Submitted by Almir Azevedo (barbio1313@gmail.com) on 2014-08-08T13:19:37Z No. of bitstreams: 1 MAURICIO ALVES DA CUNHA E AGHINA_D_2012.pdf: 3967079 bytes, checksum: 4967d16b666f01de68d7c1b24d6591d4 (MD5) / Made available in DSpace on 2014-08-08T13:19:37Z (GMT). No. of bitstreams: 1 MAURICIO ALVES DA CUNHA E AGHINA_D_2012.pdf: 3967079 bytes, checksum: 4967d16b666f01de68d7c1b24d6591d4 (MD5) Previous issue date: 2012 / Devido as normas muito rígidas de segurança de operação de uma usina nuclear, os operadores, tem que ser muito bem treinados, para que possam opera-la dentro dos procedimentos de segurança necessários. Este treinamento é feito através de simuladores, que possibilitam a operação do usuário, a mais próxima possível de uma mesa de controle real, e que possam ser inseridas situações de acidentes, para que eles treinem, como voltar a usina para uma condição normal de funcionamento. Normalmente são utilizados dois tipos de simulador.: O desejado é o full scope que é um simulador computacional da dinâmica da usina usado em conjunto com a réplica física da mesa de controle, mas este tipo de simulador envolve um alto custo de construção. O segundo tipo é o que usa janelas sinópticas de varias regiões da mesa de controle original, o seu custo de construção é menor, mas tem uma fidelidade menor com a aparência da mesa original. Atualmente, com o uso da realidade virtual as mesas de controle podem ser modeladas em 3D, fazendo com que a interface do simulador seja bem parecida com a aparência da mesa de controle original e com um baixo custo de construção. Este trabalho mostra o uso de interfaces naturais para interação do operador com a mesa virtual, com o intuito que ele não use nenhum dispositivo mecânico para a visualização e atuação com a mesa virtual. Para isto foram usados procedimentos, tais como: visão computacional para reconhecimento da posição de observação do operador, de suas mãos para a atuação dos controles da mesa e reconhecimento de voz. / Due to very strict standards of safe operation of a nuclear power plant operators must be well trained so they can operate it within the necessary safety procedures. This is done through training simulators, which enable the user operation, as close as possible to the real control desk, and can be inserted accident situations, so they train, how to return the plant to a normal operating condition. Normally is used two types of simulator. Preferred is the full scope simulator, what is a computational dynamics program of the plant used in conjunction with a physical replica of the control desk, but this type of simulator involves a high construction cost. The second type is what uses synoptic windows of various regions of the original control desk, its construction cost is smaller, but it have a little fidelity to the original appearance of the table. Currently, with the use of virtual reality, control desks can be modeled in 3D, making the simulator interface is very similar to the appearance of the real control desk with a low cost construction. This work shows the use of natural interfaces for operator interaction with the virtual control desk, in order that it does not use any mechanical device for displaying and acting with it. For procedures that were used, such as: computer vision to recognize the position of the operator's and observation of their hands to the work of the desk controls and voice recognition.

