Nowadays, new computers generation provides a high performance that enables to build computationally expensive computer vision applications applied to mobile robotics. Building a map of the environment is a common task of a robot and is an essential part to allow the robots to move through these environments. Traditionally, mobile robots used a combination of several sensors from different technologies. Lasers, sonars and contact sensors have been typically used in any mobile robotic architecture, however color cameras are an important sensor due to we want the robots to use the same information that humans to sense and move through the different environments. Color cameras are cheap and flexible but a lot of work need to be done to give robots enough visual understanding of the scenes. Computer vision algorithms are computational complex problems but nowadays robots have access to different and powerful architectures that can be used for mobile robotics purposes. The advent of low-cost RGB-D sensors like Microsoft Kinect which provide 3D colored point clouds at high frame rates made the computer vision even more relevant in the mobile robotics field. The combination of visual and 3D data allows the systems to use both computer vision and 3D processing and therefore to be aware of more details of the surrounding environment. The research described in this thesis was motivated by the need of scene mapping. Being aware of the surrounding environment is a key feature in many mobile robotics applications from simple robotic navigation to complex surveillance applications. In addition, the acquisition of a 3D model of the scenes is useful in many areas as video games scene modeling where well-known places are reconstructed and added to game systems or advertising where once you get the 3D model of one room the system can add furniture pieces using augmented reality techniques. In this thesis we perform an experimental study of the state-of-the-art registration methods to find which one fits better to our scene mapping purposes. Different methods are tested and analyzed on different scene distributions of visual and geometry appearance. In addition, this thesis proposes two methods for 3d data compression and representation of 3D maps. Our 3D representation proposal is based on the use of Growing Neural Gas (GNG) method. This Self-Organizing Maps (SOMs) has been successfully used for clustering, pattern recognition and topology representation of various kind of data. Until now, Self-Organizing Maps have been primarily computed offline and their application in 3D data has mainly focused on free noise models without considering time constraints. Self-organising neural models have the ability to provide a good representation of the input space. In particular, the Growing Neural Gas (GNG) is a suitable model because of its flexibility, rapid adaptation and excellent quality of representation. However, this type of learning is time consuming, specially for high-dimensional input data. Since real applications often work under time constraints, it is necessary to adapt the learning process in order to complete it in a predefined time. This thesis proposes a hardware implementation leveraging the computing power of modern GPUs which takes advantage of a new paradigm coined as General-Purpose Computing on Graphics Processing Units (GPGPU). Our proposed geometrical 3D compression method seeks to reduce the 3D information using plane detection as basic structure to compress the data. This is due to our target environments are man-made and therefore there are a lot of points that belong to a plane surface. Our proposed method is able to get good compression results in those man-made scenarios. The detected and compressed planes can be also used in other applications as surface reconstruction or plane-based registration algorithms. Finally, we have also demonstrated the goodness of the GPU technologies getting a high performance implementation of a CAD/CAM common technique called Virtual Digitizing.

This work was carried out in collaboration with the Department of Adaptive Biotechnologies, Global Change Research Centre AS CR. It deals with the quantitative analysis of photosynthetic cell cultures. It uses images captured by a confocal fluorescent microscope to the automatic determining the number of cells in the sample. The work consists of a theoretical analysis, which briefly describes fluorescence and confocal microscopy. It also concisely introduces a microscope Leica TCS SP8 X, which I used to scan data. One capture is devoted to the theory of digital image processing. The second part deskribes the development of algorithm for processing 3D data and simplified algorithm for processing 2D data and its program implementations in a programming environment MATLAB R2013b. Grafical user interface is explained in detail. Done measurement are presented at the conclusion. It mentions compiled sample preparation protocol. The results of the program are compared with manual counting. Number of cells per 1 ml are determined by created program in samples of cell cultures Chenopodium rubrum (Cr) and Solanum lycopersicum (To).

Robust detection, localization, and tracking of objects are essential for autonomous driving. Computer vision has largely driven development based on camera sensors in recent years, but 3D localization from images is still challenging. Sensors such as LiDAR or RADAR are used to compute depth; each having its own advantages and drawbacks. The main idea of the project is to be able to mix images from the camera and RADAR detections in order to estimate depths for the objects appearing in the images. Fusion strategies can be considered the solution to give a more detailed description of the environment by utilizing both the 3D localization capabilities of range sensors and the higher spatial resolution of image data. The idea is to fuse 3D detections from the RADAR onto the image plane, this requires a high level of synchronization of the sensors and projections of the RADAR data on the required image. / Robust detektering, lokalisering och spårning av objekt är avgörande för autonom körning. Datorseende har till stor del drivit utvecklingen baserad på kamerasensorer de senaste åren, men 3D-lokalisering från bilder är fortfarande utmanande. Sensorer som LiDAR eller RADAR används för att beräkna djup; var och en har sina egna fördelar och nackdelar. Huvudtanken med projektet är att kunna blanda bilder från kameran och RADAR-detektioner för att uppskatta djup för de objekt som förekommer i bilderna. Fusionsstrategier kan anses vara lösningen för att ge en mer detaljerad beskrivning av miljön med både 3D-lokaliseringsförmågan hos avståndssensorer och den högre rumsliga upplösningen av bilddata. Tanken är att smälta samman 3D-detektioner från RADAR till bildplanet, detta kräver en hög nivå av synkronisering av sensorerna och projektioner av RADAR-data på den önskade bilden.

Deep learning has shown to be successful on the task of semantic segmentation of three-dimensional (3D) point clouds, which has many interesting use cases in areas such as autonomous driving and defense applications. A common type of sensor used for collecting 3D point cloud data is Light Detection and Ranging (LiDAR) sensors. In this thesis, a time-correlated single-photon counting (TCSPC) LiDAR is used, which produces very accurate measurements over long distances up to several kilometers. The dataset collected by the TCSPC LiDAR used in the thesis contains two classes, person and other, and it comes with several challenges due to it being limited in terms of size and variation, as well as being extremely class imbalanced. The thesis aims to identify, analyze, and evaluate state-of-the-art deep learning models for semantic segmentation of point clouds produced by the TCSPC sensor. This is achieved by investigating different loss functions, data variations, and data augmentation techniques for a selected state-of-the-art deep learning architecture. The results showed that loss functions tailored for extremely imbalanced datasets performed the best with regard to the metric mean intersection over union (mIoU). Furthermore, an improvement in mIoU could be observed when some combinations of data augmentation techniques were employed. In general, the performance of the models varied heavily, with some achieving promising results and others achieving much worse results.

Light Detection and Ranging (LiDAR) sensors have many different application areas, from revealing archaeological structures to aiding navigation of vehicles. However, it is challenging to interpret and fully use the vast amount of unstructured data that LiDARs collect. Automatic classification of LiDAR data would ease the utilization, whether it is for examining structures or aiding vehicles. In recent years, there have been many advances in deep learning for semantic segmentation of automotive LiDAR data, but there is less research on aerial LiDAR data. This thesis investigates the current state-of-the-art deep learning architectures, and how well they perform on LiDAR data acquired by an Unmanned Aerial Vehicle (UAV). It also investigates different training techniques for class imbalanced and limited datasets, which are common challenges for semantic segmentation networks. Lastly, this thesis investigates if pre-training can improve the performance of the models. The LiDAR scans were first projected to range images and then a fully convolutional semantic segmentation network was used. Three different training techniques were evaluated: weighted sampling, data augmentation, and grouping of classes. No improvement was observed by the weighted sampling, neither did grouping of classes have a substantial effect on the performance. Pre-training on the large public dataset SemanticKITTI resulted in a small performance improvement, but the data augmentation seemed to have the largest positive impact. The mIoU of the best model, which was trained with data augmentation, was 63.7% and it performed very well on the classes Ground, Vegetation, and Vehicle. The other classes in the UAV dataset, Person and Structure, had very little data and were challenging for most models to classify correctly. In general, the models trained on UAV data performed similarly as the state-of-the-art models trained on automotive data.

[ES] La Gestión del Patrimonio Cultural (GPC) es una operación muy compleja cuyo objetivo es preservar la integridad física de los Bienes Culturales y, al mismo tiempo, difundir los valores históricos y permitir el disfrute del Patrimonio. Debido a las múltiples fases que componen la GPC (documentación, intervención, conservación preventiva, uso), el gestor se ve sometido a un gran esfuerzo de coordinación de las interacciones creadas por figuras profesionales muy diferentes, tanto por formación como por necesidades específicas en el ámbito de la gestión. En los últimos años, la aplicación de las tecnologías digitales al Patrimonio Cultural se ha convertido en una parte indispensable de la GPC. Las metodologías más utilizadas para la adquisición de datos, como el escaneo láser terrestre y la fotogrametría digital, también se han convertido en una práctica habitual en las actividades profesionales. Sin embargo, el uso de modelos 3D para la gestión se limita hasta ahora a algunas investigaciones académicas, que a menudo no tienen continuidad tras la finalización del proyecto. Además, hasta la fecha existen pocas normas supranacionales que guíen a las instituciones en el proceso de creación y uso de modelos 3D para la GPC. Por tanto, la falta de herramientas para controlar la calidad de los datos y productos digitales adquiridos afecta negativamente a la interacción entre el sector de la investigación académica, el sector de la gestión y el mundo profesional. La investigación propone el uso de los modelos 3D como una herramienta válida de apoyo en todas las fases de la gestión, ya sea utilizando los datos tridimensionales como base del archivo digital, o explotando todos los productos obtenidos a partir de los datos básicos para las múltiples acciones de cada fase. Por lo tanto, el objetivo de la tesis doctoral es desarrollar directrices para la producción de modelos 3D con el fin de gestionar, introducir y preservar eficazmente los datos. Estas directrices investigan todos los aspectos del proceso que va desde la adquisición de datos, pasando por su catalogación y archivo, hasta su tratamiento y la creación de un sistema de información simplificado para su gestión. Cada directriz guía al usuario a través de una fase específica del tratamiento y el uso de los datos digitales, y proporciona indicaciones adaptadas al nivel de conocimientos respecto a las tecnologías y metodologías digitales. De este modo, el gestor puede utilizar los modelos 3D para su gestión y controlar su calidad y sus estándares mínimos. Se ha optado por un enfoque interdisciplinar e internacional con el fin de elaborar directrices que se adapten al mayor número posible de Bienes Culturales, desarrollando la tesis en el marco de un acuerdo de cotutela entre la Universidad de Bolonia y la Universitat Politècnica de València. Con el fin de obtener unas pautas universales, las metodologías analizadas en el estudio del estado del arte se aplicaron a una serie de casos de estudio. Los principales son los monumentos paleocristianos de Rávena (Italia), pertenecientes a la Lista del Patrimonio Mundial de la UNESCO, y un panteón neogótico situado en Castellón de la Plana (España). Las experiencias realizadas sobre la GPC en los dos países han contribuido a la elaboración de directrices y normas universales que mejoren las interacciones entre el mundo académico, los gestores y el sector profesional. La investigación, al poner de manifiesto los problemas inherentes a la GPC, ha permitido identificar las principales cuestiones abiertas que se deben explorar en futuras líneas de investigación, como la aplicación de estándares a un gran número de Bienes Culturales para conducir a la puesta a punto de los mismos estándares; la búsqueda de sistemas para la clasificación automática de los datos brutos; el tratamiento de los datos recogidos para la creación de relaciones, estrategias y métodos de clasificación, integración y optimización de datos heterogéneos. / [CA] La Gestió del Patrimoni Cultural (GPC) és una operació molt complexa l'objectiu de la qual és preservar la integritat física del els Béns Culturals i, al mateix temps, difondre els valors històrics i permetre el gaudi del Patrimoni. A causa de les múltiples fases que componen la GPC (documentació, intervenció, conservació preventiva, ús) , el gestor es veu sotmés a un gran esforç de coordinació de les interaccions creades per figures professionals molt diferents, tant per formació com per necessitats específiques en l'àmbit de la gestió. En els últims anys, l'aplicació de les tecnologies digitals al Patrimoni Cultural s'ha convertit en una part indispensable de la GPC, des de les fases de documentació fins a les d'intervenció. Les metodologies més utilitzades per a l'adquisició de dades, com l'escaneig làser terrestre i la fotogrametria digital, també s'han convertit en una pràctica habitual en les activitats professionals. No obstant això, l'ús de models 3D per a la gestió es limita fins ara a algunes investigacions i aplicacions acadèmiques, que sovint no tenen continuïtat després de la finalització del projecte. A més, fins a la data hi ha poques normes supranacionals que guien a les institucions en el procés de creació i ús de models 3D. Per tant, la falta de ferramentes per a controlar la qualitat de les dades i productes digitals adquirits afecta negativament la interacció entre el sector de la investigació acadèmica, el sector de la gestió i el món professional. La investigació proposa l'ús dels models 3D com una ferramenta vàlida de suport en totes les fases de la gestió, ja siga utilitzant les dades tridimensionals com a base de l'arxiu digital, o explotant tots els productes obtinguts a partir de les dades bàsiques per a les múltiples accions de cada fase. Per tant, l'objectiu de la tesi doctoral és desenrotllar directrius per a la producció de models 3D a fi de gestionar, introduir i preservar eficaçment les dades. Estes directrius investiguen tots els aspectes