FLAME 3 Dataset: Unleashing the Power of Radiometric Thermal UAV Imagery for Wildfire Management

TL;DR

FLAME 3 addresses the scarcity of radiometric UAV wildfire data by delivering a synchronized visual-RGB and radiometric thermal dataset collected across six prescribed burns, including radiometric TIFFs and nadir thermal plots. The data processing pipeline and alignment workflow enable per-pixel temperature ground truth and automated labeling, supporting robust multimodal modeling. Empirical evaluations demonstrate that radiometric TIFF data yield superior Fire/No-Fire classification performance, surpassing prior FLAME iterations, especially in TIFF-only and RGB-TIFF configurations. The work aims to democratize access to radiometric wildfire data and empower AI-driven detection, segmentation, and burn assessment with practical implications for wildfire management.

Abstract

The increasing accessibility of radiometric thermal imaging sensors for unmanned aerial vehicles (UAVs) offers significant potential for advancing AI-driven aerial wildfire management. Radiometric imaging provides per-pixel temperature estimates, a valuable improvement over non-radiometric data that requires irradiance measurements to be converted into visible images using RGB color palettes. Despite its benefits, this technology has been underutilized largely due to a lack of available data for researchers. This study addresses this gap by introducing methods for collecting and processing synchronized visual spectrum and radiometric thermal imagery using UAVs at prescribed fires. The included imagery processing pipeline drastically simplifies and partially automates each step from data collection to neural network input. Further, we present the FLAME 3 dataset, the first comprehensive collection of side-by-side visual spectrum and radiometric thermal imagery of wildland fires. Building on our previous FLAME 1 and FLAME 2 datasets, FLAME 3 includes radiometric thermal Tag Image File Format (TIFFs) and nadir thermal plots, providing a new data type and collection method. This dataset aims to spur a new generation of machine learning models utilizing radiometric thermal imagery, potentially trivializing tasks such as aerial wildfire detection, segmentation, and assessment. A single-burn subset of FLAME 3 for computer vision applications is available on Kaggle with the full 6 burn set available to readers upon request.

Figures (8)

• Figure 1: Three Images from Willamette prescribed burn on September 23rd, 2022. From left to right, images taken at 2:38 PM, 2:39 PM, and 2:40 PM respectively.
• Figure 2: RGB-IR Image Pairs from Hanna Hammock prescribed burn. From top to bottom, time progresses between images with images taken at 5:48:30 PM, 5:48:42 PM, and 5:48:50 PM.
• Figure 3: Raw RGB-IR image Pair collected from each prescribed burn site found in Table II.
• Figure 4: TIFF Creation Process: Raw Thermal RJPEG Image (Input), Thermal TIFF File and Thermal JPEG with copied exif metadata (Output); Images show visualization of data at each step. Extracted radiometric data shown in greyscale. Red = Input Data, Yellow = Data Processing, Green = Output Data.
• Figure 5: FOV Corrections Process: Raw RGB Image (Input), FOV Corrected (thermal-resolution aligned) RGB image with original exif metadata (Output); Images show visualization of data at each step. Red = Input Data, Yellow = Data Processing, Blue = Decision Process, Green = Output Data.