Submitted by Marcele Costal de Castro (costalcastro@gmail.com) on 2017-08-30T16:41:39Z No. of bitstreams: 1 MAURICIO ALVES DA CUNHA E AGHINA_ M.pdf: 5627813 bytes, checksum: 216957f5a8919957ecacf28fdcd5ccf6 (MD5) / Made available in DSpace on 2017-08-30T16:41:39Z (GMT). No. of bitstreams: 1 MAURICIO ALVES DA CUNHA E AGHINA_ M.pdf: 5627813 bytes, checksum: 216957f5a8919957ecacf28fdcd5ccf6 (MD5) Previous issue date: 2009-03 / Uma usina nuclear é uma instalação de geração de energia elétrica. Devido ao seu alto grau de complexidade e normas rígidas de segurança é extremamente necessário que seus operadores sejam muito bem treinados para a sua operação, pois caso haja uma falha humana, esta irá acarretar um desligamento da planta com conseqüentes prejuízos econômicos para a operadora e para a população em geral, devido a possibilidade de um black out na rede elétrica. Para evitar essa possibilidade de falha a operadora, normalmente, possui um simulador full scope da mesa de controle da planta, que é a réplica física da mesa de controle original. O controle deste simulador é um programa de computador que pode gerar o funcionamento igual ao normal ou vários cenários de acidentes para treinarem os seus operadores nas varias condições de operação da planta. Devido ao simulador ser a réplica física da mesa de controle, acarreta um custo muito elevado para a construção das instalações prediais e dos componentes da mesa. A proposta deste trabalho é apresentar um projeto de um simulador virtual com a modelagem em 3D estéreo da mesa de controle da planta nuclear e com as mesmas funções de operação do simulador original. Este simulador virtual terá um custo muito menor e serve para um pré-treinamento de operadores com o intuito de se familiarizar com o original. / Nuclear Power Plant (NPP) is a facility for electrical energy generation. Because of its high degree of complexity and very rigid norms of security it is extremely necessary that operators are very well trained for the NPP operation. A mistaken operation by a human operator may cause a shutdown of the NPP, incurring in a huge economical damage for the owner and for the population in the case of a electric net black out. To reduce the possibility of a mistaken operation, the NPP usually have a full scope simulator of the plant´s control room, which is the physical copy of the original control room. The control of this simulator is a computer program that can generate the equal functioning of the normal one or some scenarios of accidents to train the operators in many abnormal conditions of the plant. A physical copy of the control room has a high cost for its construction, not only of its facilities but also for its physical components. The proposal of this work is to present a project of a virtual simulator with the modeling in 3D stereo of a control room of a given nuclear plant with the same operation functions of the original simulator. This virtual simulator will have a lower cost and serves for pretraining of operators with the intention of making them familiar to the original control room.

Submitted by Amanda Silva (amanda.osilva2@ufpe.br) on 2015-03-05T14:30:00Z No. of bitstreams: 2 DISSERTAÇÃO Adauto Pereira Costa.pdf: 7019224 bytes, checksum: cd4204a7ece60af83bad8f34def2ad94 (MD5) license_rdf: 1232 bytes, checksum: 66e71c371cc565284e70f40736c94386 (MD5) / Made available in DSpace on 2015-03-05T14:30:00Z (GMT). No. of bitstreams: 2 DISSERTAÇÃO Adauto Pereira Costa.pdf: 7019224 bytes, checksum: cd4204a7ece60af83bad8f34def2ad94 (MD5) license_rdf: 1232 bytes, checksum: 66e71c371cc565284e70f40736c94386 (MD5) Previous issue date: 2014 / A intervisibilidade é a visibilidade mútua entre dois ou mais pontos. Não é baseada apenas no espectro eletromagnético do visível. Ela é fundamental para inúmeras atividades que vão desde arqueologia a fins militares. Nas Ciências Geodésicas tem larga aplicação em levantamentos topográficos, os quais necessitam de intervisibilidade entre as estações. Esta dissertação aborda o tema da intervisibilidade geométrica utilizando conceitos da Álgebra Linear e Geometria Analítica e trata não somente a clássica linha de visada como sendo um segmento de reta, como também casos mais complexos como o cone de intervisibilidade e lançamento balístico sobre o MDS. Demonstram-se neste trabalho casos particulares de intersecção entre superfícies. As equações desenvolvidas permitem respostas relevantes utilizando-se estruturas vetoriais triangulares do Modelo Digital de Superfície. São realizados testes com dados sintéticos para avaliar as equações desenvolvidas. Também é realizada uma avaliação da intervisibilidade entre os marcos geodésicos da rede geodésica cadastral do Campus Joaquim Amazonas da Universidade Federal de Pernambuco.

