一种基于D_S推理的异源信息目标识别方法.pdf

29 5 Vol.29 No. 5 2007 5 Systems Engineering and Electronics May 2007 : 1001506X(2007) 0507 03 DS 1 2 3 3 于 昕, 韩崇昭, 潘 泉 , 谢明志 ( 1. 西安电子科技大学智能信息处理研究所, 陕西 西安710071; 2. 西安交通大学电信学院综合自动化所, 陕西 西安71006 ; 3. 西北工业大学自动化学院, 陕西 西安710072) : 讨论了多帧异类信源的目标识别问题, 将DS 证据推理 用于多传感器的数据融合, 将异类辨识框 架下的证据体组合得到新置信指派, 利用多个传感器共同或联合操作的优势来提高目标识别的有效性和精确性 同时给出基于基本概率赋值决策的目标识别时序累积方法最后通过仿真验证了该方法的有效性, 可信度及准 确性得到提高, 具有比累积前更可信的特点 : 信息融合; 目标识别; 证据理论; 时序累积法 :T P391.4 : A Method based on evidence theory for multisource target recognition 1 2 3 3 YU Xin , HAN Chongzhao , PAN Q uan , XIE M ingzhi (1. Inst. of Intellig ent I nf ormation P rocessing , X idian Univ . , X i an 710071, China; 2. Dep t. of utomatic Control, X ianJ iaotong Univ . , X i an 710068, China; 3. Dep t. of utomatic Control, N orthwesternP olytechnical Univ . , X i an 710072, China) Abstract: The problems of target identification based on multiple and asynchronous inf ormation sources are discussed. DempsterShafer evidence reasoning is applied to combine different evidences coming from various sensors. Then the new confidence function can be gotten. A t the same time, a method of consequent target identif ication based on the decision from basic belief assignment is put forward. Finally the effectiveness of this method is demonstrated. T he results of simulation can prove that the new method is more reliable and accurate. Keywords: information fusion; target recognition; evidence theory; consequent target identif ication 0 , ,