Robust Point Cloud Segmentation of Rodents using Close Range Depth Cameras in Controlled Environments

Abstract

Rodent behavior analysis is a extremely important task for pre-clinical testing of new drugs and neurodegenerative diseases. Some of the underlying mechanisms regulating natural interactions among multiple animals require long term interaction, which implies the usage of automated, objective and systematic video analysis system. However, current video analysis systems are limited by idiosyncratic features of the 2D video technology, and requiring a great number of parameters. This early stage work focuses on the first step of a depth-based tracking system, offering some methods to robustly perform rodent segmentation in a controlled environment, fully exploiting its geometrical properties, providing a qualitative overview of the results attained so far.