Anca Popescu - Academia.edu (original) (raw)
Papers by Anca Popescu
This paper presents a comparative assessment of Synthetic Aperture Radar interferometric techniqu... more This paper presents a comparative assessment of Synthetic Aperture Radar interferometric techniques (InSAR) that allow the detection of deformation models along the line-of-sight of the radar. Given the susceptibility of these methods to several limitations that act as noise effects in the interferograms, known as decorrelation phenomena, multi-temporal InSAR techniques have been used for the exploiting of phase information acquired over long time intervals.
SAR image understanding for students
Abstract Synthetic Aperture Radars (SAR) have gained great importance over the last few years, ... more Abstract Synthetic Aperture Radars (SAR) have gained great importance over the last few years, being used in many areas of applications and science. Unlike Real Aperture Radars, the SAR system synthetically increases the size of the antenna in order to obtain a higher ...
This paper addresses the problem of extracting content-based image parameters from high resolutio... more This paper addresses the problem of extracting content-based image parameters from high resolution SAR data. The information extraction is achieved through a spectral approach in which space-variant measures are defined in the frequency domain. In order to thoroughly capture the informational content from the TerraSAR-X Spotlight single look complex data, a preliminary phase correction algorithm is applied to the focused SSC image to compensate the drift of the Doppler centroid frequency in azimuth direction, caused by variation of the squint angle during spotlight imaging. The proposed spectral features are computed using a large window size and capture not only reflectance differences between the different scattering targets, but are also sensitive to the scene geometry, thus it is possible to discriminate between different scattering mechanisms present in the SAR scene and to point out similar regions or objects, as for instance when mining large data archives. The extracted features compose a feature vector and thus a vector classification algorithm is necessary to perform the classification of the SSC image and to label the different informational objects.
High resolution TerraSAR-X image mining using RELAX
This paper compares the performance of two feature extraction methods applied on data acquired by... more This paper compares the performance of two feature extraction methods applied on data acquired by the ESA satellite Sentinel-1. The feature extraction methods used in our experiments were previously tested on high and very high resolution SAR data and were reported to be able to discriminate between a relevant numbers of land cover classes. Based on the available resolution (10x10m) of Sentinel-1 Interferometric Wide images compared to the resolution of TerraSAR-X (1m) the number of detected classes is expected to be much lower. The results are quantitatively assessed by employing optical data (Urban Atlas) as reference and for visual support.
Progress in Spatial Analysis and Fusion of Remote Sensing Satellite Data for Generating High-Level Semantic Geo-Information in Support of GMES Applications
Enhanced classification of land cover through joint analysis of Sentinel 1 and Sentinel 2 data
The new European missions Sentinel 1 and Sentinel 2 bring added value to the Earth Observation ma... more The new European missions Sentinel 1 and Sentinel 2 bring added value to the Earth Observation market for land applications by providing timely high quality products with complementary information. The efficient and effective exploitation of the vast amounts of data already delivered by Sentinel 1, combined with the data delivered by Sentinel 2 is mandatory in order to get improved capabilities for the understanding and modelling of land cover classes dynamic and synergy. In this context, this paper discusses a potential methodology for Sentinel 1 and Sentinel 2 joint analysis and highlights the benefits brought by the increased information yielded by putting together these two kinds of data. To this end, the proposed approach follows the next steps: data pre-processing and registration, data representation and combined feature extraction, data classification and interpretation. The results are emphasized on urban and agricultural areas of high heterogeneity.
Traditional approaches for SAR image classification and target recognition in high resolution acq... more Traditional approaches for SAR image classification and target recognition in high resolution acquisitions discard phase information and are solely based on the detected data. When one makes use of several SLC image acquisitions, the phase variation is the main source of information. We propose a method for feature definition and extraction which combines single SAR acquisitions with interferometric information. Not only that phase becomes less random, but it describes the changes that occur on the received scene reflectivity. In the hypothesis that two successive acquisitions are taken on a stationary scene (tandem acquisitions), combined information from these data can be employed to describe the scene content. Thus, we propose the usage of spectral based descriptors that make use of the full complex signal from two or more SLC acquisitions to construct robust image descriptors. The experiments show that the interferometric information can be very valuable for the recognition of c...
