火力与指挥控制2023，Vol.48Issue(11)：72-80,9.DOI:10.3969/j.issn.1002-0640.2023.11.011
基于改进MAE的装甲车辆目标前景遮挡部分补全方法
Foreground Occlusion Complementation Method for Armored Vehicle Targets Based on Improved Masked Auto Encoder(MAE)
摘要
Abstract
Target images taken by unmanned equipment during reconnaissance operations are often obscured by foreground objects in large areas,which seriously affects the accuracy and reliability of intelligent target recognition.A method based on improved Masked Autoencoders(MAE)to complement the foreground obscured part of suspected targets is proposed,which effectively solves the practical problems such as high integration of military targets into the environment,high resolution of reconnaissance images and insufficient volume of military training samples,and the target complementation results are more accurate and more consistent with human recognition in terms of image semantics.cognition.Through quantitative and qualitative analysis and comparison experiments,the complementation effect of this method is verified and its practical value is demonstrated.关键词
装甲车辆目标/图像补全/掩码自编码器/前景遮挡/目标识别Key words
armored vehicle targets/image complementation/masked auto encoder/foreground mask-ing/target recognition分类
军事科技引用本文复制引用
余晓晗,毛绍臣,綦秀利,王家宝,张所娟..基于改进MAE的装甲车辆目标前景遮挡部分补全方法[J].火力与指挥控制,2023,48(11):72-80,9.基金项目
江苏省自然科学基金资助项目(BK20200581) （BK20200581）
