Dissertation > Excellent graduate degree dissertation topics show

Research on Feature Correspondences for Fisheye Images

Author: TianYaPing
Tutor: LiXiaoMing
School: Shanxi University
Course: Pattern Recognition and Intelligent Systems
Keywords: Fisheye image Feature matching Motion Estimation Contour feature RANSAC
CLC: TP391.41
Type: Master's thesis
Year: 2010
Downloads: 81
Quote: 0
Read: Download Dissertation

Abstract


Corresponding to an image point of the match is determined between the different images corresponding to the image point of the same spatial points. In almost all computer vision problems (such as motion estimation, object recognition, object detection, three-dimensional reconstruction, etc.), the image corresponding point matching is a critical step, Therefore, this issue is in the computer vision core issue. At home and abroad in this field has made fruitful research results, but most focus on the perspective image. The fisheye image a larger field of view, has important applications in the real world, but due to the severe nonlinear deformation, making it more difficult to match the corresponding points. This paper focuses on how to match the corresponding points in the fisheye image did some research, and have achieved a certain amount of research. The completed work:. Robust estimation strategy linear parameter estimation problem and propose a practical robust estimation method. Away from the center of the image area than the area near the center of the image deformation is more serious, data noise is greater, but the data in these regions is very important parameter estimation. Based on this observation, we reasonably assume that the different regions of the data noise, and a noise data focused strategy of gradually more and more choices within. Use this strategy, on the one hand, the more data points can be obtained, this is very advantageous to the noise data for the linear estimation problem, on the other hand, the data set includes more away from the center of the image data points. These two aspects make the final estimates of the motion parameters have been significantly improved in accuracy and stability. Programming literature three typical contour feature point extraction method and experimental analysis and comparison of these three methods. These three methods are contour extrema of curvature point detection method, the method for the detection of the minimum value point of contour elastic Multiscale visual detection method of the point of curvature of the contour. Correctness from detection, parameter setting, algorithm execution efficiency analyzed, the the visual curvature point detection algorithm is better than the other two detection algorithms. 3 a fish-eye image corresponding point matching frame, and is given a specific embodiment method. Due to the severe deformation of the fisheye image directly using traditional methods are often based on affine invariant only get less corresponding match, in general, can not meet the actual needs of the application. The main idea of ??our proposed matching framework: First initial matching affine invariant-based approach, and preliminary estimates of geometric constraint model; Although this model is not very accurate, but can guide the follow-up process to get more features matching, and can gradually update the the constraint model ultimately get stable results. The experimental results show that this matching method can be obtained a large number of reliable, stable image matching feature points, so that the final motion estimation and geometric estimation precision can be significantly improved.

Related Dissertations

  1. Optimizing and Realising Research on Vedio Compression in TV Guidance System,TN919.81
  2. SURF algorithm based on improved electronic image stabilization technology,TP391.41
  3. Exam cheating Intelligent Analysis Method,TP391.41
  4. Image Matching Based on Local Invariant Features,TP391.41
  5. Real-Time Image Mosaic Based on CUDA,TP391.41
  6. Statistical Analysis the Collection and Distribution Function Structure of Comprehensive Passenger Hub,U115
  7. Design of Electronic Stabilization System for Video Images Based on FPGA,TP391.41
  8. Lunar Soft-landing Vision Navigation and Descending Sequential Imagery Match Research,TP391.41
  9. UAV image automatic registration and stitching methods,TP751
  10. The Research of Panoramic Images Mosaic Based on Image Sequences,TP391.41
  11. Image Matching Algorithm Based on SIFT in the Vehicle Navigation,TP391.41
  12. Three Dimensional Vessel Tree Reconstruction by Multi-view Stereo Reconstruction,TP391.41
  13. The Design and Implementation of Obstacles Detection Algorithm by 3-D Reconstruction Based on Monocular Vision,TP391.41
  14. Research on Environmental Modeling Technology for Auto-Parking System,TP391.41
  15. X-ray Medical Images Stitching Based on Corner Matching Using Two-way Asynchronous Voting Strategy,TP391.41
  16. CCD Three-Dimensional Measurement of Binocular Vision,TP391.41
  17. Research on Vision Based Odometer for Indoor Mobile Robot,TP242.6
  18. Study on Video Mosaic for Large View,TP391.41
  19. Research on Feature Correspondences for Fisheye Images Based on Dynamic Programming,TP391.41
  20. Application Research on Networks Video Server Based on H.264,TN919.81
  21. Research on Feature-driven Image Registration and Image Mosaic of CT Images,TP391.41

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