3d-skärmar är en spännande teknik som har fått en ny vind på senaste tiden med artiklar som 3DeeShell till iPhone och Wazabees 3d-lösningar till PC. När dessa 3d-skärmar används till att visa spel i 3d så stöter spelutvecklare på problem när de ska utveckla visuella effekter såsom anti-aliasing och billboarding, då dessa tekniker kanske inte är anpassade till 3d-skärmens funktionalitet. Undersökningen i detta arbete söker att finna några svar på hur dessa tekniker fungerar tillsammans med dessa 3d-skärmar och det kommer att redogöras för några åtgärder för att förbättra kvalitéen på dessa visuella tekniker.

SammanfattningDen här rapporten beskriver arbetet att ta fram försöksmaterial i form av 3D-modeller med tillhörande animationer till det Internetbaserade rådgivningsverktyget för läkemedelsprodukter AssistansPlus*. Utöver detta ger rapporten en inblick i hur framtagning av 3D-medierat material för Internet, inom området rådgivning och presentation, kan se ut generellt.Viktigt att tänka på vid arbetet med modeller och animationer till AssistansPlus är att de måste anpassas till W3D formatet, som 3D spelaren i verktyget är utvecklad i. De viktigaste var här att veta vad som följde med vid en export till formatet W3D (animerandet måste ske på ett relativt analogt sett, vissa bildformat för texturerna fungerar bättre än andra etc.). Rapporten innehåller en checklista som ska fungera som en guide vid framtida framtagning av material till verktyget. Filstorlek är av stor vikt i arbete med 3D för Internet och var därför en viktig del också i uppdraget. Att begränsa filernas storlek genom att använda så få ytor som möjligt, genom att stilisera modellerna och genom att göra texturer så bra som möjligt i förhållande till filstorleken, var viktiga delar arbetet. För och nackdelar med denna rika mediatyp diskuteras i rapporten. Aspekter som tas upp är navigering och interaktivitet samt synen på 3D som medietyp. Många Internetanvändare har aldrig interagerat med 3D, varför en väl genomtänkt navigering är viktig. 3D kopplas gärna ihop med spel och animerad film. Att behandle 3D som en mediatyp bland de på Internet mer etablerade (som text bild och film) är viktigt. Det finns relativt många format och visningsverktyg att välja mellan i arbete med 3D på Internet. Det finns inget enskilt format som idag kan anses vara att rekomendera. En framtida Flash-version med implementerat stöd för 3D hantering tycks vara den mest sannolika allmäna standarden i framtiden.3D-material för användning i webbläsare är rapportens fokus och innehållet begränsas därför till detta. Vidare begränsas innehållet till webbaserat 3D med syfte att informera, instruera eller marknadsföra kring en produkt. 3D-applikationer inom området nöje behandlas inte. Trots att utvecklingen av 3D-grafik i hög grad har drivits framåt av spel- och filmindustrin, är detta områden som det inte läggs fokus på. Dock är det så att områdena ofta går in i varandra; paralleller är därför förekommande.

The increasingly personal and ubiquitous capabilities of computing—everything from smartphones to virtual reality—are enabling us to build a brave new world in the digital realm. Despite these advances in the virtual world, our ability as end-users to transform the physical world still remains limited. The emergence of low-cost fabrication technology (most notably 3D printing) has brought us a dawn of making, promising to empower everyday users with the ability to fabricate physical objects of their own design. However, the technology itself is oblivious of the physical world—things are, in most cases, assumed to be printed from scratch in isolation from the real world objects they will be attached to and work with. To bridge this ‘gulf of fabrication’, my thesis research focuses on developing fabrication techniques with design tool integration to enable users to expressively create designs that can be attached to and function with existing real-world objects. Specifically, my work explores techniques that leverage the 3D printing process to create attachments directly over, onto and around existing objects; a design tool further enables people to specify and generate adaptations that can be attached to and mechanically transform existing objects in user-customized ways; a user-driven approach allows people to express and iterate structures that are optimized to support existing objects; finally, a library of ‘embeddables’ demonstrate that existing objects can also augment 3D printed designs by embedding a large variety of material to realize different properties and functionalities. Overall my thesis aspires to make fabrication real—enabling people to express, iterate and fabricate their designs that closely work with real-world objects to augment one another.

