirror surface and orthogonal to the vanishing points. The camera internal parameters can be directly calculated by using the two pairwise orthogonal vanishing points. It should be noted that if the camera parameters are unknown and there are no orthogonal vanishing points (which results in no estimation of camera parameters), the mirror normal constraint in the previous step is not applied.
The invention has the beneficial effects that: according to the invention, a new virtual visual angle is obtained from a single picture through the mirror surface, a new mirror symmetry constraint is provided, the depth ambiguity of monocular human body three-dimensional posture estimation can be eliminated, and the geometric characteristics of the mirror surface can be recovered at the same time. In addition, the invention also fully utilizes human body semantics and scene information, obtains a sufficient number of vanishing points from the picture to recover the mirror normal direction and camera parameters, and can be widely applied to network pictures.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
Fig. 1 is a schematic overall flow chart of an embodiment of the present invention, where (a) is an input image, (b) is an initial result, and (c) is a reconstruction result.
FIG. 2 is a mirror symmetry constraint diagram of an embodiment of the invention.
Fig. 3 is a schematic diagram of acquiring a plurality of orthogonal vanishing points by using human semantic key points and scene information according to an embodiment of the present invention.
Detailed Description
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in detail below.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, but the present invention may be practiced in other ways than those specifically described and will be readily apparent to those of ordinary skill in the art without departing from the spirit of the present invention, and therefore the present invention is not limited to the specific embodiments disclosed below.
As shown in fig. 1, an embodiment of the present invention provides a method for reconstructing a three-dimensional human body by using a mirror, which specifically includes the following steps:
1. a picture containing a mirror, a human body, and a mirror image of the human body is input, and the main body part of the human body in the image is visible outside the mirror inside the mirror, as shown in fig. 1 (a). A two-dimensional enclosure frame of all people in the image is detected by using a human body detector, and the people and mirror images of the people are taken as two people. And inputting each detected frame to a two-dimensional human body posture estimator, such as HRNet, so as to obtain two-dimensional key points of each person. In addition, the initial three-dimensional pose parameters may also be estimated for the detected human body using existing methods (e.g., SPIN), which is not necessary. The two-dimensional human body pose and the initial three-dimensional pose parameter estimation are shown in fig. 1 (b).
2. Considering that a person and its mirror image should have the same pose (undergo left-right transformation), the three-dimensional pose parameters θ corresponding to their two models should be shared. The three-dimensional attitude parameters can be relative coordinates of three-dimensional key points, rotation angles among joints of the skeleton model, attitude parameters of the human body model and the like. The three-dimensional attitude parameter theta for constraining the real human body is consistent with the three-dimensional attitude parameter theta' of the mirror image human body after left-right transformation.
θ′＝S(θ)
The three-dimensional posture parameters referred to herein can be mapped to obtain the three-dimensional posture of the human body through a predefined mapping. For skeletal models, the mapping is a process of positive kinematics; for a human model, the mapping is a regressor of model parameters to human patches.
3. As shown in fig. 1 (c), the line between two people can be used to establish a mirror symmetry constraint, as will be explained later in fig. 2. The three-dimensional human body and the two-dimensional key points on the image plane can be used for establishing the reprojection constraint. Further a priori terms are not shown in the figure. The mathematical expression of the reprojection term is:
where n is the projection operation, p is the Geman-McClure robust loss function, WiIs a two-dimensional key point, ciIs the corresponding confidence, R and T are the global rotation and translation, respectively, representing the orientation and position of the person, J (θ)iIs the mapping from the three-dimensional attitude parameter theta to the three-dimensional key points of the human body. The following is a mathematical expression for the prior term:
wherein
Is the initial three-dimensional attitude parameter, L, under the model's standard motionpThe purpose of (2) is to make the optimized parameters not deviate too far from the initial values, thereby playing a certain regularization role.
