A Fusion approach for Bird Classification by Machine learning Techniques Ensembling Contour and Statistical Features

Abstract

Bird classification approach is a mind perplexing work especially for human vision due to the presence of enormous number of classes with identical features. The approach of sorting is more tedious in traditional methods compared to machine learning techniques because approximately there are ten thousand species grouped into thirty sets of birds from albatrosses to woodpeckers. Birds play a vital role in the landscape helping agriculturalists and to maintain the food chain of nature. But still there is no proper classification methods available to group the birds and ornithologists are in research for a standard scheme since birds accompany human race from the evolution. This paper deliberates novel method for bird classification using machine learning techniques. The classification process is enriched by feature extraction methods namely statistical features along with shape and edge detection methods. The features extracted are used for categorizing the birds and the basic four groups are considered for this method. Totally three classification techniques namely Random Forest, Support Vector Machine and KNN methods are implemented and their accuracy is compared for best selection. The precise method is Random Forest method based on the performance metrics and the accuracy is proved to be highest.