摘要
为了有效融合待识别系统的多源证据信息,提高模式识别的准确性,提出了一种多源证据信息加权融合的模式识别方法。该方法基于不同来源证据对辨识框架中各命题识别具有不同可靠性这一事实,将各证据对各命题识别的正确率转换成加权系数,通过研究证据理论的加权改进,构建加权融合的识别体系,保证各证据在模式识别过程中存在的不确定性经过融合后能够最大限度削弱,从而从理论上降低了模式识别的不确定性。实例分析表明,多源信息加权融合后的识别结果可信度明显增大、识别正确率显著提高,充分验证了该融合识别方法能够有效提高模式识别的准确性。
To efficiently combine the multi-source evidence information on a recognition system and improve its pattern recognition accuracy,a weighted fusion method for recognizing the patterns of multi-source information is proposed. The evidence source of each proposition in the recognition framework is different,and the reliability resulting from the recognition of the proposition isn＇t the same. The correct rate of recognizing all the propositions can be transformed into the weighted coefficient; thus the recognition system of the weighted fusion method is constructed by improving the weighted evidence theory to reduce the uncertainty to its minimum that certainly exists during pattern recognition and to enhance the determinacy and creditability of the weighted evidence theory. The study results show that the creditability of the recognition system is obviously enhanced and that its correct rate is distinctively increased.It is verified that the weighted fusion method can effectively improve the pattern recognition accuracy.
出处
《机械科学与技术》
CSCD
北大核心
2016年第3期381-385,共5页
Mechanical Science and Technology for Aerospace Engineering
基金
国家自然科学基金项目(51475049)
校人才引进科研基金项目(12003
12004)
湖南省"十二五"重点建设学科资助项目
湖南省教育厅资助科研项目(15C0123
12A016
14C0094)资助
关键词
证据理论
加权融合
多源信息
可靠性
模式识别
不确定性
evidence theory
multi-source information
pattern recognition
reliability
uncertainty
weighted fusion
