Showing papers by "Peter Reinartz published in 2013"

Airborne Vehicle Detection in Dense Urban Areas Using HoG Features and Disparity Maps

Abstract: Vehicle detection has been an important research field for years as there are a lot of valuable applications, ranging from support of traffic planners to real-time traffic management. Especially detection of cars in dense urban areas is of interest due to the high traffic volume and the limited space. In city areas many car-like objects (e.g., dormers) appear which might lead to confusion. Additionally, the inaccuracy of road databases supporting the extraction process has to be handled in a proper way. This paper describes an integrated real-time processing chain which utilizes multiple occurrence of objects in images. At least two subsequent images, data of exterior orientation, a global DEM, and a road database are used as input data. The segments of the road database are projected in the non-geocoded image using the corresponding height information from the global DEM. From amply masked road areas in both images a disparity map is calculated. This map is used to exclude elevated objects above a certain height (e.g., buildings and vegetation). Additionally, homogeneous areas are excluded by a fast region growing algorithm. Remaining parts of one input image are classified based on the `Histogram of oriented Gradients (HoG)' features. The implemented approach has been verified using image sections from two different flights and manually extracted ground truth data from the inner city of Munich. The evaluation shows a quality of up to 70 percent.

Region-based automatic building and forest change detection on Cartosat-1 stereo imagery

Abstract: In this paper a novel region-based method is proposed for change detection using space borne panchromatic Cartosat-1 stereo imagery. In the first step, Digital Surface Models (DSMs) from two dates are generated by semi-global matching. The geometric lateral resolution of the DSMs is 5 m × 5 m and the height accuracy is in the range of approximately 3 m (RMSE). In the second step, mean-shift segmentation is applied on the orthorectified images of two dates to obtain initial regions. A region intersection following a merging strategy is proposed to get minimum change regions and multi-level change vectors are extracted for these regions. Finally change detection is achieved by combining these features with weighted change vector analysis. The result evaluations demonstrate that the applied DSM generation method is well suited for Cartosat-1 imagery, and the extracted height values can largely improve the change detection accuracy, moreover it is shown that the proposed change detection method can be used robustly for both forest and industrial areas.

Assessment of Cartosat-1 and WorldView-2 stereo imagery in combination with a LiDAR-DTM for timber volume estimation in a highly structured forest in Germany

Abstract: Stereo satellites provide height information of the earth’s surface with increasing accuracy. High temporal resolution and wide regional coverage are the great advantages of satellites compared with aerial surveys. There is currently little experience of how accurate forest attributes can be modelled using high-resolution stereo satellite data, especially for highly structured forests in Central Europe. Thus, the potential of Cartosat-1 and WorldView-2 was assessed for timber volume estimation in a complex forest in Germany. Digital surface models were generated using Semi-Global Matching. Canopy height models (CHMs) were computed by subtracting a Light detection and ranging (LiDAR) terrain model. The CHMs were co-registered with field plots of a forest inventory. Explanatory variables were derived from the CHMs for timber volume estimation using regressions. Accuracies were evaluated at plot and stand levels. Results were compared with estimations based on a LiDAR-CHM. At plot level the following root mean squared errors (RMSEs) for timber volume estimation were obtained: 50.26 per cent for Cartosat-1, 44.40 per cent for WorldView-2 and 38.02 per cent for LiDAR. The RMSEs were smaller than the standard deviation of the observed timber volume. The RMSEs at a stand level yielded 21.49 per cent for Cartosat-1, 19.59 per cent for WorldView-2 and 17.14 per cent for LiDAR. The study demonstrates the potential of satellite stereo images for regionalization of sample plot inventories.

Building Reconstruction Using DSM and Orthorectified Images

Abstract: High resolution Digital Surface Models (DSMs) produced from airborne laser-scanning or stereo satellite images provide a very useful source of information for automated 3D building reconstruction. In this paper an investigation is reported about extraction of 3D building models from high resolution DSMs and orthorectified images produced from Worldview-2 stereo satellite imagery. The focus is on the generation of 3D models of parametric building roofs, which is the basis for creating Level Of Detail 2 (LOD2) according to the CityGML standard. In particular the building blocks containing several connected buildings with tilted roofs are investigated and the potentials and limitations of the modeling approach are discussed. The edge information extracted from orthorectified image has been employed as additional source of information in 3D reconstruction algorithm. A model driven approach based on the analysis of the 3D points of DSMs in a 2D projection plane is proposed. Accordingly, a building block is divided into smaller parts according to the direction and number of existing ridge lines for parametric building reconstruction. The 3D model is derived for each building part, and finally, a complete parametric model is formed by merging the 3D models of the individual building parts and adjusting the nodes after the merging step. For the remaining building parts that do not contain ridge lines, a prismatic model using polygon approximation of the corresponding boundary pixels is derived and merged to the parametric models to shape the final model of the building. A qualitative and quantitative assessment of the proposed method for the automatic reconstruction of buildings with parametric roofs is then provided by comparing the final model with the existing surface model as well as some field measurements.

