Imaging system for classification of local flora of Uttarakhand region

Many plants are facing the risk of extinction due to unplanned urbanization and over growth of population. Digital databases of plants should be maintained for proper tracking of local flora and making data-driven policies/decisions for their preservation. Plant identification is important for medical as well as educational purposes but maintaining an exhaustive digital database is a challenging task due to the presence of large number of plant species. This paper proposes a system for building a digital database of local flora and recognizing different plants using their leaf images. The system proposed in this paper involves four steps: 1) image acquisition, 2) image preprocessing, 3) feature extraction, and 4) classification. Images are acquired using commonly available general purpose desktop scanner with white paper as a background. In the image-preprocessing module, the system applies several image-processing techniques to prepare a leaf image for the feature extraction process. Then twelve leaf-shape based features are estimated and IBI classifier is used to classify the plant species. The proposed system was used to build a dataset of local flora of Uttarakhand region, consisting of 1684 images of thirty-two different plant species. The database contains around fifty leaves of each plant species. The proposed system gives promising results with an average classification accuracy of 79% for these thirty species of plants.