TNO多波段图像数据集

该数据集由荷兰TNO研究所公开发布，主要面向红外与可见光融合、检测与跟踪等多模态视觉研究。原始数据涵盖多种典型场景与环境条件，每组数据由一张红外图像与一张可见光图像配对，用于评估算法在不同光照、天气以及目标分布下的性能表现。在本研究中，为了支持复杂动态系统中的多智能体超图建模感知，我们基于Matlab脚本对TNO数据集进行筛选与预处理，包括对原始图像进行裁剪、配准、灰度归一化等操作，以保证不同模态间的对齐精度。最终从中选取了21对红外与可见光图像（共42张），全部保存为PNG格式，数据容量约为5.3 MB。基于该数据集，可在多模态数据融合与分析方面开展进一步实验，包括多智能体协同感知、目标检测识别与跟踪等下游任务研究。

This dataset was publicly released by TNO, the Netherlands Organization for Applied Scientific Research, and is primarily intended for multimodal vision research such as infrared and visible light fusion, object detection and tracking. The raw data covers various typical scenarios and environmental conditions, with each data sample consisting of a paired infrared image and visible light image, which is used to evaluate algorithm performance under different lighting, weather conditions and target distributions. In this study, to support multi-agent hypergraph modeling and perception in complex dynamic systems, we filtered and preprocessed the TNO dataset using Matlab scripts, including operations such as cropping, registration, and gray-scale normalization of the original images to ensure alignment accuracy between different modalities. Finally, 21 pairs of infrared and visible light images (42 images in total) were selected, all saved in PNG format, with a total data size of approximately 5.3 MB. Based on this dataset, further experiments can be carried out in multimodal data fusion and analysis, including downstream task research such as multi-agent collaborative perception, object detection, recognition and tracking.

搜集汇总

背景与挑战

背景概述

TNO多波段图像数据集是一个面向多模态视觉研究的数据集，包含21对红外与可见光图像，适用于红外与可见光融合、检测与跟踪等任务。数据集经过预处理，确保不同模态间的对齐精度，可用于多智能体协同感知和目标检测识别等研究。