Abstract Detecting airborne dust in standard RGB images presents significant challenges. Nevertheless, the monitoring of airborne dust holds substantial potential benefits for climate protection, environmentally sustainable construction, scientific research, and various other fields. To develop an efficient and robust algorithm for airborne dust monitoring, several hurdles have to be addressed. Airborne dust can be opaque or translucent, exhibit considerable variation in density, and possess indistinct boundaries. Moreover, distinguishing dust from other atmospheric phenomena, such as fog or clouds, can be particularly challenging. To meet the demand for a high-performing and reliable method for monitoring airborne dust, we introduce DustNet++, a neural network designed for dust density estimation. DustNet++ leverages feature maps from multiple resolution scales and semantic levels through window and grid attention mechanisms to maintain a sparse, globally effective receptive field with linear complexity. To validate our approach, we benchmark the performance of DustNet++ against existing methods from the domains of crowd counting and monocular depth estimation using the Meteodata airborne dust dataset and the URDE binary dust segmentation dataset. Our findings demonstrate that DustNet++ surpasses comparative methodologies in terms of regression and localization capabilities.

Machine learning methods have proven to be a powerful tool for the classification of Polarimetric Synthetic Aperture Radar (PolSAR) images. Since well-established pixel- or image-based classifiers expect real-valued input data, the information content of complex-valued PolSAR data needs to be represented by real-valued descriptors. This paper proposes a method to enable the generation of suitable descriptors without the explicit extraction of hand-crafted and preselected polarimetric features. For this purpose, the neighbor graph based dimension reduction method Uniform Manifold Approximation and Projection (UMAP) is applied to translate pixel similarity measured by the revised Wishart distance to real-valued 3-dimensional feature descriptors. Land cover classification results performed on a real-world dataset, show that the resulting 3-dimensional feature descriptor which allows a fast and memory efficient training, provides a similar level of information compared to an exhaustive feature set composed of 30 polarimetric features.

Airborne dust has a broad impact from climate to human health. Extensive dust monitoring can lead to identifying environmental hazards and developing mitigation strategies. However, conventional dust measuring devices are usually expensive and limited for the monitoring of the spatial characteristics of dust. Available RGB camera systems might be a potential tool for the measurement of these spatial characteristics, but the automatic detection of airborne dust within these images is not well-researched. The challenges for the required algorithm for such an automatic detection are manifold, including the opaqueness, the wide range of possible density levels, the visual similarity to effects like smoke or clouds, and the fuzzy boundaries of airborne dust. In order to face these challenges in the underexplored research field of detecting airborne dust in terrestrial RGB images, we propose DeepDust. DeepDust is a dust density estimation neural network and exploits convolutional-based multi-level embeddings to merge features from different resolutions and semantic levels. Due to the absence of existing methods in our research field, we compare results achieved by our DeepDust with techniques from the crowd counting and monocular depth estimation domain on the Meteodata dust dataset. Our DeepDust outperforms the other evaluated approaches regarding regression ability by a wide margin.

The Vietnamese Mekong Delta (VMD) has been affected by land subsidence for more than one decade, which is the result of different drivers, including groundwater over-exploitation and natural compaction of holocene layers. The exact share of the contributions is so far unknown. In this work, we systematically study the initial settlement of newly built constructions and built-up areas in two cities in the VMD by incorporating Temporary Persistent Scatterers (TPS) into an existing PS-InSAR approach. We find that the portion of appearing TPS with high initial subsidence rates is larger than the portion of PS and fading TPS with high subsidence rates. We furthermore cluster appearing TPS by their location and appearing date to identify and analyse new constructions and built-up areas and find that they can have really high initial settlement rates which statistically decrease over time. We attribute this signal to the loading-induced initial settlement of constructions which lack a proper foundation. We assume settlement of this kind to contribute noticeably to the measured total subsidence all across the VMD, especially due to the recent high building activity.

