基于SSD框架的红外弱小目标检测技术研究  

Research of infrared dim and small targets detection technology based on SSD framework

作　　者：马玉莹 黄成章 黄静颖 王伟丞 MA Yu-ying;HUANG Cheng-zhang;HUANG Jing-ying;WANG Wei-cheng(North China Research Institute of Electro-Optics,Beijing 100015,China)

机构地区：[1]华北光电技术研究所,北京100015

出　　处：《激光与红外》2021年第10期1342-1347,共6页Laser & Infrared

摘　　要：针对红外序列图像中点状机动目标的检测问题,利用卷积特征对图像具有更细致的描述能力、可以有效地感知弱小目标与干扰的差异的特点,提出了一种基于SSD框架的红外弱小目标检测方法。本文设计了针对弱小目标的SSD检测框架,且克服了传统方法缺少大量实验数据的困难,利用真实红外图像数据库进行了仿真验证,实验结果表明,该方法对多种不同场景中的信噪比较低的目标具有良好的检测效果和虚警抑制能力。Aiming at the detection problem of point-like maneuvering targets in infrared sequential images,an infrared dim and small targets detection method based on SSD framework is proposed,which can effectively perceive the difference between small and small targets and interference by taking advantage of the convolutional feature,which has a more detailed description ability to the images.In this paper,the SSD detection framework for dim small targets is designed,and it overcomes the difficulty that traditional methods lack a large amount of experimental data.Simulation verification is carried out by using real infrared image database.Experimental results show that this method has good detection effect and false alarm suppression ability for targets with low SNR in various scenes.

关 键 词：弱小目标检测 卷积神经网络 杂波抑制 

分 类 号：TN391.41[电子电信—物理电子学]