Spatial Quality Assessment of Pan-sharpened High Resolution Satellite Imagery Based on an Automatically Estimated Edge Based Metric

Abstract: Most of the existing pan-sharpening quality assessment methods consider only the spectral quality and there are just few investigations, which concentrate on spatial characteristics. Spatial quality of pan-sharpened images is vital in elaborating the capability of object extraction, identification, or reconstruction, especially regarding man-made objects and their application for large scale mapping in urban areas. This paper presents an Edge based image Fusion Metric (EFM) for spatial quality evaluation of pan-sharpening in high resolution satellite imagery. Considering Modulation Transfer Function (MTF) as a precise measurement of edge response, MTFs of pan-sharpened images are assessed and compared to those obtained from the original multispectral or panchromatic images. Spatial quality assessment of pan-sharpening is done by comparison of MTF curves of the pan-sharpened and reference images. The capability of the proposed method is evaluated by quality assessment of two different residential and industrial urban areas of WorldView-2 pan-sharpened images. Obtained results clearly show the wide spatial discrepancy in quality of Pan-sharpened images, resulting from different fusion methods, and confirm the need for spatial quality assessment of fused products. The results also prove the capability of the proposed EFM as a powerful tool for evaluation and comparison of different image fusion techniques and products.

A fuzzy decision making system for building damage map creation using high resolution satellite imagery

Abstract: . Recent studies have shown high resolution satellite imagery to be a powerful data source for post-earthquake damage assessment of buildings. Manual interpretation of these images, while being a reliable method for finding damaged buildings, is a subjective and time-consuming endeavor, rendering it unviable at times of emergency. The present research, proposes a new state-of-the-art method for automatic damage assessment of buildings using high resolution satellite imagery. In this method, at the first step a set of pre-processing algorithms are performed on the images. Then, extracting a candidate building from both pre- and post-event images, the intact roof part after an earthquake is found. Afterwards, by considering the shape and other structural properties of this roof part with its pre-event condition in a fuzzy inference system, the rate of damage for each candidate building is estimated. The results obtained from evaluation of this algorithm using QuickBird images of the December 2003 Bam, Iran, earthquake prove the ability of this method for post-earthquake damage assessment of buildings.

A Multiple SVM System for Classification of Hyperspectral Remote Sensing Data

Abstract: With recent technological advances in remote sensing sensors and systems, very high-dimensional hyperspectral data are available for a better discrimination among different complex land-cover classes. However, the large number of spectral bands, but limited availability of training samples creates the problem of Hughes phenomenon or ‘curse of dimensionality’ in hyperspectral data sets. Moreover, these high numbers of bands are usually highly correlated. Because of these complexities of hyperspectral data, traditional classification strategies have often limited performance in classification of hyperspectral imagery. Referring to the limitation of single classifier in these situations, Multiple Classifier Systems (MCS) may have better performance than single classifier. This paper presents a new method for classification of hyperspectral data based on a band clustering strategy through a multiple Support Vector Machine system. The proposed method uses the band grouping process based on a modified mutual information strategy to split data into few band groups. After the band grouping step, the proposed algorithm aims at benefiting from the capabilities of SVM as classification method. So, the proposed approach applies SVM on each band group that is produced in a previous step. Finally, Naive Bayes (NB) as a classifier fusion method combines decisions of SVM classifiers. Experimental results on two common hyperspectral data sets show that the proposed method improves the classification accuracy in comparison with the standard SVM on entire bands of data and feature selection methods.

