Object recognition and tracking using Haar-like Features Cascade Classifiers: Application to a quad-rotor UAV

Abstract

Verbatim from arXiv. Not paraphrased, not summarized.

In this paper, we develop a functional Unmanned Aerial Vehicle (UAV), capable of tracking an object using a Machine Learning-like vision system called Haar feature-based cascade classifier. The image processing is made on-board with a high processor single-board computer. Based on the detected object and its position, the quadrotor must track it in order to be in a centered position and in a safe distance to it. The object in question is a human face; the experiments were conducted in a two-step detection, searching first for the upper-body and then searching for the face inside of the human body detected area. Once the human face is detected the quadrotor must follow it automatically. Experiments were conducted which shows the effectiveness of our mythology; these results are showing in a video.

Authors

• Luis Arreola
• Gesem Gudiño
• Gerardo Flores

Keywords

• cs.RO