Deep Person Re-identification with the Combination of Physical Biometric Information and Appearance Features

Abstract In this paper, we propose a novel Person Re-identification model that combines physical biometric information and traditional appearance features. After manually obtaining target human ROI from detection results, the skeleton points of person will be automatically extracted by OpenPose algorithm. Combining with (height, shoulder width.) calculated vision-based geometric estimation, further (stride length, swinging arm.) could estimated. order to improve re-identification performance, an improved triplet loss function has been applied in framework [1] where both feature are utilized full connection layer (FCL). Through experiments carried out on public datasets real school surveillance video, effectiveness efficiency proposed algorithm have confirmed.