1,721,220 research outputs found
Dalle immagini allo spazio 3D: il ruolo dei punti chiave semantici per la percezione 3D
No full textUno degli obiettivi della Computer Vision è quello di comprendere la percezione 3D umana attraverso rappresentazioni 2D come immagini e video. Estrarre informazioni 3D robuste da quest’analisi è una sfida significativa. Questa tesi si concentra su rappresentazioni 3D basate su punti chiave, esplorando applicazioni in differenti scenari reali. Differentemente dai tradizionali descrittori puntuali come ORB o SIFT, i punti chiave semantici stabiliscono correlazioni tra specifici punti 3D appartenenti a oggetti rigidi o articolati. I recenti sviluppi in Deep Learning, in particolar modo nel rilevamento di punti chiave 2D, hanno aperto la strada per affrontare problemi complessi di visione 3D. Questa tesi dimostra l’applicazione di questi metodi nella guida autonoma e nella videosorveglianza, evidenziando la robustezza e la precisione nel ridurre il divario tra il piano immagine 2D e il mondo 3D.
Nel conteso automotive, la nostro indagine si concentra sui problemi di sintetizzazione di nuove viste e di ricostruzione 3D di veicoli in ambienti urbani. La rappresentazione 3D di un veicolo in una scena può essere utile per l’analisi del traffico e la prevenzione di incidenti. Perciò, progettiamo un metodo che sfrutta la localizzazione di punti chiave 2D in modo da aumentare le caratteristiche visuali per l’accurata classificazione di modelli 3D di veicoli. Assicurando una classificazione robusta, studiamo come migliorare la generazione di veicoli sintetici da punti di vista non visti attraverso un sistema di Deep Learning trainato su un insieme di immagini da singoli punti di vista. In aggiunta, per esplorare tecniche più sofisticate di ricostruzione 3D di oggetti da immagini, introduciamo un’architettura di Deep Learning capace di ricostruire oggetti di diverse categorie. Questo approccio è allenato su un insieme di immagini da singoli punti di vista e comporta la deformazione di rappresentazioni 3D esplicite.
La seconda area di ricerca si concentra sulla predizione di scheletri 3D di persone e robot osservati da una prospettiva esterna come le camere di videosorveglianza. I punti chiave in questo contesto sono integrati nella definizione di scheletro rappresentato come un grafo di punti semantici. Il focus iniziale è sul dominio robotico dove un sistema intelligente che predice scheletri 3D può essere cruciale per la sicurezza in ambienti collaborativi condivisi da persone e robot. Considerando le difficoltà nell’ottenere dataset reali in robotica, enfatizziamo il ruolo della simulazione. Il nostro approccio comporta la raccolta di un dataset sintetico e reale, affrontando il problema della stima della posa 3D attraverso una rappresentazione a due mappe di calore. Esploriamo il divario tra il dominio sintetico e reale utilizzando le mappe di profondità per aumentare l’accuratezza. Introducendo informazioni temporali, il nostro sistema sposa il nuovo paradigma di Pose Nowcasting, in cui predire le pose future rappresenta un problema ausiliario per raffinare la precisione della posa corrente. Passando allo scenario umano, proponiamo un sistema di raffinamento della posa basato sull’analisi di mappe di profondità. Contemporaneamente, la nostra indagine si estende all’interazione uomo-computer, in cui presentiamo un metodo non supervisionato per rilevare e classificare gesti delle mani dinamici usando dati di un sensore che traccia il movimento.