Persistent scatterers detection on synthetic aperture radar images acquired by Sentinel-1 satellite
Advanced Topics in Optoelectronics, Microelectronics, and Nanotechnologies VIII, 2016
Persistent Scatterers Interferometry (PS-InSAR) has become a popular method in remote sensing bec... more Persistent Scatterers Interferometry (PS-InSAR) has become a popular method in remote sensing because of its capability to measure terrain deformations with very high accuracy. It relies on multiple Synthetic Aperture Radar (SAR) acquisitions, to monitor points with stable proprieties over time, called Persistent Scatterers (PS)[1]. These points are unaffected by temporal decorrelation, therefore by analyzing their interferometric phase variation we can estimate the scene’s deformation rates within a given time interval. In this work, we apply two incoherent detection algorithms to identify Persistent Scatterers candidates in the city of Focșani, Romania. The first method studies the variation of targets’ intensities along the SAR acquisitions and the second method analyzes the spectral proprieties of the scatterers. The algorithms were implemented on a dataset containing 11 complex images of the region covering Buzău, Brăila and Focșani cities. Images were acquired by Sentinel-1 satellite in a time span of 5 months, from October 2014 to February 2015. The processing chain follows the requirements imposed by the new C-band SAR images delivered by the Sentinel-1 satellite (launched in April 2014) imaging in Interferometric Wide (IW) mode. Considering the particularities of the TOPS (Terrain Observation with Progressive Scans in Azimuth) imaging mode[2], special requirements had to be considered for pre-processing steps. The PS detection algorithms were implemented in Gamma RS program, a software which contains various function packages dedicated to SAR images focalization, analysis and processing.
Dual-band Bistatic SAR system with satellite emitter of opportunity and ground-based receiver
2016 International Conference on Communications (COMM), 2016
Space-borne-ground-based Bistatic SAR systems are becoming increasingly interesting, and the moti... more Space-borne-ground-based Bistatic SAR systems are becoming increasingly interesting, and the motivation is threefold: they call for specific signal processing algorithms and allow for a better characterization of the signal properties and propagation path; angular diversity introduced by the bistatic angle and the receiver attitude opens new perspectives for target characterization; modular and configurable receiver hardware permits diverse applications with little effort. In this context, this paper presents the main hardware characteristics of a dual-band Bistatic SAR system developed at the University Politehnica of Bucharest. The system consists of a C/X-band SAR receiver with satellite emitter of opportunity. We discuss the system properties, applications and we shortly revisit the ongoing experiments and measurements.
HR Sar Iim
High resolution SAR images show specifc properties that can be used for image information mining.
High resolution SAR classification using Ränyi Entropy constrained spectrum estimates
This paper discusses a feature extraction procedure based on the estimation of spectral component... more This paper discusses a feature extraction procedure based on the estimation of spectral components of the complex SAR image for scene classification. The estimation is performed in an iterative manner, using the periodogram of SAR image patches of fixed size. A Ranyi Entropy constraint is introduced in the process of selection of best spectral components for content characterization. Results are presented on a database consisting of 1650 image patches covering urban areas. Results show that by selecting the spectral components that maximize the Ranyi entropy of the spectrum with α=3 can improve the accuracy of the classification.
Proceedings of Fringe 2015: Advances in the Science and Applications of SAR Interferometry and Sentinel-1 InSAR Workshop, 2015
The paper evaluates new Staring Spotlight mode capabilities to monitor the mining activities impa... more The paper evaluates new Staring Spotlight mode capabilities to monitor the mining activities impacts on the environment to ensure an effective management and to prevent possible natural and technological hazards. The societal and environmental impacts are huge such as: the topographic alteration, changes in the soil structure and vegetation coverage, influence on the underground water resources and on the rain water draining regime and air pollution. Rosia Jiu opencast test site is affected by subsidence phenomena caused by the closing of the hollows remained from the underground exploitation of lignite and by altering of the hidro-geological conditions, due to the applying of a forced and high intensity dewatering of the aquifer system within the area. A methodology based on deformation maps is designed for monitoring of the elastic deformation, early warning stage and detection of the risk occurrence. Intense mining activities from the summer -autumn seasons implied as interferometric pairs to have very low coherence making almost impossible to find PS candidates.
