Tracking of small animals in laboratory An application of CUDA technology for the optimization of computer vision algorithms

Abstract

This article describes a system that aims to track small animals in the context of laboratory tests in all lighting conditions. The proposed system consists of a 3D sensor and a GPU accelerated computing unit, equipped with CUDA (Compute Unified Device Architecture) technology. The data acquired by the 3D sensor, i.e. the camera, is processed by an algorithm that uses parallelization techniques to detect small animals in real time. To assess the benefits of such parallelism, it was compared with a non-parallel algorithm. Although the research is still at an early stage, the preliminary results demonstrate that the proposed method has potential to be applied in a laboratory environment.