Through-Foliage Surface Temperature Reconstruction for Early Wildfire Detection

Abstract We present a method to reconstruct surface temperatures through forest vegetation by combining signal processing and machine learning, enabling fully automated aerial wildfire monitoring with drones for early fire detection. Synthetic aperture (SA) sensing reduces canopy occlusion but introduces thermal blur. To overcome this, we train a visual state space model to recover subtle thermal signals of partially occluded soil and fire hotspots from blurred data. To address limited real-world training data, we generate realistic surface temperature simulations using a latent diffusion model, temperature augmentation, and procedural thermal forest modeling. On simulated datasets, our method reduces RMSE by 2-2.5 versus conventional thermal and uncorrected SA imaging; in field experiments on hotspots, RMSE improved by 12.8-fold and 2.6-fold, respectively. Our approach also generalizes to other thermal signals, including human signatures, capturing morphology and extent --critical where simple thresholding fails-- while conventional imaging struggles with partial occlusion. Training Data for 1D Grid The dataset folder contains simulated procedural forest data generated under different ambient temperature conditions. Each subfolder is named according to the ambient temperature it represents. Within each ambient temperature folder, multiple simulated scenes are provided, corresponding to different configurations as described in Table 1 of the main manuscript. Scene-specific parameters are detailed inscene_parameters.txt. --- Ambient Environment Folder ├── Scene Name Folder ├── images/ # Contains single images and their corresponding temperature data │ ├── [image].png # Single image │ ├── [image]_min_max_temp.txt # Minimum and maximum temperature (Kelvin) for each image ├── GT.tiff # Unoccluded surface temperature ground truth (Kelvin, 32-bit image) ├── GT_pose_0_thermal.png # 8-bit conversion of the unoccluded surface temperature ├── GT_pose_0_thermal_min_max_temp.txt # Min/max temperature (Kelvin) corresponding to GT surface temp ├── center.tiff # Central perspective image (occluded by forest, temp in Kelvin) ├── integrall_normalized_0.png # AOS integral image ├── global_min_max_temp.txt # Global min/max temperature (Kelvin) extracted from all single images, used for recovering temperature from the 8-bit AOS └── pose.txt # Camera poses of the single images Training Data for 2D Grid The dataset folder contains simulated procedural forest data generated under different ambient temperature conditions. Each subfolder is named according to the ambient temperature it represents. Within each ambient temperature folder, multiple simulated scenes are provided, corresponding to different configurations as described in Table 1 of the main manuscript. Scene-specific parameters are detailed inscene_parameters.txt. --- Ambient Environment Folder ├── Scene Name Folder ├── images/ # Contains single images and their corresponding temperature data │ ├── [image].png # Single image │ ├── [image]_min_max_temp.txt # Minimum and maximum temperature (Kelvin) for each image ├── GT.tiff # Unoccluded surface temperature ground truth (Kelvin, 32-bit image) ├── GT_pose_0_thermal.png # 8-bit conversion of the unoccluded surface temperature ├── GT_pose_0_thermal_min_max_temp.txt # Min/max temperature (Kelvin) corresponding to GT surface temp ├── center.tiff # Central perspective image (occluded by forest, temp in Kelvin) ├── integrall_normalized_0.png # AOS integral image ├── global_min_max_temp.txt # Global min/max temperature (Kelvin) extracted from all single images, used for recovering temperature from the 8-bit AOS └── pose.txt # Camera poses of the single images Grafenberg Alm wildfire data A dataset of 543 thermal images was captured by drone during a wildfire at Grafenberg Alm, Austria (1,780 m above sea level, ASL) on 17th of October 2023 at approximately 2pm local time. The ambient temperature was approximately 9◦C. Procedural Forest Simulation Here you can find the code used to generate the procedural forest for both 1D and 2D grid. Generating Surface Temperature The training and inference code used to generate the surface temperature data is provided, together with a README.md file describing the installation steps and usage instructions. Reconstructing Surface Temperature The training and inference code used to recounstructing the surface temperature data is provided, together with a README.md file describing the installation steps and usage instructions.