The whole optimization process is constructed as follows:
wherein L is2dAnd L'2dRepresenting the reprojection constraints of a person and its image, respectively. L ispAnd L'pRepresenting the a priori constraints of the person and its image, respectively. L issAnd LnAre constraints associated with the mirror, as will be described later. Lambda [ alpha ]p、λs、λnRepresenting the weight coefficients. Variables Θ ═ { θ, R, T } and Θ '═ θ', R ', T' } that need to be optimized belong to the real human body and its mirror image, respectively.
4. And adopting a certain optimization method, such as L-BFGS, to the established optimization representation to finally obtain optimized parameters. The reconstruction result is shown in fig. 1 (c).
FIG. 2 illustrates the mirror symmetry constraint, considering any pair of three-dimensional keypoints { J } on a person and its mirrori，JjAnd { J'i，J′jIs theoretically connected with a line JiJ′iAnd JjJ′jAre parallel and perpendicular to the mirror plane. Let p be taken into account that the geometrical information of the mirror is unknowniAnd pjRespectively, the midpoints of two connecting lines, which should be located at the mirror if the position estimate is accurate, thus relaxing the constraint that the line is perpendicular to the mirror as pipj. The mirror symmetry constraint is as follows:
wherein n isiAnd njRespectively represent a connecting line JiJ′iAnd JjJ′jIn the direction of (a). For any pair of keypoints, the one perpendicular and one parallel constraint can be constructed, and the sum is the whole mirror symmetry constraint. It is noted that the geometric parameters of the mirror are not explicitly required here. But if the normal to the mirror is obtainable (whether estimated or true), a mirror normal constraint can be added:
where n denotes the mirror normal.
Fig. 3 illustrates a process of acquiring a plurality of orthogonal vanishing points by using human semantic key points and scene information, wherein the upper graph is an enlargement of an image plane in the lower graph. In the picture, v0Is obtained from two-dimensional key points, v1Are obtained from the edge lines of the mirror. Specifically, the intersection point of the two-dimensional key point connecting lines corresponding to the human body and the mirror image thereof, namely v is calculated0The direction of the vanishing point is perpendicular to the mirror surface. Furthermore, considering that in many scenarios the whole body mirror is square, the intersection of the parallel edge lines at the mirror edges can provide one to two vanishing points parallel to the mirror plane, v in FIG. 31Is such a vanishing point, and is associated with v0Are orthogonal. From the knowledge of the projective geometry, three orthogonal vanishing points can recover camera parameters from a single picture, if the camera principal point is assumed to be at the image center, only two orthogonal vanishing points are needed. Therefore, the camera parameters can be recovered through the plurality of orthogonal vanishing points obtained here. Once the camera parameters are acquired, the mirror normal n may be passed through v0And obtaining a camera internal parameter K:
n＝K-1v0
wherein v is0Are homogeneous coordinates. The mirror normal can be added to the optimization process as shown in figure 2. It should be noted that this step is not necessary, and if a plurality of orthogonal vanishing points cannot be conveniently obtained, this constraint may not be added, and the optimization effect is not significantly affected.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments.
The foregoing is only a preferred embodiment of the present invention, and although the present invention has been disclosed in the preferred embodiments, it is not intended to limit the present invention. Those skilled in the art can make numerous possible variations and modifications to the present teachings, or modify equivalent embodiments to equivalent variations, without departing from the scope of the present teachings, using the methods and techniques disclosed above. Therefore, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention are still within the scope of the protection of the technical solution of the present invention, unless the contents of the technical solution of the present invention are departed.
Claims (10)
1. A method for reconstructing a three-dimensional body using mirrors, the method comprising the steps of:
(1) inputting a picture containing a mirror, a human body and a human body mirror image to obtain a human body two-dimensional surrounding frame, taking the human body and the human body mirror image as two persons, and respectively detecting two-dimensional key points of the two persons;
(2) estimating three-dimensional attitude parameters of a real human body and a mirror image, wherein the parameters can obtain the three-dimensional attitude of the human body through predefined mapping relation transformation, and restrict the real human body and the mirror image to share a group of three-dimensional attitude parameters;
(3) and constructing an objective function for optimization based on the sharing constraint of the three-dimensional attitude parameters, and finally obtaining the optimized human body three-dimensional attitude parameters.
