基于运动显著图和光流矢量分析的目标分割算法
DOI:
作者:
作者单位:

火箭军工程大学502教研室西安710025

作者简介:

通讯作者:

中图分类号:

TP391TH741

基金项目:


Video object segmentation algorithm based on motion saliency map and optical flow vector analysis
Author:
Affiliation:

Staff Room 502, Rocket Force University of Engineering, Xi′an 710025, China

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:

    为提高运动目标分割算法对多种复杂场景的自适应能力和分割精度,提出一种基于运动显著图和光流矢量分析的目标分割算法。该算法首先基于运动显著图提取运动目标的大致区域,然后利用光流矢量获得运动目标和背景区域的运动边界,并结合点在多边形内部原理得到运动目标内部精确的像素点,最后以超像素为基本分割单元,通过引入置信度的概念实现最终像素一级的目标分割。通过与典型算法进行多场景实验对比,表明该算法能够有效实现多种复杂场景下的运动目标分割,并且较现有算法具有更高的分割精度。

    Abstract:

    In order to improve the adaptive ability and segmentation accuracy of video object segmentation algorithm in various complex scenes, an object segmentation algorithm based on motion saliency map and optical flow vector analysis is proposed in this paper. Firstly, the proposed algorithm extracts the rough target region based on motion saliency map. Then, the motion boundaries of the motion object and background region are determined based on the optical flow vector between pairs of subsequent frames. The above information is combined to acquire the accurate pixels inside the moving objects with the pointinpolygon principle from the computational geometry. Finally, to refine the spatial accuracy of object segmentation in the previous step, per frame superpixels are acquired with oversegmenting method. And these superpixels are labeled as foreground or background based on confidence level and statistical model. The proposed algorithm was compared with typical algorithms in different scenes, and the results indicate that the proposed algorithm can effectively deal with the moving object segmentation on a variety of challenging scenes, and has higher segmentation accuracy than other existing algorithms.

    参考文献
    相似文献
    引证文献
引用本文

崔智高,李艾华,王涛,李辉.基于运动显著图和光流矢量分析的目标分割算法[J].仪器仪表学报,2017,38(7):1792-1798

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: 2017-08-07
  • 出版日期: