Camera observation has become an important field in robotics and autonomous vehicle control. Nowadays, it is not only used for monitoring the operation of the robot and possibly stop it in an emergency situation, but the robot may also use the obtained information for controlling itself. The most useful information are the position and orientation of the robot, which can be obtained via visual feedback. The camera can be installed onboard, or placed at fixed ground locations around the vehicle. The first step of this method is to run image processing algorithms on the recorded pictures or video stream. These algorithms must have minimal latency. In case of using onboard camera, its size, weight and possible costs have to be minimal. Nevertheless, it has to provide as good pictures as possible (both in terms of resolution and refresh rate), because this is the basis of image processing algorithms.
The project’s goal was to make a general purpose camera module which can be used onboard, or at fixed positions for an autonomous indoor quad rotor helicopter, which is developed at the Department of Control Engineering and Informa-tion Technology of the Budapest University of Technology and Economics.
In my thesis I show how to connect a LeopardBoard 365 Global Shutter camera module to a BeagleBoard xM embedded system, how to integrate its Aptina sensor’s driver into the used Ångström kernel, and finally how to grab images from it. I walk through how to start the PowerVR SGX GPU built into the BeagleBoard, and how to run OpenGL ES offscreen rendering without any windowing system. At last I demon-strate some image processing algorithms on the system, based on OpenCV.