Combining Leaf Salient Points and Leaf Contour Descriptions for Plant Species Recognition

Manual Plant identification done by experts is tedious and time consuming. This process needs to be automatic and easy to handle by the different stakeholders. In this paper, we propose an original method for plant species recognition, based on the leaf observation. We consider two sources of information: the leaf margin and the leaf salient points. For the leaf shape description, we investigate the shape context descriptor and two multiscale triangular approaches: the well-known triangle area representation (TAR) and the triangle side length representation (TSL). We propose then their combination with a shape-context based descriptor that represents the spatial correlation between the leaf salient points and the leaf margin. Experiments are carried out on three public leaf datasets. Results show that our approach achieves a high retrieval accuracy and outperforms state-of-art methods.