一种基于序贯三支决策的图像识别方法.

The image recognition method based on sequential three-way decisions only considers the signed distance between the testing samples and the support vector machine（SVM） separating hyperplane, ignoring the information of the existing granular layer, which can further improve the performance of image recognition by this means. This paper proposed a new dual confidence score and ensembling（DCE）, which synthesizes signed distance with the distance between the testing samples and the center of each class of training samples in the feature space as the basic confidence score. The current granularity layer integrated with the confidence score of coarse-granularity layer which insufficient to recognized images. DCE improved the accuracy of recognition by the existing information between granularity layers. Meanwhile, DCE enhanced the recognition efficiency as making decisions at the coarser granularity layer. Finally, this paper utilizes two classical pre-trained models for experimental verification on Caltech-101 and Caltech-256 datasets. The experimental results show that this approach can effectively improve the recognition accuracy and efficiency. [ABSTRACT FROM AUTHOR]