del procés que va des de l'adquisició de dades, passant per la seua catalogació i arxiu, fins al seu tractament i la creació d'un sistema d'informació simplificat per a la seua gestió. Cada directriu particular guia l'usuari a través d'una fase específica del tractament i l'ús de les dades digitals, i proporciona indicacions adaptades al nivell de coneixements respecte a les tecnologies i metodologies digitals. D'esta manera, el gestor pot utilitzar els models 3D per a la seua gestió i controlar la seua qualitat i els seus estàndards mínims. S'ha optat per un enfocament interdisciplinari i internacional a fi d'elaborar directrius que s'adapten al nombre més gran possible de tipus de Béns Culturals Cultural, desenrotllat la tesi en el marc d'un acord de cotutela entre la Universitat de Bolonya i la Universitat Politècnica de València. A fi d'obtindre unes pautes universals, les metodologies analitzades en l'estudi de l'estat de l'art es van aplicar a una sèrie de casos d'estudi. Els principals són els monuments paleocristians de Ravenna (Itàlia), pertanyents a la Llista del Patrimoni Mundial de la UNESCO, i una capella neogòtica situada a Castelló de la Plana (Espanya). Les experiències realitzades sobre la GPC en els dos països han contribuït a l'elaboració de directrius i normes universals que milloren les interaccions entre el món acadèmic, els gestors i el sector professional. La investigació, al posar de manifest els problemes inherents a la GPC, permet identificar les principals qüestions obertes que s'han d'explorar en futures línies d'investigació, com l'aplicació d'estàndards a un gran nombre de Béns Culturals per a conduir a la posada al punt dels mateixos estàndards; la busca de sistemes per a la classificació automàtica de les dades brutes; el tractament de les dades arreplegats per a la creació de relacions, estratègies i mètodes de classificació, integració i optimització de dades heterogènies. / [EN] The Management of Cultural Heritage (MCH) is a very complex operation aimed at protecting the physical integrity of Cultural Heritage assets, while promoting their historical value and development of tourism industry. Composed by distinct phases (documentation, intervention, monitoring and use), MCH implies a great effort for the project manager to coordinate the interactions among very different professional figures. In recent years, the use of digital technologies has become an essential part of the MCH delicate process, from early documentation to late intervention phases. The most commonly used methodologies for digital data acquisition, such as terrestrial laser scanning and digital photogrammetry, have become common practice in a broad range of professional activities. On the contrary, the use of 3D models for MCH is still limited to few academic research to date, often lacking continuity and wide application after the end of specific projects. Furthermore, very few supra-national standard guidelines regulating their use are available to date. As a consequence, the operator who decides to use a 3D model as a basis for management is faced with the scarcity and fragmentation of standards and guidelines. Moreover, the lack of standard on quality of acquired data and digital products negatively influences the interaction between the academic research sector, the managers and the professional world. The focus is on the use of 3D models as a valid support tool in the MCH process, highlighting their advantages in all the distinct phases of the management. As an example, 3D data can constitute themselves the basis for the digital database, gathering all available information concerning a Cultural Heritage site, exploitable for restoration works or for scientific dissemination. In particular, the aim of this PhD research is to develop guidelines to produce 3D models for MCH, with the purpose to efficiently entry, store and manage digital data. The here provided guidelines investigate every aspect of the process leading from data acquisition to cataloguing and archiving, processing and creation of a simplified information system. Each recommendation guides the user through the management of digital data, by adapting to his/her level of knowledge with respect to digital technologies and methodologies. In this way, the manager can efficiently use 3D models in MCH projects. In order to elaborate guidelines that could be suitable for as many typologies of Cultural Heritage as possible an international approach was chosen, developing the thesis in joint supervision under the University of Bologna and the Universitat Politècnica de València. We decided to apply state-of-the-art technologies and methodologies to a variety of case studies. The main ones are the early Christian monuments of Ravenna (Italy) belonging to the UNESCO World Heritage List, and a small neogothic chapel located in Castellón de la Plana (Spain). The fruitful collaboration between two different countries allowed an invaluable exchange of MCH expertise and, more broadly, contributed to the elaboration of standardized and universally applicable MCH guidelines that will allow a better interaction between managers, the academic research world and the professional one. The investigation, by highlighting the problems inherent to the MCH, made it possible to identify the main open issues that need to be explored in future lines of research, such as the application of standards to a large number of cultural assets in an iterative, continuous and automatic way, in order to perfecting the standards; the search for automatic classification of raw data; the processing of collected data for the creation of relations, strategies and methods for the classification, integration and optimisation of heterogeneous data. / Bertacchi, G. (2022). Guidelines for the Management of Cultural Heritage Using 3D Models for the Insertion of Heterogeneous Data [Tesis doctoral]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/182419 / TESIS