电子与信息学报
電子與信息學報
전자여신식학보
JOURNAL OF ELECTRONICS & INFORMATION TECHNOLOGY
2015年
9期
2040-2046
,共7页
信息融合%证据理论%Dezert-Smarandache理论%近似推理%拆分映射
信息融閤%證據理論%Dezert-Smarandache理論%近似推理%拆分映射
신식융합%증거이론%Dezert-Smarandache이론%근사추리%탁분영사
Information fusion%Evidence theory%Dezert-Smarandache Theory (DSmT)%Approximate reasoning%Splitting mapping
该文对Dempster-Shafer(DS)理论以及Dezert-Smarandache理论(DSmT)进行了深入研究,为了能够在仅需较低计算复杂度的前提下得到更加精确的融合结果,提出一种新的快速 DSmT-DS 近似推理融合方法。该方法针对超幂集空间仅单子焦元具有信度赋值的情况,将超幂集空间拆分映射成元素为各单子焦元和其补集的二元集合的新的超幂集空间,并求出每个补集的信度赋值;再运用Dezert-Smarandache框架中的第5条比例冲突分配规则(DSmT+PCR5)在新的超幂集空间的二元集合子空间下对多证据源进行融合,得到各单子焦元的融合结果;然后通过归一化处理求得各单子焦元的信度赋值。通过理论分析得出该文方法的融合结果是介于 Dezert-Smarandache框架中的第5条比例冲突分配规则(DSmT+PCR5)及Dempster-Shafer(DS)框架下的 Dempster组合规则之间。该文方法在需要较低计算复杂度的前提下,可以得到优于Dempster组合规则的近似融合结果。最后通过多个角度与已有方法进行对比,验证了该文方法的优越性。
該文對Dempster-Shafer(DS)理論以及Dezert-Smarandache理論(DSmT)進行瞭深入研究,為瞭能夠在僅需較低計算複雜度的前提下得到更加精確的融閤結果,提齣一種新的快速 DSmT-DS 近似推理融閤方法。該方法針對超冪集空間僅單子焦元具有信度賦值的情況,將超冪集空間拆分映射成元素為各單子焦元和其補集的二元集閤的新的超冪集空間,併求齣每箇補集的信度賦值;再運用Dezert-Smarandache框架中的第5條比例遲突分配規則(DSmT+PCR5)在新的超冪集空間的二元集閤子空間下對多證據源進行融閤,得到各單子焦元的融閤結果;然後通過歸一化處理求得各單子焦元的信度賦值。通過理論分析得齣該文方法的融閤結果是介于 Dezert-Smarandache框架中的第5條比例遲突分配規則(DSmT+PCR5)及Dempster-Shafer(DS)框架下的 Dempster組閤規則之間。該文方法在需要較低計算複雜度的前提下,可以得到優于Dempster組閤規則的近似融閤結果。最後通過多箇角度與已有方法進行對比,驗證瞭該文方法的優越性。
해문대Dempster-Shafer(DS)이론이급Dezert-Smarandache이론(DSmT)진행료심입연구,위료능구재부수교저계산복잡도적전제하득도경가정학적융합결과,제출일충신적쾌속 DSmT-DS 근사추리융합방법。해방법침대초멱집공간부단자초원구유신도부치적정황,장초멱집공간탁분영사성원소위각단자초원화기보집적이원집합적신적초멱집공간,병구출매개보집적신도부치;재운용Dezert-Smarandache광가중적제5조비례충돌분배규칙(DSmT+PCR5)재신적초멱집공간적이원집합자공간하대다증거원진행융합,득도각단자초원적융합결과;연후통과귀일화처리구득각단자초원적신도부치。통과이론분석득출해문방법적융합결과시개우 Dezert-Smarandache광가중적제5조비례충돌분배규칙(DSmT+PCR5)급Dempster-Shafer(DS)광가하적 Dempster조합규칙지간。해문방법재수요교저계산복잡도적전제하,가이득도우우Dempster조합규칙적근사융합결과。최후통과다개각도여이유방법진행대비,험증료해문방법적우월성。
In this paper, Dempster-Shafer (DS) theory and Dezert-Smarandache Theory (DSmT) are conducted thorough reasearch, and in order to obtain more accurate fusion results in the premise of needing less computation complexity, a fast DSmT-DS approximate reasoning method is proposed. This method is only fit for the case that there are only singleton focal elements with assignments in hyper-power set. The hyper-power set is splitted and mapped to a new hyper-power set which consists of the binary sets of the focal element and its complementary set to the assignments of the complementary sets are computed. Proportional Conflict Redistribution No.5 within Dezert-Smarandache framework (DSmT+PCR5) is applied to fuse the multi-source evidence in the binary sets of the new hyper-power set to get the fusion results of singleton focal elements. Then the assignments of singleton focal elements are obtained by normalization. Through the theoretical analysis, the conclusion is drawn that the fusion results of the mothod in this paper is between the results of DSmT+PCR5 and Dempster’s combination rule based on DS model, and the fusion results of the method in this paper which is better than the rusults of Dempster’s combination rule can be obtained in the premise of minimal computation complexity. Finally, by comparing the method in this paper with the existing methods from different views, the superiority of new one is testified well.