Dissertation > Excellent graduate degree dissertation topics show

Study on 3D Model Retrieval Technology Based on 3D Reconstruction

Author: ZhangJiaWei
Tutor: CaoKui
School: Henan University
Course: Computer Applications
Keywords: Two-dimensional feature point selection Optical flow SFM algorithm EM algorithm EGI
CLC: TP391.3
Type: Master's thesis
Year: 2009
Downloads: 187
Quote: 0
Read: Download Dissertation

Abstract


The three-dimensional model compared to the two-dimensional image with richer and more intuitive. With the development of 3D retrieval technology, it was found that the use of three-dimensional retrieval technology combined with three-dimensional reconstruction can solve some recognition and retrieval problems can not be solved on the plane, and the increasing emphasis and development has been based on three-dimensional reconstruction of 3D model retrieval technology. How to make efficient the 3D retrieval algorithm based on three-dimensional reconstruction has become a cutting-edge and challenging topic. Based on the three-dimensional reconstruction of the three-dimensional model retrieval technology is the use of computer technology to extract the three-dimensional model of the target object from a video scene then classified and retrieval technology model in accordance with the characteristics of the model. The work process is divided into feature point selection; reconstruct the 3D point cloud model based on feature points; reconstructed 3D point cloud model are classified according to the different characteristics for easy retrieval. Feature point selection stage, the existing methods require user participation in the selection process of the feature points can not be achieved automatically select feature points;, traditional SFM method is strictly dependent on the correspondence between 2D feature points in the three-dimensional reconstruction phase, requirements feature points of the two images correspond exactly, but the exact correspondence between the characteristic points is very difficult; retrieval phase of the three-dimensional model, because the reconstructed 3D point cloud model point cloud model of similarity is very difficult. To solve the above problem, this paper is summarized as follows: First, feature point selection stage, this project presents a new 2D feature points based on the Horn-Schunck optical flow algorithms and canny operator calibration algorithm to automatically obtain feature points . The first to use the canny operator to extract the edge of the object in the first image in the image sequence, and Horn-Schunck optical flow algorithm to calculate all the pictures of the optical flow field; Finally, the object edge glazing flows to the amount of the maximum of a set of points as feature points . The second in a three-dimensional reconstruction phase, the subjects using the SFM algorithm based on the EM algorithm, this algorithm characteristics do not need to give an accurate correspondence between the original image feature points, instead of using a space three-dimensional coordinate points to two dimensional feature point of the \3D point cloud model of \. In 3D model retrieval stage, adopted by the subject the Crust algorithm from the point cloud model generated surface mesh of the object, and then using the extended Gaussian image (EGI, Extended Gaussian Image) An improved calculation method with the kernel density estimation (Kernel Density Estimation) to complete the three-dimensional model matching and retrieval. Finally, the experiments show high classification accuracy and efficiency in the implementation of the algorithm.

Related Dissertations

  1. Research on Visual Detection and Tracking of Mobile Robots,TP242.62
  2. Research and Implementation of Non-rigid Medical Image Registration Method Based on Improved Optical Flow Model,TP391.41
  3. Research & Implementation of Obstacle Detection Algorithm Based on Feature Points Matching,U463.6
  4. The Analysis of the Data of Mental Health Status Based on GPCM Model,G444
  5. Energy Saving and Intelligent Surveillance System for Escalator Based on Panorama Vision,TP277
  6. Using multi- gene mapping data with genotype error,Q78
  7. Counting interval mapping of quantitative trait,O212.7
  8. Moving Video Object Segmentation Based on Spatio-temporal Information,TP391.41
  9. Optical Flow Aided Navigation Inspired by Honeybee Vision,TP391.41
  10. Research on Co-clustering and Application,TP311.13
  11. Research of the Relationship between Respiratory and Lung Tumor Displacement,TP391.41
  12. Study on Tracking Algorithm Based on Semi-supervised Online Learning,TP391.41
  13. The Statistical Model Recognition of the Radio Channel and Its Application in MIMO Systems,TN919.3
  14. Of speckle noise removal in medical ultrasound image,TP391.41
  15. Distributed sound source localization and tracking algorithm,TN912.3
  16. Rearch on Optical Flow Algorithm and the Application in Video Object Detection,TP391.41
  17. Research and Implementation of Moving Obstacle Detection Technology Based on Monocular Vision,TP391.41
  18. Mobile Robot Object Recognition and Localization Based on Monocular Vision,TP242.6
  19. FPGA-based Video Motion Tracking System,TP391.41
  20. OFDM-based cognitive radio (CR) systems research narrowband interference suppression,TN925
  21. A High Fidelity Mosaic Technology for Large Mural Image Based on SFM,TP391.41

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Retrieval machine
© 2012 www.DissertationTopic.Net  Mobile