Questa tesi punta a dare un valido contributo all’intersezione tra la 3D Computer Vision e il Deep Learning in vari domini. Dopo uno sguardo sullo stato dell’arte esistente sui problemi di ricostruzione 3D e stima della posa 3D, presentiamo i nostri metodi con una spiegazione tecnica esaustiva supportata da indagini dettagliate dei risultati condotte su dataset ampiamente riconosciuti in letteratura.One of the goals of the Computer Vision community is to comprehend human 3D perception through 2D representations like images and videos. Extracting robust 3D insights from these analyses is a significant challenge. This dissertation focuses on the keypoint-based 3D representation, exploring applications in different real-world scenarios. Unlike traditional pointwise feature descriptors like ORB or SIFT, semantic keypoints establish correlations between specific 3D points belonging to a rigid or articulated object. Recent advances in Deep Learning, particularly in 2D keypoints detection, have paved the way for addressing complex 3D vision problem. This thesis demonstrates the application of these methods in autonomous driving and video surveillance, showcasing their robustness and precision in bridging the gap between 2D image planes and the 3D world.
In the automotive context, our investigation centers on the tasks of novel view synthesis and 3D reconstruction of vehicles within urban scenes. A 3D representation of a vehicle in a scene can be valuable for traffic analysis and accident prevention. To achieve this, we design a method leveraging a 2D keypoint localization network to augment visual features for accurate classification of 3D vehicle models. Ensuring a robust classification, we study how to improve the generation of synthetic vehicles from unseen novel views through a deep learning pipeline trained on a collection of single-view images. Additionally, to explore more sophisticated techniques for 3D object reconstruction from images, we introduce a deep learning architecture capable of reconstructing objects across multiple categories. This approach is trained on a dataset of single-view images and involves the deformation of explicit 3D representations.
The second research area is focused on predicting the 3D skeletons of both humans and robots, observed from an external perspective, such as a video surveillance camera. The keypoints in this context are integrated into the definition of a skeleton, depicted as a graph of semantic points. Our initial focus is on the robotics domain, where an intelligent system for predicting 3D skeletons can be crucial for safety in collaborative environments shared by humans and robots. Given the challenges of obtaining real datasets in robotics, we emphasize the role of simulation. Our approach involves collecting a synthetic and real dataset, addressing the 3D pose estimation task through a double heatmap-based representation. We explore the domain gap between the synthetic and real data, utilizing depth maps to enhance accuracy. Introducing temporal cues, our pipeline embraces the novel Pose Nowcasting paradigm, where predicting future poses serves as an auxiliary task to refine current pose precision. Shifting to the human scenario, we propose a pose refinement framework based on depth map analysis. Simultaneously, our investigation extends to Human-Computer Interaction, where we present an unsupervised method for detecting and classifying dynamic hand gestures using data from a motion tracking sensor.
This thesis seeks to make a valuable contribution to the intersection of 3D Computer Vision and Deep Learning across various domains. Following an overview of the existing state-of-the-art in 3D reconstruction and 3D pose estimation tasks, we present our proposed methods with a comprehensive technical explanation supported by a detailed experimental investigation conducted on benchmark datasets widely acknowledged in the literature
Groundwater and ground displacement monitoring in the source area of the Montecchi earthflow (Northern Apennines, Italy)
No full textThe importance of pore water pressure regime for landslide activity is generally accepted. In the case of earthflows, generalized failures (reactivations) are rare and sustained slow movements can proceed for decades. The relationship between precipitation, pore water pressure responses and movement is not straightforward. We document rainfall, pore pressure regime and displacements in the source area of an active earthflow. Pressure heads at shallow depth are clearly related to infiltration from the surface and can be satisfactorily reproduced by a diffusive model while the hypothesis of gravity-dominated flow can be rejected based on the short delay between rainfall and pressure response. Displacement rates are very small to zero during the summer and increase a couple of months later than the onset of the precipitation of the wet season. Only late in the wet season, velocities attain peak values (up to 4 mm/day) and show a remarkable correlation to rainfall episodes. Higher displacement rates correspond to unexceptional pressure head values, we therefore believe that alternative mechanisms of water pressure build-up may exist. Fractures likely act as a preferential flow system and influence both the hydrological responses to rainfall and the deformation behavior of the landslide
Lithologic and morphologic controls on debris flow dynamics in the Dolomites (Italian Alps, Italy)
No full textIn high-relief landscapes, debris flows represent the most efficient mean of erosion and sediment transport across
spatial scales ranging from hectares to tens of square kilometers.
