Enhance Rotation Invariant Through Moment-Based Function For A Better Representation Of 3D Object Shape Descriptor

Muda, Azah Kamilah and Syed Ahmad, Syarifah Sakinah and Muda, Noor Azilah and Mohd Yusoh, Zeratul lzzah and Darus, Maslina (2016) Enhance Rotation Invariant Through Moment-Based Function For A Better Representation Of 3D Object Shape Descriptor. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

Text Enhance Rotation Invariant Through Moment-Based Function For A Better Representation Of 3D Object Shape Descriptor.pdf - Submitted Version Restricted to Registered users only Download (671kB)

Abstract

With the fast development of 3D technology,recognizing a 3D shape object becomes vital in various disciplines. Recognition performance also depends on the preeminent feature extraction technique.It caused an increased interest in development of 30 modeling,30 object representation and identification.Generally,20 and 30 object can be identified based on its shape.2D shape descriptors describe boundary and area based.Meanwhile,3D shape descriptor which has additional properties of volume and surface description is more powerful in recognizing an object.However,the complexity of acquiring features vector that invariant to rotation for 30 shape object leads to recognition problem.This is due to 30 object arises from various shape and not only in rounded shape object,including in curve surface.Geometric transformations of 30 object such as translation, rotation,reflection,and isotropic re-scaling,are often viewed as nuisance effects that must be eliminated as they do not contribute to the semantic classification of the shape.Besides the invariant with respect to translation,rotation,scaling,and reflection,basic requirements of shape descriptors also should fulfill robustness property with respect to level-ofdetail and multiple levels of abstraction.Thus,this research aims to formulate a new Moment-based Function of 30 object shape descriptor that invariant to rotation as a 30 invariant feature extraction technique.The experimental approach of pattern recognition framework will be adopted as the research methodology.Invarianceness analysis as measurement properties of geometric transformation for 30 object shape representation will be carried out.Followed by constructing a new Moment-based formulation for a 3D model descriptor.Lastly,an algorithm will be developed and evaluated to analyze its performance.

Actions (login required)

View Item