MEDIAN AND RECURSIVE FILTRATION
A. A. Moiseev · Radio Industry (Russia) · 2018
Possible recognition method for objects is associating of their characteristics with etalon objects ones. To overcome estimation error that locks this associating sliding statistical filtration can be used. To avoid estimation distortion it was proposed to use complex filter comprised median and recursive ones. Such a filtration suppresses estimations fluctuations and provides precision augmentation. Median and recursive filter is most effective and hence most available among other sliding filters. As filter effectiveness indicator the relation of variations was used for unfiltered and filtered evaluations. This indicator allows estimating filter influence on evaluations and can be used as optimal filter criterion. It was shown that complex filter is most effective and as a result most preferable method of sliding filtration. It was proposed to use it as base filter variant for evaluations processing.