Search

Modern spaceborne synthetic aperture radar (SAR) sensors, such as TerraSAR-X/TanDEM-X and COSMO-SkyMed, can deliver very high resolution (VHR) data beyond the inherent spatial scales (on the order of 1m) of buildings, constituting invaluable data source for large-scale urban mapping. Processing this VHR data with advanced interferometric techniques, such as SAR tomography (TomoSAR), enables the generation of 3-D (or even 4-D) TomoSAR point clouds from space. In this paper, we present a novel and generic workflow that exploits these TomoSAR point clouds in a way that is capable to automatically produce benchmark annotated (buildings/non-buildings) SAR datasets. These annotated datasets (building masks) have been utilized to construct and train the state-of-the-art deep Fully Convolution Neural Networks with an additional Conditional Random Field represented as a Recurrent Neural Network to detect building regions in a single VHR SAR image. The results of building detection are illustrated and validated over TerraSAR-X VHR spotlight SAR image covering approximately 39 km2 - almost the whole city of Berlin - with mean pixel accuracies of around 93.84%.

Reactive systems interact with their environment by reading inputs and computing and feeding back outputs in reactive cycles that are also called ticks. Often they are safety critical systems and are increasingly modeled with highlevel modeling tools. The concepts of the corresponding modeling languages are typically aimed to facilitate formal reasoning about program constructiveness to guarantee deterministic output and are explicitly abstracted from execution time aspects. Nevertheless, the worst-case execution time of a tick can be a crucial value, as it decides the frequency in which the system can interact with its surroundings. An excessive tick Worst Case Execution Time (WCET) can lead to lost inputs or tardy reaction to critical events. Thus, the modeler has to make sure that the timing behaviour of the system under development meets the specifications. This thesis proposes a general approach to interactive timing analysis, which enables the feedback of detailed timing values directly in the model representation to support timing aware modeling. The concept is based on a generic timing interface that enables the exchangeability of the modeling as well as the timing analysis tool for the flexible implementation of varying tool chains. This aims at enhancing the comparability of tools and facilitates the sharing of benchmark model suites. The introduced approach is applicable not only to dataflow-based systems, but also to state-based systems. The latter is enabled by a concept for communicating analysis requests and responses for arbitrary code parts instead of a restriction to function granularity. The proposed timing analysis approach includes visual highlighting and modeling pragmatics features to guide the user to WCET hotspots for timing related model revisions. The approach is practically evaluated with an open-source Eclipse-based example implementation for the modeling language SCCharts, which includes a user study. Reaktive Systeme interagieren mit ihrer Umgebung, indem sie in Zyklen, sogenannten Ticks, Eingaben einlesen und Ausgaben errechnen und zurückgeben. Meistens sind reaktive Systeme sicherheitskritisch. Sie werden typischerweise in spezialisierten Modellierungstools entworfen und die dazugehörigen Modellierungssprachen sind auf die formale Analyse der Konstruktivität optimiert, die eindeutige Ausgaben garantiert. Dies wird unter anderem durch eine bewusste Abstraktion von Laufzeitaspekten erreicht. Dennoch ist die maximale Laufzeit eines Ticks ein bedeutender Aspekt, denn sie bestimmt die mögliche Interaktionsfrequenz zwischen Umwelt und System und damit die Rechtzeitigkeit von Reaktionen. Dementsprechend muss bei der Systementwicklung darauf geachtet werden, dass die Spezifikationen bezüglich der maximalen Ausführungszeit eines Ticks gewahrt bleiben. In dieser Arbeit wird ein genereller Ansatz der interaktiven Ausführungszeitanalyse vorgeschlagen, der diese Aufgabe durch die Anzeige aktueller und detaillierter Zeitwerte in der Modelldarstellung erleichtert. Das Konzept hierzu beruht auf einer übergreifenden Schnittstelle, über die wechselnde Modellierungs- und Analysewerkzeuge flexibel miteinander verbunden werden können, welches Vergleiche und die Wiederverwendung von Testmodellen erleichtert. Der vorgestellte Ansatz ist sowohl für datenflussbasierte als auch für zustandsbasierte Systeme geeignet. Dies wird durch das Konzept ermöglicht, dass anstelle von Funktionen beliebige Codeabschnitte unter Analyse gestellt werden können. Der vorgeschlagene Ansatz umfasst zusätzlich die optische Hervorhebung von zeitkritischen Modellbereichen und den Einsatz von Konzepten aus der Modellierungspragmatik. Das Konzept wurde praktisch mit einer Eclipse-basierten Implementierung in Open-Source-Software ausgewertet; dies schließt die Anwendung in einer Benutzerstudie ein.

To improve the classification of hyperspectral images, this paper proposes an approach for multi-sensor data fusion of LiDAR and hyperspectral data using extinction profiles and Orthogonal Total Variation Component Analysis (OTVCA). Results on the benchmark Houston data indicate the superior performance of the proposed approach compared to other approaches used in the experiments based on classification accuracies.

Most saliency estimation methods aim to explicitly model low-level conspicuity cues such as edges or blobs and may additionally incorporate top-down cues using face or text detection. Data-driven methods for training saliency mod- els using eye-fixation data are increasingly popular, par- ticularly with the introduction of large-scale datasets and deep architectures. However, current methods in this lat- ter paradigm use loss functions designed for classification or regression tasks whereas saliency estimation is evalu- ated on topographical maps. In this work, we introduce a new saliency map model which formulates a map as a generalized Bernoulli distribution. We then train a deep ar- chitecture to predict such maps using novel loss functions which pair the softmax activation function with measures designed to compute distances between probability distri- butions. We show in extensive experiments the effective- ness of such loss functions over standard ones on four pub- lic benchmark datasets, and demonstrate improved perfor- mance over state-of-the-art saliency methods.

The Second Workshop on Benchmark Problems for Airframe Noise Computations, BANC-II, was held on 7-8 June 2012 in Colorado Springs, CO, USA. The objective of this workshop was to assess the present computational capability in the area of physics-based prediction of dierent types of airframe noise problems and to advance the state-of-the-art via a combined effort. This paper summarizes the results from workshop category 1 which focuses on the prediction of broadband turbulent boundary-layer trailing-edge noise and related source quantities. 2D airfoil sections, namely a NACA0012 and DU-96-180, served as test cases. Code-to-code comparisons in this category were mainly restricted to relatively fast RANS-based methods applying statistical noise theory. Overall, the prediction capability was sufficient to capture the principal trailing-edge farfield noise scaling behavior in the mid-frequency range (about 1 kHz