CN102129711A - GPU (Graphics Processing Unit) frame based three-dimensional reconstruction method of dotted line optical flow field - Google Patents
GPU (Graphics Processing Unit) frame based three-dimensional reconstruction method of dotted line optical flow field Download PDFInfo
- Publication number
- CN102129711A CN102129711A CN 201110072376 CN201110072376A CN102129711A CN 102129711 A CN102129711 A CN 102129711A CN 201110072376 CN201110072376 CN 201110072376 CN 201110072376 A CN201110072376 A CN 201110072376A CN 102129711 A CN102129711 A CN 102129711A
- Authority
- CN
- China
- Prior art keywords
- flow field
- optical flow
- dimensional reconstruction
- straight line
- dimensional
- 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.)
- Pending
Links
Landscapes
- Image Generation (AREA)
Abstract
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.
Description
Technical field
The present invention relates to a kind of dotted line optical flow field three-dimensional rebuilding method, relate in particular to a kind of dotted line optical flow field three-dimensional rebuilding method based on the GPU framework.
Background technology
Real world is three-dimensional, owing to be subjected to the restriction of scientific technological advance level, we can access and can be 2-D datas to it overwhelming majority who effectively handles and analyze, and make the three-dimensional information of a large amount of objects lose for a long time.Along with development of computer, people obtain from traditional two-dimensional image information, have turned to three-dimensional image.Three-dimensional reconstruction can obtain profile, colouring information and the texture information of object fast, is the effective way that people understand external environment.In recent years, the fast development of GPU technology, the OpenGL display technique is increasingly mature, and the fast processing image accurately obtains the motion of 3D object and structure and true true feeling 3D scene becomes possibility.
From two dimensional image optical flow field reconstruction 3D motion and structure is a challenging research.3D based on the pixel optical flow field rebuilds the details that can obtain object, and calculated amount is very big, almost can not handle in real time.Rebuild the profile information that 3D motion and structure can obtain object based on the straight line optical flow field.Because the parameter dimension that pixel and image straight line are expressed is different, therefore can rebuilds the unified model of object point information and profile information simultaneously and not set up as yet.In addition, former three-dimensional reconstruction calculates and all is to use CPU, and increasingly mature along with the GPU technology can overcome based on the three-dimensional reconstruction of GPU framework technology and to rebuild the big difficulty of calculated amount, realizes that real-time three-dimensional rebuilds.
Summary of the invention
The object of the present invention is to provide a kind of dotted line optical flow field three-dimensional rebuilding method, can finish three-dimensional reconstruction fast based on the GPU framework; Energy is the details of handled object accurately; Can rebuild by complete realization object dimensional.
The present invention is achieved like this, method is: at first set up the unified model based on pixel and motion of profile straight line optical flow field reconstruction of three-dimensional and structure, the image straight line is expressed dimension to be arranged to consistent with the pixel dimension, therefore can rebuild pixel and image straight line optical flow field unified under a three-dimensional reconstruction computation model, for outline of straight line, use above-mentioned Model Reconstruction, the curve that curvature is bigger can be similar to match with the segmentation straight line.Design derivation algorithm then, with rotation matrix to be asked
And translation vector
Be arranged to unknown quantity, pixel and image straight line optical flow field are known quantity, because the quantity of optical flow field has surpassed unknown quantity number to be asked, so this model is an overdetermined equation group, can find the solution with least square method.Next sets the three-dimensional reconstruction strategy, low resolution (for example 5*5 pixel) calculation level light stream down at initial yardstick, increase along with the number of plies, resolution is more and more higher, the result of calculation that obtains is added the initial value that initial value regards to descend one deck to calculate, the substitution next one is more put and the calculating of straight line light stream three-dimensional reconstruction on the high resolution graphics layer, so repeatedly, and till the three-dimensional reconstruction that reaches on the full resolution image number of plies.At last,, adopt the image operation of GPU framework technology to quicken the computing velocity of three-dimensional reconstruction, use OpenGL to show reconstructed results because the image operation amount is big.The present invention adopts VC++ Programming with Pascal Language implementation algorithm.
Technique effect of the present invention is: can finish three-dimensional reconstruction fast; Energy is the details of handled object accurately; Can rebuild by complete realization object dimensional.
Embodiment
The present invention is achieved in that the unified model of at first setting up based on pixel and motion of profile straight line optical flow field reconstruction of three-dimensional and structure.Calculation level light stream under low resolution, calculated line light stream under high resolving power.The image straight line is expressed dimension be arranged to consistently, therefore can rebuild pixel and image straight line optical flow field unified under a three-dimensional reconstruction computation model with the pixel dimension.The three-dimensional reconstruction of while considered pixel point and image straight line optical flow field, the partial points information that can rebuild object is as details such as flex point, angle point and projectioies.For outline of straight line, use above-mentioned Model Reconstruction, the bigger curve of curvature can be with the approximate match of segmentation straight line, so the profile information that this method can also 3 d objects reconstruction.Design derivation algorithm then.With rotation matrix to be asked
And translation vector
Be arranged to unknown quantity, pixel and image straight line optical flow field are known quantity.Because the quantity of optical flow field has surpassed unknown quantity number to be asked, so this model is an overdetermined equation group, can find the solution with least square method.Adopt VC++ Programming with Pascal Language implementation algorithm at last.The important link of the present invention is the view data computing, and this part is consuming time more.The strategy of this method is directly view data to be loaded into the GPU arithmetic element.Because the image operation based on GPU framework technology can quicken the computing velocity of three-dimensional reconstruction, and use the OpenGL display technique can show reconstructed results real-time, and can provide the texture and the illumination of three-dimensional body, therefore this method make above-mentioned both build the three-dimensional reconstruction platform jointly, improved three-dimensional reconstruction speed in a large number.
Claims (1)
1. dotted line optical flow field three-dimensional rebuilding method based on the GPU framework, it is characterized in that method is: at first set up unified model based on pixel and motion of profile straight line optical flow field reconstruction of three-dimensional and structure, the image straight line is expressed dimension to be arranged to consistent with the pixel dimension, therefore can rebuild pixel and image straight line optical flow field unified under a three-dimensional reconstruction computation model, for outline of straight line, use above-mentioned Model Reconstruction, the curve that curvature is bigger can be similar to match with the segmentation straight line, design derivation algorithm then, with rotation matrix to be asked
And translation vector
Be arranged to unknown quantity, pixel and image straight line optical flow field are known quantity, because the quantity of optical flow field has surpassed unknown quantity number to be asked, therefore this model is an overdetermined equation group, can find the solution with least square method, next sets the three-dimensional reconstruction strategy, calculation level light stream under the low resolution of initial yardstick, along with the increase of the number of plies, resolution is more and more higher, and the result of calculation that obtains is added the initial value that initial value regards to descend one deck to calculate, the substitution next one is more put and the calculating of straight line light stream three-dimensional reconstruction on the high resolution graphics layer, so repeatedly, till the three-dimensional reconstruction that reaches on the full resolution image number of plies, last, because the image operation amount is big, adopt the image operation of GPU framework technology to quicken the computing velocity of three-dimensional reconstruction, use OpenGL to show reconstructed results, adopt VC++ Programming with Pascal Language implementation algorithm.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 201110072376 CN102129711A (en) | 2011-03-24 | 2011-03-24 | GPU (Graphics Processing Unit) frame based three-dimensional reconstruction method of dotted line optical flow field |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 201110072376 CN102129711A (en) | 2011-03-24 | 2011-03-24 | GPU (Graphics Processing Unit) frame based three-dimensional reconstruction method of dotted line optical flow field |
Publications (1)
Publication Number | Publication Date |
---|---|
CN102129711A true CN102129711A (en) | 2011-07-20 |
Family
ID=44267784
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN 201110072376 Pending CN102129711A (en) | 2011-03-24 | 2011-03-24 | GPU (Graphics Processing Unit) frame based three-dimensional reconstruction method of dotted line optical flow field |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102129711A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2013159272A1 (en) * | 2012-04-23 | 2013-10-31 | Hewlett-Packard Development Company | Statistical analysis using graphics processing unit |
