Application of fusion of two classifiers based on principal component analysis method and time series comparison to recognize maritime objects upon FLIR images

摘要

This paper presents a method of recognition of maritime objects based on FLIR (forward looking infra-red) sensor images using two methods: Principal Component Analysis (PCA) and Dynamic Time Warping (DTW). A combination of the Principal Component Analysis PCA with the eigenimages analysis method reduces the dimensionality of the recognition problem. DTW method finds the shortest distance between two time series allowing a transformation of time for both compared series. In the presented maritime objects FLIR images classifier the DTW method is used to compare the vertical brightness projection histograms of silhouettes for the recognized object and the object pattern. To determine the silhouette of a maritime object the Otsu thresholding algorithm is used. The paper describes the eigenimages method, the DTW method of comparing time series and the data fusion method combining conclusions both classifiers. In the final part of the paper are presented preliminary test results of the classification method for a set of maritime objects FLIR images registered in the Baltic Sea.