L’holographie est souvent considérée comme la technologie de visualisation 3D la plus prometteuse, puisqu'elle fournit l'illusion du relief la plus naturelle et la plus réaliste possible. Toutefois, pour trouver application dans le domaine de la visioconférence ou de la téléprésence, les méthodes de génération numérique d'hologrammes doivent produire des scènes réalistes avec une forte illusion de profondeur, et ce en temps réel. Cette thèse se situe dans ce contexte et est organisé en deux parties. Dans la première partie de ce travail, nous avons proposé deux algorithmes de synthèse d'hologrammes permettant de se rapprocher du temps réel. Tout d'abord, nous avons développé une méthode combinant deux approches complémentaires: les approches point-source et wave-field. Alors que les méthodes de l'état de l'art réduisent la complexité de calcul de ces deux approches indépendamment, notre méthode tire parti de chacune d'entre elles. De cette manière, le temps de calcul de l'hologramme a été réduit de plus de 65% par rapport aux méthodes point-source et wave-field conventionnelles. Deuxièmement, nous avons cherché à accélérer cette méthode hybride en supprimant les redondances temporelles entre les trames consécutives d'une vidéo 3D. Pour chaque image, l'algorithme détecte les changements dans la scène et met à jour l'hologramme des points affectés. Étant donné que seules de petites régions de l'hologramme sont mises à jour, la charge de calcul est considérablement réduite, permettant le calcul d’hologrammes couleur à 60 images par seconde. Dans la deuxième partie de ce travail, nous avons proposé deux algorithmes afin d'améliorer la qualité visuelle des scènes. Tout d'abord, nous avons perfectionné la méthode hybride pour tenir compte des occultations dans la scène. Pour cela, nous avons conçu une méthode efficace pour le calcul de l'occultation d'une onde lumineuse par un point. Cette méthode reproduit les occultations sans augmenter de manière significative le temps de calcul de la méthode hybride originale. Enfin, nous avons proposé une méthode pour le calcul d'un hologramme à partir de données multivue-plus-profondeur (MVD) avec prise en compte des réflexions spéculaires. Selon cette méthode, la géométrie de la scène est reconstruite à partir des données MVD sous la forme d'un nuage de points, ce qui permet de n'utiliser que quelques projections de la scène. En outre, afin de tenir compte de réflexions spéculaires, chaque point de la scène est considéré avoir une émission différente selon les directions. Enfin, la lumière émise par la scène est propagée dans le plan de l'hologramme. Les résultats expérimentaux montrent que cette méthode reproduit tous les indices de la profondeur et l'illumination précise de la scène avec une complexité de calcul réduite. / Holography is often considered as the most promising 3D visualization technology, since it can produce the most realistic and natural depth illusion to the naked eye. However, in order to have application in the field of videoconferencing or telepresence systems, hologram synthesis methods should be able to produce realistic 3D scenes with strong depth illusion in real-time. This thesis falls within this context and is organized into two parts. In the first part of this work, we investigated two novel algorithms in order to get closer to real-time computation. First, we designed a fast hologram calculation method by combining two approaches which complement one another: the point-source and wave-field approaches. Whereas previously proposed methods reduced the computational complexity of these approaches independently, our method takes advantages from both of them. By this way, the hologram calculation time has been reduced by more than 65% compare to the conventional point-source and wave-field methods. Second, we further accelerated this hybrid method by removing temporal redundancies between consecutive frames of a 3D video. For each video frame, the algorithm detects changes in the scene and updates the hologram of only affected scene points. Since only small regions of the hologram are updated at each video frame, this method allows the computational burden to be dramatically reduced, enabling the computation of colorful video holograms at 60 frames per second. In the second part of this work, we proposed two algorithms in order to enhance the visual quality of displayed scenes. First, we improved the hybrid method to take into account occlusions between objects in the scene. To this end, we designed an efficient algorithm for light shielding between points and light waves. Experimental results revealed that this method provides occlusion effect without significantly increasing the hologram calculation time of the original hybrid method. Finally, we proposed a hologram computation method from Multiview-plus-depth (MVD) data with rendering of specular reflections. In this method, the 3D scene geometry is first reconstructed from the MVD data as a layered point-cloud, enabling the use of only a few perspective projections of the scene. Furthermore, in order to take into account specular reflections, each scene point is considered to emit light differently in all the directions. Finally, light scattered by the scene is numerically propagated towards the hologram plane in order to get the final hologram. Experimental results show that the proposed method is able to provide all the human depth cues and accurate shading of the scene with reduced computational complexity.

