Robust real-time detection of multiple balls on a mobile robot

In this paper, we present a robust method for detecting multiple balls without using color information. The method builds upon the boosted Haar classifier approach introduced by Viola et al. [11] and is applied onto an image from a catadioptric camera without performing rectification of the image. We compare the method with one based on a classical color segmentation approach. Originally both methods have been developed for the participation on the 2012 “SICK robot day”, an international robotics competition. The task was to collect balls with a robot and put them into dedicated perches within a certain amount of time. Both methods showed to be robust during the competition. Our team placed second out of fifteen contenders.