2. The method for reconstructing a three-dimensional human body by using a mirror as claimed in claim 1, wherein the step 1 uses a human body detector to obtain a two-dimensional enclosure frame of the human body; two-dimensional keypoints are detected using a two-dimensional human pose estimator.
3. A method for reconstructing a three-dimensional body using mirrors as claimed in claim 1, wherein the sharing of the three-dimensional pose parameters in step 2 is as follows: and (4) replacing the three-dimensional attitude parameters of the real human body according to the bilateral symmetry relationship of the human body, and constraining the three-dimensional attitude parameters to be consistent with the three-dimensional attitude parameters of the mirror image.
4. The method of claim 1, wherein the optimization process of step 3 further comprises mirror symmetry constraints, wherein there is a parallel term and a perpendicular term;
parallel terms: by utilizing mirror symmetry information, a person and key points corresponding to the mirror image are connected into a straight line, and any two connected straight lines are constrained to be parallel to each other, namely the cross product in the two straight line directions is small enough;
vertical terms: the point product constraining any two connecting lines perpendicular to their midpoints, i.e. the direction of the line and the line connecting the midpoints of the two lines, should be small enough.
5. The method for reconstructing a three-dimensional human body by using a mirror as claimed in claim 1, wherein the optimization process in step 3 further comprises a reprojection constraint, the reprojection constraint enables the projection of the three-dimensional key points to be consistent with the two-dimensional key points, and the mathematical expression is as follows:
where n is the projection operation, p is the Geman-McClure robust loss function, WiIs a two-dimensional key point, ciIs the corresponding confidence, R and T are the global rotation and translation, respectively, representing the orientation and position of the person, J (θ)iIs the mapping from the three-dimensional attitude parameter theta to the three-dimensional key points of the human body.
6. The method for reconstructing a three-dimensional human body by using a mirror as claimed in claim 1, wherein the optimization process in step 3 further comprises a priori constraint, the priori constraint has a regularization effect, so that the optimized parameters do not deviate too far from the initial estimation, and the mathematical expression is as follows:
wherein
Is the initial three-dimensional attitude parameter under the standard action of the model.
7. The method for reconstructing a three-dimensional human body by using a mirror as claimed in claim 1, wherein the optimization process in step 3 further comprises mirror normal constraint: connecting the two-dimensional key points corresponding to the human body and the mirror image into a straight line, wherein the connecting lines of all the corresponding key points are parallel to the normal direction of the mirror surface.
8. Method for reconstructing a three-dimensional body using mirrors according to claim 7, characterized in that said estimation of the normal direction of the mirror surfaces is in particular: and calculating the intersection point of a plurality of straight lines formed by connecting a plurality of groups of two-dimensional key points, wherein the intersection point is a vanishing point vertical to the mirror surface direction, and the vanishing point can be combined with camera internal parameters to calculate the normal direction of the mirror surface.
9. The method for reconstructing a three-dimensional human body by using a mirror as claimed in claim 8, wherein the estimation of the camera parameters is specifically as follows: if the mirror surface is rectangular, calculating the intersection point of the mirror edge on the image plane through two parallel edges of the mirror edge, thereby obtaining one or two vanishing points parallel to the mirror surface direction and orthogonal to the vanishing points vertical to the mirror surface direction, and calculating the camera internal parameter by using the two-two orthogonal vanishing points.