Embankment dams are often used to supply their surrounding region with freshwater, hydro-energy or manage floods and are thus of great importance for the region. However, should the dam body burst, it would be an economic loss for the re-gion and threaten the environment, infrastructure, and human life. Therefore, monitoring embankment dams is crucial, es-pecially during the first filling. It is the time period in which the structure experiences the load of the impounded water for the first time. The technique Persistent Scatterer Interferom-etry is well-positioned for this task. Persistent Scatterer Inter-ferometry is a remote sensing technique and thus facilitates infrastructure monitoring in hardly accessible areas. In addition, the technique provides two-dimensional information on the surface deformation instead of sporadic pointwise infor-mation. One end product is a map of the mean velocity of the surface deformation of the observation target in line of sight of the sensor for a specific observation period. In this paper, we present the challenges of creating a surface deformation map for the Parapeiros-Peiros dam. The dam body is excepted to experience a significant amount of defor-mation within a small area during its first filling, which makes identifying persistent scatterers challenging. Hence, we also discuss different approaches described in the literature to select persistent scatterers and their suitability for this specific application.

The persistent scatterer interferometry (PSI) is a powerful remote sensing technique to monitor displacements of the Earth’s surface. It is based on identifying and analyzing phase-stable scatterers undergoing marginal decorrelation over time. Most PSI approaches constrain the analysis to scatterers which are coherent over the whole considered time series (PS) and neglect scatterers which are only temporary persistent (TPS). Here, we propose a method to fully integrate TPS into an existing PSI approach which is characterized by an iterative parameter estimation with subsequent phase unwrapping. We build on a previously published method to identify TPS and estimate their coherent lifetime based on amplitude statistics. A phase-based likelihood ratio test is proposed to iteratively refine the appearing/fading dates of TPS during the parameter estimation. Finally, we jointly unwrap the phase observations of PS and TPS to receive displacement time series. The temporal datum of TPS is re-defined if their lifetime does not cover the selected master scene. Experimental results based on Sentinel-1 data in the Vietnamese city of Ca Mau show that the change date refinement significantly increases the average coherence and number of identified TPS. The densification of the observation point network by incorporating TPS helps to better detect and understand displacement phenomenons. The displacement time series of TPS are beneficial to analyze nonlinear motion in connection with urban development, like initial settlement of newly constructed buildings.

Land subsidence in urban areas is frequently characterized by high spatial heterogeneity, caused by variability in the local lithology, land-use history as well as different load and foundation depths of infrastructure. This effect can, for example, be observed in the Mekong Delta, which is subsiding on large scale with rates up to several centimeters per year. We estimated subsidence rates using PS-InSAR and systematically analyzed differential subsidence between bridges, which are usually built with pile foundations, and the surrounding land surface. We used GIS information on bridge locations from OpenStreetMap and automatically determined the differential subsidence rates at these GIS objects on a delta-wide scale. We show that most of the considered bridges are affected by lower subsidence rates than their surroundings. These results can help in future studies to constrain the depth of soil compaction and better understand the drivers of land subsidence.

In this paper, we focus on the use of multi-modal data to achieve a semantic segmentation of aerial imagery. Thereby, the multi-modal data is composed of a true orthophoto, the Digital Surface Model (DSM) and further representations derived from these. Taking data of different modalities separately and in combination as input to a Residual Shuffling Convolutional Neural Network (RSCNN), we analyze their value for the classification task given with a benchmark dataset. The derived results reveal an improvement if different types of geometric features extracted from the DSM are used in addition to the true orthophoto.

The detection of changes based on remote sensing data is a highly frequented field of research with several of important applications. In [1], a concept for change analysis in urban areas was proposed. The main part of this concept is the categorization of detected changes. Aiming on the changes itself, it can be distinguished between objects, which are static at the time of image acquisition and objects, which are in motion. In SAR image analysis, the detection of moving objects is commonly known as Ground Moving Target Indication (GMTI). For two-channel SAR systems like TerraSAR-X (TSX), a traditional method is Along-Track Interferometry (ATI). In our method [1], changes caused by static objects have been investigated using repeat-pass SAR imagery with a time gap of at least 11 days. This dataset excludes the application of ATI. Furthermore, ATI is only capable of detecting moving objects with motion in across-track direction [5]. In this paper, a GIS based detection scheme is proposed, which considers objects moving in along-track as well as in across-track direction at the time of image acquisition.

