Tiziana D’Orazio

Moving object segmentation by background subtraction and temporal analysis

In this paper, we address the problem of moving object segmentation using background subtraction. Solving this problem is very important for many applications: visual surveillance of both in outdoor and indoor environments, traffic control, behavior detection during sport activities, and so on. All these applications require as a first step, the detection of moving objects in the observed scene before applying any further technique for object recognition and activity identification. We propose a reliable foreground segmentation algorithm that combines temporal image analysis with a reference background image. We are especially careful of the core problem arising in the analysis of outdoor daylight scenes: continuous variations of lighting conditions that cause unexpected changes in intensities on the background reference image. In this paper, a new approach for background adaptation to changes in illumination is presented. All the pixels in the image, even those covered by foreground objects, are continuously updated in the background model. The experimental results demonstrate the effectiveness of the proposed algorithm when applied in different outdoor and indoor environments.

Object Tracking by Non-overlapping Distributed Camera Network

People Tracking is a problem of great interest for wide areas video surveillance systems. In these large areas, it is not possible for a single camera to observe the complete area of interest. Surveillance systems architecture requires algorithms with the ability to track objects while observing them through multiple cameras. We focus our work on multi camera tracking with non overlapping fields of view (FOV). In particular we propose a multi camera architecture for wide area surveillance and a real time people tracking algorithm across non overlapping cameras. In this scenario it is necessary to track object both in intra-camera and inter-camera FOV. We consider these problems in this paper. In particular we have investigated different techniques to evaluate intra-camera and inter-camera tracking based on color histogram. For the intra-camera tracking we have proposed different methodologies to extract the color histogram information from each object patches. For inter-camera tracking we have compared different methods to evaluate the colour Brightness Transfer Function (BTF) between non overlapping cameras. These approaches are based on color histogram mapping between pairs of images of the same object in different FOVs. Therefore we have combined different methodology to calculate the color histogram in order to estimate different colour BTF performances. Preliminary results demonstrates that the proposed method combined with BTF outperform the performance in terms of matching rate between different cameras.

Helipad detection for accurate UAV pose estimation by means of a visual sensor

In this article, we tackle the problem of developing a visual framework to allow the autonomous landing of an unmanned aerial vehicle onto a platform using a single camera. Specifically, we propose a vision-based helipad detection algorithm in order to estimate the attitude of a drone on which the camera is fastened with respect to target. Since the algorithm should be simple and quick, we implemented a method based on curvatures in order to detect the heliport marks, that is, the corners of character H. By knowing the size of H mark and the actual location of its corners, we are able to compute the homography matrix containing the relative pose information. The effectiveness of our methodology has been proven through controlled indoor and outdoor experiments. The outcomes have shown that the method provides high accuracies in estimating the distance and the orientation of camera with respect to visual target. Specifically, small errors lower than 1% and 4% have been achieved in the computing of measurements, respectively.

Design of a Low-Cost Vision System for Laser Profilometry Aiding Smart Vehicles Movement

We present a fast and accurate method to derive the pose of a mobile vehicle moving within bounded paths. A triangulation-based vision system made of a laser source, able to generate a line pattern, and a high speed camera is applied on the front side of an autonomous vehicle, namely the Smoov ASRV platform, which is able to store and retrieve pallets in smart warehouses. The presented system extracts the properties of the emitted laser line on the camera plane and transfers these information to the vehicle reference system. Then, the presence of constitutive landmarks along the path, i.e., holes and bends, permit the estimation of other parameters, such as vehicle speed, enabling the exact control of the vehicle. Further validations have returned accuracies lower than 2 and 3.2 % in distance and tilt measurements with respect to the rail border, respectively.

A Likelihood-Based Background Model for Real Time Processing of Color Filter Array Videos

One of the first tasks executed by a vision system made of fixed cameras is the background (BG) subtraction and a particularly challenging context for real time applications is the athletic one because of illumination changes, moving objects and cluttered scenes. The aim of this work is to extract a BG model based on statistical likelihood able to process color filter array (CFA) images taking into account the intrinsic variance of each gray level of the sensor, named Likelihood Bayer Background (LBB). The BG model should be not so computationally complex while highly responsive to extract a robust foreground. Moreover, the mathematical operations used in the formulation should be parallelizable, working on image patches, and computationally efficient, exploiting the dynamics of a pixel within its integer range. Both simulations and experiments on real video sequences demonstrate that this BG model approach shows great performances and robustness during the real time processing of scenes extracted from a soccer match.

