摘要
本文提出了一种用于快速图象匹配的序贯判决方法,它以大偏差样本作为统计量,在给定配准概率P_D和伪配准概率P_F的条件下导出了门限序列的表达式。在一幅实际地形图上进行的仿真匹配结果表明这种方法比常用的MAD方法至少快一个数量级以上,且在低信噪比(SNR)下性能优于MAD。达到了某些实际系统中实时匹配的要求。因此在低SNR下,它是一种比较好的初始匹配定位方法。
Using the number of large deviation samples as statistics, a new sequential decision method for fast image matching is proposed. Under the condition that registration probability PD and false registration probability PF are given, the threshold sequence expression is derived. The method is at least over one order faster than common MAD and its performance is better than MAD in low SNR. Monte-Carlo simulation has shown that the method satisfies the demand of some systems for real time matching. Therefore it is a better initial location method under low SNR condition.
出处
《电子学报》
EI
CAS
CSCD
北大核心
1989年第3期29-35,共7页
Acta Electronica Sinica