الفهرس 
INTRODUCTIONOnly 14 pages are availabe for public view
 |  | from | 154 |  |  |
| | | from | 154 | | |
Abstract
This work presents a new system, proper for the plant classification. It depends on the leaf texture, neither on its color nor its shape which are naturally changeable over seasons. In the present system, statistical approaches for feature extraction phase including Local Binary Pattern are used, and it is found that the histogram properties are a very powerful tool for describing the characteristics of the texture image. Also Combined Classifier Learning, Vector Quantization and Radial Function for classification phase are also satisfactorily used. Perhaps the most important factor affecting the computational classification is the length of feature vectors. Computational simplicity is another advantage of the proposed method as the features can be obtained with a few and simple calculations and comparisons. The system is compared with previous widely used classification approaches, and it is found that the present approach reaches an accuracy 98.7% for the classification process. Since the present approach depends on the texture leaf, and since the microscopic internal structure is regularly repeated on a very small microscopic scale in the leaf, this enables the botany researcher to satisfactorily assign the correct plant position in the classification, using a small bit of the leaf, even the leaf is damaged.