Reinhold Huber

Road extraction from high-resolution airborne SAR using operator fusion

Extraction of roads from high-resolution airborne X-band SAR data is described. The method employs identification of regions of interest, followed by fusion of basic road feature detectors and is complemented by a higher level road model. Identification of regions of interest employs a significance test for the local coefficient of variation. In regions of interest a road feature detector provides a road score through fusion of functions for road edge presence and road center continuity. Finally, an active contour model which is optimized by a genetic algorithm is applied to the fused image.

Adaptive aperture control for image acquisition

Image processing strongly relies on the quality of the input images, as images of good quality can significantly decrease the development effort for image processing and analysis algorithms. A flexible acquisition system for image enhancement, which is able to operate in real time under changing brightness conditions, is suggested The system is based on controlling the aperture of the lens, which makes it useable in combination with all types of image sensors. The control scheme is based on an adaptive image quality estimator and can be used for full images and regions of interest within images. We demonstrate the real-time performance of our approach for different static and dynamic in- and outdoor test scenarios with and without region of interest.

Group Decision Making on Urban Planning Using Desktop Multimedia Conferencing

Recent developments on multimedia systems and networking technology show that using desktop multimedia conferencing for group decision making on wide area networks such as the Internet is possible. In this paper we review the design, hardware and software requirements and organizational issues in a desktop multimedia conferencing system. We draw on our experiences from multiple multimedia conferences on the Internet and in particular we focus on a case study on urban planning using desktop multimedia conferencing on the Internet. Further we discuss implications for further research on desktop multimedia conferencing.

Improving Statistical Measures of Feature Subsets by Conventional and Evolutionary Approaches

In this paper we compare recently developed and highly effective sequential feature selection algorithms with approaches based on evolutionary algorithms enabling parallel feature subset selection. We introduce the oscillating search method, employ permutation encoding offering some advantages over the more traditional bitmap encoding for the evolutionary search, and compare these algorithms to the often studied and well-performing sequential forward floating search. For the empirical analysis of these algorithms we utilize three well-known benchmark problems, and assess the quality of feature subsets by means of the statistical Bhattacharyya distance measure.

Feature selection for ERS-1/2 InSAR classification: high dimensionality case

A systematic way of selection and assessment of the performance of a large number of texture features extracted from spaceborne interferometric SAR data and classified with different types of classifiers is presented. Multi-seasonal ERS-1 and ERS-2 SAR data of the Czech Republic is used to classify into four different land-cover classes. A multistage search method in the space of all possible feature subsets taken from local statistics, fractal analysis and co-occurrence matrices is proposed and tested. In the early stages of the method, features are ranked according to its discriminatory power measured by a ranking coefficient based on subset performance measured by Jeffreis-Matusita-distance. Best ranked features are chosen and a new set is formed and evaluated using the hold-out method employing maximum-likelihood, nearest neighbor and multilayer perceptron classifiers.

Mixed size tournament selection

Abstract In genetic algorithms, tournament schemes are often applied as selection operators. The advantage is simplicity and efficiency. On the other hand, major deficiencies related to tournament selection are the coarse scaling of the selection pressure and the poor sampling accuracy. We introduce a new variant of tournament selection which provides an adjustable probability distribution, a fine-tuning facility for the selection pressure and an improved sampling accuracy at the cost of a minimal increase of the complexity and with almost no loss of efficiency.

Image acquisition using aperture control adapted to spatio-temporal properties

Abstract.Image processing strongly relies on the quality of the input images, as images of appropriate quality can significantly decrease the development effort for image processing and computer vision algorithms. A flexible acquisition system for image enhancement, which is able to operate in real time under changing brightness conditions, is suggested. The system is based on controlling the aperture of the acquisition camera lens, which makes it useable in combination with all types of image sensors. The control scheme is based on an adaptive image quality estimator and can be used to enhance a variety of spatio-temporal properties. Those properties are either characterized by a time-varying or spatial characteristic, or both, i.e. spatio-temporal characteristics of the imaged scene. A region of interest is derived from the more abstract spatio-temporal property. We present results for aperture control adapted to regions of interest characterized by 2D and 3D spatio-temporal properties. We investigate control implemented in software and aimed towards different spatio-temporal properties. Hardware configuration and real-time acquisition capability for static and dynamic changing image contents is demonstrated, and adaptation time and improvement of image quality are measured and compared.

Adaptive aperture control for image enhancement

Image processing strongly relies on the quality of the input images, as images of good quality can significantly decrease the development effort for image processing and analysis algorithms. A flexible acquisition system for image enhancement, which is able to operate in real time under changing brightness conditions, is suggested. The system is based on controlling the aperture of the lens, which makes it useable in combination with all types of image sensors. The control scheme is based on an adaptive image quality estimator and can be used for full images and regions of interest within images. We demonstrate the realtime performance of our approach for different static and dynamic in-and outdoor test scenarios with and without region of interest

Recognition of obstacles on structured 3D background

A stereo vision system for recognition of 3D-objects is presented. The method uses a stereo camera pair and is able to detect objects located on a structured background constituting a repetitive 3D pattern, e.g. a staircase. Recognition is based on differencing stereo pair images, where a perspective warping transform is used to overlay the left onto the right image, or vice versa. The 3D camera positions are obtained during a learning phase where a 3D background model is employed. Correspondence between images and stereo disparity are derived based on the estimated pose of the background model. Disparity provides the necessary information for a perspective warping transform used in the recognition phase. The demonstrated application is staircase surveillance. Recognition itself is based on a pyramidal representation and segmentation of image intensity differences.

Airborne InSAR image interpretation towards cartographic mapping

The capabilities of the airborne high-resolution AeS-1 X-Band InSAR data for automatic cartographic map production are investigated. The cartographic interpretation task becomes feasible by using coherence images and elevation information provided by InSAR processing additionally to the SAR backscatter image. Moreover, as in visual interpretation, features in SAR images are classified upon their textural appearance and background knowledge. Special attention is paid to SAR foreshortening and shadow ambiguities and their impact onto classification. Those ambiguity areas are derived from the DEM and taken into account in training example selection and a semantic classification step supporting a textural multilayer perceptron classifier. A contextual classification operating on MLP results incorporates class context and a-priori knowledge on class composition by Bayesian information fusion.