Shape-only Features for Plant Leaf Identification

TL;DR

This paper introduces a new shape-only feature set for plant leaf identification that achieves high accuracy and invariance to rotation and scale, suitable for mobile deployment.

Contribution

The paper proposes a novel shape-only feature set including local area integral invariants for plant leaf identification, demonstrating competitive accuracy without relying on color or texture features.

Findings

Over 90% classification accuracy on multiple datasets

Top-four accuracy exceeds 98% on several datasets

Features are rotation and scale invariant

Abstract

This paper presents a novel feature set for shape-only leaf identification motivated by real-world, mobile deployment. The feature set includes basic shape features, as well as signal features extracted from local area integral invariants (LAIIs), similar to curvature maps, at multiple scales. The proposed methodology is evaluated on a number of publicly available leaf datasets with comparable results to existing methods which make use of colour and texture features in addition to shape. Over 90% classification accuracy is achieved on most datasets, with top-four accuracy for these datasets reaching over 98%. Rotation and scale invariance of the proposed features are demonstrated, along with an evaluation of the generalisability of the approach for generic shape matching.