Recognition of Leaf Image Set Based on Manifold-Manifold Distance

Abstract

Recognizing plant leaves has so far been a difficult and important work. In this paper, we formulate the problem of classifying leaf image sets rather than single-shot images, each of sets contain leaf images pertain to the same class. We compute the distance between two manifolds by modeling each leaf image set as a manifold. Specifically, we apply a clustering procedure in order to express a manifold by a collection of local linear models which are depicted by a subspace. Then the distance is measured between local models which come from different manifolds constructed above. Finally, the problem is transformed to integrate the distances between pairs of subspaces from one of the involved manifolds. Experiment based on the leaves (ICL) from intelligent computing laboratory of Chinese academy of sciences shows the method has great performance.