Searcharxiv⌕ Search

arXiv subjects

Irena Hajnsek

Publications and source records attributed to Irena Hajnsek.

6 recordsLinked to original sources

Spatial phase coherence and its applications to masking and adaptive neighborhood sizing in multitemporal InSAR

We examine spatial phase coherence (SPC), a measure of the correlation between the wrapped phase time series of two pixels in a stack of single-look synthetic aperture radar acquisitions. A pixel and its adjacent neighbors share speckle through the point spread function, so that decorrelation alone does not lower their SPC below a floor. We find that masking pixels whose mean SPC to their eight neighbors is well below that floor, as in parts of layover and shadow areas, is an inexpensive complement to geometry-derived layover and shadow masks. For distributed scatterer pixels, averaging SPC over a fixed set of neighbors that share no speckle yields a proxy for the local coherence, the phase coherence indicator (PCI), that is available before the selection of statistically homogeneous pixels (SHPs). In phase estimation, averaging more pixels reduces noise but may also increase the bias in heterogeneous areas. To relax this trade-off, we propose to enhance SHP selection algorithms by scaling the search window size according to the PCI and truncating the set of selected pixels where a smaller nested set agrees clearly better with the phases at the center. In simulations of a TerraSAR-X and a Sentinel-1 stack with 44 and 79 acquisitions, we apply this adaptive neighborhood sizing (ANS) to established amplitude- and covariance-based SHP selection algorithms with thresholds that adapt to the PCI. Without ANS, the largest tested search window size has 13 to 29 % more phase linking error than the best of the tested sizes. With ANS, the largest window has the lowest error of the tested sizes, 11 % lower than that of the same selection at its best tested size without ANS. On a real TerraSAR-X stack, the mask based on SPC flags pixels in building shadows that a terrain model misses. On real Sentinel-1 stacks, ANS preserves dense fringes on a slowly deforming slope and in part on a fast landslide.

eess.SP↗

Hybrid AI-Physical Modeling for Penetration Bias Correction in X-band InSAR DEMs: A Greenland Case Study

Digital elevation models derived from Interferometric Synthetic Aperture Radar (InSAR) data over glacial and snow-covered regions often exhibit systematic elevation errors, commonly termed "penetration bias." We leverage existing physics-based models and propose an integrated correction framework that combines parametric physical modeling with machine learning. We evaluate the approach across three distinct training scenarios - each defined by a different set of acquisition parameters - to assess overall performance and the model's ability to generalize. Our experiments on Greenland's ice sheet using TanDEM-X data show that the proposed hybrid model corrections significantly reduce the mean and standard deviation of DEM errors compared to a purely physical modeling baseline. The hybrid framework also achieves significantly improved generalization than a pure ML approach when trained on data with limited diversity in acquisition parameters.

cs.AI↗

Hybrid Machine Learning Model for Forest Height Estimation from TanDEM-X and Landsat Data

Integrating machine learning (ML) with physical models (PM) has emerged as a promising way of retrieving geophysical parameters from remote sensing data. In this context, a ML model for estimating forest height from TanDEM-X interferometric coherence measurements has recently been proposed, that constrains the learning process through a PM. While the features used for training and inversion where selected to ensure the physical consistency of the solutions, they could not resolve all height / structure and baseline / terrain slope ambiguities in the data. To improve this, the extension of the feature space with optical Landsat data is proposed able to provide complementary information on forest type or structure. The extended model is applied and validated on several TanDEM-X acquisitions over the Gabonese Lopé national park site and assessed against airborne LiDAR measurements. Results show a 13.5% reduction in RMSE and a 16.6% reduction in MAE compared to the original hybrid model, confirming the added value of multispectral inputs.

cs.CV↗

Heuristic Quality Coefficients for Interferometric Phase Linking

In multitemporal InSAR, phase linking (PL) refers to the estimation of a single-reference interferometric phase history for distributed scatterers (DS) from the information contained in the sample coherence matrix. Because the phase information in this matrix is typically inconsistent, DS processing needs practical reliability indicators to decide whether a pixel's PL estimate is sufficiently supported by the data for subsequent deformation analysis. For maximum-likelihood estimation, uncertainty can be quantified via Fisher-information-based covariance estimates, but no analogous, generally applicable uncertainty quantification is available for the broad range of non-ML methods. We propose three heuristic quality coefficients within a unified mathematical framework that covers common PL methods: (1) a method-specific goodness-of-fit coefficient that normalizes the achieved PL objective between a method-consistent upper bound and an empirically modeled noise floor level; (2) a closure phase coefficient computed from the sample coherence matrix in advance; and (3) an ambiguity coefficient that compares the obtained PL estimate with the best alternative in its orthogonal complement in the solution space. All coefficients are normalized to the interval $[0,1]$, where 1 indicates maximum reliability and 0 matches the behavior expected under pure noise. Simulations under exponential and seasonal decorrelation models show that the goodness-of-fit coefficient tracks the normalized absolute phase error most consistently, whereas the closure phase coefficient provides an a priori indicator for pre-screening. Experiments on a TerraSAR-X stack over Visp, Switzerland, reveal plausible spatial patterns across urban and vegetated areas and show that the ambiguity coefficient provides complementary information, especially in regions with temporally varying scattering mechanisms.

eess.SP↗

Target Characteristics on Rotation in Euclidean Space Using Full Polarimetric SAR Data

In radar polarimetry, the target characteristics significantly depend on the rotation of the target concerning the radar line of sight. Hence, several attempts have been made in the literature to compensate for the rotational dependency or to derive a complete roll-invariant parameter using full polarimetric SAR data. However, the degree of dependency of the targets on the rotation domain has yet to be well explored. Hence, in this study, we have proposed a new parameter that characterizes the targets concerning rotation in Euclidean space. The parameter shows similar values for the targets, which can be related to each other through a unitary transformation. The advantage of this parameter has been demonstrated over different canonical targets. Further, the characteristics of the natural targets have been shown using the Single Look Complex data of C band Radarsat-2 and ALOS PALSAR over the San Francisco Bay area and Cuba, respectively.

physics.app-ph↗

Self-Supervised Pretraining and Controlled Augmentation Improve Rare Wildlife Recognition in UAV Images

Automated animal censuses with aerial imagery are a vital ingredient towards wildlife conservation. Recent models are generally based on deep learning and thus require vast amounts of training data. Due to their scarcity and minuscule size, annotating animals in aerial imagery is a highly tedious process. In this project, we present a methodology to reduce the amount of required training data by resorting to self-supervised pretraining. In detail, we examine a combination of recent contrastive learning methodologies like Momentum Contrast (MoCo) and Cross-Level Instance-Group Discrimination (CLD) to condition our model on the aerial images without the requirement for labels. We show that a combination of MoCo, CLD, and geometric augmentations outperforms conventional models pre-trained on ImageNet by a large margin. Crucially, our method still yields favorable results even if we reduce the number of training animals to just 10%, at which point our best model scores double the recall of the baseline at similar precision. This effectively allows reducing the number of required annotations to a fraction while still being able to train high-accuracy models in such highly challenging settings.

cs.CV↗