The Method of Guaranteeing the Separation Between the Recognised Object and Background

Abstract

The aim of the following study was to develop a procedure which guarantees the separation between the part of an image where we have the recognised object and the part of the image which corresponds to the terrain where the object moves. This research is conducted for grey scale images. The authors have presented the method which uses moment invariants for creating feature vectors which define the features of the recognised object and the features of the background. The presented method is based on calculating the distance between the values of invariant functions calculated for an object and the background. The distances were calculated for all moment invariants. These moment invariants were elements of the feature vector. In the next step the elements of the feature vector were ordered according to the values of these distances—from lowest to highest. Finally, the moment invariants, for which the distances were highest, were chosen as elements of a new—shorter feature vector. Furthermore, the algorithm of creating features vector was presented in the following paper. The developed algorithm allows to assess if a given invariant function is useful for the classification of the elements of a given set of classes. Owing to this approach, it was possible to choose properly the invariant functions which constitute the features vector. On the one hand, we can decrease the size of the features vector by choosing the invariant functions which separate particular classes in the best way. On the other hand, we know which function is the most proper to be added to the features vector when the size of the features vector is too small. On top of that, this study presents the example of recognising the object moving in some kind of a terrain.