Comprehensive study and investigation of ROS for computer vision applications using Raspberry Pi

Jignesh Patoliya, Hiren Mewada

The machine in the form of the robots has a very large community which makes impressive progress in recent trends. Progressive examplesof these types of robots are land based mobile robots, quadcopters, humanoid, etc. Motion tracking and object recognition is the base processin major robotic applications. For better flexibility and integration of robots with video processing applications, the ROS framework is largelyused. The major issue with ROS is its latency and integrity. This paper investigates the integration of the ROS framework with OpenCVlibraries on the Raspberry PI processor for video processing applications. In the proposed experiment setup, a camera node interfaced withthe raspberry PI captures images and publishes it in ROS message form on a specific topic. The subscriber node converts ROS message intoan image using cvbridge. Converted image is processed again using OpenCV library on the raspberry Pi board. The extracted informationcan be used to actuate peripheral devices interfaced with the raspberry Pi. An investigation of the raspberry Pi based implementation revealsthat ROS introduces 0.63% overhead and optimum implementation on raspberry Pi can avoid the high configured computer and raspbeery Pican process the video at 13 frames per second at most.