You are here
Person Re-Identification in Camera Networks
Person re-identification consists in matching observations of individuals across disjoint views in a network of surveillance cameras (this task is some times also referred to as multi-camera single person tracking).
This is a non-trivial problem because the appearance of individuals varies greatly through the scenes, due to possibly different acquisition devices and ambient illuminant, changes in viewpoints, illumination conditions, shadows, occlusions, different pose/orientation of the person that has to be searched for, as well as the presence of other similar individuals that populate the scenes.
Reidentification methods can be roughly divided into single-shot and multiple-shot approaches. The former have only one occurrence of the individual to be searched, while the latter integrate information over time using multiple views of the subject by tracking she in the video-stream where she is indicated as suspect by an operator (or by an intelligent module of the surveillance platform). The features to describe the suspect can be biometric (face, gait, height) and/or appearance-based (clothes, pieces of clothes, case). The selection depend on the resolution of the images and the filed of view. In any case, features are used to build a signature of the person. Then they are extracted from the frames of the video streams captured by the surveillance cameras, possibly in restricted regions, e.g. only where people move, to be compared with the signature of the suspect, therefore detecting possible locations of her presence.
TeV contributed to the development of a video-surveillance platform with re-identification modules designed to assist security operators, therefore providing hypothesis of re-identifications.
ReferenceS. Messelodi, C.M. Modena
Boosting Fisher Vector based Scoring Functions for Person Re-Identification
Image and Vision Computing, Vol. 44, pp. 44-58, 2015
For performance comparison we provide all the random partitions of four databases (VIPeR, 3DPeS, PRID2011, iLIDS119) that we used to train and test the our method, called BFiVe.
- Person Re-ID 2011 Dataset (Austrian Institute of Technology). Specifically created for testing person re-identification approaches
- caviar4reid extracted from the CAVIAR dataset (University of Edinburgh) for evaluating person re-identification algorithms. From CAVIAR clips from Shopping Center in Portugal recorder from two different points of view
- VIPeR (UC Santa Cruz) Couple of images of the same person
- GRID Queen Mary University of London) underGround Re-Identification Dataset
- 3D PeS (University of Modena and Reggio Emilia, Italy) A dataset for people tracking and reidentification
- Outdoor (University of Modena and Reggio Emilia, Italy) Multicamera disjoint views
- i-LIDS UK government's benchmark for for Video Analytics systems developers. Not for free.
- CUHK01 and CUHK02 (Chinese University of Hong Kong) Person Re-identification Dataset. Require via e-mail.