Building damage assessment after the earthquake in Haiti using two post-event satellite stereo imagery and DSMs

Abstract: In this paper, a novel disaster building damage monitoring method is presented. This method combines the multispectral imagery and DSMs from stereo matching to obtain three kinds of changes. The proposed method contains three basic steps. The first step is to segment the panchromatic images to get the smallest possible homogeneous regions. In the second step, based on a rule based classification using change information from Iteratively Reweighted Multivariate Alteration Detection (IR-MAD) and height, the changes are classified to ruined buildings, new buildings, and changes without height change (mainly temporary residential area, etc. tents). In the last step, a region based grey level co-occurrence matrix texture measurement is used to refine the third change class. The method is applied to building change detection after the Haiti earthquake.

Alphabet-Based Multisensory Data Fusion and Classification Using Factor Graphs

Abstract: The way of multisensory data integration is a crucial step of any data fusion method. Different physical types of sensors (optic, thermal, acoustic, or radar) with different resolutions, and different types of GIS digital data (elevation, vector map) require a proper method for data integration. Incommensurability of the data may not allow to use conventional statistical methods for fusion and processing of the data. A correct and established way of multisensory data integration is required to deal with such incommensurable data as the employment of an inappropriate methodology may lead to errors in the fusion process. To perform a proper multisensory data fusion several strategies were developed (Bayesian, linear (log linear) opinion pool, neural networks, fuzzy logic approaches). Employment of these approaches is motivated by weighted consensus theory, which lead to fusion processes that are correctly performed for the variety of data properties. As an alternative to several methods, factor graphs are proposed as a new approach for multisensory data fusion. Feature extraction (data fission) is performed separately on different sources of data to make an exhausting description of the fused multisensory data. Extracted features are represented on a finite predefined domain (alphabet). Factor graph is employed for the represented multisensory data fusion. Factorization properties of factor graphs allow to obtain an improvement in accuracy of multisensory data fusion and classification (identification of specific classes) for multispectral high resolution WorldView-2, TerraSAR-X SpotLight, and elevation model data. Application and numerical assessment of the proposed method demonstrates an improved accuracy comparing it to well known data and image fusion methods.

Automatic Crowd Analysis from Very High Resolution Satellite Images

Abstract: . Recently automatic detection of people crowds from images became a very important research field, since it can provide crucial information especially for police departments and crisis management teams. Due to the importance of the topic, many researchers tried to solve this problem using street cameras. However, these cameras cannot be used to monitor very large outdoor public events. In order to bring a solution to the problem, herein we propose a novel approach to detect crowds automatically from remotely sensed images, and especially from very high resolution satellite images. To do so, we use a local feature based probabilistic framework. We extract local features from color components of the input image. In order to eliminate redundant local features coming from other objects in given scene, we apply a feature selection method. For feature selection purposes, we benefit from three different type of information; digital elevation model (DEM) of the region which is automatically generated using stereo satellite images, possible street segment which is obtained by segmentation, and shadow information. After eliminating redundant local features, remaining features are used to detect individual persons. Those local feature coordinates are also assumed as observations of the probability density function (pdf) of the crowds to be estimated. Using an adaptive kernel density estimation method, we estimate the corresponding pdf which gives us information about dense crowd and people locations. We test our algorithm usingWorldview-2 satellite images over Cairo and Munich cities. Besides, we also provide test results on airborne images for comparison of the detection accuracy. Our experimental results indicate the possible usage of the proposed approach in real-life mass events.

Towards automatic 3-D change detection through multi-spectral and digital elevation model information fusion

Abstract: Various 2-D and 3-D change detection techniques have been developed in the literature in order to monitor changes inside urban areas. Nevertheless, most of these techniques require the interaction of the user either to input data, set parameters or to train classifiers. Automatic unsupervised processes have been seldom tackled since they are very difficult to develop if high accuracies are necessary. This article provides a fully automatic change detection procedure for urban areas monitoring. It exploits at best the information provided by multi-spectral images and Digital Elevation Model (DEM) from two different epochs. A fusion, both at feature and decision levels, is thus proposed in order to automatically detect for each epoch the following land cover classes: buildings, shadowed areas, water bodies, ground, low and high vegetation. Applying such fusion on Ikonos stereo data acquired over an Asian urban area in spring 2006 and winter 2010 and their ensuing DEMs has proved both the efficiency and wort...

