Detection of Compound Leaves for Plant Identification

Automatic plant identification has been an important issue in the last years. Most of the state-of-the-art methods for this purpose use leaf features to predict the species. Despite there are many methods to extract different leaf features, just few of them are focused on discriminating between simple and compound leaves. In this work, we introduce a method to detect compound leaves. Our method uses concentric circles to explore the surface of the leaf to count the changes of color in binary images, then, the changes are analyzed to detect compound leaves. The method predicts correctly more than 96% of the leaves in the Flavia data set. We also tested the method with some images of leaves available on the Internet, with 100% of correctness. The information on whether a leaf is ​​or not compound, was used on experiments to observe whether this improves the performance of classifiers.