基于递归标记的红外图像弱小点目标自动识别方法  

Automatic Recognition of Weak Point Targets in Infrared Images Based on Recursive Tagging

作　　者：潘文 周波[3] 曹志浩 PAN Wen;ZHOU Bo;CAO Zhi-hao(Information and Finance Department,Xuancheng Vocational&Technical College,Xuancheng 242000,China;Education and Management Department,Xuancheng Vocational&Technical College,Xuancheng 242000,China;Computer Information Department,Hefei University of Technology,Hefei 230000,China)

机构地区：[1]宣城职业技术学院信息与财经学院,安徽宣城242000 [2]宣城职业技术学院教学与管理学院,安徽宣城242000 [3]合肥工业大学计算机信息系,安徽合肥230000

出　　处：《遵义师范学院学报》2024年第4期85-89,共5页Journal of Zunyi Normal University

基　　金：2020年度安徽省质量工程项目教学研究重点项目(2020jyxm2235)。

摘　　要：为准确识别红外图像中的弱小点目标,以递归标记算法为核心提出了红外图像弱小点目标自动识别方法。采用平滑滤波器和区域生长分割法,完成待识别红外图像的预处理;应用基于递归的二值图像标记模式,扫描所有的红外图像像素点并定义相应的标记值,形成多个图像连通域;针对每个连通域分别提取几何特征和形心特征,将其应用到稀疏表示分类器和卷积神经网络分类器中,自动生成弱小点目标识别结果。实验结果显示:应用所提方法识别海面红外图像、云层红外图像和纯净天空红外图像的mAP值分别为0.95、0.96与0.91,表明其具有较好的红外图像弱小点目标自动识别效果。To accurately identify weak point targets in infrared images,an automatic recognition method of weak point targets in infrared images is proposed based on recursive marking algorithm.Smoothing filter and region growing segmentation method are used to complete the preprocessing of the infrared image to be recognized.The binary image marking mode based on recursion is applied to scan all infrared image pixels and define corresponding marking values to formmultiple image connected regions.Geometric features and centroid features are extracted for each connected domain,and applied to sparse representation classifier and convolutional neural network classifier to automatically generate weak point target recognition results.The experimental results show that the mAP values of sea surface infrared image,cloud layer infrared image and pure sky infrared image are 0.95,0.96 and 0.91 respectively,which shows that the proposed method has a good automatic recognition effect for weak point targets in infrared image.

关 键 词：递归方法 像素标记 红外图像 弱小点 目标识别 形心特征 

分 类 号：TN911.73[电子电信—通信与信息系统]