论文部分内容阅读
为缩小图像的低层特征与高层语义之间的语义鸿沟,基于支持向量机的相关反馈机制受到越来越广泛的关注,但这种方法并没有利用未标记样本的隐含信息.为更好地利用这些信息,提出将直推式支持向量机作为反馈过程中的学习算法.通过分析其所用特征向量的特点,设计一种颜色稀疏特征,并将其与纹理特征结合作为图像描述的特征.实验结果表明该方法较令人满意,同时也说明直推式支持向量机可在文本分类以外的领域取得较好结果.
In order to reduce the semantic gap between low-level features and high-level semantics, the related feedback mechanism based on SVM has drawn more and more attention, but this method does not use the implicit information of unlabeled samples. Using these information, we propose a direct push support vector machine (SVM) as the learning algorithm in the feedback process.An analysis of the characteristics of the eigenvectors used in this paper designs a color sparse feature and combines it with the texture features as the features of the image description. The results show that this method is more satisfactory, and also shows that the direct support vector machine can achieve better results in the field except text classification.