Abstract: Leaf vein is inherent property of plant leaf and a significant feature for plant recognition.Extraction of leaf vein feature is good for automatic leaf recognition.It is difficult because of complicated and diversified characteristic of leaf vein.An approach based on Top-Hat transform is proposed.The direction of each pixel on leaf image is calculated by Gauss gradient.And the Top-Hat transform is carried by multidirectional Top-Hat operators.Finally，leaf vein is extracted by threshold segmentation.Compared with conventional threshold-based methods，experimental results demonstrate that it is capable of extracting more accurate modality of leaf venation for subsequent leaf recognition.

Key words: Top-Hat transform, leaf vein extraction, Gauss gradient, mathematical morphology

摘要：

叶脉是植物叶的内在特征和识别植物的重要特征，叶脉特征的提取有利于植物叶自动识别。针对叶脉复杂多变使得叶脉难于提取的特点，提出了多方向Top-Hat变换的叶脉特征提取方法。该方法采用Gauss梯度计算叶图像上各点方向，采用多方向Top-Hat算子进行Top-Hat变换，最后阈值分割提取叶脉。实验结果表明，与传统方法相比，该方法提取叶脉更精确，为下一步植物叶识别打下了良好的基础。

关键词: Top-Hat变换, 叶脉提取, Gauss梯度, 数学形态学