May 2026· arXiv.org· Vol abs/2605.22556· 0 citations· 65 references
Computer Science
TL;DR
This work presents ImplicitTerrainV2, which advances terrain INRs toward a compact, efficient neural terrain data format by combining a spectral control mechanism with wavelet-guided spatial adaptivity, derivative-aware supervision, and post-training model compression.
Abstract
Digital elevation models (DEMs) underpin terrain analysis in Geographic Information Systems (GIS), but commonly as raster representation, they rely on interpolation for off-grid sampling and finite-difference operators for derivative-based analysis. Implicit neural representations (INRs) offer a continuous alternative, but prior terrain INRs lack explicit frequency control, neglect the gradient structure of terrain, and remain too large and costly to train for practical deployment. We present ImplicitTerrainV2, which advances terrain INRs toward a compact, efficient neural terrain data format by combining a spectral control mechanism with wavelet-guided spatial adaptivity, derivative-aware supervision, and post-training model compression. At its core, a wavelet complexity field (WCF) derives spatially-adaptive frequency masks from analytically computed wavelet coefficients, localizing high-frequency capacity to complex terrain regions. The same field guides complexity-aware adaptive sampling that concentrates training in high-complexity regions, while gradient matching applies extra supervision to enforce the smooth manifold structure of terrain DEMs for improved derivative fidelity. Post-training mixed-precision quantization and entropy coding reduce storage to 1.23 bpp with a 0.28 dB PSNR drop. On 50 morphologically diverse Swiss terrain tiles, ImplicitTerrainV2 reaches 66.25 dB end-to-end PSNR, improving over the prior work by 5.70 dB while using 3.2x fewer parameters and training in 55 s per tile on a single GPU. Our compressed neural format is competitive with several established DEM codecs in rate-distortion performance, while additionally supporting off-grid point queries, closed-form derivative evaluation, and resolution-independent reconstruction, which may benefit many downstream GIS applications.
Accurate 3D terrain maps are essential for emergency response when assessing wildfire hazards. However, wildfire-prone regions often span vast areas where conventional reconstruction methods underperform. Airborne LiDAR systems provide high-resolution terrain data, but they are expensive and infrequently updated. Image-based methods offer a lower-cost alternative, but struggle due to sparse visual features and limited image overlap. We propose a multi-modal reconstruction framework leveraging outdated Digital Elevation Models (DEMs) as geometric priors for image-based 3D reconstruction. Our key innovation is physics-based pixel-pixel alignment between images and DEM data, dramatically reducing computational complexity by eliminating expensive feature matching procedures. To validate our approach, we developed a large-terrain simulator based on a real wildfire-prone area, generating realistic images enabling a comprehensive evaluation. Given posed images and legacy DEMs, our method produces high-fidelity depth maps while maintaining real-time performance. We find significant improvements in reconstruction accuracy and computational efficiency over existing techniques, offering a scalable solution for wildfire response.
Abstract. High-precision digital terrain models (DTMs) are important for Mars explorations and research, providing indispensable spatial information for landing site assessment, rover path planning, and surface environment analysis. However, challenges such as high-resolution stereo data scarcity and complex atmospheric conditions on the Martian surface result in traditional terrain reconstruction methods suffer from limitations in accuracy, coverage and resolution. To enhance the model’s ability to recover fine-grained topography, we present a diffusion-based monocular terrain reconstruction method, which progressively recovers Martian terrains from single-view high-resolution optical images. We employed a multi-scale U-Net denoising network with attention mechanisms and introduced an additional end-to-end depth constraint. To improve terrain reconstruction efficiency, we implemented a diffusion model in the latent space and adopted a skipping sampling mechanism. We employed the proposed method to reconstruct terrain in different regions. Experimental results demonstrate that the reconstructed terrain achieves an accuracy of 2 m. Furthermore, compared to photogrammetric terrain, the shaded relief generated by our method exhibits greater similarity to the input imagery.
Jiarui Cao, Rong Huang, Yusheng Xu et al.· The International Archives o...· 0 citations
Newly triggered landslides rarely carry immediate annotations, so cross-domain transferability determines the value of landslide mapping for emergency response and regional risk assessment. Vision foundation models have strengthened representational transfer, yet on unseen regions, events, and data sources they still generate high-confidence false alarms. Terrain, material, and rainfall triggering can constrain such errors, but their supports are local, regional, and event-scale, so that resampling onto a 10~m grid misaligns them with the segmentation decision unit and compounds the uncertain geographic context problem (UGCoP). We propose GeoPhysAdapter, which anchors on a frozen vision foundation model, restricts terrain, material, and triggering to dense spatial guidance, regional modulation, and event-timing forcing, and applies bounded adaptation at two decision units, the pixel and the candidate landslide body, reverting exactly to the visual prediction where support is insufficient. On an event-isolated PILD dataset of four public sources, 55 global landslide events, and 7,890 test samples, 70.3% of cross-domain false-positive mass lies in near-pure spurious bodies of median equivalent diameter 207m, matching coarse-prior support rather than the pixel. Pixel-level adaptation removes a net 507,817 erroneous pixels and reduces error by 7.76%, whereas raising the decision unit to the candidate body, under identical samples, anchor, and baseline, increases error reduction to 23.99%, approximately 3.1 times the pixel-level effect, improves IoU by 0.031 (14.2% relative), and corrects 9.92 pixels per pixel harmed. The data and code are publicly available at: https://github.com/Liu-Zhihang/geophysadapter.
Zhihang Liu, Mei-Po Kwan, Jinlin Wu et al.· 0 citations
High-resolution geospatial grid generation requires interpolation methods that are accurate but still practical for large datasets. This paper presents an interpolation framework that couples a Gaussian-weighted quadratic surface fitting kernel with KD-Tree neighbor search and CPU multithreading. The fitting kernel is used to represent local non-linear variation, whereas the Gaussian weights reduce the influence of distant samples. To reduce the cost of repeated neighborhood queries and independent target-cell calculations, the implementation builds a shared KD-Tree index and assigns non-overlapping target-cell blocks to worker threads. In experiments using EGM2008 gravity anomaly data, the proposed framework achieved lower errors than three conventional methods (Moving Surface Fitting, Natural Neighbor Interpolation, and Spline Interpolation), with reductions in Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) of up to 63.6% and 61.2%, respectively. The fitting-model evaluation produced a coefficient of determination (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$R^2$$\end{document}) of 0.998611. The optimized implementation reduced the runtime from 240 s to 10 s while keeping the reported accuracy metrics unchanged. These results indicate that the proposed combination of local surface fitting, spatial indexing, and parallel execution is effective for the tested high-resolution grid interpolation task.
Autonomous vessel operation relies on X-band radar for navigation because it offers greater robustness than optical sensors under adverse weather and sea conditions. Place recognition and map-based navigation algorithms assume a static world, yet terrain, vessel, and sea-clutter returns share overlapping intensities and morphologies in real radar scans with unreliable separation. Terrain-only radar data are therefore required for training and evaluating maritime autonomy algorithms, but public datasets remain unavailable, and electronic navigational charts (ENCs) cannot reproduce real radar scattering characteristics. We propose m2rDM, a map-conditioned diffusion model with two complementary applications: m2rDM-RS generates synthetic terrain-accurate X-band radar images from ENC map masks without at-sea data collection, and m2rDM-DS suppresses dynamic-object returns in real radar scans to yield terrain-only outputs for downstream perception. The model is conditioned on geolocation-aligned map masks, radar-specific embeddings from a pretrained segmentation encoder, and temporal context from consecutive radar frames. On the MOANA dataset, m2rDM outperforms existing generative baselines in pixel-level accuracy [peak signal-to-noise ratio (PSNR), root-mean-square error (RMSE)] and gray-level co-occurrence matrix (GLCM)-based texture fidelity, with m2rDM-DS achieving the highest reconstruction accuracy and m2rDM-RS producing the closest texture distributions from map masks alone. m2rDM-DS also improves place recognition accuracy on dense-traffic sequences.
Jinbum Park, Dongje Lee, Yejin Kang et al.· IEEE Transactions on Radar S...· 0 citations
This article explores the performance of analytical and neural-based hillshading methods in a dense urban environment using high-resolution digital elevation model (DEM) and digital surface model (DSM) data for downtown Calgary. The study compares single-direction and multi-direction analytical hillshading with relief shading generated in Eduard, a machine-learning system originally developed to emulate Swiss-style shaded relief trained primarily on mountainous landscapes. Because Eduard was not designed for buildings, bridges, streets, trees, and other urban infrastructures, the central question is not whether it perfectly reproduces urban morphology, but whether parameter tuning can nevertheless produce visually strong, cartographically useful, and in some cases superior results when compared with conventional analytical methods. The analysis focuses especially on terrain type, micro and macro generalization, and flat-area detail parameters, while keeping the large-scale shading style constant throughout the neural experiments. The article is structured as an exploratory comparison rather than a benchmark of universal best practice. It aims to identify where analytical hillshading remains more reliable, where Eduard offers unexpected strengths, and where neural shading fails because of its training bias toward alpine terrain. The study contributes to current work on terrain representation by testing whether a neural approach designed for natural landforms can be adapted to a highly built urban setting, and it concludes by arguing for future model training and evaluation specifically targeted at urban relief shading.
A new machine-learning framework aims to improve the success rate of computational protein design while moving away from results that reproduce sequences found in nature.
MIT News · Artificial Intelligence· news.mit.eduAug 24, 2026
