Real-Time Color Coded Object Detection Using a Modular Computer Vision Library

Abstract

In this paper we present a novel computer vision library called UAVision that provides support for different digital cameras technologies, from image acquisition to camera calibration, and all the necessary software for implementing an artificial vision system for the detection of color-coded objects. The algorithms behind the object detection focus on maintaining a low processing time, thus the library is suited for real-world real-time applications. The library also contains a TCP Communications Module, with broad interest in robotic applications where the robots are performing remotely from a basestation or from an user and there is the need to access the images acquired by the robot, both for processing or debug purposes.  Practical results from the implementation of the same software pipeline using different cameras as part of different types of vision systems are presented. The vision system software pipeline that we present is designed to cope with application dependent time constraints. The experimental results show that using the UAVision library it is possible to use digital cameras at frame rates up to 50 frames per second when working with images of size up to 1 megapixel. Moreover, we present experimental results to show the effect of the frame rate in the delay between the perception of the world and the action of an autonomous robot, as well as the use of raw data from the camera sensor and the implications of this in terms of the referred delay.