This paper presents the innovative Compressive Sensing (CS) concept for tomographic reconstruction of 3D neutrospheric water vapor fields using data from Global Navigation Satellite Systems (GNSS) and Interferometric Synthetic Aperture Radar (InSAR). The Precipitable Water Vapor (PWV) input data are derived from simulations of the Weather Research and Forecasting modeling system. We apply a Compressive Sensing based approach for tomographic inversion. Using the Cosine transform, a sparse representation of the water vapor field is obtained. The new aspects of this work include both the combination of GNSS and InSAR data for water vapor tomography and the sophisticated CS estimation: The combination of GNSS and InSAR data shows a significant improvement in 3D water vapor reconstruction; and the CS estimation produces better results than a traditional Tikhonov regulari-zation with l2norm penalty term.

Remote sensing data have been increasingly used to measure the content of water vapor in the atmosphere and to characterize its temporal and spatial variations. In this paper, we use observations from Global Navigation Satellite System(s) (GNSS) to estimate time series of precipitable water vapor (PWV) by applying the technique of precise point positioning. For an accurate quantification of the absolute PWV, it is necessary to combine the GNSS observations with meteorological data measured directly or inferred at the GNSS site. In addition, measurements of the surface temperature are used to calculate the empirical constant required to convert the GNSS-based delay into water vapor. Our results show strong agreement between the total precipitable water estimated based on GNSS observations and that measured by the sensor MEdium Resolution Imaging Spectrometer with a mean RMS value of 0.98 mm. In a similar way, we compared the GNSS-based total PWV estimates with those produced by the Weather Research and Forecasting (WRF) Modeling System. We found that the WRF model simulations agree well with the GNSS estimates with a mean RMS value of 0.97 mm.

The advantages of SAR sensors for remote sensing applications have been proven many times already. Especially due to the high resolutions that the new generation of SAR sensors (e.g. TerraSAR-X, COSMO-Skymed) can achieve, those are becoming very interesting for the analysis of urban areas. Up to now, mainly interferometric data are used for retrieving the 3D information of building, due to their precise phase information. However, such methodology suffers from the relatively long time span that is required to obtain the data (e.g. TerraSAR-X repeat-pass: 11 days). SAR radargrammetry, on the contrary, has the advantage that the required acquisitions are obtained in shorter time spans, what can be helpful in cases where rapid response is on demand. By matching corresponding points of a stereo image pair, surface height can be retrieved. Our approach propose to determine 3D building parameters, especially height information, relying on the radargrammetric analysis of building layover areas. In this paper, we newly combined a hierarchical matching approach with a matching criterion based on the coefficient of variation. Our method shows improvement if compared to previous works.

The distribution of coherent targets takes a key role in the assessment of the performance and the evaluation of the processing results of multi-temporal Synthetic Aperture Radar interferometry (InSAR) approaches. In previous studies, the evaluation of the spatial distribution of coherent targets was largely based on rather simple parameters such as the spatial point density. While these parameters provide information on the total number of coherent points in an area, they are often insufficient to fully describe their spatial distribution. Hence, with this paper, new descriptors are introduced that better characterize the spatial distribution of coherent targets in an interferogram and serves as a better indicator of InSAR performance. A quantitative study is established both via simulated and real SAR data and the performance of the new parameters are discussed to demonstrate the capability of the developed parameters in describing the spatial distribution of coherent targets.

The dominant error source for short wavelength spaceborne radar signals is due to water vapor present in the neutral atmosphere (neutrosphere). This distortion signal is characterized by high variations in time and space, and can be exploited as a valuable source for quantifying the water vapor content of the Earth's atmosphere. Available water vapor measurements provided by Envisat Medium Resolution Imaging Spectrometer (MERIS) and simulations from numerical weather prediction models are still limited in observing rapid fluctuations of water vapor. Therefore, we are investigating Interferometric Synthetic Aperture Radar (InSAR) for water vapor mapping. In this paper, water vapor maps derived from Persistent Scatterer InSAR (PSI), MERIS, and the Weather Research and Forecasting (WRF) model are presented with comparative analyses.

The high-resolution space borne sensor TerraSAR-X shows good ability for extracting height information by use of repeat-pass interferometry. Indeed, the interferometric phase is still noisy due to temporal decorrelation. The new TanDEM-X mission offers for the first time the possibility of performing high resolution single-pass space borne interferometry, providing good coherence, and thus allowing a better mapping of the 3D shape of objects. These data are still affected by noise though and a filtering is prerequisite in order to enhance object recognition. In previous work, we presented several filtering methods in order to reduce the interferometric noise at building location. We particularly showed the benefits of introducing GIS information such as building footprints into the filtering. In this paper, we present a new filtering method, consisting of combining GIS-supported and area filters for better consideration of large building shape. Results of the GIS supported filter and of the new combined filter are presented on TanDEM-X data.

The research into automatic cartographic mapping is a current topic due to today's availability of high resolution remote sensing data. In order to get as much reliable information as possible, it is recommendable to fuse different image data of the same scene. No matter if the images are acquired by different sensors, from different directions (i.e. multi-aspect data), or are multi-temporal, a careful fusion is required. In this paper we present a high-level decision fusion based on Bayesian network theory developed for automatic road extraction from multi-aspect SAR data. First, the Bayesian network theory is briefly introduced, followed by the process of developing the fusion for the road extraction: 1) Formulating the problem by means of a Bayesian network 2) Learning by estimating up conditional probabilities. Results of the fusion tested on TerraSAR-X data are presented. In the end the potential of the Bayesian network fusion for automatic mapping of cartographic features are discussed.

Fast mapping of storm-damaged forest areas is in great demand. In general, airborne platforms are called into action to get a quick impression and to record high-resolution data. However, such storm events come often along with bad weather conditions that limit acquisition of optical data as well as flying by airplane. In this case, the new generation of high-resolution spaceborne SAR sensors (e.g., TerraSAR-X) can be used to acquire rapidly image data. The new generation of high-resolution spaceborne sensors increases the expectation of more promising results. In this paper, we focus first on the border line extraction of forest areas to enable a fast estimation of wind-thrown areas, whereby the pre-event forest border is derived from multi-spectral data. Second, clean-up operations are monitored in the affected forest area by applying a change detection operator.

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.

Active remote sensing techniques, like SAR tomography and full-waveform LIDAR, are able to capture the 3D reflectance function at or inside objects. They are therefore of special interest for analyzing forest environments. Research goals are the derivation and characterization of the different physical measurement aspects of data taken over forested areas, as well as establishing mutual relations in such a way that LIDAR data can be used to calibrate and correct 3D density data of SAR tomography. The paper outlines the phenomenology of forested areas in SAR tomograms and full-waveform LIDAR data and sketches a simple mathematical methodology for linking SAR and LIDAR reflection density profiles.

Today space-borne high resolution SAR sensors (e.g., TerraSAR-X, TanDEM-X, SAR-Lupe or Cosmo-SkyMed) provide SAR images up to spatial resolutions of 1-3m and even better in spotlight modes. Hence, one major issue of these missions is the development of methods to automatically derive detailed cartographic information from their data. Especially, the analysis of rural and urban areas is on demand in case of disasters (e.g., earthquakes), where active remote sensing systems are highly attractive. Here, an important issue is the development of automatic methods for damage assessment or change detection, in general. For that it is advisable to combine existing GIS data with current SAR data. In this paper an approach for 3D building reconstruction is presented, based on information fusion by utilizing GIS and InSAR data. Thereby, the GIS data are providing the 2D building footprints and the acquired InSAR data the height information. An InSAR simulation step and the subsequent assessment between the real and simulated InSAR phase data enables the extraction of the current building shape.

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.

Current traffic research is mostly based on data from fixed-installed sensors like induction loops, bridge sensors, and cameras. Thereby, the traffic flow on main roads can partially be acquired, while data from the major part of the entire road network are not available. Today's optical sensor systems on satellites provide large-area images with 1-m resolution and better, which can deliver complement information to traditional acquired data. In this paper, we present an approach for automatic vehicle detection from optical satellite images. Therefore, hypotheses for single vehicles are generated using adaptive boosting in combination with Haar-like features. Additionally, vehicle queues are detected using a line extraction technique since grouped vehicles are merged to either dark or bright ribbons. Utilizing robust parameter estimation, single vehicles are determined within those vehicle queues. The combination of implicit modeling and the use ofa prioriknowledge of typical vehicle constellation leads to an enhanced overall completeness compared to approaches which are only based on statistical classification techniques. Thus, a detection rate of over 80% is possible with very high reliability. Furthermore, an approach for movement estimation of the detected vehicle is described, which allows the distinction of moving and stationary traffic. Thus, even an estimate for vehicles' speed is possible, which gives additional information about the traffic condition at image acquisition time.

The side-looking geometry of SAR sensors hampers the interpretation of SAR images of urban areas. Simulation tools for illuminating 3D models of man-made objects by means of a virtual sensor support the interpretation of scattering effects by providing artificial images in the azimuth-range plane. In this paper, a simulation approach is presented which extends SAR simulation to three dimensions in order to focus detected intensity contributions in azimuth, range and elevation. Based on the simulation output, a concept for creating scatterer histograms displaying the number of scatterers within one resolution cell is introduced. Methods for analyzing simulated elevation data by means of selected slices are presented for an urban test site. Eventually, the number of scatterers extracted for a selected pixel by tomographic analysis, using a stack of spotlight TerraSAR-X images, is confirmed by results provided by the simulator.

In this paper, we describe and evaluate an image analysis system for the automated verification of intact roads from multi-sensorial spaceborne imagery. The system is designed to support rescue teams during the reaction on natural disasters such as floodings, landslides or earthquakes. Hence, verification of intact roads comprises both the identification of roads as well as the classification of accessible/intact and inaccessible/damaged roads. Depending on the type and complexity of the input data, the system can be run in a fully automatic or semi-automatic mode, where a human operator can edit intermediate results to ensure the required quality of the final results.

This paper presents an approach for using backward ray tracing for simulating radar reflectivity maps. After explaining the simulation concept which consists of three parts - modeling, sampling and image generation - two applications are presented for showing the performance of the simulator. At first reflection contributions like single or multiple bounce are simulated for a modeled building and compared to a real TerraSAR-X image. It shows how the capability of separating different bounce levels in different layers can support the interpretation of the SAR image. Afterwards reflection effects caused by quasi-perfect specular reflection are detected for another building model by means of geometrical analysis. This is thought as input for advanced Persistent Scatterer (PS) Analysis, since PS typically appear due to such reflection effects.

The Persistent Scatterer (PS) technique is a well known method for estimating object and/or surface deformation from space with high accuracy and high spatial coverage and sampling by analyzing estimation frameworks generated from stacks of interferograms [1]. It is especially suitable for urban areas which, in general, show a high density of PS to detect movements with a wide area deformation pattern as well as changes within small scales, e.g. subsidence of isolated buildings. The new high resolution class satellites like TerraSAR-X, Radarsat-2 or Cosmo-SkyMed put forth new potentials and challenges as the density of PS increases dramatically. As several PS can be found on one single object like a building or a bridge, a advanced motion estimation for this group of points can be introduced, accounting for the coupled movement. In this paper we show how the estimation can be carried out restricting the movement to certain models we employ in order to increase the accuracy of the displacement for the whole group of PS. The results obtained for the proposed estimation technique are based on simulated data to thoroughly investigate the potentials of this approach. They are the basis for real datasets available from TerraSAR-X.

In this paper, we describe an extension of an automatic road extraction procedure developed for single SAR images towards multi-aspect SAR images. Extracted information from multi-aspect SAR images is not only redundant and complementary, in some cases even contradictory. Hence, multi-aspect SAR images require a careful selection within the fusion step. In this work, a fusion step based on probability theory is proposed. During fusion each extracted line primitive is assessed by means of Bayesian probability theory. The assessment is based on the attributes of the line primitive (i.e. length, straightness, etc), global context and sensor geometry. The fusion and its integration into the road extraction system are tested in a sub-urban SAR scene.

