信息准则在信息融合中的应用  

Application of Information Criteria in Information Fusion

作　　者：夏飞[1] 张玲芳[1] 

机构地区：[1]上海电力学院电力与自动化工程学院,上海200090

出　　处：《计算机仿真》2008年第3期218-220,274,共4页Computer Simulation

基　　金：上海市教委青年基金项目(kz2004-17);上海市重点学科建设项目资助(P1301)

摘　　要：在图像识别时,如果样本较小,往往会造成识别结果的不准确。采用信息融合的方法可以改善这一问题。在这一过程中,通过引入信任度系数对信息源的质量作出评价,可以提高图像识别的准确性。提出了一种利用信息准则来确定信息源信任度系数的方法,并以此作为衡量信息源质量的依据。其中着重比较了根据两种不同的信息准则所确定的信任度系数对于识别结果的影响。实验结果表明利用BIC准则确定的信任度系数可以得到更好的图像识别结果,同时也证明了该方法应用在信息融合中的可行性。Information fusion is used in image recognition with a small quantity of samples to ameliorate the result of classification. If the quality of information sources is considered in the information fusion process, the accuracy of recognition will be improved. A method for defining the coefficient of confidence of each information source by using information criteria was proposed in this paper. Especially, the results of the recognition of image with the different confidence coefficients which were obtained by the different information criteria were compared. The result of recognition with the method of BIC information criteria was better. It was proved in the experiment. The results of experiment showed also the feasibility of the proposed method.

关 键 词：信息准则 信息融合 信息源 信任度系数 图像识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]