Abstract: In order to realize automatically identifying the register state of color printing，and achieve online detecting overprint deviation，the feature extraction method of the registering mark image is studied.The color feature extraction method based on the human visual model is presented to aim at reducing the impact of light；the texture feature extraction method is presented.The average value of the gray co-occurrence matrix in the 0°，45°，90° and 135° directions is the final characteristic parameters，which makes the texture characteristics independent of the direction.It also designs a minimum distance classifier.The identifying experiments are separately tested with applications of the different distance measure；the results indicate that it is feasible to identify register state using the features parameters of color moment and grey level co-occurrence matrix，and the experimental results indicate that the identifications through texture feature are quite accurate.

Key words: register state, feature extraction, recognition

摘要： 套准状态的自动识别是实现套印偏差在线检测的关键一步，研究了套印标志图像的特征提取方法，在图像颜色特征提取方面，提出了一种接近人的视觉模型（HSI）的颜色特征提取方法，减少了光照的影响；在图像纹理特征提取方面，将灰度共生矩阵在0°、45°、90°和135°四个方向上的特征参数求均值作为最终参数，抑制了方向分量，使得到的纹理特征与方向无关。设计了最小距离分类器进行分类识别，并应用不同距离测度进行分类对比实验，分类实验结果表明：提出的方法可行，基于纹理特征参数的分类效果优于颜色特征的分类。

关键词: 套准状态, 特征提取, 识别方法