Position stability system for unmanned aerial vehicle based on video stream data analysis

There are many smallunmanned aerial vehicles or copters, which are very sensitive to air flow influence, the main problems are flying at low altitudes and taking off. When UAV takes off the ground, it creates a turbulent airflow that reflects from the surfaces and makes the copter move chaotically, so it becomes hard to control, it is also very dangerous for the flying machine itself and for the people around.

One of the possible solutions could be a stability system based on machine vision methods. Computer vision provides not only a way to setup a hovering mode, but a technique, that allows us to create multipurpose sensors, which are similar to a human eye. This kind of a stability system consists of an onboard computing unit and a camera. Anything connected with computer vision could be done with a desktop computer, but thecopter based system should be lightweight and have the ability to analyze large amount of data, running on low power sources. In this case computer boards similar to Raspberry Pi could be useful.

The main idea is to create a system that functions like an optical computer mouse, and in this case it is necessary to use a phase correlation methodimplemented in the software. This technique gives us an opportunity to estimate the relative translative offset between two similar images. The camera is attachedto the bottom side of UAV, and the whole system is connected to the flight controller via serial interface.The program, written in Python, compares every two captured frames with each other in order to find out were there any offset or not, and to get the actual value of this offset. The software computes the direction in which correction actions should be done and sends these data to the flight controller to perform roll and pitch.

The copter with this kind of a stability system will follow a helical trajectory above the limited area. This makes UAV more controllable and gives the ability to perform complex tasks such as completely autonomous flight.

This system has a lot of improvement potential. A camera attached to the bottom could be a multipurpose sensor to measure a traveled distance or to recognize patterns, painted on the ground, and a main board could be an additional computing resource for other applications.

It should be noted, that this kind of a stability system has some disadvantages. It is impossible to predict how this system will function without light, but this problem could be solved by adding infrared lights to the UAV, also the algorithm is sensitive to the altitude of flight.

References
1. Welcome to OpenCV-Python Tutorials’s documentation! [Электронный ресурс]. URL: https://opencv-python-tutroals.readthedocs.io/en/latest/, (Дата обращения: 19.09.2015).
2. Яа8рЬеггуР1в России [Электронный ресурс]. URL: http://raspberrypi.ru/, (Дата обращения: 21.12.2015).
3. WelcometoPython.org // Python Software Foundation (US) [Электронныйресурс]. URL: https://www.python.org / (Дата обращения: 21.12.2015).

Abstract. The article describes a computer vision-based method of UAV stabilization. Phase correlation algorithm is explained. The functionality of this system is also reported, and possible problems are considered.
Keywords: unmanned aerial vehicle, stabilization, computer vision, camera.

А. О. Лекарев