In the Dolomite region of Italy, the landscape is dominated by steep massifs mainly made of limestone and
dolostone rocks. Abundant talus and fan deposits connect the bottom of the valleys to the rocky massifs. Thick
debris talus was deposited in post-glacial climatic conditions and is actively fed by steep dolomitic rock walls.
Debris flows are widespread over the territory. They are commonly triggered by water discharge concentrated on
steep headwater catchments and delivered to talus slopes. Headwater catchments are typically very steep (45°-60°
on the average) and mostly consist of exposed bedrock with no vegetation and sparse to none soil cover. Debris
flow fans are relatively steep (10 to 30°)
Our study area extends over 250 km2 and includes about one-hundred debris flow catchments. We take advantage
of the remarkable morphological similarities of the catchments to characterize their lithology, morphometry and
investigate scale-relationships. We use LiDAR, field surveys and aerial photo interpretation to describe the
principal topographic conditions associated with debris-flow initiation, transportation, and deposition. Debris
flow-prone catchments associated to recent activity has been selected to investigate the role of geology and
morphology as influencing factor at basin scale. They display a decreasing specific sediment yield with increasing
size. Depending on the abundance of loose debris deposits, volumes of sediment delivered to the fan span about
one order of magnitude. Dependence of deposited sediment volumes on the basin scale is clear and translates into
an inverse proportionality when specific yield is considered
High intensity convective rainfalls and corresponding hydrological response in rocky head water basins
No full textDeformation responses of slow moving landslides to precipitation in the Northern Apennines (Italy).
No full textSlow moving landslides are a frequent feature in the Northern Apennines of Italy and one of the main landscape forming agents. Among the most common landslide types are complex earth flows that occur in chaotic clayshales and complex rock slides in highly fractured flysch. We present the results of an InSAR-based survey in the Reno and Panaro river catchments, which are located South of Bologna and Modena in the Northern Apennines of Italy. We processed Envisat and Cosmo-SkyMed radar data using the Stanford Method of Persistent Scatterers (StaMPS) and documented movement on 62 deep-seated landslides. These landslides were compared to theregional landslide inventory that contains information about the type of landslide, its state of activity and the lithological characteristics of the host rock. Of the landslides found using InSAR, 42 % correspond to active landslide bodies in the regional inventory, while 48 % are mapped as dormant and 10% are not previously mapped. InSAR derived landslides often do not correspond to the exact extent of mapped landslide bodies. InSAR results show two recurring styles of deformation: (1) earthflows involving chaotic clay-shale units that exhibit steady state displace ment, or exhibit long-term (multi-annual) accelerations and decelerations, or (2) complex landslides in flysch units that are characterized by distinct increases in displacement rate following periods (weeks-to-months) of intense precipitation. Such differences in behavior might be due to inherent differences in the mechanical and hydraulic characteristics of the material in relation to the specific climatic forcing experienced during the observation period (2003 and 2015). Flysch units are relatively more fragile and have higher permeability when compared to clayshale units. Hence the deformation response to the precipitation is likely faster and characterized by more abrupt accelerations and decelerations
Measurements of runoff discharges at the feet of a rocky channel incised on the Dimai Peak (Venetian Dolomites, Northeast Italy)
No full textThe direct shear strength and dilatancy of sand-gravel mixtures
No full textA total of 87 direct shear tests in a large direct shear-box apparatus have been used to investigate the strength and dilatancy of sand-gravel mixtures. This paper focuses on the differences in behaviour between a silica sand (yellow Leighton Buzzard sand) and sand-gravel mixtures obtained by adding fractions of two kinds of gravel to the sand. The purpose is to find a relation between the grain-size characteristics of the materials and the shearing resistance. Experimental results are analysed in terms of the frictional and dilatant contributions to the strength of mixtures as a function of their relative density, and are compared with dilatancy theories and empirical equations. The addition of gravel to the mixtures, even at low fractions (less than 0.1 by volume), causes an increase in