基于LIG的红外视频小目标检测方法

LIG-based Infrared Video Small Target Detection Method

准确检测远程红外视频中的小目标具有一定难度,提出一种面向远距离红外视频小目标的高性能检测方法,设计一个灵活的模块化处理流程,包括视频超分辨率模块、无监督目标检测器、连通区域分析模块和轨迹关联模块。该方法由视频分辨率增强、局部梯度小目标检测、多帧检测轨迹关联构成,并采用远距离红外视频样本进行了实验验证。对3种视频超分辨率算法进行了研究,结果表明:目标关联算法具有较好的提升效果,证明了所提方法的有效性。

To overcome the difficulty in determining small targets with accuracy on a remote ingrained video,a small-target high-resolution detection method of remote intra-video is proposed.And a flexible modularized processing flow,which includes video super-resolution module,unsupervised target detector,connected area analysis module and track correlation module,is established.The proposed method,composed of video resoluton increase,local gradient small together detection and multi frame detection track correlation,is experimentally verified by adapting remote infrared video samples.Reserch comparison is made among the three super-resolution algorithms with the results indicating that the target correlation algorithm has favourable improvement,which verifies the effectiveness of the proposed method.