3B Zernike Momentleri Kullanılarak İnsan Hareketlerinin Tanınması

Abstract

Because they are invariant under rotation, Zernike moments are widely used for shape recognition in two-dimensional images. Similarly, magnitudes of 3D Zernike moments have the analogous property in three dimensional images. In recent years, capturing 3D images of humans has become feasible as a result of advances in optical technologies. In this work, we propose a new method to classify human actions which have been recorded as three dimensional images by using 3D Zernike moments. We have applied our method to the i3DPost Multi-view Human Action Dataset. We were able to classify the actions in the dataset into main activities such as walking, running, jumping, bending, hand-waving, jumping in place and sitting with an accuracy of greater than 99 %.