为降低异常证据对合成结果的影响,提出了基于投影分解与 k 最近邻距离的异常证据检测算法。该算法在对证据集中所有证据进行焦元单一元素投影分解的基础上,重新构造证据的基本概率赋值,然后利用证据之间形成的欧式距离,采用 k 最近邻距离算法对异常证据进行检测。无线传感器网络应用实验分析表明:该算法可有效地对异常证据进行检测。对检测前后的证据利用证据合成规则进行融合对比结果发现,剔除了异常证据的合成结果并具有良好的峰值性和可分辨性,合成结果有利于融合决策。%The concept of abnormal evidence and its detection algorithm are analyzed in detail.Thus, a new algorithm for detecting such abnormal evidence is presented based on projection decomposition and k nearest neighbors distance.Firstly,the real evidence is decomposed with a single focal element vector to reconstruct the basic probability assignment of evidence.Then,the Euclidean distance be-tween the evidences is calculated and the abnormal evidence is detected by use of the algorithm based on k nearest neighbors distance.The applied and experimental analysis through the wireless sensor network show that the proposed algorithm is feasible and effective.Through the comparison of the combination between all pieces of evidence and the remained ones,the latter is found to be good at peak value and in resolving power,which is better for fusion decision.
展开▼