摘要
以中医舌像为研究对象,针对单一特征无法很好表述舌像的缺点,提出了基于多特征融合的中医舌像检索方法以提高中医舌像的识别率。由于检索结果的排序直接关系到整个系统的性能,重点研究了在训练集上学习一组合理的多特征融合参数来提高排序性能,从而达到提高中医舌像识别率的目的。该算法以最小化预测排序与目标排序之间的反序对为优化目标,最终得出一个有效的排序学习算法。在该舌图像数据集上的实验结果验证了该算法的有效性。
This paper used tongue images as study object, as the shortcomings of single feature couldn＇ t describe efficiently the information of the tongue images, proposed a traditional Chinese medical tongue images retrieval method based on multi-feature fusion to improve the recognition rate of tongue images. Since the results ranking was an important issue in the field of information retrieval, focused on learning a set of reasonable parameters of the multi-feature fusion to improve retrieval performance on the training set, thus to improve the recognition rate of traditional Chinese medicine tongue images. The optimization target of this algorithm minimized the number of the anti-order between prediction rank and true rank. Finally, worked out an effective rank learning algorithm. The experimental results on the tongue image set verify the effectiveness of the algorithm.
出处
《计算机应用研究》
CSCD
北大核心
2010年第2期791-793,共3页
Application Research of Computers
关键词
特征融合
中医
舌像检索
multi-features fusion
traditional Chinese medical
tongue image retrieval
