Aiming at the problem that the uncertainty caused by the multiple features of color image data is easy to appear in the object feature mining, a new mining method for the target feature data in the color image database was proposed. The color image data were clustered with the subtractive clustering method, and the clustered data were classified with the outlier detection technique. In addition, the optimal target image feature data were selected with the quantum behaved particle swarm optimization method. In combination with the structural similarity calculation method, the mining of optimal target image feature data was realized. The results show that compared with the traditional mining method, the recall rate of proposed method reduces by about 17%, while the mining accuracy increases by about 28. 6%.%针对由于彩色图像数据特征较多使得目标特征挖掘容易出现不确定性的问题,提出一种新的彩色图像数据库中目标特征数据挖掘方法.采用减法聚类算法对彩色图像数据进行聚类,采用离群点检测技术对聚类数据进行分类处理,采用量子行为粒子群优化方法选取最优目标图像特征数据,并与结构相似度计算方法相结合,实现对最优目标图像特征数据的挖掘.结果证明,该方法相比传统的挖掘方法,其挖掘召回率降低了约17%,挖掘精确度提高了约28.6%.
展开▼
