A Web-based Plant Identification Application Using Multi-Organ Images Query

In this paper, we describe a plant identification system based on various image organs query. The proposed system handles several issues such as preprocessing, feature extraction and description, classification method and fusion techniques. For the preprocessing step, we aims to separate the Regionof-interests (e.g., fruits, flowers, leaves) from complicated backgrounds. The relevant preprocessing techniques for each type of organ are selected based on its characteristics. For the feature extraction and description, we propose to use kernel-based descriptor (KDES). For each type of organ, we proposed to use a suitable type of kernel. Then, we investigate an efficient fusion scheme that combines classification results from different organ types in order to build the retrieved observation list. Finally, we have deployed the proposed system in form of a web application for easy using