Building detection using local features and DSM data

Abstract: Detecting and locating buildings in satellite images has various application areas. Unfortunately, manually detecting buildings is hard and very time consuming. Therefore, in the literature several methods are proposed to automatically detect buildings. These methods can be divided into two main groups. In the first group, researchers used panchromatic or multispectral information to detect buildings. In the second group, researchers used DSM data to detect buildings. In this study, we propose two novel methods to detect buildings by combining the panchromatic and DSM data. The first method uses corner points extracted by Harris corner detection method. These corner points are used jointly with DSM data. Using a kernel based density estimation method, possible building locations are detected. In the second method, shadow of buildings are used in a similar way. We tested both methods on WorldView-2 images and DSM data generated from them.

Band Grouping versus Band Clustering in SVM Ensemble Classification of Hyperspectral Imagery

Abstract: Due to the dense sampling of spectral signatures of land covers, hyperspectral images have a better discrimination among similar ground cover classes than traditional remote sensing data. However, these images are usually composed of tens or hundreds of spectrally close bands, which result in high redundancy and great amount of computation time in hyperspectral image classifi cation. In addition, the large number of spectral bands, but limited availability of training samples creates the problem of Hughes phenomenon. Consequently, traditional classifi cation strategies have often limited performance in classifi cation of hyperspectral imagery. Referring to the limitation of single classifi ers in these situations, classifi er ensemble system may exhibit better performance. This paper presents a method for classifi cation of hyperspectral data based on two concepts of Band Clustering (BC) and Band Grouping (BG) through a Support Vector machine (SVM) ensemble system. The proposed method uses the BC\BG strategies to split data into few band portions. After this step, we applied SVM on each band cluster\group that is produced in previous step. Finally, Naive Bayes as a classifi er fusion method combines the decisions of SVM classifi ers. Experimental results show that the proposed method improves the classification accuracy in comparison to the standard SVM and to feature selection methods.

Feature analysis for detecting people from remotely sensed images

Abstract: We propose a novel approach using airborne image sequences for detecting dense crowds and individuals. Although airborne images of this resolution range are not enough to see each person in detail, we can still notice a change of color and intensity components of the acquired image in the location where a person exists. Therefore, we propose a local feature detection-based probabilistic framework to detect people automatically. Extracted local features behave as observations of the probability density function (PDF) of the people locations to be estimated. Using an adaptive kernel density estimation method, we estimate the corresponding PDF. First, we use estimated PDF to detect boundaries of dense crowds. After that, using back- ground information of dense crowds and previously extracted local features, we detect other people in noncrowd regions automatically for each image in the sequence. To test our crowd and people detection algorithm, we use airborne images taken over Munich during the Oktoberfest event, two different open-air concerts, and an outdoor festival. In addition, we apply tests on GeoEye-1 satellite images. Our experimental results indicate possible use of the algo- rithm in real-life mass events. © The Authors. Published by SPIE under a Creative Commons Attribution 3.0 Unported License. Distribution or reproduction of this work in whole or in part requires full attribution of the original publication, including its DOI. (DOI: 10.1117/1.JRS.7.073594)

A Classification Algorithm for Hyperspectral Images Based on Synergetics Theory

Abstract: This paper presents a classification methodology for hyperspectral data based on synergetics theory. Pattern recognition algorithms based on synergetics have been applied to images in the spatial domain with limited success in the past, given their dependence on the rotation, shifting, and scaling of the images. These drawbacks can be discarded if such methods are applied to data acquired by a hyperspectral sensor in the spectral domain, as each single spectrum, related to an image element in the hyperspectral scene, can be analyzed independently. The spectrum is first projected in a space spanned by a set of user-defined prototype vectors, which belong to some classes of interest, and then attracted by a final state associated to a prototype. The spectrum can thus be classified, establishing a first attempt at performing a pixel-wise image classification using notions derived from synergetics. As typical synergetics-based systems have the drawback of a rigid training step, we introduce a new procedure which allows the selection of a training area for each class of interest, used to weight the prototype vectors through attention parameters and to produce a more accurate classification map through plurality vote of independent classifications. As each classification is in principle obtained on the basis of a single training sample per class, the proposed technique could be particularly effective in tasks where only a small training data set is available. The results presented are promising and often outperform state-of-the-art classification methodologies, both general and specific to hyperspectral data.

Fusion of multi-spectral bands and DSM from WorldView-2 stereo imagery for building extraction

Abstract: This paper describes a building extraction approach by fusion panchromatic image, multispectral image and Digital Surface Model (DSM) generated from WorldView-2 stereo imagery. First, the DSM is used to indicate a rough building location, and then Hough-transform is followed to generate 2D segments from panchromatic images. Hough lines in the two main directions of each building are kept. The probability of each segment belonging to a building is calculated from random forest estimations. An automatic training data selection strategy is designed for this supervised classification. Finally, the 2D segments and classification results are combined to get the final building boundaries. This method has been tested in the city center of Munich, Germany.

