基于YOLO与图像修复的仿真场景等效构设研究
作者:
作者单位:

作者简介:

通讯作者:

中图分类号:

基金项目:

江苏省自然科学基金青年项目(No.BK20210290)资助。


Research on construction method for scenes based onYOLO and image repair algorithms
Author:
Affiliation:

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    基于典型场景环境实测数据的仿真场景的等效构设,其相较于传统辐射渲染下的仿真模型搭建,展现了更高的准确度和逼真度,同时等效构设过程中形成的特征库也能支撑仿真建模校验。然而,场景等效构设过程中目标与背景的解耦问题一直是制约其逼真度提升的关键难题。为此,本文提出了一种结合改进YOLOv8和Criminisi算法的背景解耦技术。首先,利用YOLOv8精确提取目标掩膜;接着,用改进Criminisi算法补全背景,保持结构连贯性。最后,通过泊松融合算法将目标与背景融合,提升仿真场景的逼真度。实验结果表明,用该方法等效构设的仿真环境与真实采集到的图像相似,解决了在仿真场景实时渲染过程中场景目标单一、构设逼真度不高等问题。

    Abstract:

    The equivalently construction of the simulation scenes based on the measured data from typical environmental scenarios yields greater accuracy and realism than traditional radiosity based simulation model construction.Concurrently,the feature library generated in the process of equivalent construction process can further support the verification of simulation models.Nonetheless,the persistent challenge of decoupling targets from backgrounds during scene construction has been a pivotal obstacle impeding fidelity enhancement.To this end,a novel background decoupling technique that synergizes the enhanced YOLOv8 with the Criminisi algorithm is introduced in this paper.The process begins with the precise extraction of the target mask utilizing YOLOv8.Subsequently,the refined Criminisi algorithm is employed to reconstruct the background while preserving structural integrity.Ultimately,the Poisson blending algorithm is applied to amalgamate the target and background,thereby augmenting the simulation scene′s realism.Experimental results demonstrate that the simulation environment crafted via this methodology exhibits a higher resemblance to actual captured images,effectively resolving issues related to the singularity of scene targets and the diminished realism in the real time rendering of simulation scenes.

    参考文献
    相似文献
    引证文献
引用本文

周佳乐,宋敏敏,雷昊,刘建旭,曹卫卫,施瑶瑶,董大兴,刘友文.基于YOLO与图像修复的仿真场景等效构设研究[J].激光与红外,2025,55(1):145~154
ZHOU Jia-le, SONG Min-min, LEI Hao, LIU Jian-xu, CAO Wei-wei, SHI Yao-yao, DONG Da-xing, LIU You-wen. Research on construction method for scenes based onYOLO and image repair algorithms[J]. LASER & INFRARED,2025,55(1):145~154

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:
  • 最后修改日期:
  • 录用日期:
  • 在线发布日期: 2025-01-17
  • 出版日期:
×
最新公告
根据主办单位安排,编辑部2025年春节放假时间为1月26日~2月9日,2月10日起正常上班。放假期间投稿系统正常运行,其他业务暂缓办理。
    考虑寒假和快递物流等影响,为避免信件丢失,2025年第一、二期的杂志样刊等相关信件拟在3月份前后通过邮政快递发出,可通过本刊平台浏览、下载当期封面、目录、文章电子版。
    另外,本刊电子书架现已上线,可点击平台首页“电子书架”或扫码在线阅读。