Search

In this paper we propose to face the problem of event detection from single images, by exploiting both background information often containing revealing contextual clues and details, which are salient for recognizing the event. Such details are visual objects critical to understand the underlying event depicted in the image and were recently defined in the literature as “event-saliency”. Adopting the Multiple-Instance Learning (MIL) paradigm we propose a hierarchical approach analyzing first the entire picture and then refining the decision on the basis of the event-salient objects. Validation of the proposed method is carried out on two benchmarking datasets and it demonstrates the effectiveness of the proposed hierarchical approach to event discovery from single images.

The requirements of the water transport system of small herbaceous species differ considerably from those of woody species. Despite their ecological importance for many biomes, knowledge regarding herb hydraulics remains very limited. We compared key hydraulic features (vulnerability to drought-induced hydraulic decline, pressure-volume relations, onset of cellular damage, in situ variation of water potential, and stomatal conductance) of three Ranunculus species differing in their soil humidity preferences and ecological amplitude. All species were very vulnerable to water stress (50% reduction in whole-leaf hydraulic conductance [kleaf] at -0.2 to -0.8 MPa). In species with narrow ecological amplitude, the drought-exposed Ranunculus bulbosus was less vulnerable to desiccation (analyzed via loss of kleaf and turgor loss point) than the humid-habitat Ranunculus lanuginosus Accordingly, water stress-exposed plants from the broad-amplitude Ranunculus acris revealed tendencies toward lower vulnerability to water stress (e.g. osmotic potential at full turgor, cell damage, and stomatal closure) than conspecific plants from the humid site. We show that small herbs can adjust to their habitat conditions on interspecific and intraspecific levels in various hydraulic parameters. The coordination of hydraulic thresholds (50% and 88% loss of kleaf, turgor loss point, and minimum in situ water potential) enabled the study species to avoid hydraulic failure and damage to living cells. Reversible recovery of hydraulic conductance, desiccation-tolerant seeds, or rhizomes may allow them to prioritize toward a more efficient but vulnerable water transport system while avoiding the severe effects that water stress poses on woody species.

Researchers of different fields have been involved in human behavior analysis during the last years. The successful recognition of human activities from video analysis is still a challenging problem. Within this context, applications targeting elderly care are of considerable interest both for public and industrial bodies, especially considering the aging society we are living in. Ambient intelligence (AmI) technologies, intended as the possibility of automatically detecting and reacting to the status of the environment and of the persons, is probably the major enabling factor. AmI technologies require suitable networks of sensors and actuators, as well as adequate processing and communication technologies. In this paper we propose an innovative solution based on a real time analysis of video with application in the field of elderly care. The system performs anomaly detection and proposes the automatic reconfiguration of the camera network for better monitoring of the ongoing event. The developed framework is tested on a publicly available dataset and has also been deployed and evaluated in a real environment.

An approach based on a stochastic optimization technique is proposed for medical microwave imaging. The approach is based on the integral equations of the electromagnetic inverse scattering. After discretization of the continuous model, the problem solution is recast as a global optimization problem. A functional is constructed on the basis of a Markov random field model and minimized by a genetic algorithm. In order to reduce the computational load, a model of the cross section of the biological body is considered. In this way, the investigation area is limited by separating the scattering contribution of a fixed region under test from those of other parts of the model. Some preliminary results concerning a two-dimensional model of a human thorax are reported. Such a biological structure is inspected by the proposed tomographic approach in order to detect and localize the presence of an "object" modeling a tumor.

Events emerged in the last years as a driving concept to efficiently index and retrieve media. Several approaches have been proposed to analyse the relationship between events and related media, to enable event discovery, event-based tagging, or event-based media retrieval. Notwithstanding the outstanding work done by several researchers in this area, a major still unsolved problem is how to understand the inherent link between visual concepts and events, and in particular, which are the key elements that allow a human being to perceive from a media collection what is the underlying event. In this demo, we address this problem in an original way by exploiting human knowledge using a game with a hidden purpose. Users are engaged in a competition, where they can alternatively try to mask a photo collection to prevent other users to recognize the related event or they can see a photo partially masked and try to guess the event connected. A set of rules and an adequate score system force players to focus on real important details, thus allowing the detection of most event-related salient parts of the media. The final result is a saliency map that, differently from the typical concept of saliency, highlights the visual concepts that are considered fundamental to perceive the nature of an event.

