On-line adaptive visual tracking

Visual tracking is the problem of estimating some variables related to a target given a video sequence depicting the target. Visual tracking is key to the automation of many tasks, such as visual surveillance, robot or vehicle autonomous navigation, automatic video indexing in multimedia databases. Despite many years of research, long term tracking in real world scenarios for generic targets is still unaccomplished. The main contribution of this thesis is the definition of effective algorithms that can foster a general solution to visual tracking by letting the tracker adapt to mutating working conditions. In particular, we propose to adapt two crucial components of visual trackers: the transition model and the appearance model. The less general but widespread case of tracking from a static camera is also considered and a novel change detection algorithm robust to sudden illumination changes is proposed. Based on this, a principled adaptive framework to model the interaction between Bayesian change detection and recursive Bayesian trackers is introduced. Finally, the problem of automatic tracker initialization is considered. In particular, a novel solution for categorization of 3D data is presented. The novel category recognition algorithm is based on a novel 3D descriptors that is shown to achieve state of the art performances in several applications of surface matching.

The calcite(10-14) surface: a versatile substrate for molecular self-assembly

This thesis presents a detailed and successful study of molecular self-assembly on the calcite CaCO3(10-14) surface. One reason for the superior applicability of this particular surface is given by reflecting the well-known growth modes. Layer-by-layer growth, which is a necessity for the formation of templated two-dimensional (2D) molecular structures, is particularly favoured on substrates with a high surface energy. The CaCO3(10-14) surface is among those substrates and, thus, most promising. rnrnAll experiments in this thesis were performed using the non-contact atomic force microscope (NC-AFM) under ultra-high vacuum conditions. The acquisition of drift-free data became in this thesis possible owing to the herein newly developed atom-tracking system. This system features a lateral tip-positioning precision of at least 50pm. Furthermore, a newly developed scan protocol was implemented in this system, which allows for the acquisition of dense three-dimensional (3D) data under room-temperature conditions. An entire 3D data set from a CaCO3(10-14) surface consisting of 85x85x500 pixel is discussed. rnrnThe row-pairing and (2x1) reconstructions of the CaCO3(10-14) surface constitute most interesting research subjects. For both reconstructions, the NC-AFM imaging was classified to a total of 12 contrast modes. Eight of these modes were observed within this thesis, some of them for the first time. Together with literature findings, a total of 10 modes has been observed experimentally to this day. Some contrast modes presented themselves as highly distance-dependent and at least for one contrast mode, a severe tip-termination influence was found. rnrnMost interestingly, the row-pairing reconstruction was found to break a symmetry element of the CaCO3(10-14) surface. With the presence of this reconstruction, the calcite (10-14) surface becomes chiral. From high-resolution NC-AFM data, the identification of the enantiomers is here possible and is presented for one enantiomer in this thesis. rnrnFive studies of self-assembled molecular structures on calcite (10-14) surfaces are presented. Only for one system, namely HBC/CaCO3(10-14), the formation of a molecular bulk structure was observed. This well-known occurence of weak molecule-insulator interaction hinders the investigation of two-dimensional molecular self-assembly. It was, however, possible to force the formation of an island phase for this system upon following a variable-temperature preparation. rnFor the C60/CaCO3(10-14) system it is most notably that no branched island morphologies were found. Instead, the first C60 layer appeared to wet the calcite surface. rnrnIn all studies, the molecules arranged themselves in ordered superstructures. A templating effect due to the underlying calcite substrate was evident for all systems. This templating strikingly led either to the formation of large commensurate superstructures, such as (2x15) with a 14 molecule basis for the C60/CaCO3(10-14) system, or prevented the vast growth of incommensurate molecular motifs, such as the chicken-wire structure in the trimesic acid (TMA)/CaCO3(10-14) system. rnrnThe molecule-molecule and the molecule-substrate interaction was increased upon choosing molecules with carboxylic acid moieties in the third, fourth and fifth study, using terephthalic acid, TMA and helicene molecules. In all these experiments, hydrogen-bonded assemblies were created. rnrnDirected hydrogen bond formation combined with intermolecular pi-pi interaction is employed in the fifth study, where the formation of uni-directional molecular "wires" from single helicene molecules succeeded. Each "wire" is composed of heterochiral helicene pairs, well-aligned along the [01-10] substrate direction and stabilised by pi-pi interaction.

