Recognition of Periodic Motions Using One-Dimensional Contour Based Features

In this article, a new method for detecting periodic motions in image sequences based on one-dimensional features is presented. The method starts by the detection of moving object in the consecutive frames. Then the contours of the moving target are extracted and one-dimensional features are constituted. The features are based on the calculated distance from target contour to its bounding box sides in four directions. By applying the autocorrelation to the calculated distances, four one-dimensional curves are obtained and the periodicity of the target motion is evaluated. The proposed algorithm was tested with a database of 100 videos and experimental results showed the efficiency of the proposed algorithm.