27 research outputs found
A framework for interpreting, modeling and recognizing human body gestures through 3D eigenpostures
No full textAccurate characterization of embedded Structure from Motion
No full textTrajectory estimation and 3d scene reconstruction from single camera, e.g. Structure from Motion, is going to have a central role in the future of automotive industry. Typical appliance fields will be: Collisions avoidance with any kind of object (people included), parking assisted maneuvers and many more. Indeed various countries are becoming more and more concerned about road traffic safety and therefore through its 'Advanced Program', EuroNCAP rewards vehicle manufacturers who employ Advanced Safety Technologies that assists the driver. This paper had mainly two different goals: (1) to describe the implementation of a state of art Structure from Motion pipeline able to run in real time with embedded fish-eye camera, which includes nonlinear optimization (i.e. local bundle adjustment); (2) to demonstrate quantitatively its performances on a synthetic test space specifically designed for its characterization in term of accuracy
Visual Search of multiple objects from a single query
No full textHundreds of millions of images are uploaded to the cloud every day. Innovative applications able to analyze and extract efficiently information from such a big database are needed nowadays more than ever. Visual search is an application able to retrieve information of a query image comparing it against a large image database. In this paper a Visual Search pipeline implementation is presented able to retrieve multiple objects depicted in a single query image. Quantitative and qualitative precision results are shown on both real and synthetic datasets
Early Detection of Partially Emerged Large Scale Marine Debris Based on Laser Pulses
No full textThe Large Scale Marine Debris (LSMD), or drifting objects, pose a serious threat to navigation safety, be they containers, dispersed cargo, large trunks, marine animals, small boats or other large-sized scattered materials. Specifically, in the case of containers, it is estimated that over 10,000 of the approximately 100 million that cross seas and oceans are dispersed each year. Several reported incidents have caused significant dents on ships hulls and in some cases, breaches leading to the sinking of vessels. In many cases, on-board devices such as radar or sonar are able to effectively detect completely emerged or submerged obstacles, respectively, but the presence of partially emerged floating bodies is more difficult to detect due to the significant disturbance introduced by the wave motion. In this article, we present a method based on the photothermal effect which allows for the detection of floating objects even at great distances (500m) through the collimated light of a laser, enabling turning operations for even large vessels and avoiding collision with the obstacle
Eyewear with eye-tracking system
No full textThe eyewear comprises:
A frame with a front, two temples coupled to the front, and at least one lens coupled to the front.
An eye-tracking system integrated into the frame, featuring:
At least one infrared (IR) radiation source positioned to emit IR radiation toward the user's eye when the eyewear is worn.
Multiple IR detectors placed along the edge of the lens to detect optical signals reflected by the eye and generate corresponding electric detection signals.
A processing and control unit linked to the IR source and detectors. This unit:
Activates the IR source.
Processes the electric signals from the detectors.
Estimates the gaze direction based on the received signals.
A battery to power the processing unit, IR detectors, and IR source.
Key Advantages:
Seamless integration into the eyewear’s design.
Simplified and cost-effective production compared to existing solutions.
Real-time gaze tracking for applications like augmented reality (AR), accessibility, or user interaction
Fast Skin Segmentation on Low Resolution Grayscale Images for Remote PhotoPlethysmoGraphy
No full textFacial skin segmentation is an important preliminary task in many applications, including remote PhotoPlethysmoGraphy (rPPG), which is the problem of estimating the heart activity of a subject just by analysing a video of their face. By selecting all the subjects skin surface, a more robust pulse signal could be extracted and analyzed in order to provide an accurate heart activity monitoring. Single-Photon Avalanche Diode (SPAD) cameras have proven to be able to achieve better results in rPPG than traditional cameras. Altought this kind of cameras produces accurate photon counts at high frame rate they are able to capture just grayscale low resolution images. For this reason, in this work, we propose a novel skin segmentation method based on deep learning that is able to precisely localize skin pixels inside a low resolution grayscale image. Moreover since the proposed method makes use of depthwise separable convolutional layers it could achieve real time performances even when implemented on a small low powered IoT device
Deep Learning Coronary Artery Centerlines Mapping from Contrast-Enhanced CT Images of the Heart
No full textAccurate omnidirectional multi-camera embedded structure from motion
