基于PCR-DSmT的序列帧融合景像匹配算法

针对INS/SMNS组合导航系统中,单帧景像匹配难以判断匹配结果是否正确并给出准确的匹配置信度的情况,提出一种基于PCR-DSmT的序列帧融合景像匹配算法。算法分为单帧粗匹配和序列帧融合精匹配两步:首先提取图像的相位一致性特征并采用快速归一化互相关算法初步匹配;然后建立序列帧时空约束关系,利用相关阵中极大峰构建辨识框架,采用层次分析法自适应计算置信指派并利用适配因子进行折扣运算,最后采用证据推理组合规则融合并根据判决准则输出匹配位置及置信度或对错误匹配结果报警。针对Dempster组合规则在高冲突序列帧融合时出现错误以及DSmT组合规则在多证据融合时正确位置置信指派难以增大并收敛的问题,提出一种PCR-DSmT组合规则。采用真实航拍图像和对应的Google earth卫星基准图像的仿真实验验证了匹配算法的有效性。

A Novel and Effective Scene Matching Algorithm for INS/SMNS Integrated Navigation of UAV

A novel scene matching algorithm based on sequential frames fused by PCR-DSmT is put forward for INS/SMNS integrated navigation system of UAV(unmanned aerial vehicle),where PCR stands for Proportional Conflict Redistribution,DSmT stands for Dezert-Smarandache Theory,INS stands for Inertial Navigation System,and SMNS stands for Scene Matching Navigation System.Section 1 of the full paper explains this algorithm,which we believe is novel and effective.The core of section 1 consists of: "Single frame course matching and sequential frame precise matching are involved in this algorithm.Phase congruency image is extracted and fast NCC(Normalized Cross Correlation) matching is processed in the first step,and then spatial-temporal correlation of sequential frames is founded and identification framework is constructed with the maximal peaks of the matching correlation matrix.Then the basic belief assignment of the peaks is calculated self-adaptively and discounted by suitable-matching factors.A PCR-DSmT combination rule is proposed to fuse the data of sequential frames because of the illogical fusion of Dempster rule and DSmT.At last matching confidence or error alarm is outputted according to the decision criterion." Simulation results,presented in Tables 1 through 4,and their analysis show preliminarily that our novel algorithm is indeed feasible and effective.