41199 research outputs found
Sort by
Lossless compression of LiDAR frames
Full text linkLiDARs are sensors which obtain a three-dimensional mapping of the surrounding environment. In addition to provide accurate localization and mapping information, LiDARs can be used in object detection and in autonomous driving. These sensors produce a large amount of data and so it is necessary to implement compression techniques, especially if the data has to be processed in real time, e.g., for safety-critical services. In this thesis two different techniques for LiDAR frame compression were compared and implemented, 2D- and 3D-oriented. The performance had been analyzed in terms of compression efficiency, quality of the decompressed frame compared to the original one and time spent to do the compression. Only lossless compression methods have been examined, limited to intraframe operations. We demonstrated that, thanks to the form in which LiDAR frames are saved, compression methods already known like those for two-dimensional images have given equivalent results, if not better, than those designed for three-dimensional point clouds. This thesis also presented some advanced strategies based on spherical coordinates to
improve image-based compression results
Quantum Computer Simulation via Tensor Networks
Full text linkIn recent years lots of efforts have been spent in the realization of quantum computers able to reproduce quantum circuits involving increasing number of qubits with the
greatest possible accuracy. The final goal is to reach the limit, the so-called quantum supremacy, where a classical computer is no longer able to reproduce the results of
a quantum machine. Indeed, simulating quantum many-body systems is very computationally demanding due to the exponential scaling of the Hilbert space with the number of qubits.
In order to perform a classical simulation of a quantum circuit acting on a qubits register, one must choose between two possible approaches: the first is an exact description
of the qubits’ state, possible up to a maximum reachable number of qubits. The second, instead, consists of representing the state approximately. But, even quantum processors
are not able to reproduce exactly a given quantum circuit: their coupling with the environment, which is minimized but not removed by the experimental implementation, induces
errors through quantum channels like decoherence or bit-flip. Therefore an approximate representation of the qubits' state is acceptable as long as its errors are comparable
with the experimental ones. The tensor network methods allow one to approximate a quantum state by efficiently compressing its information, introducing a controllable error.
In this thesis, these methods will be used to simulate a quantum computer on a large computational cluster, to push as far as possible the classical simulation framework
Realizzazione e test di un prototipo dimostratore per la misura online della radioattività nelle acque
Full text linkLo studio di questa tesi si incentra attorno al problema del monitoraggio
dell’acqua destinata al consumo umano che riveste grande importanza per la
sicurezza e la salute della popolazione. Si inserisce all’interno dello sviluppo del
progetto TAWARA_RTM che propone una tecnica di monitoraggio real time dei
bacini di acqua. Ciò comporta il problema della contaminazione dei rivelatori
che sono in continuo contatto con l’acqua radioattiva. Lo scopo della tesi è
quello di testare nuovi rivelatori innovativi creati nei laboratori del Dip. Di
Fisica e Astronomia dell’Università di Padova, caratterizzati da un basso livello
di contaminazione. Per il lavoro di questa tesi sono stati testati preliminarmente
in aria con sorgenti di particelle alfa e beta, studiandone l’efficienza
