A Content-based Plant Leaf Identification System Based On Android

To automatically identify plant species is very useful for ecologists, amateur botanists, educators, and so on. Previous identification systems of plant leaves mostly work on heavy equipments. With the development of computer technology, the handheld devices have the rapid momentum. Therefore, the development of plant leaf recognition on a handheld device will have a large market and social values. However, there are still many problems.In this application, the outputs of the demo system to users are the top several species which match the query leaf image the best, as well as the textual descriptions and additional images about plant leaves, flowers, etc. Our system works well with state-of-the-art identification performance.In this paper, a method for plant image identification is proposed which is operating on handheld terminal devices. The main tasks include:(1) Considering the low hardware performance of handheld devices, we begin with the extraction and selection of image features to solve the problem of conflict between accuracy and efficiency.(2) We propose to extract several simple but complementary features to earn a high efficiency. And we select and fuse the extracted features to improve the accuracy.(3) To further improve the efficiency, we cluster the features in the database to form several initial clusters and pick the centers for identification.(4) Base on the works above, the development of a plant leaf for Android identification system.