10. A method for reconstructing a three-dimensional body using mirrors as claimed in any one of claims 1 to 9, wherein in the optimization procedure of step 3, an optimization algorithm L-BFGS is used for optimization, and convergence is performed to obtain the final result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110146923.8A CN112837409B (en) | 2021-02-02 | 2021-02-02 | Method for reconstructing three-dimensional human body by using mirror |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110146923.8A CN112837409B (en) | 2021-02-02 | 2021-02-02 | Method for reconstructing three-dimensional human body by using mirror |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112837409A true CN112837409A (en) | 2021-05-25 |
| CN112837409B CN112837409B (en) | 2022-11-08 |
Family
ID=75932565
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110146923.8A Active CN112837409B (en) | 2021-02-02 | 2021-02-02 | Method for reconstructing three-dimensional human body by using mirror |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112837409B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113887319A (en) * | 2021-09-08 | 2022-01-04 | 北京达佳互联信息技术有限公司 | Three-dimensional attitude determination method and device, electronic equipment and storage medium |
| CN114627491A (en) * | 2021-12-28 | 2022-06-14 | 浙江工商大学 | Single three-dimensional attitude estimation method based on polar line convergence |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CA2716257A1 (en) * | 2010-10-01 | 2012-04-01 | Martin Habbecke | System and method for interactive painting of 2d images for iterative 3d modeling |
| US8165403B1 (en) * | 2010-11-19 | 2012-04-24 | Mitsubishi Electric Research Laboratories, Inc. | Method and system for determining poses of specular objects |
| CN106447763A (en) * | 2016-07-27 | 2017-02-22 | 扬州大学 | Face image three-dimensional reconstruction method for fusion of sparse deformation model and principal component regression algorithm |
| CH711803A2 (en) * | 2015-11-23 | 2017-05-31 | Kiodia Sàrl | Interactive mirror surface interaction method |
| US20180150962A1 (en) * | 2016-11-30 | 2018-05-31 | Canon Kabushiki Kaisha | Image registration method |
| CN111126269A (en) * | 2019-12-24 | 2020-05-08 | 京东数字科技控股有限公司 | Three-dimensional target detection method, device and storage medium |
| CN111583386A (en) * | 2020-04-20 | 2020-08-25 | 清华大学 | Multi-view Human Pose Reconstruction Method Based on Label Propagation Algorithm |
| CN111798486A (en) * | 2020-06-16 | 2020-10-20 | 浙江大学 | Multi-view human motion capture method based on human motion prediction |
| US20210012558A1 (en) * | 2018-08-28 | 2021-01-14 | Tencent Technology (Shenzhen) Company Limited | Method and apparatus for reconstructing three-dimensional model of human body, and storage medium |
-
2021
- 2021-02-02 CN CN202110146923.8A patent/CN112837409B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CA2716257A1 (en) * | 2010-10-01 | 2012-04-01 | Martin Habbecke | System and method for interactive painting of 2d images for iterative 3d modeling |
| US8165403B1 (en) * | 2010-11-19 | 2012-04-24 | Mitsubishi Electric Research Laboratories, Inc. | Method and system for determining poses of specular objects |
| CH711803A2 (en) * | 2015-11-23 | 2017-05-31 | Kiodia Sàrl | Interactive mirror surface interaction method |
| CN106447763A (en) * | 2016-07-27 | 2017-02-22 | 扬州大学 | Face image three-dimensional reconstruction method for fusion of sparse deformation model and principal component regression algorithm |
| US20180150962A1 (en) * | 2016-11-30 | 2018-05-31 | Canon Kabushiki Kaisha | Image registration method |
| US20210012558A1 (en) * | 2018-08-28 | 2021-01-14 | Tencent Technology (Shenzhen) Company Limited | Method and apparatus for reconstructing three-dimensional model of human body, and storage medium |
| CN111126269A (en) * | 2019-12-24 | 2020-05-08 | 京东数字科技控股有限公司 | Three-dimensional target detection method, device and storage medium |
