科研之友
微信
新浪微博
Facebook
分享链接
摘要
Since there are many misclassification samples in image classification process, it results in low classification accuracy. For the purpose of effectively correcting misclassification samples, this paper presents an approach called collaborative evolution computation (CEC) strategy. On the basis of collaborative representation, we take the synergy of support samples and competition samples around misclassification samples into consideration, gradually make misclassification samples pulled into correct decision area, and finally obtain the achievement of making misclassification sample correctly classified. The experiments on simulation data set and real image set provide the validation that a misclassification sample could converge into a unique point by using CEC strategy. In the combination with ISOMAP manifold, we propose collaborative evolution model which is proved to be a stable and convergent model. Compared with state-of-the-art image classification methods, this model achieves better efficiency and classification performance on low dimensions.
- 出版日期2014
- 单位合肥工业大学