Classifier fusion of hyperspectral and lidar remote sensing data for improvement of land cover classifcation

Abstract: The interest in the joint use of remote sensing data from multiple sensors has been remarkably increased for classification applications. This is because a combined use is supposed to improve the results of classification tasks compared to single-data use. This paper addressed using of combination of hyperspectral and Light Detection And Ranging (LIDAR) data in classification field. This paper presents a new method based on the definition of a Multiple Classifier System on Hyperspectral and LIDAR data. In the first step, the proposed method applied some feature extraction strategies on LIDAR data to produce more information in this data set. After that in second step, Support Vector Machine (SVM) applied as a supervised classification strategy on LIDAR data and hyperspectal data separately. In third and final step of proposed method, a classifier fusion method used to fuse the classification results on hypersepctral and LIDAR data. For comparative purposes, results of classifier fusion compared to the results of single SVM classifiers on Hyperspectral and LIDAR data. Finally, the results obtained by the proposed classifier fusion system approach leads to higher classification accuracies compared to the single classifiers on hyperspectral and LIDAR data.

Object-based change detection for individual buildings in SAR images captured with different incidence angles

Abstract: Change detection of two SAR images captured with different incidence angles is a difficult task but may be important in urgent situations like earthquakes. This paper presents a simulation based algorithm to detect negative changes of buildings in two high resolution SAR images captured with different incidence angles. The analysis is supported by LiDAR data where individual wall models are extracted and are simulated to predict their shape in the SAR images. Afterwards, point signatures within the layover areas are extracted, converted to the same geometry, and are compared with a buffer change detection algorithm. The proposed method is tested for several buildings (in Munich city center) imaged in TerraSAR-X spotlight mode.

Line-Based Registration of DSM and Hyperspectral Images

Abstract: . Data fusion techniques require a good registration of all the used datasets. In remote sensing, images are usually geo-referenced using the GPS and IMU data. However, if more precise registration is required, image processing techniques can be employed. We propose a method for multi-modal image coregistration between hyperspectral images (HSI) and digital surface models (DSM). The method is divided in three parts: object and line detection of the same object in HSI and DSM, line matching and determination of transformation parameters. Homogeneous coordinates are used to implement matching and adjustment of transformation parameters. The common object in HSI and DSM are building boundaries. They have apparent change in height and material, that can be detected in DSM and HSI, respectively. Thus, before the matching and transformation parameter computation, building outlines are detected and adjusted in HSI and DSM. We test the method on a HSI and two DSM, using extracted building outbounds and for comparison also extracted lines with a line detector. The results show that estimated building boundaries provide more line assignments, than using line detector.

Quality assessment of TanDEM-X Raw DEMs oriented to a fusion with CartoSAT-1 DEMs

Abstract: This paper addresses a quality assessment of TanDEM-X standard Raw DEMs, with a resolution of 12 meters, for three different terrain configurations: urban areas, moderate topography and complex topography. The analysis is performed in the geospatial and in the spectral domain. Beside TanDEM-X, the same analysis is also carried out for CartoSAT-1 and LiDAR DEMs. The latter one is used as a reference. Nevertheless, the focus is centered on TanDEM-X, whose geometric limitations and their impacts on the DEM are analyzed here in detail. How the DEM appears in layover, shadow and phase unwrapping error areas is one of the objectives of the paper. The chosen test site is around Terrassa/Barcelona (Spain), offering all kinds of terrain variations. The final scope of the analysis is to learn about the potentials and the limitations of the two systems, radar (TanDEM-X) and optical (CartoSAT-1), in a way to optimally fuse them and to create an enhanced DEM. A simple fusion processing chain, based on a weighted average depending on the quality of the DEMs adapted to the local geometry, is tested. First results show that in urban and complex terrain areas the improvements are limited, mainly due to the previously analyzed geometrical issues, whereas in moderate terrain areas the enhancement is significant, with a drop in the RMSE of about 25% for TanDEM-X and 30% for CartoSAT-1.

