Automated plant recognition system based on multi-objective parallel genetic algorithm and neural network
Plant recognition system is a system that recognizes the species of plants automatically. The applications of this system are in medicine, botanical research and agriculture. In the recent years, lack of sufficient botanist increases the need for computerized system. Also, it can be seen that workin...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2014
|
| Subjects: | |
| Online Access: | http://psasir.upm.edu.my/id/eprint/60076/1/FK%202014%2067IR.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Plant recognition system is a system that recognizes the species of plants automatically. The applications of this system are in medicine, botanical research and agriculture. In the recent years, lack of sufficient botanist increases the need for computerized system. Also, it can be seen that working with these systems are more convenient and quick when dealing with huge data. The problem with the existing plant recognition system is the lack of method to find the best structure for their classifiers. This work presents some contributions to plant recognition system. Number of samples involving Flavia, Citrus and Coleus were collected. Then, suitable features including texture and shape were extracted from the dataset. Texture features involved the middle energy and the middle entropy and shape features involved statistical characterizations including variance, median, standard deviation and mean. Next, the classification was carried out. First, the best set of structures for feed forward neural network were found by multi objective parallel genetic algorithm. This approach regarded three criteria involving mean square error, Akaike information criterion and minimum description length to rate different feed forward neural network structures and to select the best set of them. Lastly, feed forward neural network with the best structures were applied to classify the dataset. This method resulted around 99% of classification rate. To conclude, multi objective parallel genetic algorithm can automatically tune feed forward neural network to classify the dataset with a good classification rate. |
|---|