An assessment of feature extraction methods for SENTINEL-1 images on urban areas
2015 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), 2015
This paper makes a comparative assessment of the observable landcover classes visible in the data... more This paper makes a comparative assessment of the observable landcover classes visible in the data provided by the newly launched SENTINEL-1 (S-1) satellite. The analysis focuses on two feature extraction methods previously reported in the literature to be able to distinguish between a relatively large number of classes in high and very high resolution SAR data. The analysis of the S-1 medium resolution data makes use of the Gabor filtering and Fourier spectral coefficients. Moreover, we consider the opportunity of speckle reduction before feature extraction, considering that the texture analysis is sensitive to correlated speckle. Furthermore, the comparison takes into account the adaptation of the window size to the resolution and pixel spacing of the data. In order to make a quantitative assessment of the results, we perform a joint evaluation of the detection probability at pixel and patch level, with respect to an expert annotated dataset.
A comparison of feature extraction methods for Sentinel-1 images: Gabor and Weber transforms
SAR Image Analysis, Modeling, and Techniques XV, 2015
The purpose of this paper is to compare the performance of two feature extraction methods when ap... more The purpose of this paper is to compare the performance of two feature extraction methods when applied on high resolution Synthetic Aperture Radar (SAR) images acquired with the new ESA mission SENTINEL-1 (S-1). The feature extraction methods were previously tested on high and very high resolution SAR data (imaged by TerraSAR-X) and had a good performance in discriminating between a relevant numbers of land cover classes (tens of classes). Based on the available spatial resolution (10x10m) of S-1 Interferometric Wide (IW) Ground Range Detected (GRD) images the number of detectable classes is much lower. Moreover, the overall heterogeneity of the images is much lower as compared to the high resolution data, the number of observable details is smaller, and this favors the choice of a smaller window size for the analysis: between 10 and 50 pixels in range and azimuth. The size of the analysis window ensures the consistency with the previous results reported in the literature in very high resolution data (as the size on the ground is comparable and thus the number of contributing objects in the window is similar). The performance of Gabor filters and the Weber Local Descriptor (WLD) was investigated in a twofold approach: first the descriptors were computed directly over the IW GRD images and secondly on the sub-sampled version of the same data (in order to determine the effect of the speckle correlation on the overall class detection probability).
This paper addresses the problem of extracting content-based image parameters from high resolutio... more This paper addresses the problem of extracting content-based image parameters from high resolution SAR data. The information extraction is achieved through a spectral approach in which space-variant measures are defined in the frequency domain. In order to thoroughly capture the informational content from the TerraSAR-X Spotlight single look complex data, a preliminary phase correction algorithm is applied to the focused SSC image to compensate the drift of the Doppler centroid frequency in azimuth direction, caused by variation of the squint angle during spotlight imaging. The proposed spectral features are computed using a large window size and capture not only reflectance differences between the different scattering targets, but are also sensitive to the scene geometry, thus it is possible to discriminate between different scattering mechanisms present in the SAR scene and to point out similar regions or objects, as for instance when mining large data archives. The extracted features compose a feature vector and thus a vector classification algorithm is necessary to perform the classification of the SSC image and to label the different informational objects.
Mining Very High Resolution InSAR Data based on Complex-GMRF Cues and Relevance Feedback
With the increase in number of remote sensing satellites, the number of image-data scenes in our ... more With the increase in number of remote sensing satellites, the number of image-data scenes in our repositories is also increasing and a large quantity of these scenes are never received and used. Thus automatic retrieval of desired image-data using query by image content to fully utilize the huge repository volume is becoming of great interest. Generally different users are interested in scenes containing different kind of objects and structures. So its important to analyze all the image information mining (IIM) methods so that its easier for user to select a method depending upon his/her requirement. We concentrate our study only on high-resolution SAR images and we propose to use InSAR observations instead of only one single look complex (SLC) images for mining scenes containing coherent objects such as high-rise buildings. However in case of objects with less coherence like areas with vegetation cover, SLC images exhibits better performance. We demonstrate IIM performance comparis...
HR Sar Iim
High resolution SAR images show specifc properties that can be used for image information mining.
High resolution TerraSAR-X image ming using RELAX
We present a feature extraction procedure for SAR images based on the RELAX algorithm.
