Shindong Wang

Habitat classification acts as the crucial role for structuring and understanding of the natural world. Natural habitats have been defined as “terrestrial or aquatic areas distinguished by geographic abiotic and biotic features, whether entirely natural or semi-natural” in the European Union Habitats Directive. The classification for natural habitats has been carried out for a long time and attracted attentions of environmental agencies to monitor and maintain habitats. There are multiple scheme have been developed to categorize characterization, and to reduce the complexity present of habitats. One of the most widely employed approaches is the Phase 1 Habitat Survey scheme. This hierarchical classification divides habitats into ten categories and provides detailed records of vegetation for a determined area. By following Phase 1 scheme, we propose to use machine learning, more specifically, deep learning and “Convolutional Neural Networks” (CNN) techniques to classify and annotate automatically for habitat images