Image Registration for Gait Analysis of Farm Animals

概要

論文の詳細を見る
Development of gait recognition technology has opened new realm of application, gait diagnosis, which aims to diagnose disease related with gait. This challenge, if successfully applied to farm animals, will drastically reduce the cost for managing farm animals. As the first step to analyzing gait, silhouettes of subjects are essential from which most of gait features are extracted. Importantly, a sequence of images from a static perspective is desirable for background subtraction-based extraction. However, a static camera is unavailable in the farms due to probable infection of diseases from animals and only hand-held cameras are being allowed. In this study, we propose a method that registers images from a hand-held camera in order to overcome such constraints in the farms. For this purpose, we exploited fences to estimate the position of the camera and aligned frames to based on the initial frame. An experiment showed that registration worked well to some extent in terms of adjacent frames, although accumulation errors still remained.
2014-02-24