Person re-identification consists in recognizing an individual in diverse locations and time over different non-overlapping camera views, considering a large set of candidates. The re-identification task is fundamental for a set of surveillance applications specially when dealing with large and structured environments such as museums, shopping malls, airports, convention centers, universities, etc. The task is challenging because it must be robust to changes in perspective, occlusions, human pose deformation, illuminations variance.
The project is mainly focused on investigating new descriptors and methodologies to deal with the re-identification task. As the cameras do not provide resolution high enough to work with facial or iris recognition, the classical solutions normally relies on appearance information, i.e., clothings and accessories. These appearance-based methods build the person signature by extracting features from specific body parts of the human silhouette, like a-symmetric parts and pictorial structures.
The novel technology of RGBD cameras such as Microsoft Kinect® and Asus Xtion Pro® are utilized to acquire depth information simultaneously to the classical RGB Images. This sanctions the re-identification solution to work with shapes and other geometrical information and therefore produces solutions which can be robust the vicissitudes in appearance, sanctioning the re-identification task to be done in spawn of days or weeks instead of intra-day.
References:
- M. Farenzena, L. Bazzani, A. Perina, V. Murino, M. Cristani
"Person re-identification by symmetry-driven accumulation of local features
23rd IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2010, San Francisco, USA - D. S. Cheng, M. Cristani, M. Stoppa, L. Bazzani, V. Murino
"Custom pictorial structures for re-identification"
22nd British Machine Vision Conference (BMVC), 2011, Dundee, UK
