surveillance

BVTsurveillance

The project aims at building a flexible active video surveillance platform able to automatically collect data and detect events or the potential threats, designed to assist security operators in their decisions. The platform has to handle all tasks related to the automatic management of a video surveillance system, thus relieving the operator from routine checks and permitting her/him to concentrate solely on abnormal events reported by the system. The goal is to implement a modular, flexible, and scalable architecture able to adapt to different operational scenarios.

TeV role is the development and implementation of various computer vision modules to be integrated in the video surveillance platform: background updating, people detection and tracking, people re-identification in different camera streams.

Duration: from July 2011 to September 2013

Partners: 

• BV Tech Ricerca

• Università degli Studi di Trento

• FBK (TeV research unit)

Person re-identification in camera networks consists in matching observations of individuals across disjoint views in a network of surveillance cameras. 

The appearance of individuals varies greatly through the scenes, due to possibly different acquisition devices and ambient illumination, 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.

Re-identification 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 tracked in the video-stream upon the first indication as suspect given by the operator. The appearance-based features to describe the suspect are used to build a "signature" of the person. Then the frames of the video streams captured by other surveillance cameras are analyzed, possibly only in compatible times and restricted regions (e.g. when the person might be where motion is detected), generating local signatures as well. Signatures are compared with the descriptor of the suspect and if they are similar likely locations are suggested. The main challenge are: which features to consider and how to define the similarity between them.

In this project TeV developed a single-shot re-identification module and integrated it in the video surveillance system, in order to propose hypothesis of re-identifications. It is a supervised method to compute a scoring function that, when applied to a pair of images, provides a score expressing the likelihood that they depict the same individual. 

Using the implemented person detection and tracking modules we obtain multiple descriptors of the same person, therefore producing a more robust output in the re-identification task. 

related publications

S. Messelodi, C.M. Modena, Boosting Fisher Vector based Scoring Functions for Person Re-Identification, Image and Vision Computing, Vol. 44, pp. 44-58, 2015

N. Conci, F.G.B. De Natale, S. Messelodi, C.M. Modena, M. Verza, and R. Fioravanti. An integrated framework for video surveillance in complex environments. IEEE International Smart Cities Conference - ISC2,  2016