This paper presents a comparative assessment of Synthetic Aperture Radar interferometric techniqu... more This paper presents a comparative assessment of Synthetic Aperture Radar interferometric techniques (InSAR) that allow the detection of deformation models along the line-of-sight of the radar. Given the susceptibility of these methods to several limitations that act as noise effects in the interferograms, known as decorrelation phenomena, multi-temporal InSAR techniques have been used for the exploiting of phase information acquired over long time intervals.
SAR image understanding for students
Abstract Synthetic Aperture Radars (SAR) have gained great importance over the last few years, ... more Abstract Synthetic Aperture Radars (SAR) have gained great importance over the last few years, being used in many areas of applications and science. Unlike Real Aperture Radars, the SAR system synthetically increases the size of the antenna in order to obtain a higher ...
This paper addresses the problem of extracting content-based image parameters from high resolutio... more This paper addresses the problem of extracting content-based image parameters from high resolution SAR data. The information extraction is achieved through a spectral approach in which space-variant measures are defined in the frequency domain. In order to thoroughly capture the informational content from the TerraSAR-X Spotlight single look complex data, a preliminary phase correction algorithm is applied to the focused SSC image to compensate the drift of the Doppler centroid frequency in azimuth direction, caused by variation of the squint angle during spotlight imaging. The proposed spectral features are computed using a large window size and capture not only reflectance differences between the different scattering targets, but are also sensitive to the scene geometry, thus it is possible to discriminate between different scattering mechanisms present in the SAR scene and to point out similar regions or objects, as for instance when mining large data archives. The extracted features compose a feature vector and thus a vector classification algorithm is necessary to perform the classification of the SSC image and to label the different informational objects.
High resolution TerraSAR-X image mining using RELAX
This paper compares the performance of two feature extraction methods applied on data acquired by... more This paper compares the performance of two feature extraction methods applied on data acquired by the ESA satellite Sentinel-1. The feature extraction methods used in our experiments were previously tested on high and very high resolution SAR data and were reported to be able to discriminate between a relevant numbers of land cover classes. Based on the available resolution (10x10m) of Sentinel-1 Interferometric Wide images compared to the resolution of TerraSAR-X (1m) the number of detected classes is expected to be much lower. The results are quantitatively assessed by employing optical data (Urban Atlas) as reference and for visual support.
Progress in Spatial Analysis and Fusion of Remote Sensing Satellite Data for Generating High-Level Semantic Geo-Information in Support of GMES Applications
Enhanced classification of land cover through joint analysis of Sentinel 1 and Sentinel 2 data
The new European missions Sentinel 1 and Sentinel 2 bring added value to the Earth Observation ma... more The new European missions Sentinel 1 and Sentinel 2 bring added value to the Earth Observation market for land applications by providing timely high quality products with complementary information. The efficient and effective exploitation of the vast amounts of data already delivered by Sentinel 1, combined with the data delivered by Sentinel 2 is mandatory in order to get improved capabilities for the understanding and modelling of land cover classes dynamic and synergy. In this context, this paper discusses a potential methodology for Sentinel 1 and Sentinel 2 joint analysis and highlights the benefits brought by the increased information yielded by putting together these two kinds of data. To this end, the proposed approach follows the next steps: data pre-processing and registration, data representation and combined feature extraction, data classification and interpretation. The results are emphasized on urban and agricultural areas of high heterogeneity.
Traditional approaches for SAR image classification and target recognition in high resolution acq... more Traditional approaches for SAR image classification and target recognition in high resolution acquisitions discard phase information and are solely based on the detected data. When one makes use of several SLC image acquisitions, the phase variation is the main source of information. We propose a method for feature definition and extraction which combines single SAR acquisitions with interferometric information. Not only that phase becomes less random, but it describes the changes that occur on the received scene reflectivity. In the hypothesis that two successive acquisitions are taken on a stationary scene (tandem acquisitions), combined information from these data can be employed to describe the scene content. Thus, we propose the usage of spectral based descriptors that make use of the full complex signal from two or more SLC acquisitions to construct robust image descriptors. The experiments show that the interferometric information can be very valuable for the recognition of c...
