Abstract: Recognition of plants based on plant leaves is of important aid for biological and ecological sciences. An algorithm for leaf classification based on shape and texture features is presented. Following the preprocessing of image denoising, the leaf region is obtained through segmentation and mathematical morphological operations. Shape features are extracted from the segmented binary region image, and texture features are extracted from the gray-scale image. A BP forward neural network with the features as inputs is adopted for classification. Experimental results on real-world images show that higher classification accuracy can be achieved by the proposed method compared with existing algorithms.

Key words: leaf classification, shape feature, texture, Back Propagation（BP） neural network

摘要： 根据植物叶片识别植物种类对于生物科学与生态科学具有重要的辅助作用。针对叶片分类，提出了一种基于形状与纹理特征的分类算法。在进行了去噪等预处理后，通过阈值分割和数学形态学方法获取叶片区域；在分割得到的二值区域图像上提取了形状特征，在灰度图像上提取了纹理特征；在所得特征的基础上，利用BP网络对叶片进行分类。在实际图片上的实验结果表明，相比于已有算法，该方法可以达到更高的正确分类率。

关键词: 叶片分类, 形状特征, 纹理, 反向传播（BP）神经网络