Influence of the reference image formation method on the efficiency of the onboard correlation-extreme tracking system for tracking ground objects

Abstract

To improve the efficiency of the onboard correlation-extremal tracking system in the presence of isomorphic transformations of the current object image (changes in the brightness level, scale, shift and rotation of the image), as well as natural and artificial noise during the observation process, a new method of adaptive formation of the reference image is proposed. It consists in using an Interactive Multiple Model (IMM), which includes Kalman filters of the 0th and 1st order and a Singer filter of the 0th order. The proposed method of adaptation is compared with the known methods of frame-by-frame change and “exponential smoothing”. To carry out comparative modeling of a mathematical modeling complex, in which an input action simulator based on the Unity3D cross-platform environment was used, taking into account the kinematics of the movement of ground objects and the flight of an unmanned aerial vehicle in a natural coordinate system. The effectiveness of the evaluation of the tracking of ground objects of each of the evaluations was estimated by the value of the average coefficient of tracking. The obtained research results can be used in the development of reconnaissance UAVs operating in real time.