Automatic Model Selection for 3D Reconstruction of Buildings from Satellite Imagary

Abstract: Through the improvements of satellite sensor and matching technology, the derivation of 3D models from space borne stereo data obtained a lot of interest for various applications such as mobile navigation, urban planning, telecommunication, and tourism. The automatic reconstruction of 3D building models from space borne point cloud data is still an active research topic. The challenging problem in this field is the relatively low quality of the Digital Surface Model (DSM) generated by stereo matching of satellite data comparing to airborne LiDAR data. In order to establish an efficient method to achieve high quality models and complete automation from the mentioned DSM, in this paper a new method based on a model-driven strategy is proposed. For improving the results, refined orthorectified panchromatic images are introduced into the process as additional data. The idea of this method is based on ridge line extraction and analysing height values in direction of and perpendicular to the ridgeline direction. After applying pre-processing to the orthorectified data, some feature descriptors are extracted from the DSM, to improve the automatic ridge line detection. Applying RANSAC a line is fitted to each group of ridge points. Finally these ridge lines are refined by matching them or closing gaps. In order to select the type of roof model the heights of point in extension of the ridge line and height differences perpendicular to the ridge line are analysed. After roof model selection, building edge information is extracted from canny edge detection and parameters derived from the roof parts. Then the best model is fitted to extracted facade roofs based on detected type of model. Each roof is modelled independently and final 3D buildings are reconstructed by merging the roof models with the corresponding walls.

Operational Generation of High-Resolution Digital Surface Models from Commercial Tri-Stereo Satellite Data

Abstract: High resolution stereo satellite imagery is now well suited for the creation of digital surface models (DSM) in urban areas due to recent developments in data resolution, quality and collection capabilities. A system for highly automated and operational DSM and orthoimage generation based on WorldView-2 imagery is presented using dense matching methodology, with emphasis on the usage of tri-stereo data for the generation of optimized DSMs. Due to constraints given by the three images, which allow six different pair-wise matchings (including left and right matching of each pair), robust results containing only very few outliers can be generated. The proposed system processes level-1 stereo scenes using the rational polynomial coefficients (RPC) universal sensor model. The RPC are derived from orbit and attitude information and exhibit a lower absolute accuracy than the ground resolution of approximately 0.5 m. In order to use the images for orthorectification or DSM generation, a bias or affine RPC correction is required, which can be achieved through a bundle adjusment using further scenes of the covered area from different dates. Furthermore these DSMs can be used to generate higher level products like digital terrain models (DTM), extracted single 3D objects like buildings and for automatic 3D change detection analysis. DLR-IMF and GAF AG developed and implemented within a close co-operation an operational workflow which now provides operational services based on multi source tri-stereo satellite data. The DSM processing is shortly described, some results of generated DSMs are shown and also examples for higher level products are given in the paper.

A Decision Level Fusion Method for Object Recognition Using Multi-Angular Imagery

Abstract: Spectral similarity and spatial adjacency between various kinds of objects, shadow and occluded areas behind high rise objects as well as complex relationships lead to object recognition difficulties and ambiguities in complex urban areas. Using new multi-angular satellite imagery, higher levels of analysis and developing a context aware system may improve object recognition results in these situations. In this paper, the capability of multi-angular satellite imagery is used in order to solve object recognition difficulties in complex urban areas based on decision level fusion of Object Based Image Analysis (OBIA). The proposed methodology has two main stages. In the first stage, object based image analysis is performed independently on each of the multi-angular images. Then, in the second stage, the initial classified regions of each individual multi-angular image are fused through a decision level fusion based on the definition of scene context. Evaluation of the capabilities of the proposed methodology is performed on multi-angular WorldView-2 satellite imagery over Rio de Janeiro (Brazil).The obtained results represent several advantages of multi-angular imagery with respect to a single shot dataset. Together with the capabilities of the proposed decision level fusion method, most of the object recognition difficulties and ambiguities are decreased and the overall accuracy and the kappa values are improved.

