A leaf vein detection scheme for locating individual plant leaves

Abstract

Individual leaf detection from natural condition is a fundamental task for many agricultural automation systems. Individual leaf detection is challenging because of the complicity of shape variation and pose changing of living plant leaves. In this paper, we proposed a leaf detection scheme by examining leaf veins. Individual leaves could be located in the plant images and their direction could be estimated. Initially, background is removed by examining luminance and smoothness in G channel of RGB color space. Accordingly, a SKEDET method is proposed to extract candidate skeleton of leaves. The longest skeleton of a leaf is selected as the main leaf vein. Subsequently, the direction of leaf is estimated according to thickness of the vein. The experiments were carried out with sweet potato leaves. The experimental results demonstrated that the proposed method could stably detect individual leaves and their directions. The proposed method could be applied to many agricultural applications, such as plant inspection system, agricultural robotics.