OPTIMIZATION PARAMETERS OF INTELLECTUAL IDENTIFICATION SYSTEM OF OBJECTS ON THE TERRAIN

Abstract

Operation algorithm of object identification system on the ground with the optimization the set of features and control tolerances of their value is considered. Global descriptor for describe object image that is invariant to rotation and takes into account local features and their spatial distribution is proposed. The algorithm to calculate the proposed descriptor lies in search a vector of frequencies occurrence of characteristic points of the object in the search window and a histogram of occurrences of pairs of neighboring cell search window in which both contain cue points with the same label. The algorithm for the formation of parameterized training samples to adapt to the viewing distance when using a lens without focus-managed. The proposed algorithms to solve the problem of identification of land vehicles in the monitored area are modeled in Unity 3D virtual environment.

Keywords

• машинне навчання, розпізнавання образів, навчальна матриця, глобальний дескриптор, оптимізація, інформаційний критерій