peak friction angle (φ′peak) which results both from higher dilatancy at failure (ψmax) and higher constant volume friction angle (φ′cv). Use of the minimum voids ratio (emin) of the materials allows the data for the two families of mixtures to be normalized and interpreted in terms of φ ′cv and the ratio (φ′peak -φ′cv/ψmax. The relationships between relative density (Dr), ψmax and φ′peak-φ′cv are only partly explained on a physical basis, so we develop empirical equations to predict the peak shear resistance of sand-gravel mixtures (up to gravel contents of 0.5) on the basis of easily measurable quantities. Such equations constitute a practical tool to overcome the problems arising from the impracticality of testing coarse material in the standard shear-box apparatus. © Springer 2006
Deformation responses of slow moving landslides to seasonal rainfall in the Northern Apennines, measured by InSAR
Full text linkSlow moving landslides are widespread geomorphological features in the Northern Apennines of Italy where they represent one of the main landscape forming processes. The lithology of the Northern Apennines fold and thrust belt is characterized by alternations of sandstone, siltstone and clayshales, also known as flysch, and clay shales with a chaotic block in matrix fabric, which are often interpreted as tectonic or sedimentary mélanges. While flysch rocks with a high pelitic fraction host earthslides that occasionally evolve into flow like movements, earthflows are the dominant landslide type in chaotic clay shales. In the present work, we document the kinematic response to rainfall of landslides in these different lithologies using radar interferometry. The study area includes three river catchments in the Northern Apennines. Here, the Mediterranean climate is characterized by two wet seasons during autumn and spring respectively, separated by dry summers and winters with moderate precipitation. We use SAR imagery from the X-band satellite COSMO SkyMed and from the C-band satellite Sentinel 1 to retrieve spatial displacement measurements between 2009 and 2016 for 25 landslides in our area of interest. We also document detailed temporal and spatial deformation signals for eight representative landslides, although the InSAR derived deformation signal is only well constrained by our dataset during the years 2013 and 2015. In spring 2013, long enduring rainfalls struck the study area and numerous landslide reactivations were documented by the regional authorities. During 2013, we measured higher displacement rates on the landslides in pelitic flysch formations compared to the earthflows in the clay shales. Slower mean velocities were measured on most landslides during 2015. We analyse the temporal deformation signal of our eight representative landslides and compare the temporal response to precipitation. We show that earthslides in pelitic flysch formations accelerate faster than earthflows in chaotic clay shales and reach higher velocities, while the kinematic behaviour of the earthflows can be described as rather steady with only minor accelerations. Although we have no detailed pore pressure measurements for the period of interest, the observed behaviour can be explained in our view by the morphological and hydrological characteristics of the different landslide types. On the one hand landslide material and bedrock in the pelitic flysch rocks are more resistant, which is why slope angles are higher in this lithology. On the other hand, landslides in the pelitic flysch formations have often deeper slip surfaces and landslide material is more permeable. This is why long persistent rainfall is necessary to saturate the landslide material and induce pore pressures that are high enough to trigger displacement
Going Beyond Counting First Authors in Author Co-citation Analysis
Full text linkThe present study examines one of the fundamental aspects of author co-citation analysis (ACA) - the way co-citation
counts are defined. Co-citation counting provides the data on which all subsequent statistical analyses and mappings
are based, and we compare ACA results based on two different types of co-citation counting - the traditional type that
only counts the first one among a cited work's authors on the one hand and a non-traditional type that takes into
account the first 5 authors of a cited work on the other hand. Results indicate that the picture produced through this non-traditional author co-citation counting contains more coherent author groups and is therefore considerably clearer. However, this picture represents fewer specialties in the research field being studied than that produced through the traditional first-author co-citation counting when the same number of top-ranked authors is selected and analyzed. Reasons for these effects are discussed
- …
