Abstract: Based on analysis of Gabor-filters parameters，select filters with multi-scale and multi-orientation parameters，extract texture features from the convolution results that use these filters and shard images，and reduce the dimensions of feature-data based on principal component analysis.At last，use nearest neighbor method to automatically classify and recognize these ancient shards.

摘要： 通过对Gabor滤波器相关参数的分析，提出了一组具有多尺度和多方向特征的Gabor滤波器对古瓷碎片图像进行处理，获取图像的纹理特征数据。采用主分量分析方法对所获取的纹理特征向量进行降维，利用最近邻法进行分类，实现了对古瓷碎片的自动分类识别。