You need JavaScript to view this

Orientation of a 3D object: implementation with an artificial neural network using a programmable logic device;Orientacion de un objeto 3D : implementacion de redes neuronales artificiales utilizando logica programable

Thesis/Dissertation:

Abstract

Complex information extraction from images is a key skill of intelligent machines, with wide application in automated systems, robotic manipulation and human-computer interaction. However, solving this problem with traditional, geometric or analytical, strategies is extremely difficult. Therefore, an approach based on learning from examples seems to be more appropriate. This thesis addresses the problem of 3D orientation, aiming to estimate the angular coordinates of a known object from an image shot from any direction. We describe a system based on artificial neural networks to solve this problem in real time. The implementation is performed using a programmable logic device. The digital system described in this paper has the ability to estimate two rotational coordinates of a 3D known object, in ranges from -80{sup 0} to 80{sup 0}. The operation speed allows a real time performance at video rate. The system accuracy can be successively increased by increasing the size of the artificial neural network and using a larger number of training examples;La extraccion de informacion compleja a partir de imagenes es una habilidad clave en las maquinas inteligentes con vasta aplicacion en los sistemas automatizados, la manipulacion robotica y la interaccion humano-computadora. Sin embargo, resulta una tarea extremadamente dificil de  More>>
Authors:
Carnevale, Federico J [1] 
  1. Universidad Nacional de Cuyo, Instituto Balseiro, Centro Atomico Bariloche (Argentina)
Publication Date:
Jul 01, 2010
Product Type:
Thesis/Dissertation
Report Number:
INIS-AR-V-254
Resource Relation:
Other Information: Thesis(Ms.D)
Subject:
97 MATHEMATICAL METHODS AND COMPUTING; IMAGES; MATHEMATICAL LOGIC; NEURAL NETWORKS; PATTERN RECOGNITION; THREE-DIMENSIONAL CALCULATIONS
OSTI ID:
21392622
Research Organizations:
Comision Nacional de Energia Atomica, Centro Atomico Bariloche, San Carlos de Bariloche (Argentina). Laboratorio de cavitacion y biotecnologia
Country of Origin:
Argentina
Language:
Spanish
Other Identifying Numbers:
TRN: AR10V0254120058
Availability:
Available from: Repositorio Institucional, Centro Atomico Bariloche, Comision Nacional de Energia Atomica (AR), http://ricabib.cab.cnea.gov.ar/
Submitting Site:
INIS
Size:
92 p. pages
Announcement Date:
Feb 04, 2011

Thesis/Dissertation:

Citation Formats

Carnevale, Federico J. Orientation of a 3D object: implementation with an artificial neural network using a programmable logic device;Orientacion de un objeto 3D : implementacion de redes neuronales artificiales utilizando logica programable. Argentina: N. p., 2010. Web.
Carnevale, Federico J. Orientation of a 3D object: implementation with an artificial neural network using a programmable logic device;Orientacion de un objeto 3D : implementacion de redes neuronales artificiales utilizando logica programable. Argentina.
Carnevale, Federico J. 2010. "Orientation of a 3D object: implementation with an artificial neural network using a programmable logic device;Orientacion de un objeto 3D : implementacion de redes neuronales artificiales utilizando logica programable." Argentina.
@misc{etde_21392622,
title = {Orientation of a 3D object: implementation with an artificial neural network using a programmable logic device;Orientacion de un objeto 3D : implementacion de redes neuronales artificiales utilizando logica programable}
author = {Carnevale, Federico J}
abstractNote = {Complex information extraction from images is a key skill of intelligent machines, with wide application in automated systems, robotic manipulation and human-computer interaction. However, solving this problem with traditional, geometric or analytical, strategies is extremely difficult. Therefore, an approach based on learning from examples seems to be more appropriate. This thesis addresses the problem of 3D orientation, aiming to estimate the angular coordinates of a known object from an image shot from any direction. We describe a system based on artificial neural networks to solve this problem in real time. The implementation is performed using a programmable logic device. The digital system described in this paper has the ability to estimate two rotational coordinates of a 3D known object, in ranges from -80{sup 0} to 80{sup 0}. The operation speed allows a real time performance at video rate. The system accuracy can be successively increased by increasing the size of the artificial neural network and using a larger number of training examples;La extraccion de informacion compleja a partir de imagenes es una habilidad clave en las maquinas inteligentes con vasta aplicacion en los sistemas automatizados, la manipulacion robotica y la interaccion humano-computadora. Sin embargo, resulta una tarea extremadamente dificil de resolver con estrategias clasicas, geometricas o analiticas. Por lo tanto, un enfoque basado en aprendizaje a partir de ejemplos parece mas adecuado. Esta tesis trata acerca del problema de orientacion 3D, cuyo objetivo consiste en estimar las coordenadas angulares de un objeto conocido, a partir de una imagen tomada desde cualquier direccion. Se describe un sistema, basado en redes neuronales artificiales, para resolver este problema en tiempo real. La implementacion, capaz de funcionar a frecuencia de video, se realiza utilizando un dispositivo de logica programable. El sistema digital final demuestra la capacidad de estimar dos coordenadas de rotacion de un objeto 3D conocido en rangos de -80{sup 0} a 80{sup 0}. Su velocidad de funcionamiento permite la operacion a frecuencia de video. La precision del sistema puede incrementarse sucesivamente aumentando el tamano de la red neuronal artificial y utilizando una mayor cantidad de ejemplos de entrenamiento}
place = {Argentina}
year = {2010}
month = {Jul}
}