计算机工程 ›› 2018, Vol. 44 ›› Issue (8): 257-262,267.doi: 10.19678/j.issn.1000-3428.0047659

• 图形图像处理 • 上一篇    下一篇

室内场景的布局估计与目标区域提取算法

吴晓秋 a,霍智勇 a,b   

  1. 南京邮电大学 a.通信与信息工程学院; b.江苏省图像处理与图像通信重点实验室,南京 210003
  • 收稿日期:2017-06-21 出版日期:2018-08-15 发布日期:2018-08-15
  • 作者简介:吴晓秋(1993—),女,硕士研究生,主研方向为图像处理、多媒体通信;霍智勇,教授。
  • 基金项目:

    国家自然科学基金(61471201,61501260);江苏省高校自然科学研究重点项目(13KJA510004);江苏省自然科学基金青年基金(BK20130867);江苏省“六大人才高峰”项目(2014-DZXX-008)。

Layout Estimation and Object Region Extraction Algorithm for Indoor Scene

WU Xiaoqiu a,HUO Zhiyong a,b   

  1. a.College of Telecommunications and Information Engineering; b.Jiangsu Provincial Key Lab of Image Processing and Image Communication,Nanjing University of Posts and Telecommunications,Nanjing 210003,China
  • Received:2017-06-21 Online:2018-08-15 Published:2018-08-15

摘要:

现有的目标提取方法在应用于复杂的室内场景图像时,容易出现小尺寸物体与平面区域中物体被忽视,以及因遮挡造成大物体提取错误等问题。为此,提出一种针对室内RGB-D场景的无监督布局估计与目标区域提取算法。利用3D点云进行平面分割与分类以完成布局估计,采用 2种图像分割方法对RGB-D图像做过分割处理,并利用4种相似度衡量方式进行层次分组。在此基础上,根据布局估计的结果,对不同类别的区域采取不同的边界框匹配策略。实验结果表明,该方法无需预训练即可改善目标区域提取效果,在产生较少目标候选区的情况下提高边 界框召回率,加快计算速度。

关键词: 深度信息, 特征融合, 室内场景, 布局估计, 图像分割, 目标提取

Abstract:

When applying most existing object proposal methods on complex indoor scenes,the results show that there are some problems such as ignoring the small size object and objects in planar regions and detection inaccuracies of big objects caused by occlusion.Aiming at above these problems,this paper proposes a layout estimation and object region extraction algorithm for indoor RGB-D scenes.Firstly,it uses the 3D point cloud for plane segmentation and classification.Secondly,it adopts two segmentation methods using RGB-D data for obtaining crude object segments and then utilizes four similarity measures for hierarchical grouping.Finally,based on the results of layout estimation,it takes diversification strategies to fit bounding boxes for different regions.Experimental result shows that the proposed algorithm can improve extraction efficiency obviously and improve bounding box proposal recall score with fewer object candidates.In addition,it does not need pre-training and has fast calculation speed.

Key words: depth information, feature fusing, indoor scene, layout estimation, image segmentation, object extraction

中图分类号: