Color-based bird image classification using Support Vector Machine

Abstract

The classification and identification of the bird species from the visual image is complex compared by using audio song. The knowledge of the features species type is very important as to ensure it is classified to the correct species. Color-based feature extraction is one of the procedure in extracting the color properties from the bird which to represent the species of the bird. However, it is a challenging task due to the environment problems which is from the background with natural habitat of the bird images. It is also difficult when the bird images come into view of different angles and sizes. Therefore, this paper proposed a solution to consumer electronic which field-portability, cost-effectiveness and easy-to-use interface that experimented on the segmented bird images as to ensure the accurate results of classification is produced. This paper investigated on nine color-based features of mean, standard deviation and skewness of each plane of red, green and blue (RGB) from the bird images. All these features are experimented on 100 images for each species of snowy owl and toucan. The bird classification using Support Vector Machine algorithm is identified as a promising method in bird classification which produced 97.14% accuracy rate for training data and 98.33% for testing data.