Document Type : Original Article
Authors
1 Department of Agricultural Biosystems Engineering, University of Mohaghegh Ardabili, Ardabil, Iran
2 Department of Computer Engineering, Sharif University of Technology, Tehran, Iran
Abstract
The identification and classification of tree leaves hold significant importance in botanical and agricultural research. This study focuses on the development of an advanced computer vision system for classifying leaves from five distinct tree species, including pomegranate, fig, almond, raspberry, and hawthorn. The proposed system utilizes a dataset comprising 525 digital images to extract diverse features from the color domain and the gray-level co-occurrence matrix (GLCM). From each image, 126 features are extracted from the color domain and 80 features from the GLCM. The feature selection process is conducted using an advanced method that combines artificial neural networks (ANN) with ant colony optimization (ACO). This approach aids in identifying key features, including the angular second moment at 0 degrees, angular contrast at 90 degrees, angular maximum probability at 45 degrees, the normalized difference index in the CMY and HSV color spaces, and the standard deviation of the first component in the YCbCr color space. Leaf classification is performed using a hybrid ANN and metaphor competition algorithm, achieving a classification accuracy of 92.5%. This system serves as an effective tool for precise and efficient leaf classification and has the potential to lead to significant advancements in botanical research and agricultural applications. Furthermore, the results of this study indicate that the use of intelligent methods in feature selection can enhance the accuracy and efficiency of classification models. Ultimately, this research emphasizes the importance of developing advanced techniques for leaf identification and classification, representing a crucial step toward improving existing methods in this field.
Keywords
)