Paralelização de um classificador contextual de imagens / Parallelimplementationofcontextualclassiferforremotesensingimages

AUTOR(ES)

Rui Nelson Taborda Almeida

DATA DE PUBLICAÇÃO

2000

RESUMO

This work presents basic concepts about digital image processing, with emphasis on classification of images from multi-spectral sensors. We have chosen two classification algorithms (MAXVER and ICM), to exemplify the application of the proposed methodologies. Before showing these methodologies, we evaluated the situation of present algorithms, in the SPRING system. We described the aspects related to parallel systems and the standard of communication by Message Passing Interface (MPI). We adjusted the current programs (sequential)to be executed outside of the SPRING system. The objective was to optimize the tests and the evaluation of the results. From the code of these programs, we developed able versions to classify images using parallel processing, based on message passing interface with MPI. During the development of the programs, the objective was to increase the classification performance, using a portable code across parallel systems. The new programs were tested in parallel systems with different architectures. We used both images with low and with high volume of information. We calculated the times of processing in regard to aspects such as: selected algorithm, communication, I/O, information volume, etc. The parallel programs were evaluated in their aspects of performance and efficiency. To assess the quality of the results, we compared the resulting images of the parallel case with the resulting images of the sequential case. We confirmed that the classification can be optimized, with reduction of processing time. Furthermore, the developed programs can be used in parallel systems with different architectures, without changes in their original code. Thus, we concluded that the methodologies used in this work are very important to the development of systems for image processing.

ASSUNTO(S)

classificação de imagens computaÇÃo aplicada programação paralela parallel programming desempenho de sistemas computacionais image classification processamento de imagens processamento paralelo image processing parallel processing (computers) computer systems performace

Documentos Relacionados

• Paralelização de um classificador contextual de imagens
• Contextual medical image viewer.
• Classificador neural híbrido para imagens obtidas por sensoriamento remoto
• Imagens digitais aplicadas para determinação de áreas agrícolas através do classificador Bayes.
• Classificador de gestos das mãos baseado em imagens para aplicação em interfaces.