On the Use of Simple Geometric Descriptors Provided by RGB-D Sensors for Re-Identification

Abstract

The re-identification problem has been commonly accomplished using appearance features based on salient points and color information. In this paper, we focus on the possibilities that simple geometric features obtained from depth images captured with RGB-D cameras may offer for the task, particularly working under severe illumination conditions. The results achieved for different sets of simple geometric features extracted in a top-view setup seem to provide useful descriptors for the re-identification task, which can be integrated in an ambient intelligent environment as part of a sensor network.

People Semantic Description and Re-identification from Point Cloud Geometry

“…The experiments carried out in the previously mentioned works using the zenithal configuration cover a reduced number of different identities, up to 10 in [10], and up to 20 in [6]. In both cases the dataset was recorded in a single day session.…”

“…As already mentioned, and similarly to the approach presented in [6], [10], we have located the sensor in a top view configuration, with the aim at easing the segmentation. However, instead of developing a background subtraction technique based on the gray level assigned to the depth image as done by [6], we have designed a solution that works on the 3D coordinates.…”

“…However, instead of developing a background subtraction technique based on the gray level assigned to the depth image as done by [6], we have designed a solution that works on the 3D coordinates. The approach makes use of the Point Cloud Library (PCL) [11], defining a background model over a regularized point cloud of 640 × 320 points.…”

“…To our knowledge, there are at least two recent papers based on a zenithal RGBD sensor for reidentification purposes [6], [10]. However, we have not located any other work, using that configuration setup, related to other semantic descriptors such as gender, etc.…”

“…Current consumer depth cameras are able to provide depth information in addition to the color images. This now cheaper and compact extra information has recently been used in a number of human interaction scenarios, with remarkable results [1], including reidentification tasks [2], [3], [4], [5], [6], [7].…”