Object Recognition Using Combined Color and Angular Spatial Matching

Author: 
Peer reviewed: 
No, item is not peer reviewed.
Scholarly level: 
Graduate student (Masters)
Date created: 
2014-01-27
Identifier: 
etd8242
Keywords: 
Object recognition
Color
Shape
Image processing
Matching
Abstract: 

In this thesis, we proposed a novel object recognition method that utilizes object's color along with its shape information to represent the appearance of it. Such representation of a known object can be used to identify a target object in unknown environments. The color and shape characteristics are first extracted from multiple images of an object and later used to search for that object within a scene. The Saturation-weighted Distributive Hue (SDH) histograms are considered to be promising color descriptors as they possess computational efficiency and view invariance properties. However, the proportion of each color on an object's surface is not included in this descriptor. To overcome this problem, we developed the Chromatically-improved SDH (CSDH) histograms based on proportional color segmentation. Additionally, a shape descriptor, called Bag of Angles (BoA), is developed that increases the recognition rate for objects with similar colors but with different shapes. The BoA shape descriptor utilizes the angles between consecutive segments of the objects contours. To identify a target object in the scene, the color and shape features are sequentially compared to those extracted from the scene image. For evaluation purpose, the proposed method is tested on the Robotic Vision Object Image Library (RVOIL) and the Amsterdam Library of Object Images (ALOI) databases. The experimental results show that the proposed method can achieve an average recognition rate of 84\%, which is better than the state-of-the-art method (80\%).

Language: 
English
Document type: 
Thesis
Rights: 
Copyright remains with the author. The author granted permission for the file to be printed and for the text to be copied and pasted.
Senior supervisor: 
Parvaneh Saeedi
Department: 
Applied Sciences: School of Engineering Science
Thesis type: 
(Thesis) M.A.Sc.
Statistics: