计算机工程与应用 ›› 2022, Vol. 58 ›› Issue (20): 212-219.DOI: 10.3778/j.issn.1002-8331.2103-0264

• 图形图像处理 • 上一篇    下一篇

机载视频空中目标实时优化稳像算法

刘玉清,刘成,敖雪聪   

  1. 1.中国科学院 国家空间科学中心 复杂航天系统综合电子与信息技术重点实验室,北京 100190
    2.中国科学院大学 计算机科学与技术学院,北京 100049
  • 出版日期:2022-10-15 发布日期:2022-10-15

Real-Time Image Stabilization Algorithm for Airborne Video

LIU Yuqing, LIU Cheng, AO Xuecong   

  1. 1.Key Laboratory of Electronics and Information Technology for Space Systems, National Space Science Center, Chinese Academy of Sciences, Beijing 100190, China
    2.School of Computer Science and Technology, University of Chinese Academy of Sciences, Beijing 100049, China
  • Online:2022-10-15 Published:2022-10-15

摘要: 针对机载飞机视频摄取与监视中,由于背景稀疏和前景的大幅度快速运动,造成实时稳像算法存在的画面不稳定的问题,提出了自适应Shi-Tomasi机载视频空中目标实时优化稳像算法。根据提取特征点分布自适应地改变Shi-Tomasi角点检测阈值,解决单一阈值不能适应空中复杂稀疏背景特征点提取的问题。构建带约束的实时优化算法,计算平滑的视频路径,解决基于滤波的算法缺少约束导致画面偏移过大的问题。比对实验结果表明,自适应Shi-Tomasi优化稳像算法能够应对各类机载稀疏背景视频的稳像,解决了快速运动目标稳像后画面大幅偏移的问题,稳定性提高,速度达到20?frame/s以上,满足实时处理需求。

关键词: 机载视频, 复杂稀疏背景, 运动目标, 自适应角点检测, 实时优化

Abstract: To solve the problem of image instability in real-time image stabilization algorithm due to the sparse background and rapid movement of foreground in airborne video intake and monitoring, an adaptive Shi-Tomasi real-time optimization image stabilization algorithm for airborne target in airborne video is proposed. Firstly, the Shi-Tomasi corner detection threshold is adaptively changed according to the distribution of extracted feature points to solve the problem that a single threshold cannot adapt to the extraction of feature points from complex and sparse background in the air. Then a real-time optimization algorithm with constraints is constructed to calculate a smooth video path and deal with excessive image offset caused by the lack of constraints in the filtering algorithm. The comparative experimental results show that the adaptive Shi-Tomasi optimized image stabilization algorithm can cope with the image stabilization of various airborne sparse background videos, and solve the problem of large image offset after image stabilization of fast-moving targets. With improved stability, the speed can reach more than 20 frame/s, which meets the real-time processing requirements.

Key words: airborne video, complex sparse background, moving object, adaptive corner detection, real time optimization