No full textTrajectory estimation and 3D scene reconstruction from multiple cameras (also referred as Structure from Motion, SfM) will have a central role in the future of automotive industry. Typical appliance fields will be: autonomous navigation/guidance, collisions avoidance against static or moving objects (in particular pedestrians), parking assisted maneuvers and many more. The work exposed in this paper had mainly two different goals: (1) to describe the implementation of a real time embedded SfM modular pipeline featuring a dedicated optimized HW/SW system partitioning. It included also nonlinear optimizations such as local and global bundle adjustment at different stages of the pipeline; (2) to demonstrate quantitatively its performances on a synthetic test space specifically designed for its characterization. In order to make the system reliable and effective, providing the driver or the autonomous vehicle with a prompt response, the data rates and low latency of the 5G communication systems appear to make this choice the most promising communication solution
Visual Odometry from Omnidirectional Images for Intelligent Transportation
No full textIn this article we use omnidirectional images obtained from equirectangular panoramas of Google MapsTM to estimate camera egomotion. The systems was also tested using a 360 camera. The goal is to provide an effective and accurate positioning system for indoor environments or in urban canyons where GPS signal could be absent. We reformulated classical Computer Vision geometrical constraints for pin-hole cameras, like epipolar and trifocal tensor, to omnidirectional cameras obtaining new and effective equations to accurately reconstruct the camera path using couples or triplets of omnidirectional images. Tests have been performed on straight and curved paths to validate the presented approaches
A novel Eye-tracking system based on multiple infrared LEDs and photo-detectors for smartglasses applications
No full textLAUREA MAGISTRALENegli ultimi decenni è stato osservato un crescente interesse nei dispositivi di Eye Tracking
(ET), ossia tecnologie in grado di misurare ed esamninare i movimenti e le posizioni degli
occhi di una persona. Analizzando cosa e dove le persone stiano guardando è possibile
studiare diversi aspetti umani legati alla cognizione, al comportamento e all’interazione
con l’ambiente. Date le sue peculiarità, l’ET è stato sfruttato in diversi settori quali psicologia,
medicina, sport o aviazione, attraverso l’utilizzo di occhiali, elettrodi o maschere.
Tuttavia, le tecniche maggiormente utilizzate attualmente sono o invasive e poco pratiche
o sono operabili solo tramite strumenti difficili da usare quotidianamente.
Questa tesi, essendo parte del Joint Research Centre Project promosso da Luxottica S.p.A.
per la realizzazione di un innovativo paio di smartglasses, propone una nuova tecnica di
ET basata sull’utilizzo di emettitori e ricevitori ad infrarossi. Considerando che questi
ultimi saranno poi posizionati sulla montatura dell’occhiale, gli scopi di questo studio
hanno riguardato la verifica della tecnica proposta e la localicazzione dei punti migliori
sulla montatura dove poter posizionare i sensori.
Lo studio è stato condotto su diversi modelli realistici di occhio umano dove test, analisi
spettrali ed elaborazione di immagini hanno permesso di definire i punti al centro degli assi
orizzontali e gli angoli inferiori della montatura attorno alla lente come le zone migliori
dove collocare i fotorivelatori. Inoltre, è stato mostrato come il possibile utilizzo di più di
quattro fotorivelatori risulterebbe in un aumento notevole in termini di punti dove poter
sistemare i sensori, lasciando quindi al produttore una certà libertà nel decidere dove
poterli posizionare.In the last decades it has been observed a growing interest in the Eye Tracking (ET)
devices, namely pieces of technologies able to measure and examine the person’s eyes’
movements and positions. By analysing where and what people are looking at, it could
be possible to study lots of aspects related to human cognition, behaviour and interaction
with the environment. Given its peculiarities, ET has been employed in a variety
of sectors such as psychology, medicine, sport or aviation, either by means of glasses,
electrodes or mask. Though, the techniques majorly used nowadays are either invasive
and unpractical or have resulted in the realisation of devices hard to use on a daily basis.
This thesis, being part of a Joint Research Centre Project promoted by Luxottica S.p.A.
for the realisation of an innovative pair of smartglasses, proposes a new ET technique
based on the exploitation of IR emitters and photodetectors. With these sensors being
placed on the glasses’ frame, the goals of this work refer to the validation of the technique
itself and the localization of the frame’s spots on where to locate the sensors.
The study has been conducted on different realistic eye models where testings, spectral
analysis and images elaboration have allowed to define the spots in the middle of the
frame’s horizontal axis and in the bottom corners as the optimal zones where to put the
photodetectors. Additionally, it has been shown that if more than four receivers are decided
to be exploited, then the number of candidates points for their localization increase
considerably, leaving the producer a certain freedom when deciding where to place them