| CN111583386A (en) * | 2020-04-20 | 2020-08-25 | 清华大学 | Multi-view Human Pose Reconstruction Method Based on Label Propagation Algorithm |
| CN111798486A (en) * | 2020-06-16 | 2020-10-20 | 浙江大学 | Multi-view human motion capture method based on human motion prediction |
Non-Patent Citations (4)
| Title |
|---|
| ABDULLAH AKAY ET AL.: "3D Reconstruction with Mirrors and RGB-D Cameras", 《2014 INTERNATIONAL CONFERENCE ON COMPUTER VISION THEORY AND APPLICATIONS》 * |
| HAI LI ET AL.: "Saliency Guided Subdivision for Single-View Mseh Reconstruction", 《2020 INTERNATIONAL CONFERENCE ON 3D VISION（3DV）》 * |
| 杨业: "基于多平面镜单相机三维扫描技术", 《中国优秀硕士学位论文全文数据库（电子期刊）信息科技辑》 * |
| 王巧丽等: "虚拟试衣中个性化三维人体建模的研究进展", 《纺织导报》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113887319A (en) * | 2021-09-08 | 2022-01-04 | 北京达佳互联信息技术有限公司 | Three-dimensional attitude determination method and device, electronic equipment and storage medium |
| CN114627491A (en) * | 2021-12-28 | 2022-06-14 | 浙江工商大学 | Single three-dimensional attitude estimation method based on polar line convergence |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112837409B (en) | 2022-11-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wong et al. | Unsupervised depth completion from visual inertial odometry | |
| CN113538667B (en) | Dynamic scene light field reconstruction method and device | |
| CN110335343B (en) | Human body three-dimensional reconstruction method and device based on RGBD single-view-angle image | |
| CN101916454B (en) | Method for reconstructing high-resolution human face based on grid deformation and continuous optimization | |
| Taylor et al. | Non-rigid structure from locally-rigid motion | |
| CN106570507B (en) | Multi-view-angle consistent plane detection and analysis method for monocular video scene three-dimensional structure | |
| CN106204718B (en) | A kind of simple and efficient 3 D human body method for reconstructing based on single Kinect | |
| US9235928B2 (en) | 3D body modeling, from a single or multiple 3D cameras, in the presence of motion | |
| Wu et al. | Fusing multiview and photometric stereo for 3d reconstruction under uncalibrated illumination | |
| CN108416840A (en) | A Dense Reconstruction Method of 3D Scene Based on Monocular Camera | |
| CN107240129A (en) | Object and indoor small scene based on RGB D camera datas recover and modeling method | |
| CN119180908A (en) | Gaussian splatter-based laser enhanced visual three-dimensional reconstruction method and system | |
| Ye et al. | Free-viewpoint video of human actors using multiple handheld kinects | |
| CN111862299A (en) | Human body three-dimensional model construction method, device, robot and storage medium | |
| CN113313828A (en) | Three-dimensional reconstruction method and system based on single-picture intrinsic image decomposition | |
| CN102768767B (en) | Online three-dimensional reconstructing and locating method for rigid body | |
| Li et al. | 3D human avatar digitization from a single image | |
| CN118037965B (en) | Human body 3D gesture analysis method based on automatic variation correction under multi-eye vision | |
| CN118196332A (en) | A multi-view human body surface reconstruction method and system | |
| CN112837409A (en) | A method for reconstructing a three-dimensional human body using mirrors | |
| Chen et al. | LiftImage3D: Lifting any single image to 3D Gaussians with video generation priors | |
| Yookwan et al. | Multimodal fusion of deeply inferred point clouds for 3D scene reconstruction using cross-entropy ICP | |
| Remondino et al. | 3D reconstruction of human skeleton from single images or monocular video sequences | |
| CN119478173B (en) | Three-dimensional scene novel view synthesis method based on matched rays | |
| CN111914790B (en) | Real-time human rotation angle recognition method in different scenarios based on dual cameras |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |