Dissertation > Excellent graduate degree dissertation topics show

The Study of 3D Reconstruction for with Regularity Based on Structural Priori

Author: JiaoZengTao
Tutor: ZhuXiuChang
School: Nanjing University of Posts and Telecommunications
Course: Signal and Information Processing
Keywords: 3D reconstruction scenes with regularity structural priori cluster Markov random field vanishing point
CLC: TP391.41
Type: Master's thesis
Year: 2012
Downloads: 87
Quote: 0
Read: Download Dissertation

Abstract


3D scene reconstruction from images has become one of the important research areas of computer vision after years of innovation and development. While scenes with regularity are closely associated with people daily life, the reconstruction of these scenes is valuable on study and practice. The concerns of this paper are the methods to utilize the priori clues of scenes with regularity for 3D reconstruction. The traditional algorithms on 3D scene reconstruction from images are ineffective for scenes with regularity. On the other hand, there are many structural features in these scenes, such as coplanar or collinear structures, which are helpful. The focus of this paper is the using of structural priori clues in scenes with regularity.Since the coplanar structures are common in the scenes with regularity, it’s of importance to estimate the dominant planes of these scenes. In this paper, the assumptions of strong structural models, such as Manhattan world model, are not adopted simplify the scenes. To estimate the normal directions and positions of dominant planes, a cluster algorithm is utilized on normal directions and positions of 3D oriented points according to the widespread existence of 3D dominant planes. This method relaxes the assumption of previous scholars without any constraints on the normals of 3D dominant planes and can handle more general structure of regular scenes.According to the priori of structural imaging in scenes with regularity, this paper constructs the Markov random field energy function and solved the function with optimization method to get a piecewise planar depth map. In this paper, the energy function is built without discrete depth values as labels and pixels as nodes, but in larger particle size with 3D dominant planes as labels and segmentation regions as nodes to estimate a more accurate distribution of the regional label map.Vanishing point is the 2D expression of 3D parallelism in scenes with regularity, and the estimation of vanishing point is very useful for 3D reconstruction. Firstly, a LSD-based line detection algorithm is utilized to extract 2D segments from images. Secondly, the structure of optimal solution for vanishing points is defined as intersection point frequency functions and intersection point neighborhoods. Finally, the estimation is done through AGS.

Related Dissertations

  1. Design and Study on Movable Packing Machine in Acrylic Fiber Production Lines,TH248
  2. Study on the Heat Transfer Characteristics of Particle Cluster in Circulating Fluidized Bed,TK124
  3. 3D Visualization Research of Medicial Ultrasound Image,TP391.41
  4. Based on statistical methods of magnetic resonance imaging of the human brain image segmentation and three-dimensional analysis of the data,R445.2
  5. Comprehensive Quality Assessment of College Students,G645.5
  6. Studieson Effects of Soybean Species on Yuba and Initial Establishment of Quality Evalution System for Yuba,TS214.2
  7. Research on the Soil Environmental Function Zoning,X321
  8. Approach for Calibration and Orientation of Digital Camera Based on Linear Features,TP391.41
  9. Research on Key Technologies for Binocular Stereo Vision,TP391.41
  10. Research on Medical Image Segmentation Method Based on Markov Random Field Model,TP391.41
  11. The Research of Non-rigid Object Motion Reconstruction Based on Dynamic Image Sequences,TP391.41
  12. Research on Automobile Industrial Cluster in Shenyang,F224
  13. Study on the Multi-fiber Combination and the Wearability and Style of the Woven Fabrics,TS941.15
  14. Research and Implementation of 3D Reconstruction Algorithm Based on Computer Vision,TP391.41
  15. Research on 3D Visualization Algorithm of Medical Image,TP391.41
  16. Computer-assisted hip replacement surgery systems research and development,TP391.41
  17. The System for Engineering Drawing Vectoring and 3D Reconstruction,TP391.41
  18. Design and Implementation of the Three-dimensional Reconstuction System of Human Body Based on Multiple Images,TP391.41
  19. Research on 3D Reconstruction Technology with Marking Points and Its Application to Dimension Measurement of Castings,TP391.41
  20. The Effect of Heterogeneous Microstructure on the Room Temperature Mechanical Properties in Bulk Metallic Glass,TG139.8
  21. Researches and Simulation of Load Control Strategies in TD-SCDMA Trunking System,TN929.533

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device
© 2012 www.DissertationTopic.Net  Mobile