Industrial and biomedical applications of microwave-imaging techniques based on inverse scattering integral relations become more and more important. In order to reduce computational costs and hence allow a quasi real-time processing, an innovative inversion procedure based on the use of a genetic algorithm and on the Sherman-Morrison-Woodbury matrix inversion method is presented. Selected numerical results concerning various scenarios and scatterer dimensions are presented in order to give some indications on the effectiveness and also current limitations of the proposed approach

This work presents a methodology to locate and characterize multiple unknown scatterers exploiting the scattered electromagnetic radiation collected on a measurement region outside the area under investigation. In many practical cases, an accurate quantitative imaging of the scenario under test is required and it can be reached by using a high resolution representation of the dielectric profile of the scatterers. Towards this aim, an enhanced iterative multi-resolution procedure that exploits a morphological processing for detecting and focusing on different non-connected regions-of-interest has been developed. A suitable set of representative numerical results will demonstrate that the proposed approach is able to efficiently detect the objects in the imaging scenario and to enhance the accuracy in reconstructing multiple scatterers. This is the author's version of the final version available at IEEE.

Subsurface imaging of civil structures is a subject of growing interest because of the need to check the integrity of buildings, roads, and bridges in a non-invasive (and certainly non-destructive way). In this framework, this study is aimed at presenting a microwave imaging technique for crack or failure detection. The proposed approach is based on a numerically computed Green's function and on an effective evolutionary optimisation technique. In order to evaluate the achieved results and to assess the effectiveness and current limitations of the proposed approach, a set of selected numerical simulation is presented.

This paper presents an innovative microwave imaging technique suitable for NDT/NDE applications. To reduce the number of problem unknowns as well a the computational load by improving the convergence rate of the process, the available a-priori information is exploited by introducing a computationally-effective procedure for the prediction of the electric field based on the updating Sherman-Morrison-Woodbury (SMW) formula. The numerical results as well as comparisons with state-of-the-art methods confirm the effectiveness, the feasibility, and the robustness of the proposed approach.

Automatic recognition of human activities and behaviors is still a challenging problem for many reasons, including limited accuracy of the data acquired by sensing devices, high variability of human behaviors, and gap between visual appearance and scene semantics. Symbolic approaches can significantly simplify the analysis and turn raw data into chains of meaningful patterns. This allows getting rid of most of the clutter produced by low-level processing operations, embedding significant contextual information into the data, as well as using simple syntactic approaches to perform the matching between incoming sequences and models. We propose a symbolic approach to learn and detect complex activities through the sequences of atomic actions. Compared to previous methods based on context-free grammars, we introduce several important novelties, such as the capability to learn actions based on both positive and negative samples, the possibility of efficiently retraining the system in the presence of misclassified or unrecognized events, and the use of a parsing procedure that allows correct detection of the activities also when they are concatenated and/or nested one with each other. An experimental validation on three datasets with different characteristics demonstrates the robustness of the approach in classifying complex human behaviors.

In the framework of NDE/NDT, this paper presents a new microwave diagnostic tool. The approach, developed in the spatial domain, is based on the numerical computation of the inhomogeneous Green's function in order to fully exploit all the available a-priori information of the domain under test. The heavy reduction of the computational complexity of the proposed procedure (with respect to standard free-space Green's function based procedures) is also achieved by means of a customized hybrid-coded genetic algorithm. In order to assess the effectiveness of the method, the results of several realistic simulations are presented and deeply discussed.

In the framework of microwave imaging techniques for NDE/NDT, this paper presents an innovative approach based on the use of the SMW inversion procedure for the electric field prediction. Starting from the integral form of inverse scattering equations, the problem of determining the presence of an unknown defect in a known host domain is recast into an optimization one by defining a suitable cost function and reducing the problem unknowns only to the flaw "descriptors". By considering the effective inversion technique based on the SMW formula, the estimation of the secondary unknowns (namely, the electric field distributions) is performed in a cost-effective way. Selected numerical results are finally presented in order to validate the approach showing current potentialities and limitations.

Industrial and biomedical applications of microwave imaging techniques based on inverse scattering integral relations have become more and more important. In order to reduce the computational costs allowing a quasi real-time processing, an innovative inversion procedure based on the use of a genetic algorithm and on the Sherman-Morrison-Woodbury matrix inversion method is presented. Selected numerical results concerning various scenarios and scatterer dimensions, are presented in order to give some indications on the effectiveness but also current limitations of the proposed approach.