目的 针对不均衡的印刷标志图像训练集构建的二分类模型，对少类的印刷套不准图像识别准确率低的问题，研究不均衡印刷标志图像套准状态的单分类模型识别方法。方法 提出支持向量数据描述方法，实现多类的印刷套准图像和少类的印刷套不准图像的准确识别。采用多类的印刷套准图像训练支持向量数据描述，构建模型。采用网格寻优方法和交叉验证方法确定模型的最佳参数 和 。利用模型对印刷标志图像套准状态进行识别。结果 采用文中提出的支持向量数据描述方法，对印刷标志图像套准状态识别获得的总体识别率a为0.9500，印刷套准图像和印刷套不准图像识别准确率的几何平均数Gmean为0.9513。结论 文中提出的方法获得的总体识别率a和识别率的几何平均数Gmean要优于实验中的其他方法。

The paper aims to research a single classification model recognition method for imbalanced printing mark images in registration state to solve the problem that the binary classifier model constructed from the imbalanced training set of printing mark images cannot recognize the minority printing misregistration images accurately. The support vector data description (SVDD) method was proposed to recognize the majority registration images and the minority misregistration images accurately. The majority printing registration images wereused for training of SVDD, and the SVDD model wasconstructed. The grid optimization method and cross validation method wereused to determine the optimal parameters and of the model. The registration of the printing mark images was identified with the model. The proposed SVDD methodachieved 0.9500 of overall recognition accuracya, and 0.9513 of geometric mean of the recognition accuracy Gmean. The overall recognition accuracy a and the geometric mean of the recognition accuracy Gmean obtained by the proposed method are superior to that of other methods in the experiment.