
Accurate three-dimensional (3D) surface deformation modeling is essential for understanding the geophysical processes of Earth, such as fault activity, subsidence, and volcanic dynamics. InSAR provides high-resolution deformation data but is limited to line-of-sight (LOS) displacements. To address the challenges of estimating three-dimensional deformation from LOS observations, this study proposes an adaptive Radial Basis Functions (RBFs) approach for RBF-constrained estimation from single- and dual-orbit InSAR observations. The number and spatial distribution of RBF centers are determined using the k-means clustering approach with a post-processing procedure inspired by the ISODATA algorithm. The network design methodology ensures the adaptive placement of RBF centers based on two user-defined thresholds. These thresholds are the minimum data number in each cluster and the minimum distance between centroids. These constraints control the density of RBF centers and thereby balance spatial flexibility against model complexity. The shape parameter for each RBF is determined using the Generalized Cross-Validation (GCV) technique where only observations in the impact area of RBF are utilized. Furthermore, the proposed methodology is integrated with GNSS measurements to assess its performance when GNSS observations are available. In each of the three study areas, six different scenarios are tested to evaluate the effectiveness of the proposed methodology. These scenarios include various combinations of single-orbit, and dual-orbit InSAR measurements, both with and without the incorporation of GNSS observations. The numerical experiments show that the proposed framework can provide spatially coherent RBF-constrained estimates of three-dimensional deformation and can incorporate dual-orbit InSAR and GNSS observations within a common inversion framework.
Acacia longifolia (Andrews) Willd is an invasive alien plant (IAP) that significantly impacts forest ecosystems. Early detection and monitoring are crucial for its control. Traditional field surveys are limited by low accuracy and efficiency, while recent advances in Unmanned Aerial Vehicles (UAV) and computer vision have enabled more effective IAP monitoring. This study presents a workflow that exploits the distinctive phenological characteristics of Acacia longifolia flowering to enable accurate detection using UAV RGB imagery and semantic segmentation. We evaluated eighteen semantic segmentation models to identify the most accurate approach for Acacia longifolia flowering detection and to assess its generalisation across different phenological conditions and an independent study site. Among the models, PSPNet with a ResNet-101 backbone achieved the highest performance on the internal validation dataset, with an F1-score of 91.97
Hyperspectral images, with their hundreds of narrow and highly correlated spectral bands, provide rich spectral information essential for accurate classification of land-cover and terrestrial objects. However, the high dimensionality of hyperspectral data introduces redundancy and incur in the curse of dimensionality, often reducing classification performance when limited training samples are available. To address these challenges, dimensionality reduction, particularly hyperspectral band selection, is an essential preprocessing step, aimed to retain discriminative spectral channels while significantly reducing the computational burden of downstream classification and mitigating the curse of dimensionality problem. Recent advances in deep learning have significantly enhanced hyperspectral band selection, enabling a more precise extraction of informative bands through advanced feature representation and optimization methods. This article presents a comprehensive review of hyperspectral band selection methods, discusses the technical challenges, including issues arising from non-differentiable band selection operations, and outlines potential solutions and future research directions. The survey serves as a guide for researchers and practitioners in selecting appropriate band selection methodologies, facilitating improved classification performance across diverse real-world hyperspectral imaging applications.
Accurate knowledge of vegetation structure is crucial for many aspects of the scientific study of ecosystems. LiDAR is a leading technology for efficiently capturing vegetation structure and plays an important role in forestry and forest science. LiDAR systems can operate from a variety of platforms, including tripods, vehicles, UAVs (uncrewed aerial vehicles), airplanes, and satellites. Depending on the platform, the spatial resolution, data acquisition effort, and coverage area vary significantly. Since 2019, satellite-based LiDAR data from NASA’s Global Ecosystem Dynamics Investigation (GEDI) mission has offered a freely available, global alternative for assessing vegetation structure on regional or global scales. This paper investigates the potential of GEDI data for describing vertical vegetation structure. To assess GEDI’s capabilities, the three key forest metrics canopy height, canopy cover, and plant area volume density (PAVD) obtained from GEDI waveforms were compared to metrics and pseudo-waveforms derived from airborne laser scanning (ALS, both full-waveform (FW) and discrete return (DR)) and terrestrial laser scanning (TLS) data for eight GEDI shots situated in the study area near Dresden, Germany. The comparison of canopy height and canopy cover estimates between GEDI and the reference datasets (ALS and TLS) indicated varying levels of agreement across the shots. RMSE values for canopy height were 5.97 m (GEDI vs. ALS-FW), 6.17 m (GEDI vs. ALS-DR), and 6.21 m (GEDI vs. TLS), while canopy cover RMSE values were 0.34, 0.32, and 0.32, respectively. The comparison of canopy metrics, waveform and PAVD representations at the shot level highlights limitations of GEDI, including geolocation uncertainty and vertical accuracy, for fine scale applications.
Reliable and safe navigation of autonomous systems is an active field of research, particularly in the context of complex urban environments. A key challenge is navigation integrity, defined as the trust in the correctness of the estimated system state; the transfer of this concept from aviation to collaborative multi-agent systems remains an open research challenge. Addressing this gap, this paper presents the methodological developments and key findings of the DFG-funded i.c.sens research project, which focused on navigation integrity in dynamic collaborative sensor networks. To account for key influencing factors and to structure the research activities accordingly, three central research directions are defined: Observations (sensor characteristics and uncertainties), Methodology (state estimation and integrity monitoring), and Map (dynamic, uncertainty-aware environmental representations). To evaluate the developed approaches and assess their performance under realistic conditions, extensive experiments were conducted. Based on experimental results obtained from publicly available datasets as well as datasets collected by the i.c.sens group, the paper summarizes key achievements of i.c.sens in integrity-aware state estimation under realistic urban conditions. These results show improved navigation performance through multi-agent information exchange, through rigorous uncertainty modeling and through the conception and usage of sophisticated map concepts. Based on these findings, we highlight remaining integrity challenges in dynamic sensor networks and point out promising directions for future research.