Glioblastoma multiforme (GBM) is among the most aggressive and mortal cancers of the central nervous system. Maximal safe surgical resection, followed by radiotherapy accompanied with chemotherapy is the standard of care for GBM patients. Despite this intensive treatment with conventional approaches, the management of GBM remains poor. The infiltrative nature of cancer cells makes the complete tumour removal by surgery virtually impossible. In addition, the blood-brain barrier’s (BBB) lack of permeability limits the number of effective chemotherapy drugs for GBM. Temozolomide (TMZ) is the most widely used chemotherapeutic agent for GBM because of its ability to pass the BBB. However, high systemic doses required to achieve brain therapeutic level, resulting in numerous side effects. The recurrence of GBM is almost inevitable due to the aforementioned shortcomings of conventional methods of treatment. Therefore, a great deal of effort has been focused on the development of new treatment methods capable of providing a high concentration of chemotherapy drug at the tumour site. Microspheres made from biodegradable polymers hold great potential to keep the chemotherapeutic agent intact within the carrier and locally deliver the drug over an extended period. However, the encapsulation of amphiphilic drug molecules such as TMZ within poly (d, l-lactide-co-glycolide) (PLGA) microspheres with conventional emulsion methods, oil-in-water (o/w), water-in-oil-in-water (w/o/w), is a major challenge. The extremely low encapsulation efficiencies obtained for TMZ-loaded PLGA microspheres using the aforementioned techniques (<7%) hampers the ability to scale up this process. Additionally, the injected microspheres to the tumour site tend to dislocate due to the cerebral flow which reduces the effectiveness of this localized drug delivery strategy. This study has focused on the development of a 3D bioprinted hydrogel-based mesh containing TMZ-loaded PLGA microspheres with high encapsulation efficiency (GlioMesh). To accomplish this, oil-in-oil (o/o) emulsion solvent evaporation technique was used to prepare PLGA microspheres loaded with TMZ. The poor solubility of TMZ in the external oil phase, liquid paraffin, resulted in obtaining encapsulation efficiencies as high as 61%. We then used the 3D bioprinting technology to embed TMZ-loaded PLGA microspheres into an alginate mesh. This provides the advantage of immobilizing the microspheres at the tumour site. Additionally, the flexibility and porosity of 3D bioprinted mesh allow for easy implantation and nutrients transportation to the brain tissue. The incorporation of polymeric microspheres within alginate fibres led to achieving an extended release of TMZ over 50 days. The functionality of GlioMesh in inducing cell cytotoxicity was evaluated by performing in vitro cell viability tests on U87 human glioblastoma cells. Higher cytotoxic effects were observed in the case of treatment with GlioMesh compared to the free drug because of the sustained release properties of our mesh. These data suggest that GlioMesh holds great promise to be used as an implant in the treatment of GBM. / Graduate / 2019-04-19

L’objectif de ces travaux de thèse est la reconnaissance automatique des espèces d’arbres à partir de scans laser terrestres, information indispensable en inventaire forestier. Pour y répondre, nous proposons différentes méthodes de reconnaissance d’espèce basées sur la texture géométrique 3D des écorces.Ces différentes méthodes utilisent la séquence de traitement suivante : une étape de prétraitement, une étape de segmentation, une étape d’extraction des caractéristiques et une dernière étape de classification. Elles sont fondées sur les données 3D ou bien sur des images de profondeur extraites à partir des nuages de points 3D des troncs d’arbres en utilisant une surface de référence.Nous avons étudié et testé différentes approches de segmentation sur des images de profondeur représentant la texture