基于树皮纹理特征识别的木材分类

Wood Classification Based on Recognition of Bark Texture Feature

为了利用树皮的纹理特征对木材的种类进行识别,分别提出了基于手工设计的方法和基于深度学习的方法.基于手工设计的方法分别提取树皮纹理图像的精细纹理特征和粗略颜色信息,并用最近子空间分类器(Nearest Subspace Classifier,NSC)对提取的特征进行分类;基于深度学习的方法利用VGG-16模型预先设定训练的参数,并利用现有的纹理图像样本对模型进行微调,从而获得最佳的树皮纹理图像识别模型.最后,在New-BarkTex树皮纹理库上进行仿真实验,实验结果表明这两种方法都取得了良好的识别效果,尤其是基于深度学习的方法具有更好的识别性能.

To implement wood classification based on the recognition of bark texture feature,this paper proposes a hand-crafted-feature based method and a deep-learning based method,respectively.The hand-crafted-feature based method first extracts the fine texture features and coarse color information,and then discriminates the extracted features by nearest subspace classifier to classify the wood.The deep-learning based method uses the pre-trained VGG-16 model,and then fine-tunes the model with the bark texture images to obtain the best model.The experimental results on the New-BarkTex database demonstrate that the proposed method can achieve high performance in terms of recognition accuracy,with superior performance of the deep learning based method.