Identification of butterfly species (Lycaenidae) with image process system

Human have needed to classify the objects and organisms around them. In the classification system called systematic or taxonomy, firstly morphological characters are used. Along with the developments in scientific studies, molecular biology, anatomy and caryology fields have also been included in systematic science. Because of the high number of species of biological organisms, similar and different characters, the classification system was supported by special programs and techniques. In the thesis, a computer vision based method has been proposed as an alternative to classical diagnostic methods for the determination of butterfly species. Characteristic butterfly wing texture images were used for the classification of butterfly species. In this study, LBP and GLCM image processing methods were used to extract effective properties from butterfly images. The detection of butterfly species was carried out in two stages. In the first stage, the attributes were obtained, and in the second stage, these attributes were performed by 10-fold cross validation test using ELM machine learning method. At the end of the study, classification success was obtained as 91.6031%. As a result, the proposed computer vision method will be supportive to alternative methods in the classification of butterfly species