复杂背景下红外弱小运动目标检测半仿真数据集

A dataset of semi-synthetic detection for small infrared moving targets under complex backgrounds

红外弱小运动目标检测是空中红外探测的核心技术之一,且一直是光电探测领域研究的热点问题,尤其是复杂背景下弱小运动目标检测,尚缺乏有效的技术解决途径。针对当前红外弱小运动目标检测领域数据规模小、场景类型丰富程度不足的问题,本数据集面向空中红外探测设备远距离探测空中飞行目标应用,借助小型无人机平台搭载红外探测设备,通过实地拍摄、数据加工及仿真目标嵌入处理等,提供了一套数据规模大、场景类型丰富的复杂背景下红外弱小运动目标检测半仿真数据集。图像数据采集中设置多种实验条件,以求尽可能地全方位模拟典型应用场景:为体现探测平台与目标的相对高度,设置了仰视、平视和俯视三种拍摄视角;为模拟不同场景类型,数据集获取场景涵盖了天空、水面、植被、建筑物等多种场景;为模拟被探测目标不同运动情况,数据集中采用分段拟合的方式,设置了不同复杂程度的目标运动轨迹;本数据集仿真了不同灰度分布、不同信号强度的弱小运动目标,并高质量地嵌入图像背景中。本数据集共计350段数据、150185帧图像,标注文件中给出目标坐标位置的详细标注信息。本数据集可为复杂背景下红外弱小运动目标检测、跟踪等研究提供基础数据。

Small infrared moving target detection is a key technology in infrared detection and remains a hot issue in photoelectric detection.Especially when it comes to identifying small infrared moving target against complex backgrounds,there is still a lack of effective technological solutions.To address the shortage of high-quality data in current research,we adopted infrared imaging equipment mounted on a small UAV platform to capture image sequences as backgrounds.Embedding synthetic targets in backgrounds properly,we produced a dataset of semi-synthetic detection for small infrared moving targets under complex backgrounds.Various conditions have been configured during image capture,including relative height(top view,horizontal view and bottom view),scenes(vegetation,water and building),platform motion,weather,time,etc.The imaging platform motion and the synthetic target motion have been taken into consideration simultaneously to ensure that the semi-synthetic image sequences resemble the real application scenarios as closely as possible.In addition,we varied the target characteristics in synthetizing,including shape,intensity,and motion.This dataset includes 350 image sequences,150,185 images,with annotation files detailing the targets positions in images.The dataset can be used for research in small infrared moving target detection and tracking.