Color-based of bird species classification using Support Vector Machine / Nur Amalina Nazery

Bird classification is an important task in computer vision problem. The problem is to classify the images from the set of training images. Birds in images may also appeared in different situation such as in different sizes, different pose and angle of view. Therefore, this project proposed a prototype of bird species classification based on color features from bird images. There are three phases involved in this project which are data collection, processing (i.e feature extraction and classification) and post processing (i.e test and evaluation). For the data collection, 200 images from two different species of birds which are snowy owl and toucan has been collected from Datasets for Computer Vision Study website. All the bird image dataset are utilized as the train and test image data. The color moment extracted from the bird images in processing phase. There are nine color features experimented which are mean, standard deviation, and skewness. These nine color features are computed from the color component of red, green, and blue. The feature vectors of mean, standard deviation and skewness are then applied in Support Vector Machine to classify two group of bird species. The results proved that it significantly works on two bird species of Snowy Owl and Toucan to classify that bird images. Hence, this prototype significantly benefits to the users who are involved in ornithology and birdwatcher. In future, more features can be added in feature extraction process to produce more accurate result of classification.