Stefan Hinz

Automatic Extraction of Urban Road Networks from Multi-view Aerial Imagery

Abstract In this paper, we present work on automatic road extraction from high-resolution aerial imagery taken over urban areas. In order to deal with the high complexity of this type of scenes, we integrate detailed knowledge about roads and their context using explicitly formulated scale-dependent models. The knowledge about how and when certain parts of the road and context model are optimally exploited is expressed by an extraction strategy. The key feature of the presented approach is the integral treatment of three essential issues of object extraction in complex scenes. (1) Specific parts of the road model and extraction strategy are automatically adapted to the respective contextual situation. (2) The extraction incorporates components for self-diagnosis that internally evaluate hypotheses indicating their relevance for further processing. (3) Multiple views on the scene are utilized in different ways. Redundancies in the extraction are exploited, occlusions are predicted and obviated, and a 3D object description is generated. The results achieved with our approach show that a stringent realization of these issues enables the extraction of roads even if their appearance is heavily affected by other objects. Based on an external evaluation of the results, we discuss advantages but also remaining deficiencies of this approach.

Ray-Tracing Simulation Techniques for Understanding High-Resolution SAR Images

In this paper, a simulation concept is presented for creating synthetic aperture radar (SAR) reflectivity maps based on ray tracing. Three-dimensional models of man-made objects are illuminated by a virtual SAR sensor whose signal is approximated by rays sent through the model space. To this end, open-source software tools are adapted and extended to derive output data in SAR geometry followed by creating the reflectivity map. Rays can be followed for multiple reflections within the object scene. Signals having different multiple reflection levels are stored in separate image layers. For evaluating the potentials and limits of the simulation approach, simulated reflectivity maps and distribution maps are compared with real TerraSAR-X images for various complex man-made objects like a skyscraper in Tokyo, the Wynn Hotel in Las Vegas, and the Eiffel Tower in Paris. The results show that the simulation can provide very valuable information to interpret complex SAR images or to predict the reflectivity of planned SAR image acquisitions.

Vehicle Detection in Aerial Images Using Generic Features, Grouping, and Context

This paper introduces a new approach on automatic vehicle detection in monocular large scale aerial images. The extraction is based on a hierarchical model that describes the prominent vehicle features on different levels of detail. Besides the object properties, the model comprises also contextual knowledge, i.e., relations between a vehicle and other objects as, e.g., the pavement beside a vehicle and the sun causing a vehicles shadow projection. In contrast to most of the related work, our approach neither relies on external information like digital maps or site models, nor it is limited to very specific vehicle models. Various examples illustrate the applicability and flexibility of this approach. However, they also show the deficiencies which clearly define the next steps of our future work.

Geometrical Fusion of Multitrack PS Point Clouds

Recent radar satellites like TerraSAR-X and COSMO-SkyMed deliver very high resolution synthetic aperture radar images at a spatial resolution of less than 1 m. Persistent scatterer (PS) positions obtained from stacks of high-resolution spotlight data show very much details of buildings and other structures in 3-D due to the enormous amount of PS obtainable from data of this resolution class. As soon as more than one stack covering the same area is available, a combination of the results is eligible. However, geocoded PSs cannot be simply united due to residual offsets in their absolute positions which stem from unknown absolute height values of the different reference points chosen when processing the individual stacks independently. In this letter, two different methods for a geometrical fusion of geocoded PSs from stacks acquired at different aspect and incidence angles are presented. The algorithms are applied to PS interferometry results of both urban and nonurban areas.

Distinctive 2D and 3D features for automated large-scale scene analysis in urban areas

We propose a new methodology for large-scale urban 3D scene analysis in terms of automatically assigning 3D points the respective semantic labels. The methodology focuses on simplicity and reproducibility of the involved components as well as performance in terms of accuracy and computational efficiency. Exploiting a variety of low-level 2D and 3D geometric features, we further improve their distinctiveness by involving individual neighborhoods of optimal size. Due to the use of individual neighborhoods, the methodology is not tailored to a specific dataset, but in principle designed to process point clouds with a few millions of 3D points. Consequently, an extension has to be introduced for analyzing huge 3D point clouds with possibly billions of points for a whole city. For this purpose, we propose an extension which is based on an appropriate partitioning of the scene and thus allows a successive processing in a reasonable time without affecting the quality of the classification results. We demonstrate the performance of our methodology on two labeled benchmark datasets with respect to robustness, efficiency, and scalability. Graphical abstractWe propose a new methodology for large-scale urban 3D scene analysis which is based on distinctive 2D and 3D features derived from optimal neighborhoods.Display Omitted HighlightsWe present a new methodology for large-scale urban 3D point cloud classification.We analyze a strategy for recovering individual 3D neighborhoods of optimal size.Our methodology involves efficient feature extraction and classification.Our methodology contains an extension towards data-intensive processing.We evaluate our methodology on two recent, publicly available point cloud datasets.

