Abstract: The intrinsic complexity of Protein Spatial Structure（PSS） brings a heavy difficulty to its structural classification.In this paper，PSS is mapped into 2-D distance matrix and further regarded as texture image of which textural directions are detected by Radon transform.Consequently，a novel method of feature extraction of PSS with low dimension is proposed based on gray level co-occurrence matrix and Radon transform project matrix.The experiments and the comparison results demonstrate that the presented method can not only product low-dimensional feature and also achieve effective classification of PSS.

Key words: distance matrix, gray level co-occurrence matrix, Radon transform, support vector machines, feature extraction

摘要： 蛋白质三维空间结构固有的复杂性给其结构分类带来了较大的困难。将蛋白质空间结构映射成为二维距离矩阵，并进一步视作灰度纹理图像，使用拉冬变换分析了该图像的纹理方向特性，基于灰度共生矩阵和Radon投影矩阵提出了一种低维的蛋白质结构特征提取方法。实验结果与对比表明，该方法不仅具有低维的特征，而且有效地实现了多类蛋白质结构分类识别。

关键词: 距离矩阵, 灰度共生矩阵, 拉冬变换, 支持向量机, 特征提取