1083 
Appearance Based 3D Object Recognition Using IPCA-ICA 
V.K Ananthashayana and Asha.V 
Image Processing and Computer Vision Lab 
Department of Computer Science and Engineering 
M.S.Ramaiah Institute of Technology 
Bangalore, India 
Commission I , ThS-2 
KEY WORDS: Principal component analysis (PCA), Independent component analysis (ICA), principal non-Gaussian directions, blind 
source separation (BSS), Appearance-based object recognition, Incremental principal component analysis (IPCAPCA). 
ABSTRACT: 
Fast incremental non Gaussian directional analysis (IPCA-ICA) is proposed as a linear technique for recognition [1]. The basic idea is to 
compute the principal components as sequence of image vectors incrementally, without estimating the covariance matrix and at the same 
time transforming these principal components to the independent directions that maximize the non-Gaussianity of the source. In this 
paper we utilize the IPCA-ICA technique for 3D Object recognition by employing its neural network architecture. We illustrate the 
potential of IPCA-ICA on a database of 1440 images of 20 different objects captured by CCD camera. The excellent recognition rates 
achieved in all the performed experiments indicates that the present method is well suited for appearance based 3D object recognition. 
1. INTRODUCTION 
A human activity relies heavily on the classification or 
identification of a large variety of visual objects. In Computer 
vision, the recognition system typically consists of sensors, and 
model database in which all the object representations and 
decision making abilities are saved. Recently, object recognition 
has been found in a great range of applications like Surveillance, 
Robot vision, Medical Imaging etc. For the view-based 
recognition, the representation takes into account the appearance 
of the object. This requires three-dimensional object recognition 
(3DOR), for which the pose of the object is of main consideration. 
The objective of the 3DOR algorithm is not only to recognize the 
object precisely but also to identify its pose as viewed. Then a 
recognition algorithm tries to find the best matched object. A 
schematic of a typical object recognition system is shown in Fig 
1. 
Recognized 
Object 
Fig 1. Block Diagram of Typical Object Recognition System 
Fig 1 represents the typical object recognition system. In this 
system, its geometric features like edges, lines, curves and 
vertices are called geometric features are extracted, since the 
edges contain more information. In contrast to the geometric 
features, appearance of an object is the combined effect of its 
shape, reflectance properties, pose and the illumination. The 
approaches that take explicitly these factors into account for use 
in object recognition have been categorized as appearance-based 
object recognition methods. The main idea in the appearance- 
based approach is to represent the images in terms of their 
projection onto a low-dimensional space is called eigenspace. 
The projection of the images onto this space is called the 
eigenimages. One popular method to obtain the low dimensional 
space is called the PCA [3], and other methods like Kemal PCA 
and independent component analysis (ICA)[2] are used in 3D 
object recognition. In these methods, the images of the objects 
are taken from various pose angles, and their compressed features 
are saved in a database. The test images are taken in similar