Intelligent Autonomous Systems (INTAS)

Laura joins our Space-Robotics team in the ViPRICE project. She is going to work on visual perception in very challenging environments.

Our work on "Challenging visual place recognition with planetary analogues of moon and mars" by Richard Fechner, Ida Germann, and Peer Neubert has been accepted for oral presentation at the Joint i-SAIRAS & iSpaRo Conference 2026.

Congratutalations to all contributers!

"Shift-Invariant Local Feature Aggregation for Visual Place Recognition“ by Ida Germann and Peer Neubert

Abstract: Visual Place Recognition (VPR) describes the task of deciding whether two camera images show the same place in the world. Existing approaches rely either on local landmarks or on holistic descriptors. The latter are usually much faster to compare, while local landmarks can be combined with epipolar geometry estimation techniques, addressing the viewpoint-dependent arrangements of the local landmarks and the illposed nature of the VPR problem. This is very challenging to address in holistic descriptors that are compared based on a single vector distance. In this work, we propose that considering the relative arrangement of landmarks not only introduces theoretical invariance to viewpoint shifts, but also offers a robust way to decouple camera motion from place identity, ensuring a location remains recognizable as long as the relative configuration of observed landmarks is preserved. Therefore, we introduce aholistic VPR approach that uses Hyperdimensional Computing (HDC) to encode local features based on their relative 3D spatial relationships rather than absolute positions. To mitigate the computational cost and rotational sensitivity of global encoding, we introduce a windowed adaptation that restricts geometric connections to local neighborhoods. This also allows for tuning the descriptor’s sensitivity to changes in the observed set of landmarks. We evaluate our method in a controlled hybrid environment, combining real visual features with simulated geometry. Our results demonstrate that relative encoding provides an isolated and effective way to analyze geometric robustness under realistic appearance noise.

"Towards Novel Keyframe Synthesis for Improved Place Recognition in Visual SLAM" by Christian Braune, Markus Weißflog, Peter Protzel, and Peer Neubert.

Abstract: SLAM systems require a loop closure mechanism to correct drift. We investigate coupling place recognition with the photorealistic map of a 3D Gaussian Splatting SLAM system to find loop closure candidates under severe camera rotations. On ReplicaMultiagent and AriaMultiagent we demonstrate the potential with improved AUPRC and no sequence degrading.

The paper "Cross-domain transfer of hyperspectral foundation models" by Nick Theisen and Peer Neubert has been accepted at the ICPR 2026 conference.

Abstract: Hyperspectral imaging (HSI) semantic segmentation typically relies on in-domain training, but limited data availability often restricts model performance in real-world applications. Current approaches to leverage foundation models in proximal sensing use cross-modality techniques, bridging RGB and HSI to exploit vision foundation models. However, these methods either discard spectral information or introduce architectural complexity. We propose cross-domain transfer as an alternative, reusing HSI foundation models – originally trained in remote sensing – for proximal sensing applications. By eliminating the need to bridge modality gaps, our approach preserves spectral information while maintaining a simple architecture. Using the HS3-Bench benchmark, we systematically evaluate and compare conventional in-domain, in-modality training, cross-modality transfer and cross-domain transfer strategies. Our results demonstrate that cross-domain transfer achieves large performance improvements over in-domain, in-modality training, reduces the performance gap to cross-modality approaches and maintains strong performance in limited data settings. Thus, this work advances more effective HSI semantic segmentation in diverse applications.

Preprint: https://arxiv.org/pdf/2604.26478

We are part of the successfull proposal in the DFG Major Research Instrumentation program by Maik Kschischo for extention of the compute server infrastructure at University of Koblenz.

Richard is going to work in the ViPRICE-Space project.

The research project "ViPRICE-Space: Visual Place recognition in changing environments" on visual localization in challenging space-analog environments achieved funding by the German Federal Ministry of Research, Technology and Space. The goal of this project is to bridge the research gap in Visual Place Recognition by transferring state-of-the-art VPR performance from terrestrial to extraterrestrial environments. These environments present unique challenges, including highly repetitive structures, sparse visual and semantic features, severe perceptual aliasing, and extreme illumination conditions. The project aims to develop robust place recognition methods capable of reliable operation under these demanding conditions while potentially exploiting different visual sensor modalities.

RA-L is one of the leading venues for publishing research in robotics and automation. It is known for its rapid review and publication process, making it popular for researchers who want timely dissemination of new results. Peer Neubert joined the editorial board in the area Visual Perception and Learning.

Congratulation to Dr. Fangming Yuan for successfully finishing his PhD thesis on "Runtime and Data-efficient Large-scale Hierarchical Visual Place Recognition for Changing Environments". He is now working at Neura Robotics.

Peer Neubert is guest editor for a Nature npj unconventional computing collection on "Hyperdimensional Computing and Vector Symbolic Architectures". This collection invites contributions reporting both theoretical and practical advances in HDC/VSA, particularly those aligned with the broad theme of unconventional computing. Submission is open until September 16, 2026 https://www.nature.com/collections/ijiiiahgdi