Persistent scatterers detection on synthetic aperture radar images acquired by Sentinel-1 satellite
Advanced Topics in Optoelectronics, Microelectronics, and Nanotechnologies VIII, 2016
Persistent Scatterers Interferometry (PS-InSAR) has become a popular method in remote sensing bec... more Persistent Scatterers Interferometry (PS-InSAR) has become a popular method in remote sensing because of its capability to measure terrain deformations with very high accuracy. It relies on multiple Synthetic Aperture Radar (SAR) acquisitions, to monitor points with stable proprieties over time, called Persistent Scatterers (PS)[1]. These points are unaffected by temporal decorrelation, therefore by analyzing their interferometric phase variation we can estimate the scene’s deformation rates within a given time interval. In this work, we apply two incoherent detection algorithms to identify Persistent Scatterers candidates in the city of Focșani, Romania. The first method studies the variation of targets’ intensities along the SAR acquisitions and the second method analyzes the spectral proprieties of the scatterers. The algorithms were implemented on a dataset containing 11 complex images of the region covering Buzău, Brăila and Focșani cities. Images were acquired by Sentinel-1 satellite in a time span of 5 months, from October 2014 to February 2015. The processing chain follows the requirements imposed by the new C-band SAR images delivered by the Sentinel-1 satellite (launched in April 2014) imaging in Interferometric Wide (IW) mode. Considering the particularities of the TOPS (Terrain Observation with Progressive Scans in Azimuth) imaging mode[2], special requirements had to be considered for pre-processing steps. The PS detection algorithms were implemented in Gamma RS program, a software which contains various function packages dedicated to SAR images focalization, analysis and processing.
Dual-band Bistatic SAR system with satellite emitter of opportunity and ground-based receiver
2016 International Conference on Communications (COMM), 2016
Space-borne-ground-based Bistatic SAR systems are becoming increasingly interesting, and the moti... more Space-borne-ground-based Bistatic SAR systems are becoming increasingly interesting, and the motivation is threefold: they call for specific signal processing algorithms and allow for a better characterization of the signal properties and propagation path; angular diversity introduced by the bistatic angle and the receiver attitude opens new perspectives for target characterization; modular and configurable receiver hardware permits diverse applications with little effort. In this context, this paper presents the main hardware characteristics of a dual-band Bistatic SAR system developed at the University Politehnica of Bucharest. The system consists of a C/X-band SAR receiver with satellite emitter of opportunity. We discuss the system properties, applications and we shortly revisit the ongoing experiments and measurements.
HR Sar Iim
High resolution SAR images show specifc properties that can be used for image information mining.
High resolution SAR classification using Ränyi Entropy constrained spectrum estimates
This paper discusses a feature extraction procedure based on the estimation of spectral component... more This paper discusses a feature extraction procedure based on the estimation of spectral components of the complex SAR image for scene classification. The estimation is performed in an iterative manner, using the periodogram of SAR image patches of fixed size. A Ranyi Entropy constraint is introduced in the process of selection of best spectral components for content characterization. Results are presented on a database consisting of 1650 image patches covering urban areas. Results show that by selecting the spectral components that maximize the Ranyi entropy of the spectrum with α=3 can improve the accuracy of the classification.
Proceedings of Fringe 2015: Advances in the Science and Applications of SAR Interferometry and Sentinel-1 InSAR Workshop, 2015
The paper evaluates new Staring Spotlight mode capabilities to monitor the mining activities impa... more The paper evaluates new Staring Spotlight mode capabilities to monitor the mining activities impacts on the environment to ensure an effective management and to prevent possible natural and technological hazards. The societal and environmental impacts are huge such as: the topographic alteration, changes in the soil structure and vegetation coverage, influence on the underground water resources and on the rain water draining regime and air pollution. Rosia Jiu opencast test site is affected by subsidence phenomena caused by the closing of the hollows remained from the underground exploitation of lignite and by altering of the hidro-geological conditions, due to the applying of a forced and high intensity dewatering of the aquifer system within the area. A methodology based on deformation maps is designed for monitoring of the elastic deformation, early warning stage and detection of the risk occurrence. Intense mining activities from the summer -autumn seasons implied as interferometric pairs to have very low coherence making almost impossible to find PS candidates.
