Classification of leaf shapes and estimation of leaf mass

The goal of classifying leaves is to find potential correlation between visually-unrelated leaves and link them with leaf mass. We apply cluster analysis to distinguish over 20 typical kinds of leaves extracted from arbor species. This approach can preferably rule out odd-shaped leaves and classify them according to the metrics. In consideration of possible effect of external factors, we also propose growth pattern model to better simulate leaves' growth, thus helping estimate leaf mass on a tree more precisely. Finally, we find function which can reflect the relationship between trunk diameter and leaf mass on an allometric tree based on pipe-model.