Dissertation > Excellent graduate degree dissertation topics show

An Eigenspace Based Approach for 3D Object Recognition

Author: HuZuo
Tutor: ZhangGuiLin
School: Huazhong University of Science and Technology
Course: Pattern Recognition and Intelligent Systems
Keywords: Target recognition Feature space Feature space update
CLC: TP391.41
Type: Master's thesis
Year: 2004
Downloads: 285
Quote: 1
Read: Download Dissertation

Abstract


Computer vision system's goal is to explain existing visual data and use these explanations to complete the task. For the needs of a large number of real-world applications of robots the workpiece crawl task, automatic navigation, automatic detection, assembly tasks and medical image analysis, and so on, target recognition has become a very active area of ??research. Many existing target recognition method is based on template matching, and geometry of the feature point extraction and classification method, these identification methods have many drawbacks, such as a direct comparison between the two images corresponding pixel gray level image matching method similar The sum of resistance. This method does not have the direction, scale, geometric distortion invariance, which requires the strong correlation of the target image, the image distortion degree of tolerance is very limited, which is a difficult point of the 3D target recognition. In this paper, for the singular value represented by algebraic features a lot of research work in 3D target recognition and pose estimation. Generation of the image characteristic reflects the nature of the image, despite its physical meaning is not very intuitive, but its mode classification still plays an important role on the target image Generation Generation transform decimation reflect the intrinsic properties of the target image characteristics to for target identification, which is the target identification of a promising direction. First-depth study of the principle of the method of feature space on two key steps in the identification process: feature representation and similarity measure specific experiments and research work to achieve target recognition feature space-based system, which uses several effective improvement effective solution to the problem of the recognition process, such as light, as well as 3D object pose estimation, to improve the recognition robustness. After the completion of the target recognition algorithm based on the feature space, taking into account the target image sample can not all simultaneously be recalculated every time when a new image to meet the requirements of the SVD (Singular Value Decomposition), which will take time (for the number of image pixels), the calculation is too high. Further feature space update method, fast and stable SVD update algorithm, to further investigate the retention of feature vectors in the update process, update speed. Finally, update the feature space and feature space-based target recognition method combining real-time learning identification scheme.

Related Dissertations

  1. ISAR Imaging Simulation of Space Targets and Target Recognition Based on ISAR Images,TN957.52
  2. Study on Virtual Detector of Infrared Hyper-Spectral Image,TP391.41
  3. Research on the Image Real-Time Acquisition, Storage and Image Processing System,TP391.41
  4. Design and Implementation of Intelligent Transportation Video Monitoring System Based on 3G Network,TP391.41
  5. Servers based on infrared image classification conditions,TP393.05
  6. Research on the Joint Classification Based on Support Vector Machine and K-nearest Neighbor,TP18
  7. Study on Target Tracking System of Mobile Robot Based on Visual Guidance,TP242
  8. The Humanoid Robot Goalkeeping System Based on the Wireless Communication,TP242
  9. Research on Robot Classification-and-grab Technology Based on Machine Vision,TP242
  10. Local learning based super -pixel image target recognition,TP391.41
  11. Shape Feature Based Aircraft Object Recognition and Extraction from High Resolution Remotely Sensed Imagery,TP751
  12. Target Recognition Based on High-resolution Radar Images,TN957.52
  13. The Object Recognition Research Based on Embedded Platform,TP391.41
  14. Based on semi-supervised integration of remote sensing image segmentation and classification,TP751
  15. The Humanoid Robot Penalty Kick System Based on the Embedded Vision,TP242.6
  16. Feature Extraction and Identification of Subsurface Objects in GPR,TN959.71
  17. Study and Related Implementation of Robust Algorithm for Adaptive Beam-forming,TN911.7
  18. Latent Semantic Index and Hierarchical Clustering Based Feature Space Reconstruction Method and Application Research,Q51
  19. Underwater Target Identification Software Design and Realization Based on Component,U666.7
  20. The laser detection automotive active anti-collision technology and systems research,TN247
  21. Research on Coastal Area Analysis and Application Based on Interpretation of Hyperspectral Data,P714.6

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