Object Recognition

Advertisement
 Download your Full Reports for Object Recognition

Object Recognition
Every day we recognize a multitude of familiar and novel objects. We do this with little effort, despite the fact that these objects may vary somewhat in form, color, texture, etc.  Objects are recognized from many different vantage points (from the front, side, or back), in many different places, and in different sizes. Objects can even be recognized when they are partially obstructed from view.


While it may be obvious that people are capable of recognizing objects under many variations in conditions, it has been thought that pigeons may not possess the same range of capabilities.  It has been proposed that pigeons act as "perceptrons," by analyzing simple features of objects and using those features to recognize objects.  If the pigeon were a perceptron, then it would not be able to  recognize an object that varied slightly in form or was seen from a novel viewpoint because the features would be altered.  Moreover, a pigeon would be unable to discriminate between two objects that contained the same features, but with a different organization.
This chapter addresses a number of fundamental issues relating to object recognition, concentrating particularly on an avian species, the pigeon.  The task is to determine whether the basic process of object recognition in pigeons is at all similar to the most probable process that has been proposed for humans.  In order to demonstrate the conditions under which object recognition may or may not occur, a number of illustrated examples will be provided.

I. Introduction
This section presents general-level background information, discusses key theoretical concepts, and provides a short statement of the significant findings of the specific experiments. More detailed descriptions can be found in the following sections, to which links are provided throughout.  Readers who are well-versed in the basics of object recognition may wish to proceed directly to the "Experiments" section. 

In this document we will introduce a novel artificial intelligence approach to object recognition. The approach is shape-based and works towards recognition under a broad range of
circumstances, from varied lighting conditions to affine transformations. The main emphasis is on its neural elements that allow the system to learn to recognize objects about which it
has no prior information. Both supervised techniques, in the form of feed-forward networks with error back-propagation, and unsupervised techniques, in the form of a self-organizing
map, are employed to deal with the interpretation of objects in images. To make the system complete, a full account is given of the necessary image processing
techniques that are applied to the images to make recognition possible. These techniques include the extraction of shapes from images and the creation of descriptors that overcome
difficulties of affine transformations.

The aim has been to create a complete object recognition system. This document gives an in-depth account about the choices made and the results the system has obtained. The
objective of this work has been to create a new system for artificial object recognition that forms a good basis for further development.

 

 Download your Full Reports for Object Recognition

Advertisement

© 2013 123seminarsonly.com All Rights Reserved.