Reproducibility and validity of video screen measurements of gait in children with spastic cerebral palsy

PURPOSE: To determine the reproducibility and validity of video screen measurement (VSM) of sagittal plane joint angles during gait. METHODS: 17 children with spastic cerebral palsy walked on a 10m walkway. Videos were recorded and 3d-instrumented gait analysis was performed. Two investigators measured six sagittal joint/segment angles (shank, ankle, knee, hip, pelvis, and trunk) using a custom-made software package. The intra- and interrater reproducibility were expressed by the intraclass correlation coefficient (ICC), standard error of measurements (SEM) and smallest detectable difference (SDD). The agreement between VSM and 3d joint angles was illustrated by Bland-Altman plots and limits of agreement (LoA). RESULTS: Regarding the intrarater reproducibility of VSM, the ICC ranged from 0.99 (shank) to 0.58 (trunk), the SEM from 0.81 degrees (shank) to 5.97 degrees (trunk) and the SDD from 1.80 degrees (shank) to 16.55 degrees (trunk). Regarding the interrater reproducibility, the ICC ranged from 0.99 (shank) to 0.48 (trunk), the SEM from 0.70 degrees (shank) to 6.78 degrees (trunk) and the SDD from 1.95 degrees (shank) to 18.8 degrees (trunk). The LoA between VSM and 3d data ranged from 0.4+/-13.4 degrees (knee extension stance) to 12.0+/-14.6 degrees (ankle dorsiflexion swing). CONCLUSION: When performed by the same observer, VSM mostly allows the detection of relevant changes after an intervention. However, VSM angles differ from 3d-IGA and do not reflect the real sagittal joint position, probably due to the additional movements in the other planes.

Haptics in forensics: the possibilities and advantages in using the haptic device for reconstruction approaches in forensic science

Non-invasive documentation methods such as surface scanning and radiological imaging are gaining in importance in the forensic field. These three-dimensional technologies provide digital 3D data, which are processed and handled in the computer. However, the sense of touch gets lost using the virtual approach. The haptic device enables the use of the sense of touch to handle and feel digital 3D data. The multifunctional application of a haptic device for forensic approaches is evaluated and illustrated in three different cases: the representation of bone fractures of the lower extremities, by traffic accidents, in a non-invasive manner; the comparison of bone injuries with the presumed injury-inflicting instrument; and in a gunshot case, the identification of the gun by the muzzle imprint, and the reconstruction of the holding position of the gun. The 3D models of the bones are generated from the Computed Tomography (CT) images. The 3D models of the exterior injuries, the injury-inflicting tools and the bone injuries, where a higher resolution is necessary, are created by the optical surface scan. The haptic device is used in combination with the software FreeForm Modelling Plus for touching the surface of the 3D models to feel the minute injuries and the surface of tools, to reposition displaced bone parts and to compare an injury-causing instrument with an injury. The repositioning of 3D models in a reconstruction is easier, faster and more precisely executed by means of using the sense of touch and with the user-friendly movement in the 3D space. For representation purposes, the fracture lines of bones are coloured. This work demonstrates that the haptic device is a suitable and efficient application in forensic science. The haptic device offers a new way in the handling of digital data in the virtual 3D space.

Accuracy of navigated surgery of the pelvis after surface matching with an a-mode ultrasound probe

Computer-aided surgery (CAS) allows for real-time intraoperative feedback resulting in increased accuracy, while reducing intraoperative radiation. CAS is especially useful for the treatment of certain pelvic ring fractures, which necessitate the precise placement of screws. Flouroscopy-based CAS modules have been developed for many orthopedic applications. The integration of the isocentric flouroscope even enables navigation using intraoperatively acquired three-dimensional (3D) data, though the scan volume and imaging quality are limited. Complicated and comprehensive pathologies in regions like the pelvis can necessitate a CT-based navigation system because of its larger field of view. To be accurate, the patient's anatomy must be registered and matched with the virtual object (CT data). The actual precision within the region of interest depends on the area of the bone where surface matching is performed. Conventional surface matching with a solid pointer requires extensive soft tissue dissection. This contradicts the primary purpose of CAS as a minimally invasive alternative to conventional surgical techniques. We therefore integrated an a-mode ultrasound pointer into the process of surface matching for pelvic surgery and compared it to the conventional method. Accuracy measurements were made in two pelvic models: a foam model submerged in water and one with attached porcine muscle tissue. Three different tissue depths were selected based on CT scans of 30 human pelves. The ultrasound pointer allowed for registration of virtually any point on the pelvis. This method of surface matching could be successfully integrated into CAS of the pelvis.

FiFi – Steuerung eines FTF durch Gesten- und Personenerkennung

Um den manuellen Transport in der Intralogistik zu erleichtern, wurde ein Fahrerloses Transportfahrzeug (FTF) entwickelt, das berührungslos vom Bediener gesteuert wird. Die Steuerung erfolgt durch Gesten- und Personenerkennung basierend auf 3D-Daten der Umgebung. Das Paper beschreibt sowohl Zielsetzung, Betriebsarten und Anwendungsmöglichkeiten als auch das Steuerungskonzept der berührungslosen Steuerung und die technische Umsetzung der Plattform. Erste Experimente bestätigen, dass ein Roboter basierend auf 3D-Daten gesteuert werden kann. Verbesserungsmöglichkeiten in der Robustheit werden aufgezeigt.

