Dissertation > Excellent graduate degree dissertation topics show

Multiclass Object Recognition Based on Features of Interest Points

Author: LiaoQiuPing
Tutor: LiuHaiHua
School: Central South University for Nationalities
Course: Biomedical Engineering
Keywords: multiclass object recognition HMAX model perception selective attention mechanism interest points
CLC: TP391.41
Type: Master's thesis
Year: 2011
Downloads: 39
Quote: 0
Read: Download Dissertation

Abstract


Human beings can easily recognize the changing objects in nature through their visual system. However, it is rather difficult for computers. Recently, how to help computers better build a fast and accurate discriminative ability as that of human beings is one of the research hotspots among many researchers. Therefore, in accordance with the research achievements on cognitive psychology and neuroscience, it is considered to be an attractive exploring field of building a Multiclass Object Recognition Model, based on a simulation of the system processing and information handling procedures of human beings’visual system. The HMAX Model, which is proposed by Poggio, et. al, is a ventral stream of visual cortex, and it has been improved a lot and proves to be great significance to object recognition and classification ability.In this thesis, based on recent study achievements on cognitive psychology and neuroscience, the author mainly focuses on analyzing HMAX’s hierarchical visual perception mechanisms and its calculating model, with a combination of studying the function and mechanism characters of the visual cortex. This study aims at improving the availability of characterized templates, as well as raising its calculating speed. Moreover, the author also implements classification and recognition purposes to eight kinds of images targets. The major findings are as follows.(1) The author firstly analyzes the physiological features and functions of the ventral stream of different visual cortexes, HMAX’s hierarchical visual perception mechanisms and its calculating model, and it is found that there are many problems such as an overweight calculating task, slow recognition speed, as well as low availability of the characterized templates, etc.(2) Due to physiological features of neurons in V4 and IT, whose visual scale is limited within certain length and range, the author defines the consistent length and range of S2 cortex of HMAX Model, so as to raise the model’s recognition speed.(3) Because of the selective attention mechanism of visual neurons, the small templates in the ROI can be selected as characterized ones, in order to avoid the problems of low availability which is caused by randomly selection. The selection of interest points is decided by the number of interest points, for the purpose of coordinating the thresholds standards among different classification targets.(4) Applying a modified model to classify and recognize eight kinds of image targets in Caltech-101 database, and then making a comparison with HMAX Model about the discrimination rate and time of C2 character in each image, etc. Experiment results and analysis show that the modified model is much better than HMAX Model in terms of recognition speed and discrimination rate.

Related Dissertations

  1. Research on Action Recognition Based on the Space-Time Interest Points,TP391.41
  2. Research and Implementation of Key Technologies in Image Mosaic,TP391.41
  3. Research of Moving Object Detection and Tracjking Based on Interest Points,TP391.41
  4. Image Retrieval Based on Combining Color and Texture Features of Interest Region,TP391.41
  5. Research on the Facial Expression Recognition Based on HMAX Model,TP391.41
  6. The Research on the Key Technologies of Content-Based Image Database Retrieval,TP391.3
  7. Relevant Algorithms for General Object Recognition Based on Feature Combination,TP391.4
  8. Image Retrieval Based on Partition for Interest Points and Text-based IR,TP391.41
  9. Content-based image retrieval key technology research,TP391.41
  10. Study on Object Recognition Models Based on Visual Pathway,TP391.41
  11. Human Pose Estimation and Action Recognition from Image Sequences,TP391.41
  12. Research on System Consistency and Optimal States Estimation of LIDAR/INS Integrated Navigation System,TN967.2
  13. Automatic generation of technical research panorama,TP391.41
  14. Research on Content-Based Image Retrieval,TP391.3
  15. Points of Interest - Based Image Retrieval,TP391.41
  16. Image Retrieval Using Salient Points and Geometric Hashing,TN911.73
  17. Based on the classification and identification of the mechanism of visual perception of the human brain,TP391.41
  18. Contour Detection Based on Primaty Visual Model,TP391.41
  19. Research on Algorithms of 2D Face Template Protection,TP391.41
  20. Research on the Classification Based on the Reconstruction of Solder Joint,TP391.41
  21. Tongue Feature Extraction and Research of Fusion Classification,TP391.41
  22. The Fatigue State Recognition of the Driver Based on Eye Detection,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