Fast identification of two-dimensional visual patterns in arbitrary position. Space of features

摘要

In the paper, we present new methods and features for identification of visual objects. The features proposed allow to apply the method of potential functions and to train the system on examples. We outline two methods for determination of the position, the angle of rotation, and the size of an object for the subsequent object identification and feature calculation. The methods applied use the input data in the form of the vector description of an image obtained by the vectorization procedure of segment borders.