This paper introduces a modularized tool for the processing of LIDAR data based on the analysis of neighbor relationships between LIDAR points with the goal to extract planes, lines, and points in 3D. The tool's functionalities will be exemplified by the application of reconstructing building roofs. Detecting buildings within digital surface models is one further step to enhance the results of fully- and semi-automatic software tools which handle huge LIDAR point clouds. The functionalities comprise the sorting of point coordinates to improve efficiency, the retrieval of LIDAR point topology by triangulation of points, the extraction of 3D planes by "plane growing" and the determination of lines, points and roof outlines based on the 3D planes by statistical estimation and hypothesis testing of their parameters.

Traffic monitoring requires mobile and flexible systems that are able to extract densely sampled spatial and temporal traffic data in large areas in near-real time. Video- based systems mounted on aerial platforms meet these requirements, however, at the expense of a limited field of view. To overcome this limitation of video cameras, we develop a system for automatic derivation of traffic flow data which is designed for commercial medium format cameras with a resolution of 25-40 cm and a rather low frame rate of only 1-3 Hz. In this paper, we emphasize the benefits of utilizing an explicit approach of knowledge-based object modeling. The knowledge necessary for automatic image interpretation is modeled by a multi-layered semantic network, and a specific extraction strategy is applied to use this knowledge during object extraction. We tested the system with data of several flight campaigns to verify its applicability.

Traffic data acquisition from space has evolved to an important task over the last years. Future SAR satellite missions will provide high resolution dual-channel SAR data and therefore a possibility to collect traffic parameters of a large area from space. In this paper a detection approach for vehicles will be presented, which considers simultaneously the effects moving objects suffer from in the SAR image. The performance of the proposed detection scheme is analyzed using experimental airborne SAR data.

In this paper we describe a fusion approach for automatic object extraction from multi-aspect SAR images. Before fusion the uncertainty of each extracted object is assessed by means of Bayesian probability theory. The assessment is performed on attribute-level and is based on predefined probability density functions learned from training data. I. INTRODUCTION Automatic extraction of man-made objects from synthetic aperture radar (SAR) images is regarded as a complicated task. Compared to optical image acquisition, SAR system is an active system and can operate during day and night. It is also nearly weather-independent and, moreover, during bad weather conditions, SAR is the only operational system available today. Extraction of man-made objects from SAR images therefore offers a suitable complement or alternative to object extraction from optical images. The recent development of new high resolution SAR systems offers new potential for automatic object extraction. Satellite SAR images up to 1 m resolution will soon be available by the launch of the German satellite TerraSAR-X (1). Airborne images already provide resolution up to 1 decimetre (2). However, the improved resolution does not automatically make automatic object extraction easier, yet it faces new challenges. Especially in urban areas, the complexity arises through dominant scattering caused by building structures, traffic signs and metallic objects in cities. These bright features hinder important extractable features. The inevitable consequences of the side-looking geometry of SAR, occlusions caused by shadow- and layover effects, is present in forestry areas as well as in built-up areas. In urban areas, the best results for the visibility of roads are obtained, when the illumination direction coincide with the main road orientations (3). Preliminary work has shown that the usage of SAR images illuminated from different directions (i.e. multi- aspect images) improves the road extraction results. This has been tested both for real and simulated SAR scenes (4)(5). Multi-aspect SAR images has appeared to be an interesting topic for automatic building extraction as well (6). In this article we present a fusion concept for object extraction based on a Bayesian statistical approach, which incorporates both global context and sensor geometry. The fusion will be implemented in a road extraction approach, (Sect. II), but can as well be applied for other man-made objects. The main focus of this paper is the proposed fusion module, which is explained in Sect. III. Some intermediate results of an uncertainty assessment of line segments based on a training step and global context are discussed in Sect IV. II. ROAD EXTRACTION SYSTEM The extraction of roads from SAR images is based on an already existing road extraction approach (7), which was originally designed for optical images with a ground pixel size of about 2m (8). The first step consists of line extraction using Steger's differential geometry approach (9), which is followed by a smoothening and splitting step. By applying explicit knowledge about roads, the line segments are evaluated according to their attributes such as width, length, curvature, etc. The evaluation is performed within the fuzzy theory. A weighted graph of the evaluated road segments is constructed. For the extraction of the roads from the graph, supplementary road segments are introduced and seed points are defined. Best- valued road segments serve as seed points, which are connected by an optimal path search through the graph. The novelty presented in this paper refers on one hand to the adoption of the fusion module to multi-aspect SAR images and on the other hand to a probabilistic formulation of the fusion problem instead of using fuzzy-functions.

