Loading...
| Friend's email | |
| Your name | |
| Your email | |
| enter code | |
This page was sent successfuly
- Type of Document: M.Sc. Thesis
- Language: English
- Document No: 48213 (58)
- University: Sharif University of Technology, International Campus, Kish Island
- Department: Science and Engineering
- Advisor(s): Khayyat, Ali Akbar; Bagheri Shouraki, Saeed
- Abstract:
- Trying both to understand the brain and to emulate some of its strengths has been one of the greatest human desires since ancient times. One of these amazing abilities is recognizing via vision. As a result, image recognition has been turned into one of the most attractive areas of research in Computer Vision field since recently. The challenging problem begins to rise where occlusion, scale, rotation and various light conditions contribute and manipulate the paradigm of image recognition. Although recognitions with these challenging problems are some of the capabilities that the brain has, but these are not all. One of the remarkable abilities of the brain is to recognize concepts through visual observation. This ability can be discussed from two point of views. First, once the brain learns an object, it is able to recognize that object in any viewpoints without having to scan an object from all viewpoints in learning phase. For example, the tire from side and front is still a tire in human point of view, although it is similar to a circle from the side view and alike a rectangle from the front view. Next, once the brain learns some models of an object; it is able to recognize untrained models of that object correctly. For example, when we see a new model of a car, it is still a car in our point of view. But how are these abilities possible? Actually, three-dimensional visualization of real-world objects has been done by stereo vision capability of the brain. In stereo vision 3-D information extracted from multiple 2-D views of a scene. These 2-D images are created by each eye according to its viewing angle of any object. The goal of this thesis is to find a solution for implementation of concept recognition through visual observation ability of the brain. To achieve this purpose, artificial neural networks, which are proposed to model the information processing capabilities of nervous systems are used as a main part of processing and the development idea of stereo vision, which is to use some viewpoints of that object in training phase is used as a part of this solution. As a result, the general idea of this solution, in the first step, is the network which should not be able to respond and recognize in untrained viewing angles if we train the network for some particular viewing angles of objects. Next, by extending the network for multiple models of an object, network would be able to respond for an untrained model of that particular object. This solution is come off by using self-organized Adaptive Resonance Theory neural network as a model of the brain in concept recognition with the combination of different image processing techniques for preparing images to better understanding and more accurate recognition. Results in final method indicated about 75% accuracy in Turntable dataset which is used to test the accuracy of the system in recognition of objects in untrained angle of view and 55% accuracy in Caltech 101 dataset which is used to test the accuracy of the system in recognition of different models of an object
- Keywords:
- Humanoid Robot ; Adaptive Resonance Theory 1 (ART1) ; Object Classification ; Body Simulation ; Concept Learning
Digital Object List-
محتواي کتاب - view
Bookmark