Publications

This paper reports an application of trajectory analysis in which forensics and video surveillance techniques are jointly employed for providing a new tool of multimedia forensics. Advanced video surveillance techniques are used to extract from a multi-camera system the trajectories of the moving people which are then modelled by either their positions (projected on the ground plane) or their directions of movement. Both these two representations can be very suitable for querying large video repositories, by searching for similar trajectories in terms of either sequences of positions or trajectory shape (encoded as sequence of angles, where positions do not care). Preliminary examples of the possible use of this approach are shown.

In this paper we present a system for automatic segmentation, annotation and image retrieval based on content, focused on illuminated manuscripts and in particular the Borso D'Este Holy Bible. To enhance the interaction possibilities with this work, full of decorations and illustrations, we exploit some well known document analysis techniques in addition to some new approaches, in order to achieve good segmentation of pages into meaningful visual objects with the relative annotation. We wanted to extend the standard keyword-based retrieval approach in a commentary with a modern visual-based retrieval by appearance similarity: an entire software user interface for exploration and visual search of illuminated manuscripts.

This paper presents a method for recognizing human actions in a multi-camera setup. The proposed method automatically extracts significant points on the human body, without the need of artificial markers. A sophisticated appearance-based tracking able to cope with occlusions is exploited to extract a probability map for each moving object. A segmentation technique based on mixture of Gaussians is then employed to extract and track significant points on this map, corresponding to significant regions on the human silhouette. The point tracking produces a set of 3D trajectories that are compared with other trajectories by means of global alignment and dynamic programming techniques. Preliminary experiments showed the potentiality of the proposed approach.

Recognizing different actions with a unique approach can be a difficult task. This paper proposes a novel holistic representation of actions that we called "action signature". This 1D trajectory is obtained by parsing the 2D image containing the orientations of the gradient calculated on the motion feature map called motion-history image. In this way, the trajectory is a sketch representation of how the object motion varies in time. A robust statistical framework based on mixtures of von Mises distributions and dynamic programming for sequence alignment are used to compare and classify actions/trajectories. The experimental results show a rather high accuracy in distinguishing quite complicated actions, such as drinking, jumping, or abandoning an object.

AD-HOC copes with the problem of multiple people tracking in video surveillance in presence of large occlusions. The main novelty is the adoption of an appearance-based approach in a formal Bayesian framework: the status of each object is defined at pixel level, where each pixel is characterized by the appearance, i.e. the color (integrated along the time) and the likelihood to belong to the object. With these data at pixel-level and a probability of non-occlusion at object-level, the problem of occlusions is addressed. The method does not aim at detecting the presence of an occlusion only, but classifies the type of occlusion at a sub-region level and evolve the status of theobject in a selective way. The AD-HOC tracking has been tested in many application for indoor and outdoor surveillance. Results on PETS2006 test set are reported where many people and abandoned objects are detected and tracked.

This paper presents the Visor (VIdeo Surveillance Online Repository) project designed with the aim of establishing anopen platform for collecting, annotating, retrieving, sharingsurveillance videos, and of evaluating the performanceof automatic surveillance systems. The main idea is to exploitthe collaborative paradigm spreading in the web communityto join together the ontology based annotation andretrieval concepts and the requirements of the computer visionand video surveillance communities. The ViSOR openrepository is based on a reference ontology which integratesmany concepts, also coming from LSCOM and MediaMillontologies. The web interface allows video browse, queryby annotated concepts or by keywords, compressed videopreview, media download and upload. The repository containsmetadata annotations, which can be either manuallycreated as ground truth or automatically generated by videosurveillance systems. Their automatic annotations can becompared each other or with the reference ground-truth exploitingan integrated on-line performance evaluator.

Recognizing and isolating cancerous cells from non pathological tissue areas (e.g. connective stroma) is crucial for fast and objective immunohistochemical analysis of tissue images. This operation allows the further application of fully-automated techniques for quantitative evaluation of protein activity, since it avoids the necessity of a preventive manual selection of the representative pathological areas in the image, as well as of taking pictures only in the pure-cancerous portions of the tissue. In this paper we present a fully-automated method based on unsupervised clustering that performs tissue segmentations highly comparable with those provided by a skilled operator, achieving on average an accuracy of 90%. Experimental results on a heterogeneous dataset of immunohistochemical lung cancer tissue images demonstrate that our proposed unsupervised approach overcomes the accuracy of a theoretically superior supervised method such as Support Vector Machine (SVM) by 8%.

This paper presents a novel and robust approach to consistent labeling for people surveillance in multi-camera systems. A general framework scalable to any number of cameras with overlapped views is devised. An off-line training process automatically computes ground-plane homography and recovers epipolar geometry. When a new object is detected in any one camera, hypotheses for potential matching objects in the other cameras are established. Each of the hypotheses is evaluated using a prior and likelihood value. The prior accounts for the positions of the potential matching objects, while the likelihood is computed by warping the vertical axis of the new object on the field of view of the other cameras and measuring the amount of match. In the likelihood, two contributions (forward and backward) are considered so as to correctly handle the case of groups of people merged into single objects. Eventually, a maximum-a-posteriori approach estimates the best label assignment for the new object. Comparisons with other methods based on homography and extensive outdoor experiments demonstrate that the proposed approach is accurate and robust in coping with segmentation errors and in disambiguating groups.