FiFi – Robuste Personenerkennung gestengesteuerter Fahrzeuge in der Intralogistik

Die Steuerung des Fahrerlosen Transportfahrzeuges „FiFi“ erfolgt berührungslos durch Gesten- und Personenerkennung basierend auf 3D-Daten der Umgebung. Die genutzten Verfahren zur Personenerkennung führen in einigen Fällen zur Falsch-Erkennung von Personen in Objekten. Das Paper beschreibt die Ursachen der Fehlerkennung und stellt die umgesetzten Lösungsansätze zur Vermeidung vor. Experimente bestätigen, dass die entwickelten Verfahren die Robustheit des Systems erhöhen.

OPTiSPIM: integrating optical projection tomography in light sheet microscopy extends specimen characterization to nonfluorescent contrasts.

Mesoscopic 3D imaging has become a widely used optical imaging technique to visualize intact biological specimens. Selective plane illumination microscopy (SPIM) visualizes samples up to a centimeter in size with micrometer resolution by 3D data stitching but is limited to fluorescent contrast. Optical projection tomography (OPT) works with fluorescent and nonfluorescent contrasts, but its resolution is limited in large samples. We present a hybrid setup (OPTiSPIM) combining the advantages of each technique. The combination of fluorescent and nonfluorescent high-resolution 3D data into integrated datasets enables a more extensive representation of mesoscopic biological samples. The modular concept of the OPTiSPIM facilitates incorporation of the transmission OPT modality into already established light sheet based imaging setups.

MorphoGraphX: A platform for quantifying morphogenesis in 4D

Morphogenesis emerges from complex multiscale interactions between genetic and mechanical processes. To understand these processes, the evolution of cell shape, proliferation and gene expression must be quantified. This quantification is usually performed either in full 3D, which is computationally expensive and technically challenging, or on 2D planar projections, which introduces geometrical artifacts on highly curved organs. Here we present MorphoGraphX (www.MorphoGraphX.org), a software that bridges this gap by working directly with curved surface images extracted from 3D data. In addition to traditional 3D image analysis, we have developed algorithms to operate on curved surfaces, such as cell segmentation, lineage tracking and fluorescence signal quantification. The software’s modular design makes it easy to include existing libraries, or to implement new algorithms. Cell geometries extracted with MorphoGraphX can be exported and used as templates for simulation models, providing a powerful platform to investigate the interactions between shape, genes and growth.DOI: http://dx.doi.org/10.7554/eLife.05864.001Author keywordsResearch organism

Systematic literature review of digital three-dimensional superimposition techniques to create virtual dental patients

PURPOSE Digital developments have led to the opportunity to compose simulated patient models based on three-dimensional (3D) skeletal, facial, and dental imaging. The aim of this systematic review is to provide an update on the current knowledge, to report on the technical progress in the field of 3D virtual patient science, and to identify further research needs to accomplish clinical translation. MATERIALS AND METHODS Searches were performed electronically (MEDLINE and OVID) and manually up to March 2014 for studies of 3D fusion imaging to create a virtual dental patient. Inclusion criteria were limited to human studies reporting on the technical protocol for superimposition of at least two different 3D data sets and medical field of interest. RESULTS Of the 403 titles originally retrieved, 51 abstracts and, subsequently, 21 full texts were selected for review. Of the 21 full texts, 18 studies were included in the systematic review. Most of the investigations were designed as feasibility studies. Three different types of 3D data were identified for simulation: facial skeleton, extraoral soft tissue, and dentition. A total of 112 patients were investigated in the development of 3D virtual models. CONCLUSION Superimposition of data on the facial skeleton, soft tissue, and/or dentition is a feasible technique to create a virtual patient under static conditions. Three-dimensional image fusion is of interest and importance in all fields of dental medicine. Future research should focus on the real-time replication of a human head, including dynamic movements, capturing data in a single step.

The virtual patient in dental medicine

OBJECTIVES The aim of this Short Communication was to present a workflow for the superimposition of intraoral scan (IOS), cone-beam computed tomography (CBCT), and extraoral face scan (EOS) creating a 3D virtual dental patient. MATERIAL AND METHODS As a proof-of-principle, full arch IOS, preoperative CBCT, and mimic EOS were taken and superimposed to a unique 3D data pool. The connecting link between the different files was to detect existing teeth as constant landmarks in all three data sets. RESULTS This novel application technique successfully demonstrated the feasibility of building a craniofacial virtual model by image fusion of IOS, CBCT, and EOS under 3D static conditions. CONCLUSIONS The presented application is the first approach that realized the fusion of intraoral and facial surfaces combined with skeletal anatomy imaging. This novel 3D superimposition technique allowed the simulation of treatment planning, the exploration of the patients' expectations, and the implementation as an effective communication tool. The next step will be the development of a real-time 4D virtual patient in motion.