This paper introduces an algorithm for fast and subpixel precise detection of small, compact image primitives ("blobs"). The algorithm is based on differential geometry and incorporates a complete scale-space description. Hence, blobs of arbitrary size can be extracted by just adjusting the scale parameter. In addition to center point and boundary of a blob, also a number of attributes are extracted. These describe the specific blob characteristics in more detail and, thus, allow for a subsequent classification of blobs. Several examples on real images illustrate the performance of the proposed algorithm.

ABSTRACT: Due to increasing traffic there is high demand in traffic monitoring of densely populated urban areas. In our approach we focus on the detection of vehicle queues and use a priori information of roads location and direction. In high resolution satellite imagery single vehicles can hardly be separated since they are merged to either dark or bright ribbons. Initial hypotheses for the queues can be extracted as lines in scale space which represent the centres of the queues. We exploit the context information that vehicle queues are composed of repetitive patterns. For discrimination of single vehicles a width function of the queues is calculated in the gradient image and the variations of the width function are analyzed. We show intermediate and final results of processing a panchromatic QuickBird image covering a part of an inner city area. ‡ A preliminary version of this article has been presented at the ISPRS-Workshop on “High Resolution Earth Imaging for Geo-Information”, Hanover, 2005.

This paper reviews the theoretical background for upcoming dual-channel Radar satellite missions to monitor traffic from space. As it is well-known, an object moving with a velocity deviating from the assumptions incorporated in the focusing process will generally appear both displaced and blurred in the azimuth direction. 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. To be able to monitor traffic under these boundary conditions in real-life situations, a specific detection scheme is proposed. This scheme integrates complementary detection and velocity estimation algorithms with knowledge derived from external sources as, e.g., road databases.

This paper analyzes the potential of the upcoming German satellite mission TerraSAR-X to monitor traffic from space. As it is well-known, an object moving with a velocity deviating from the assumptions incorporated in the focusing process will generally appear both displaced and blurred in azimuth direction. To study the impact of these (and related) distortions in focused SAR images, the analytic relations between an arbitrary moving point scatterer and its conjugate in the SAR image have been derived and adapted to the TerraSAR-X specifications. To be able to monitor traffic under these boundary conditions in real-life situations, a specific detection strategy is proposed. This strategy makes use of knowledge derived from external sources, as e.g. GIS and semantic models for traffic flow.

This paper introduces a new approach to automatic car detection in monocular large scale aerial images. The extraction is based on a hierarchical 3D-model that describes the prominent geometric features of cars or different levels of detail. Furthermore, vehicle color, windshield color, and intensity of a car's shadow area are included as radiometric features. The model automatically adapts the expected saliency of different features depending on vehicle color and the current illumination direction which are measured from the image during extraction or given a priori, respectively. Car extraction is carried out by matching the model "top-down" to the image and evaluating the support found in the image. In contrast to most of the related work, our approach does not rely on external information like digital maps or site models. Various examples illustrate the applicability of this approach. However, they also show the deficiencies which clearly define the next steps of our future work.

We give an overview of our work on automatic road extraction in urban areas. Special emphasis is on the aspect of fusing information from high resolution multiview aerial images.

We propose a scheme for road extraction in rural areas that integrates three different modules with specific strengths. The first module employs local grouping and uses multiple scales and contest to reliably extract most parts of the road network. To connect these parts, the second module exploits the network characteristics of roads for global grouping. The third module completes the network based on an analysis of path lengths within and between connected components of the network. An evaluation of the results shows that the system benefits from the integration of different types of knowledge within the road extraction scheme.