Scientific Visualization, 2018, volume 10, number 5, pages 86 - 101, DOI: 10.26583/sv.10.5.06
Development of the human face tracking algorithm based on the optical flow application
Authors: A.A. Druki1, V.G. Spitsyn2, Yu.A. Bolotova3, D. Oliva4, A.V. Gelginberg5
National Research Tomsk Polytechnic University, Russia
University of Guadalajara, Mexico
1 ORCID: 0000-0002-8221-6029, druki@tpu.ru
2 ORCID: 0000-0001-5978-1321, spvg@tpu.ru
3 ORCID: 0000-0002-8161-3575, julya.bolotova@gmail.com
4 ORCID: 0000-0001-8781-7993, inform@safety.ru
5 ORCID: 0000-0003-4800-4185, gelginberg@gmail.com
Abstract
The article is devoted to the development of algorithms and software to increase the efficiency of detection and tracking of human faces on video sequences. The review of existing methods of objects detection and tracking on video sequences indicates their advantages and disadvantages. A human face tracking algorithm based on the application of an optical flow is developed, which uses the Viola-Jones detector to objects localization. The preprocessing of the frame is carried out using the Gaussian filter, flow calculation by the pyramidal Lucas-Kanade method. The article presents the results of the algorithm software implementation with description of the main functions and classes. The developed algorithm was tested and comparison of work results with analogues was carried out.
Keywords: optical flow, object tracking, face detection, computer vision, image classification.