géométrique de l'écorce. Ces approches posent des problèmes de sur-Segmentation et d'introduction de bruit. Pour cette raison, nous proposons une nouvelle approche de segmentation des nuages de points 3D : « Burst Wind Segmentation », inspirée des lignes de partage des eaux. Cette dernière réussit, dans la majorité des cas, à extraire des cicatrices caractéristiques qui sont ensuite comparées à un dictionnaire des cicatrices (« ScarBook ») pour discriminer les espèces d’arbres.Une grande variété de caractéristiques est extraite à partir des régions segmentées par les différentes méthodes proposées. Ces caractéristiques représentent le niveau de rugosité, la forme globale des régions segmentées, la saillance et la courbure du contour, la distribution des points de contour, la distribution de la forme selon différents angles,...Enfin, pour la classification des caractéristiques visuelles, les forêts aléatoires (Random Forest) de Leo Breiman et Adèle Cutler sont utilisées dans une approche à deux étapes : sélection des variables importantes, puis classification croisée avec les variables retenues, seulement.L’écorce de l’arbre change avec l'accroissement en diamètre ; nous avons donc étudié différents critères de variabilité naturelle et nous avons testé nos approches sur une base qui présente cette variabilité. Le taux de bonne classification dépasse 96% dans toutes les approches de segmentation proposées mais les meilleurs résultats sont atteints avec la nouvelle approche de segmentation « Burst Wind Segmentation » étant donné que cette approche réussit mieux à extraire les cicatrices, utilise un dictionnaire de cicatrices et a été évaluée sur une plus grande variété de caractéristiques de forme, de courbure, de saillance et de rugosité. / The objective of the thesis is the automatic recognition of tree species from Terrestrial LiDAR data. This information is essential for forest inventory. As an answer, we propose different recognition methods based on the 3D geometric texture of the bark.These methods use the following processing steps: a preprocessing step, a segmentation step, a feature extraction step and a final classification step. They are based on the 3D data or on depth images built from 3D point clouds of tree trunks using a reference surface.We have investigated and tested several segmentation approaches on depth images representing the geometric texture of the bark. These approaches have the disadvantages of over segmentation and are quite sensitive to noises. For this reason, we propose a new 3D point cloud segmentation approach inspired by the watershed technique that we have called «Burst Wind Segmentation». Our approach succeed in extracting in most cases the characteristic scars that are next compared to those stored in a dictionary («ScarBook») in order to determine the tree species.A large variety of characteristics is extracted from the regions segmented by the different methods proposed. These characteristics are the roughness, the global shape of the segmented regions, the saliency and the curvature of the contour, the distribution of the contour points, the distribution of the shape according to the different orientations.Finally, for the classification of the visual characteristics, the Random Forest method by Leo Breiman and Adèle Cutler is used in a two steps approach: selection of the most important variables and cross classification with the selected variables.The bark of the tree changes with the trunk diameter. We have thus studied different natural variability criteria and we have tested our approaches on a test set that includes this variability. The accuracy rate is over 96% for all the proposed segmentation approaches but the best result is obtained with the «Burst Wind Segmentation» one due to the fact that this approach can better extract the scars, it uses a dictionary of scars for recognition, and it has been evaluated on a greater variety of shapes, curvatures, saliency and roughness.