CN103996210A (en) * | 2014-06-06 | 2014-08-20 | 南昌航空大学 | Image sequence motion straight line screening and tracking method based on pixel point optical flow |
CN106153977A (en) * | 2016-06-21 | 2016-11-23 | 上海交通大学 | A kind of three-dimensional flow field method of testing based on single light-field camera |
CN107784129A (en) * | 2016-08-24 | 2018-03-09 | 中国海洋大学 | Time Continuous flow field structure analytical technology based on objective Euler's coherent structure |
CN108665524A (en) * | 2018-04-28 | 2018-10-16 | 武汉大学 | A kind of a wide range of discrete flow field volume rendering method based on GPU |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101697236A (en) * | 2009-10-21 | 2010-04-21 | 南昌航空大学 | Method for three-dimensional reconstruction of straight-line optical flow field based on intelligent optimization algorithm |
-
2011
- 2011-03-24 CN CN 201110072376 patent/CN102129711A/en active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101697236A (en) * | 2009-10-21 | 2010-04-21 | 南昌航空大学 | Method for three-dimensional reconstruction of straight-line optical flow field based on intelligent optimization algorithm |
Non-Patent Citations (4)
Title |
---|
《中国图象图形学报》 20020531 陈震,高满屯,沈允文 图象光流场计算技术研究进展 第434-438页 1 第7卷, 第5期 * |
《南昌航空工业学院学报(自然科学版)》 20040930 陈震,高满屯,黎明 光流场重建线性算法的实验模拟 第1-5页 1 第18卷, 第3期 * |
《南昌航空工业学院学报(自然科学版)》 20051231 江少锋,陈震,黎明,杨素华 从二维图像直线光流场求解三维刚体旋转运动参数 第5-8页 1 第20卷, 第4期 * |
《计算机学报》 20040831 陈震,高满屯,曲仕茹,杨声云 基于直线光流场的三维运动和结构重建 第1048页第3节-第1051页第5节 1 第27卷, 第8期 * |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2013159272A1 (en) * | 2012-04-23 | 2013-10-31 | Hewlett-Packard Development Company | Statistical analysis using graphics processing unit |
GB2516192A (en) * | 2012-04-23 | 2015-01-14 | Hewlett Packard Development Co | Statistical Analysis Using Graphics Processing Unit |
CN103996210A (en) * | 2014-06-06 | 2014-08-20 | 南昌航空大学 | Image sequence motion straight line screening and tracking method based on pixel point optical flow |
CN103996210B (en) * | 2014-06-06 | 2016-08-17 | 南昌航空大学 | Image sequence line of motion based on pixel light stream screening tracking |
CN106153977A (en) * | 2016-06-21 | 2016-11-23 | 上海交通大学 | A kind of three-dimensional flow field method of testing based on single light-field camera |
CN107784129A (en) * | 2016-08-24 | 2018-03-09 | 中国海洋大学 | Time Continuous flow field structure analytical technology based on objective Euler's coherent structure |
CN108665524A (en) * | 2018-04-28 | 2018-10-16 | 武汉大学 | A kind of a wide range of discrete flow field volume rendering method based on GPU |
CN108665524B (en) * | 2018-04-28 | 2021-09-24 | 武汉大学 | Large-range discrete flow field volume rendering method based on GPU |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Kato et al. | Neural 3d mesh renderer | |
CN102306386B (en) | Method for quickly constructing third dimension tree model from single tree image | |
CN110288695B (en) | Single-frame image three-dimensional model surface reconstruction method based on deep learning | |
CN106600667B (en) | Video-driven face animation method based on convolutional neural network | |
CN101807114B (en) | Natural interactive method based on three-dimensional gestures | |
CN102129711A (en) | GPU (Graphics Processing Unit) frame based three-dimensional reconstruction method of dotted line optical flow field | |
CN101930622B (en) | Realistic modeling and drawing of shallow water wave | |
CN103514619A (en) | System and method for performing three-dimensional motion by two-dimensional character | |
CN104157000B (en) | The computational methods of model surface normal | |
CN110473284A (en) | A kind of moving object method for reconstructing three-dimensional model based on deep learning | |
CN105825545A (en) | Global illumination drawing method based on virtual light sources and adaptive sparse matrix reduction | |
CN102254346A (en) | Method for detecting augmented reality virtual-real collision based on cloud computing | |
WO2016127421A1 (en) | Real-time motion simulation method for hair and object collisions | |
CN104424658A (en) | Structure-preserving interpolation method of two-dimensional shapes | |
CN104778736A (en) | Three-dimensional garment animation generation method driven by single video content | |
Malik et al. | Simple and effective deep hand shape and pose regression from a single depth image | |
CN108734772A (en) | High accuracy depth image acquisition methods based on Kinect fusion | |
Peng et al. | Animatable implicit neural representations for creating realistic avatars from videos | |
CN102214366B (en) | High-performance rendering method of three dimensional (3D) point sampling data | |
CN1776747A (en) | GPU hardware acceleration based body drawing method for medical image | |
Zhang et al. | Avatarstudio: High-fidelity and animatable 3d avatar creation from text | |
CN115906703A (en) | GPU fluid simulation method for real-time interactive application | |
US20130278608A1 (en) | Plant Simulation for Graphics Engines | |
CN115272608A (en) | Human hand reconstruction method and equipment | |
Qi | Computer aided design simulation of 3D garden landscape based on virtual reality |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20110720 |