基于多类SVM和D-S证据理论的决策融合算法研究

Research on Decision Fusion Algorithm Based on Multiclass SVM and D-S Evidential Theory

D-S证据理论是决策融合领域研究较多的一种有效方法。然而,如何根据实际情况构造D-S证据理论中的基本概率赋值函数是必须面对的一个重要课题。本文提出了一种基于多类支持向量机和D-S证据理论的决策融合算法,将多类支持向量机作为局部判决器,构造了相应的基本概率赋值函数,然后用D-S证据理论对各初步判决结果进行融合,得出对目标的最终识别结论。最后与投票表决法对比,做出仿真,并进行分析,验证了算法的合理性和有效性。

D-S evidential theory is an effective method of decision fusion algorithms. However, constructing the basic probability assignment function（BPAF） based on practice is a problem that must be faced. In this paper, a method of decision fusion based on multiclass SVM and D-S evidential theory is presented to try to solve the problem, muhiclass SVM is used as the local classifier, and the BPAF is constructed accordingly, then the primary results are fused to obtain the ultima result using D-S evidential theory. Finally , an experiment is taken to compare it with majority vote rule to verify its rationality and validity.