La complexité des structures géologiques en Moyenne Durance rend difficile la caractérisation et l’évaluation du risque sismique dans cette région. La compréhension 3D des structures nécessite l’utilisation de techniques modernes de modélisation numérique pour réaliser des modèles 3D du substratum géologique qui soient cohérents à différentes échelles, pour effectuer des simulations des mouvements sismiques. Pour réaliser des modèles 3D géologiques, nous avons harmonisé des bases de données géologiques et géophysiques hétérogènes (forages pétroliers, sondages géotechniques, profils sismiques, profils géophysiques H/V, cartographie géologique haute et basse résolution, datations biostratigraphiques, ...) dans le géomodeleur 3D gOcad. La réalisation des modèles 3D permet de caractériser la géométrie 3D des principales structures tectoniques en Moyenne Durance, notamment la Zone de failles de la Moyenne Durance (ZFMD) ainsi que la géométrie haute résolution du substratum géologique de la vallée de Cadarache, notamment de son remplissage sédimentaire tertiaire. Notre étude permet de préciser la géométrie 3D de la ZFMD et son rôle sur les structures géologiques adjacentes. La ZFMD constitue une zone de transfert qui limite l’extension latérale et modifie la géométrie des plis et chevauchements. Les structures tectoniques en Moyenne Durance sont caractérisées par une déformation de type thin-skin. La tectonique salifère joue un rôle important dans cette déformation (niveau de décollement, formation et géométrie 3D des plis). A l’échelle de la vallée de Cadarache, l’étude géologique des affleurements couplée à l’étude géostatistique des sondages permet de réaliser une simulation en 3D de la répartition spatiale des faciès et de proposer un modèle sédimentaire 3D du remplissage tertiaire de la vallée. Il est caractérisé par une sédimentation en deux étapes : une première phase de remplissage par des apports détritiques importants en provenance du Sud-Est, liés au démantèlement du massif des Maures-Estérel, une deuxième phase de remplissage par des dépôts d’environnement fluviatile méandriforme à forte sinuosité et à dominante de sables. Enfin, l’imbrication multi-échelle des modèles 3D a permis d’expliquer la formation de la vallée de Cadarache dans le contexte géodynamique de la région de la Moyenne Durance au Tertiaire, et notamment sous l’influence de la compression pyrénéenne et de la ZFMD sur la région de la Moyenne Durance et sur la vallée de Cadarache. / The complexity of geological structures in Middle Durance region makes difficult the characterization and evaluation of seismic risk in this region. Understanding these structures in 3D requires to use modern techniques of 3D digital modelling in order to achieve the 3D geological models of the bedrock with coherence on different scales, to perform ground motion simulations. Building 3D geological models need to normalize heterogeneous geophysical and geological databases (oil boreholes, geotechnical boreholes, seismic profiles, H/V geophysical profiles, high and low resolution geological mapping, biostratigraphic dating,...) in the 3D software gOcad. The realization of 3D geological models allows to characterize the 3D geometry of main the tectonic structures in Middle Durance region, including the Middle Durance Fault Zone (ZFMD) and the high resolution geometry of geological bedrock of the Cadarache Valley, and in particular its tertiary sedimentary fill. Our study allows to specify the ZFMD 3D geometry and its role on the adjacent geological structures. The ZFMD is a transfer zone that limits lateral extension and changes the geometry of folds and thrusts. The Middle Durance tectonic structures are characterized by a thin-skin deformation. Salt tectonics plays an important role in this deformation (detachment level, training and 3D geometry of folds). At the scale of the Valley of Cadarache, outcrop geological study coupled with boreholes geostatistical study allows to perform a 3D simulation of the spatial distribution of facies and to propose a 3D sedimentary model for the Tertiary filling of the Cadarache Valley. It is characterized by sedimentation in two steps: a first step of filling by important detrital inputs from the Southeast, related to the erosion of the Maures-Esterel massif, a second step of filling by deposits of fluvial meandering sand-dominant environment with high-sinuosity. Finally, building multi-scale 3D models allows explaining the formation of the Valley of Cadarache in the geodynamic context of the Middle Durance region at Tertiary times, and particularly under the influence of the Pyrenean compression and ZFMD on the Middle Durance region and the Valley of Cadarache.