Extraction of Moving Objects in the Presence of Scale Transformations in a Sequence of Images

Abstract

The paper considers the problem of extraction of moving objects by subtracting the background in the presence of scale geometric transformations. A multi-reference algorithm is used as an algorithm for estimation of scale transformations. An algorithm for automatic extraction of moving objects is proposed, taking into account the geometric transformations caused by the movement of the image sensor. The results of experimental studies obtained using real-world videos affected by scaling using software. During the testing of the algorithm, the influence of the scale estimation parameter on the quality of object extraction was investigated. The algorithm proposed shows sufficiently higher efficiency in comparison with commonly used approach. The algorithm can be used in machine vision systems of aircrafts, mobile robots, and video surveillance systems.