An assessment of feature extraction methods for SENTINEL-1 images on urban areas
2015 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), 2015
This paper makes a comparative assessment of the observable landcover classes visible in the data... more This paper makes a comparative assessment of the observable landcover classes visible in the data provided by the newly launched SENTINEL-1 (S-1) satellite. The analysis focuses on two feature extraction methods previously reported in the literature to be able to distinguish between a relatively large number of classes in high and very high resolution SAR data. The analysis of the S-1 medium resolution data makes use of the Gabor filtering and Fourier spectral coefficients. Moreover, we consider the opportunity of speckle reduction before feature extraction, considering that the texture analysis is sensitive to correlated speckle. Furthermore, the comparison takes into account the adaptation of the window size to the resolution and pixel spacing of the data. In order to make a quantitative assessment of the results, we perform a joint evaluation of the detection probability at pixel and patch level, with respect to an expert annotated dataset.
A comparison of feature extraction methods for Sentinel-1 images: Gabor and Weber transforms
SAR Image Analysis, Modeling, and Techniques XV, 2015
The purpose of this paper is to compare the performance of two feature extraction methods when ap... more The purpose of this paper is to compare the performance of two feature extraction methods when applied on high resolution Synthetic Aperture Radar (SAR) images acquired with the new ESA mission SENTINEL-1 (S-1). The feature extraction methods were previously tested on high and very high resolution SAR data (imaged by TerraSAR-X) and had a good performance in discriminating between a relevant numbers of land cover classes (tens of classes). Based on the available spatial resolution (10x10m) of S-1 Interferometric Wide (IW) Ground Range Detected (GRD) images the number of detectable classes is much lower. Moreover, the overall heterogeneity of the images is much lower as compared to the high resolution data, the number of observable details is smaller, and this favors the choice of a smaller window size for the analysis: between 10 and 50 pixels in range and azimuth. The size of the analysis window ensures the consistency with the previous results reported in the literature in very high resolution data (as the size on the ground is comparable and thus the number of contributing objects in the window is similar). The performance of Gabor filters and the Weber Local Descriptor (WLD) was investigated in a twofold approach: first the descriptors were computed directly over the IW GRD images and secondly on the sub-sampled version of the same data (in order to determine the effect of the speckle correlation on the overall class detection probability).
This paper addresses the problem of extracting content-based image parameters from high resolutio... more This paper addresses the problem of extracting content-based image parameters from high resolution SAR data. The information extraction is achieved through a spectral approach in which space-variant measures are defined in the frequency domain. In order to thoroughly capture the informational content from the TerraSAR-X Spotlight single look complex data, a preliminary phase correction algorithm is applied to the focused SSC image to compensate the drift of the Doppler centroid frequency in azimuth direction, caused by variation of the squint angle during spotlight imaging. The proposed spectral features are computed using a large window size and capture not only reflectance differences between the different scattering targets, but are also sensitive to the scene geometry, thus it is possible to discriminate between different scattering mechanisms present in the SAR scene and to point out similar regions or objects, as for instance when mining large data archives. The extracted features compose a feature vector and thus a vector classification algorithm is necessary to perform the classification of the SSC image and to label the different informational objects.
Mining Very High Resolution InSAR Data based on Complex-GMRF Cues and Relevance Feedback
With the increase in number of remote sensing satellites, the number of image-data scenes in our ... more With the increase in number of remote sensing satellites, the number of image-data scenes in our repositories is also increasing and a large quantity of these scenes are never received and used. Thus automatic retrieval of desired image-data using query by image content to fully utilize the huge repository volume is becoming of great interest. Generally different users are interested in scenes containing different kind of objects and structures. So its important to analyze all the image information mining (IIM) methods so that its easier for user to select a method depending upon his/her requirement. We concentrate our study only on high-resolution SAR images and we propose to use InSAR observations instead of only one single look complex (SLC) images for mining scenes containing coherent objects such as high-rise buildings. However in case of objects with less coherence like areas with vegetation cover, SLC images exhibits better performance. We demonstrate IIM performance comparis...
HR Sar Iim
High resolution SAR images show specifc properties that can be used for image information mining.
High resolution TerraSAR-X image ming using RELAX
We present a feature extraction procedure for SAR images based on the RELAX algorithm.