Classification System for Wood Recognition using K-Nearest Neighbor with Optimized Features from Binary Gravitational Algorithm
Woods species recognition is a texture classification difficulty that has been studied by many researchers years ago. The species of the wood are identified by the proposed classification using the textural type that can be observed on the structural features for example the colour of the woods, we...
Saved in:
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Conference or Workshop Item |
| Language: | en |
| Published: |
2014
|
| Subjects: | |
| Online Access: | http://eprints.utem.edu.my/id/eprint/11673/1/2014_Conf._13Feb_ICRET2014_BATAM_%284%29.pdf http://eprints.utem.edu.my/id/eprint/11673/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Woods species recognition is a texture classification difficulty that has been studied by many researchers
years ago. The species of the wood are identified by the proposed classification using the textural type that can be observed on the structural features for example the colour of the woods, weight, texture and other features. Any mistakes on texture recognition will affect the economic benefits of wood production where it is an important basis for an identification of woods. Besides, to guide a person to be skilled in wood recognition, it will take a long time and the result the wood sample can be biased. These kinds of problem can be a motivation to develop a system that can recognize the wood effectively. This project will try to integrate both attempts by proposing a classification system consists of feature extractor, classifier and optimizer. The project proposes a classification system
using Gray Level Co-Occurrence Matrix (GLCM) as feature
extractor, K-Nearest Neighbor (K-NN) as classifier and Binary Gravitational Search Algorithm (BGSA) as the optimizer for GLCM’s feature selection and parameters. For this project, images of wood knot from CAIRO UTM database are used for benchmarking the proposed system performance. The result shows that the proposed approach can perform as good as previous literatures with fewer features used as input for the classifier. |
|---|