Defective areas identification in aircraft components by bivariate EMD analysis of ultrasound signals

In recent years many alternative methodologies and techniques have been proposed to perform non-destructive inspection and maintenance operations of moving structures. In particular, ultrasonic techniques have shown to be very promising for automatic inspection systems. From the literature, it is evident that the neural paradigms are considered, by now, the best choice to automatically classify ultrasound data. At the same time the most appropriate pre-processing technique is still undecided. The aim of this paper is to propose a new and innovative data pre-processing technique that allows the analysis of the ultrasonic data by a complex extension of the Empirical Mode Decomposition (EMD). Experimental tests aiming to detect defective areas in aircraft components are reported and a comparison with classical approaches based on data normalization or wavelet decomposition is also provided.

Gesture Recognition by Using Depth Data: Comparison of Different Methodologies

In this chapter, the problem of gesture recognition in the context of human computer interaction is considered. Several classifiers based on different approaches such as neural network (NN), support vector machine (SVM), hidden Markov model (HMM), deep neural network (DNN), and dynamic time warping (DTW) are used to build the gesture models. The performance of each methodology is evaluated considering different users performing the gestures. This performance analysis is required as the users perform gestures in a personalized way and with different velocity. So the problems concerning the different lengths of the gesture in terms of number of frames, the variability in its representation, and the generalization ability of the classifiers have been analyzed.

Comparative Analysis of PRID Algorithms Based on Results Ambiguity Evaluation

The re-identification of a subject among different cameras (namely Person Re-Identification or PRID) is a task that implicitly defines ambiguities. Two individuals dressed in a similar manner or with a comparable body shape are likely to be misclassified by a computer vision system, especially when only poor quality images are available (i.e. the case of many surveillance systems). For this reason we introduce a method to find, exploit and classify ambiguities among the results of PRID algorithms. This approach is useful to analyze the results of a classical PRID pipeline on a specific dataset evaluating its effectiveness in re-identification terms with respect to the ambiguity rate (AR) value. Cumulative Matching Characteristic curves (CMC) can be consequently split according to the AR, using the proposed method to evaluate the performance of an algorithm in low, medium or high ambiguity cases. Experiments on state-of-art algorithms demonstrate that ambiguity-wise separation of results is an helpful tool in order to better understand the effective behaviour of a PRID approach.

Anomalous Human Behavior Detection Using a Network of RGB-D Sensors

The detection of anomalous behaviors of people in indoor environments is an important topic in surveillance applications, especially when low cost solutions are necessary in contexts such as long corridors of public buildings, where standard cameras with long camera view would be either ineffective or costly to implement. This paper proposes a network of low cost RGB-D sensors with no overlapping fields-of-view, capable of identifying anomalous behaviors with respect a pre-learned normal one. A 3D trajectory analysis is carried out by comparing three different classifiers (SVM, neural networks and k-nearest neighbors). The results on real experiments prove the effectiveness of the proposed approach both in terms of performances and of real time application.

Design of High-Resolution Optical Systems for Fast and Accurate Surface Reconstruction

In the last few decades, virtual reconstruction of objects has grown interest in the field of quality control. As known, smart manufactures need automatic systems for the real-time investigation of production yields, i.e. techniques, methods and technologies devoted to the analysis of quality. In particular, high-resolution systems are required for the measurement of surface profiles aimed to the detection and characterization of small defects, with sizes reaching the limit of few microns. This book Chapter describes the procedure for the design of high-resolution and high-accuracy laser scanning probes based on triangulation techniques for the exhaustive reconstruction of objects. Drawbacks and limitations will be discussed, with particular focus on occlusion problems due to possible undercut surfaces within the testing objects. A complete description of novel techniques will be thus provided together with a demonstration of a resulting optical probe. Challenging metal objects, namely drilling tools, will be then investigated, proving measurement resolutions close to the physical diffraction limit.