Exploiting noisy hyperspectral bands for water analysis

Abstract: This paper proposes a novel algorithm for the recovery of noisy bands from hyperspec-tral images. The method, based on spectral unmixing, relies on the spectral behavior of the materi-als on ground composing each image element. Firstly, reference spectra related to the classes of in-terest are used to perform spectral unmixing: these exhibit negligible noise influences as they are averaged over areas for which ground truth is available. After the unmixing process, the residual vector is mostly composed by the contributions of uninteresting materials, unwanted atmospheric influences and sensor-induced noise, and is thus ignored in the reconstruction of each spectrum. Finally, the value of a pixel in a given band is predicted as a combination of the noise-free end-members, resulting in a signal with high signal-to-noise ratio in any spectral band. Experiments show that this method could be used to retrieve spectral information from corrupted bands, such as the ones placed at the edge between Ultraviolet and visible light frequencies, which are usually dominated by atmospheric effects and are thus discarded in practical applications. The proposed algorithm could then be exploited in the study of Coloured dissolved organic matter (CDOM) in natural waters.

Automatic Dem Generation And 3D Change Detection From Satellite Imagery

Abstract: In this paper we present a method for fully automatic generation of digital surface models (DSM) from very high resolution (VHR) satellite imagery and the consecutively automatic change detection from the derived 3D information. Common change detection methods are normally only based on spectral change detection. These methods will fail for e.g. comparing summer and winter scenes with the latter covered with snow. Introducing the digital elevation model into the change detection process will allow for a more detailled object modeling and also for the possibility to detect more sophisticated changes like volume estimation of mining activities. Here we present the involved methods for the generation of the high resolution surface models, the fusion and classification and finally the automatic 3D change detection. The methods are applied to some VHR stereo test data sets and the results are evaluated for quality and usefulness for automatic information derivation from large data sets.

Fusion of hyperspectral images and height models using edge probability

Abstract: We propose a method for data fusion of hyperspectral images (HSI) and digital surface models (DSM) basing on the edge probabilities from both datasets. A height discontinuity in DEM and change in material in HSI represent the high probability of an edge. Edge probabilities are computed in scale-space and combined according to the Gaussian mixture model. The reliability of the datasets can be included into this model as a prior knowledge. The method is tested on an urban area, where building boundaries represent the high probabilities of an edge in both datasets. Our results show, that the probabilistic fusion technique is advantageous where boundary detected from only one dataset are unreliable.

Comparison of two fusion based building change detection methods using satellite stereo imagery and DSMs

Abstract: Digital Surface Models (DSMs) can assist building change detection in a variety of approaches. Limited to the quality of DSMs from satellite stereo imagery, it is hard to reach precise change detection results using only DSMs. Therefore, DSMs should be used in combination with the spectral information from original stereo imagery. For that purpose, two fusion based methods, one using feature level fusion and the other using decision level fusion are proposed in our previous research. In this paper, these two methods are further evaluated and compared based on two different data sets. One test site features a typical urban environment which is captured by two pairs of very high resolution stereo imagery (IKONOS and WorldView-2). The other test site is located in an industrial area, the corresponding stereo imagery of the two dates are both from Cartosat-1. Quantitative and qualitative experiment results obtained from each dataset are analyzed in detail. Over all, the proposed feature fusion model give better results for the industrial area, while the decision fusion method works much better for the urban environment based on very high resolution imagery.

Synergetics framework for hyperspectral image classification

Abstract: . In this paper a new classification technique for hyperspectral data based on synergetics theory is presented. Synergetics – originally introduced by the physicist H. Haken – is an interdisciplinary theory to find general rules for pattern formation through selforganization and has been successfully applied in fields ranging from biology to ecology, chemistry, cosmology, and thermodynamics up to sociology. Although this theory describes general rules for pattern formation it was linked also to pattern recognition. Pattern recognition algorithms based on synergetics theory have been applied to images in the spatial domain with limited success in the past, given their dependence on the rotation, shifting, and scaling of the images. These drawbacks can be discarded if such methods are applied to data acquired by a hyperspectral sensor in the spectral domain, as each single spectrum, related to an image element in the hyperspectral scene, can be analysed independently. The classification scheme based on synergetics introduces also methods for spatial regularization to get rid of "salt and pepper" classification results and for iterative parameter tuning to optimize class weights. The paper reports an experiment on a benchmark data set frequently used for method comparisons. This data set consists of a hyperspectral scene acquired by the Airborne Visible Infrared Imaging Spectrometer AVIRIS sensor of the Jet Propulsion Laboratory acquired over the Salinas Valley in CA, USA, with 15 vegetation classes. The results are compared to state-of-the-art methodologies like Support Vector Machines (SVM), Spectral Information Divergence (SID), Neural Networks, Logistic Regression, Factor Graphs or Spectral Angle Mapper (SAM). The outcomes are promising and often outperform state-of-the-art classification methodologies.