Patrol team language identification system for DARPA RATS P1 evaluation

This paper describes the language identification (LID) system developed by the Patrol team for the first phase of the DARPA RATS (Robust Automatic Transcription of Speech) program, which seeks to advance state of the art detection capabilities on audio from highly degraded communication channels. We show that techniques originally developed for LID on telephone speech (e.g., for the NIST language recognition evaluations) remain effective on the noisy RATS data, provided that careful consideration is applied when designing the training and development sets. In addition, we show significant improvements from the use of Wiener filtering, neural network based and language dependent i-vector modeling, and fusion.

3-dimensional binaural sound rendering using acoustic measurements stored in the SOFA format

Several groups all over the world are researching in several ways to render 3D sounds. One way to achieve this is to use Head Related Transfer Functions (HRTFs). These measurements contain the Frequency Response of the human head and torso for each angle. Some years ago, was only possible to measure these Frequency Responses only in the horizontal plane. Nowadays, several improvements have made possible to measure and use 3D data for this purpose. The problem was that the groups didn't have a standard format file to store the data. That was a problem when a third part wanted to use some different HRTFs for 3D audio rendering. Every of them have different ways to store the data. The Spatially Oriented Format for Acoustics or SOFA was created to provide a solution to this problem. It is a format definition to unify all the previous different ways of storing any kind of acoustics data. At the moment of this project they have defined some basis for the format and some recommendations to store HRTFs. It is actually under development, so several changes could come. The SOFA[1] file format uses a numeric container called netCDF[2], specifically the Enhaced data model described in netCDF 4 that is based on HDF5[3]. The SoundScape Renderer (SSR) is a tool for real-time spatial audio reproduction providing a variety of rendering algorithms. The SSR was developed at the Quality and Usability Lab at TU Berlin and is now further developed at the Institut für Nachrichtentechnik at Universität Rostock [4]. This project is intended to be an introduction to the use of SOFA files, providing a C++ API to manipulate them and adapt the binaural renderer of the SSR for working with the SOFA format. RESUMEN. El SSR (SoundScape Renderer) es un programa que está siendo desarrollado actualmente por la Universität Rostock, y previamente por la Technische Universität Berlin. El SSR es una herramienta diseñada para la reproducción y renderización de audio 2D en tiempo real. Para ello utiliza diversos algoritmos, algunos orientados a sistemas formados por arrays de altavoces en diferentes configuraciones y otros algoritmos diseñados para cascos. El principal objetivo de este proyecto es dotar al SSR de la capacidad de renderizar sonidos binaurales en 3D. Este proyecto está centrado en el binaural renderer del SSR. Este algoritmo se basa en el uso de HRTFs (Head Related Transfer Function). Las HRTFs representan la función de transferencia del sistema formado por la cabeza y el torso del oyente. Esta función es medida desde diferentes ángulos. Con estos datos el binaural renderer puede generar audio en tiempo real simulando la posición de diferentes fuentes. Para poder incluir una base de datos con HRTFs en 3D se ha hecho uso del nuevo formato SOFA (Spatially Oriented Format for Acoustics). Este nuevo formato se encuentra en una fase bastante temprana de su desarrollo. Está pensado para servir como formato estándar para almacenar HRTFs y cualquier otro tipo de medidas acústicas, ya que actualmente cada laboratorio cuenta con su propio formato de almacenamiento y esto hace bastante difícil usar varias bases de datos diferentes en un mismo proyecto. El formato SOFA hace uso del contenedor numérico netCDF, que a su vez esta basado en un contenedor más básico llamado HRTF-5. Para poder incluir el formato SOFA en el binaural renderer del SSR se ha desarrollado una API en C++ para poder crear y leer archivos SOFA con el fin de utilizar los datos contenidos en ellos dentro del SSR.

Large scale SLAM with visual features

Comunicación presentada en el X Workshop of Physical Agents, Cáceres, 10-11 septiembre 2009.

JavaVis: An Integrated Computer Vision Library for Teaching Computer Vision

In this article, we present a new framework oriented to teach Computer Vision related subjects called JavaVis. It is a computer vision library divided in three main areas: 2D package is featured for classical computer vision processing; 3D package, which includes a complete 3D geometric toolset, is used for 3D vision computing; Desktop package comprises a tool for graphic designing and testing of new algorithms. JavaVis is designed to be easy to use, both for launching and testing existing algorithms and for developing new ones.