Michela Farenzena

Person re-identification by symmetry-driven accumulation of local features

In this paper, we present an appearance-based method for person re-identification. It consists in the extraction of features that model three complementary aspects of the human appearance: the overall chromatic content, the spatial arrangement of colors into stable regions, and the presence of recurrent local motifs with high entropy. All this information is derived from different body parts, and weighted opportunely by exploiting symmetry and asymmetry perceptual principles. In this way, robustness against very low resolution, occlusions and pose, viewpoint and illumination changes is achieved. The approach applies to situations where the number of candidates varies continuously, considering single images or bunch of frames for each individual. It has been tested on several public benchmark datasets (ViPER, iLIDS, ETHZ), gaining new state-of-the-art performances.

Background subtraction for automated multisensor surveillance: a comprehensive review

Background subtraction is a widely used operation in the video surveillance, aimed at separating the expected scene (the background) from the unexpected entities (the foreground). There are several problems related to this task, mainly due to the blurred boundaries between background and foreground definitions. Therefore, background subtraction is an open issue worth to be addressed under different points of view. In this paper, we propose a comprehensive review of the background subtraction methods, that considers also channels other than the sole visible optical one (such as the audio and the infrared channels). In addition to the definition of novel kinds of background, the perspectives that these approaches open up are very appealing: in particular, the multisensor direction seems to be well-suited to solve or simplify several hoary background subtraction problems. All the reviewed methods are organized in a novel taxonomy that encapsulates all the brand-new approaches in a seamless way.

Multiple-Shot Person Re-identification by HPE Signature

In this paper, we propose a novel appearance-based method for person re-identification, that condenses a set of frames of the same individual into a highly informative signature, called Histogram Plus Epitome, HPE. It incorporates complementary global and local statistical descriptions of the human appearance, focusing on the overall chromatic content, via histograms representation, and on the presence of recurrent local patches, via epitome estimation. The matching of HPEs provides optimal performances against low resolution, occlusions, pose and illumination variations, defining novel state-of-the-art results on all the datasets considered.

Structure-and-motion pipeline on a hierarchical cluster tree

This papers introduces a novel hierarchical scheme for computing Structure and Motion. The images are organized into a tree with agglomerative clustering, using a measure of overlap as the distance. The reconstruction follows this tree from the leaves to the root. As a result, the problem is broken into smaller instances, which are then separately solved and combined. Compared to the standard sequential approach, this framework has a lower computational complexity, it is independent from the initial pair of views, and copes better with drift problems. A formal complexity analysis and some experimental results support these claims.

Improving the efficiency of hierarchical structure-and-motion

We present a completely automated Structure and Motionpipeline capable of working with uncalibrated images with varying internal parameters and no ancillary information. The system is based on a novel hierarchical scheme which reduces the total complexity by one order of magnitude. We assess the quality of our approach analytically by comparing the recovered point clouds with laser scans, which serves as ground truth data.

Globally convergent autocalibration using interval analysis

We address the problem of autocalibration of a moving camera with unknown constant intrinsic parameters. Existing autocalibration techniques use numerical optimization algorithms whose convergence to the correct result cannot be guaranteed, in general. To address this problem, we have developed a method where an interval branch-and-bound method is employed for numerical minimization. Thanks to the properties of interval analysis this method converges to the global solution with mathematical certainty and arbitrary accuracy and the only input information it requires from the user are a set of point correspondences and a search interval. The cost function is based on the Huang-Faugeras constraint of the essential matrix. A recently proposed interval extension based on Bernstein polynomial forms has been investigated to speed up the search for the solution. Finally, experimental results are presented.

Social interactions by visual focus of attention in a three-dimensional environment

In human behaviour analysis, the visual focus of attention (VFOA) of a person is a very important cue. VFOA detection is difficult, though, especially in a unconstrained and crowded environment, typical of video surveillance scenarios. In this paper, we estimate the VFOA by defining the Subjective View Frustum, which approximates the visual field of a person in a three-dimensional representation of the scene. This opens up to several intriguing behavioural investigations. In particular, we propose the Inter-Relation Pattern Matrix, which suggests possible social interactions between the people present in a scene. Theoretical justifications and experimental results substantiate the validity and the goodness of the analysis performed.

Hierarchical structure-and-motion recovery from uncalibrated images

We describe a hierarchical structure-from-motion pipeline.No information is needed beside images themselves.The pipeline proved successful in real-world tasks. This paper addresses the structure-and-motion problem, that requires to find camera motion and 3D structure from point matches. A new pipeline, dubbed Samantha, is presented, that departs from the prevailing sequential paradigm and embraces instead a hierarchical approach. This method has several advantages, like a provably lower computational complexity, which is necessary to achieve true scalability, and better error containment, leading to more stability and less drift. Moreover, a practical autocalibration procedure allows to process images without ancillary information. Experiments with real data assess the accuracy and the computational efficiency of the method.

Towards a Subject-Centered Analysis for Automated Video Surveillance

In a typical video surveillance framework, a single camera or a set of cameras monitor a scene in which human activities are carried out. In this paper, we propose a complementary framework where human activities can be analyzed under a subjective point of view. The idea is to represent the focus of attention of a person in the form of a 3D view frustum, and to insert it in a 3D representation of the scene. This leads to novel inferences and reasoning on the scene and the people acting in it. As a particular application of this proposed framework, we collect the information from the subjective view frusta in an Interest Map, i.e. a map that gathers in an effective and intuitive way which parts of the scene are observed more often in a defined time interval. The experimental results on standard benchmark data witness the goodness of the proposed framework, encouraging further efforts for the development of novel applications in the same direction.

Part-based human detection on Riemannian manifolds

In this paper we propose a novel part-based framework for pedestrian detection. We model a human as a hierarchy of fixed overlapped parts, each of which described by covariances of features. Each part is modeled by a boosted classifier, learnt using Logitboost on Riemannian manifolds. All the classifiers are then linked to form a high-level classifier, through weighted summation, whose weights are estimated during the learning. The final classifier is simple, light and robust. The experimental results show that we outperform the state-of-the-art human detection performances on the INRIA person dataset.