Traffic monitoring with spaceborne SAR-Theory, simulations, and experiments

This paper reviews the theoretical background for upcoming dual-channel radar satellite missions to monitor traffic from space and exemplifies the potentials and limitations by real data. In general, objects that move during the illumination time of the radar will be imaged differently than stationary objects. If the assumptions incorporated in the focusing process of the synthetic aperture radar (SAR) principle are not met, a moving object will appear both displaced and blurred. To study the impact of these (and related) distortions in focused SAR images, the analytic relations between an arbitrarily moving point scatterer and its conjugate in the SAR image have been reviewed and adapted to dual-channel satellite specifications. Furthermore, a specific detection scheme is proposed that integrates complementary detection and velocity estimation algorithms with knowledge derived from external sources as, e.g., road databases. Results using real SAR data are presented to validate the theory.

Classification in High-Dimensional Feature Spaces—Assessment Using SVM, IVM and RVM With Focus on Simulated EnMAP Data

Support Vector Machines (SVM) are increasingly used in methodological as well as application oriented research throughout the remote sensing community. Their classification accuracy and the fact that they can be applied on virtually any kind of remote sensing data set are their key advantages. Especially researchers working with hyperspectral or other high dimensional datasets tend to favor SVMs as they suffer far less from the Hughes phenomenon than classifiers designed for multispectral datasets do. Due to these issues, numerous researchers have published a broad range of enhancements on SVM. Many of these enhancements aim at introducing probability distributions and the Bayes theorem. Within this paper, we present an assessment and comparison of classification results of the SVM and two enhancements-Import Vector Machines (IVM) and Relevance Vector Machines (RVM)-on simulated datasets of the Environmental Mapping and Analysis Program EnMAP.

Fusion of LIDAR data and aerial imagery for automatic reconstruction of building surfaces

An approach for building reconstruction based on fused information extracted from different data sources, namely LIDAR data (light detection and ranging) and aerial imagery, is proposed. The building reconstruction is performed within the scope of a general surface estimation process. This surface estimation aims at generating a DTM including buildings and vegetation removed. The buildings are reconstructed by applying polyhedral models. Thus a large variety of building types can be described with the only limitation, that the roof surface consists of planes.

Integrating pedestrian simulation, tracking and event detection for crowd analysis

In this paper, an overall framework for crowd analysis is presented. Detection and tracking of pedestrians as well as detection of dense crowds is performed on image sequences to improve simulation models of pedestrian flows. Additionally, graph-based event detection is performed by using Hidden Markov Models on pedestrian trajectories utilizing knowledge from simulations. Experimental results show the benefit of our integrated framework using simulation and real-world data for crowd analysis.

Increasing Efficiency of Road Extraction by Self-Diagnosis

The aim of self-diagnosis (internal evaluation) is to determine the geometric and semantic accuracy of the extracted objects during the extraction process. This article presents general ideas on the methodology, implementation, and representation of self-diagnosis within automatic object extraction systems. The authors demonstrate how results attached with confidence values can increase system efficiency for practical applications. They show the potential of self-diagnosis by two road extraction approaches, based on a test series of aerial images, that have been extended by a self-diagnosis component. The authors note that, in order to speed up the time- and cost-intensive inspection, the system should provide the operator with confidence values characterizing its own performance. In practice, however, this is rarely the case. The authors conclude that self-diagnosis is a very helpful tool to guide the user through the road network by pointing to such parts where problems or uncertainties occurred during the extraction. The self-diagnosis also shows great potential to improve an automatic extraction, mainly due to less sensitivity against improper parameter settings. However, there are some deficiencies in distinguishing between erroneous and uncertain, but correct, extractions.