Abstract: Deep convolutional neural network (CNN) has recently achieved great breakthroughs in many computer vision tasks. However, its application in texture classification has not been thoroughly researched. To this end, this paper carries out a systemic research on its application in image texture classification. Specifically, CNN is used to extract preliminary image feature, and subsequent PCA (principal component analysis) operation can reduce its dimensionality to obtain final texture feature which is fed into an SVM (support vector machine) classifier for prediction. This paper does comprehensive experiments and analysis on four benchmark datasets. The results show that CNN is a better texture feature representation and achieves quite good performance in most image texture datasets. However, CNN performs worse in datasets with image noise and rotation. Thus, this paper indicates the necessity to enhance the abilities of noise tolerance and rotation invariance of CNN, and it is necessary to construct a large diverse texture dataset to guarantee its best performance in image texture classification.

Key words: texture classification, convolutional neural network (CNN), computer vision

摘要： 深度卷积神经网络（convolutional neural network，CNN）在许多计算机视觉应用中都取得了突破性进展，但其在纹理分类应用中的性能还未得到深入研究。为此，就CNN模型在图像纹理分类中的应用进行了较为系统的研究。具体而言，将CNN用于提取图像的初步特征，此特征经过PCA（principal component analysis）降维后可得到最终的纹理特征，将其输入到SVM（support vector machine）分类器中便可获得分类标签。在4个常用的纹理数据集上进行了性能测试与分析，结果表明CNN模型在大多纹理数据集上均能取得很好的性能，是一种优秀的纹理特征表示模型，但其对包含旋转和噪声的纹理图像数据集仍不能取得理想结果，需要进一步提升CNN的抗旋转能力和抗噪声能力。另外，有必要构建具有足够多样性的大规模纹理数据集来保证CNN性能的发挥。

关键词: 纹理分类, 卷积神经网络（CNN）, 计算机视觉