Fast Dynamic Mosaicing and Person Following

Abstract

A system for video surveillance purposes in wide areas based on active cameras, also capable to follow a person in the scene by keeping him framed, is presented. The proposed approach is based on the so-called direction histograms to compute the ego-motion and on frame differencing for detecting moving objects. It exploits post-processing and active contours to extract precise shape of moving objects to be fed to a probabilistic algorithm to track moving people in the scene. Person following, instead, is based on simple heuristic rules that move the camera as soon as the selected person is close to the border of the field of view. Experimental results on a live active camera demonstrate the feasibility of real-time person following.

Enabling technologies on hybrid camera networks for behavioral analysis of unattended indoor environments and their surroundings

“…In literature the active camera tracking is usually faced using motion compensation, affine transformations or depth sensors [32,22,27]. We also propose an uncalibrated approach based on a kernelbased tracking (similar to [8,13]), or a particle filter tracking [34] that feeds a PTZ guidance module in order to keep the tracked person within the field of view of the PTZ camera (the so-called person following task [37]). These approaches need an object model that might be initially provided with the object visual features extracted from the fixed camera tracking, and then possibly refined during the active camera tracking.…”

“…The PTZ cameras have a double task: primarily the PTZ capability of spanning over a wide view of the environment is used to build up and then keep updated a global mosaic of the observed scene (or a portion of it) [37]. The bottom line idea is to compare the actual mosaic of the environment against a background mosaic image, in order to highlight differences that might point out changes on the infrastructure, detecting ordinary objects moves (e.g.…”

“…Beyond the people tracking, consistent labeling and entrance/exit monitoring, also more advanced tasks are performed; for instance, active tracking of moving objects through PTZ cameras: this is particularly helpful in case an object is leaving the field of view of the fixed cameras. Additionally, the PTZ cameras are used for a further task: a complete/partial mosaic of the environment [37] is built and kept updated (excluding the moving objects), and compared to the previous versions of it: this allows to detect anomalies in the environment, such as object misplacements or disappearances. The data produced by this layer is handed off to the reasoning layer.…”

“…After a manual off-line calibration, consistent labeling is performed according to [9]. The PTZbased mosaic is build with an optical-flow compensation, based on [37], and the mosaic differencing is performed with the same background modeling used over the fixed cameras. The mosaic is performed on a 180 • wide pan angle, and 90 • wide tilt angle (from horizontal to down-ward vertical orientation).…”

Cognitive visual tracking and camera control

“…The former tries to preserve the scale of the object, the latter chooses those observation parameters that yield a minimum uncertainty in a Kalman filter context. Another example is the person-following active camera system proposed by Prati et al [2], where the camera control is implemented through a simple heuristic rule that re-centres the target in the FoV whenever it is close to the image's border.…”