Please use this identifier to cite or link to this item:
http://hdl.handle.net/10603/26894
Title: | Three dimensional object recognition supported by local and global features |
Researcher: | Muralidharan, R |
Guide(s): | Chandrasekar, C |
Keywords: | Global Features Local Features Object Recognition Three Dimension |
Upload Date: | 17-Oct-2014 |
University: | Anna University |
Completed Date: | n.d. |
Abstract: | Object recognition is a technological discipline that deals with the newlineprocess of understanding design development and construction of techniques newlineto recognize the objects of interest in the image The objective of an object newlinerecognition system is to make the computer to recognize objects without the newlineassistance of the human The major problem in object recognition is newlineautomatic identification of the object with respect to its size location newlineorientation and in different illumination conditions When designing an newlineobject recognition system the following is the two problems to be addressed newlinethe first one is choosing the correct set of features to represent the image and newlinethe second one is selecting best classifier for recognition To solve the first newlineproblem it is important to extract features that should be invariant to the newlineobjects transformations like scale translation and rotation For the second newlineproblem the classifier should recognize the object with less computation time newlineand should provide high performance newlineObject Recognition is a method to identify the objects in the given newlineimage It plays a vital role in the challenging fields like surveillance systems newlinefault diagnosis systems leaf species detection etc This thesis works out for newlineincreasing the performance and efficiency of 3D object recognition with less newlinecomputational requirements newlineThe first part of this thesis is to recognize the twodimensional newlineobjects from the given image using kNearest Neighbor algorithm supported newlineby global features of the image Also in this work Hus moment invariant is newlinecomputed from the local parts of the image Once the global feature is newlineextracted the Kernel Principal Component Analysis KPCA is applied to newlinereduce the dimension of the features and the resultant eigenvalues 15 values newlineare used as a feature Using the kNearest Neighbor kNN with euclidean newlinedistance metric the object in the test image is identified newlineThe second part of the thesis is to experiment the performance of newlinethe Support Vector Machine SVM and kNearest Neighbor as a classifier for newlinerecognizin |
Pagination: | xx,162p. |
URI: | http://hdl.handle.net/10603/26894 |
Appears in Departments: | Faculty of Science and Humanities |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
01_title.pdf | Attached File | 59.27 kB | Adobe PDF | View/Open |
02_certificate.pdf | 4.26 MB | Adobe PDF | View/Open | |
03_abstract.pdf | 71.34 kB | Adobe PDF | View/Open | |
04_acknowledgement.pdf | 58.48 kB | Adobe PDF | View/Open | |
05_contents.pdf | 113.74 kB | Adobe PDF | View/Open | |
06_chapter 1.pdf | 2.17 MB | Adobe PDF | View/Open | |
07_chapter 2.pdf | 534.75 kB | Adobe PDF | View/Open | |
08_chapter 3.pdf | 170.52 kB | Adobe PDF | View/Open | |
09_chapter 4.pdf | 1.32 MB | Adobe PDF | View/Open | |
10_chapter 5.pdf | 1.6 MB | Adobe PDF | View/Open | |
11_chapter 6.pdf | 1.32 MB | Adobe PDF | View/Open | |
12_chapter 7.pdf | 79.46 kB | Adobe PDF | View/Open | |
13_references.pdf | 111.13 kB | Adobe PDF | View/Open | |
14_publications.pdf | 70.16 kB | Adobe PDF | View/Open | |
15_vitae.pdf | 52.45 kB | Adobe PDF | View/Open |
Items in Shodhganga are licensed under Creative Commons Licence Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0).
Altmetric Badge: