图像通用目标的无监督检测

宋修锐; 吴志勇

doi:10.3788/OPE.20142201.0160

您当前的位置：

首页 >

文章列表页 >

图像通用目标的无监督检测

信息科学 | 更新时间：2020-08-13

- 图像通用目标的无监督检测
- Unsupervised detection of image object with any class
- 光学精密工程 2014年22卷第1期页码：160-168
- 作者机构：
  
  1. 中国科学院长春光学精密机械与物理研究所,吉林长春,中国,130033
  2. 中国科学院大学北京,中国,100049
- 作者简介：
- 基金信息：
  
  国家863高技术研究发展计划资助项目(No.2008AA7034320B)()
- DOI：10.3788/OPE.20142201.0160
  中图分类号： TP391.4
- 收稿日期：2013-02-11，
  
  纸质出版日期：2014-01-15
- 稿件说明：
移动端阅览
宋修锐, 吴志勇,. 图像通用目标的无监督检测[J]. 光学精密工程, 2014,22(1): 160-168

SONG Xiu-rui, WU Zhi-yong,. Unsupervised detection of image object with any class[J]. Editorial Office of Optics and Precision Engineering, 2014,22(1): 160-168
宋修锐, 吴志勇,. 图像通用目标的无监督检测[J]. 光学精密工程, 2014,22(1): 160-168 DOI： 10.3788/OPE.20142201.0160.

SONG Xiu-rui, WU Zhi-yong,. Unsupervised detection of image object with any class[J]. Editorial Office of Optics and Precision Engineering, 2014,22(1): 160-168 DOI： 10.3788/OPE.20142201.0160.

摘要

为了实现对图像中多种类目标的检测

缩短目标搜索时间

本文基于图像目标的3个显著性线索(显著性检测

颜色对比

超像素跨越)

构建了一种改进的通用无监督目标检测模型。通过机器学习center-surrounding比例参数

计算各个线索的显著度得分

并在朴素贝叶斯框架下对这3个目标显著性线索进行融合

以最终确定窗口中包含图像目标的概率。实验参数在PASCAL VOC 2007图像库进行检测

检测率为28.94%

击中率达96.99%;在MSRC图片库进行检测

检测率为80.64%

击中率达99.10%;得到的结果证明了本文模型的通用性。另外

该模型对单幅图像的处理时间较Bogdan的检测模型提高了40%

改进了目标检测效率。本文模型可为后续的目标识别

图像分割提供更快、更准确的先验位置信息。

Abstract

To measure a variety of objects of an image and to reduce the detection time

an unsupervised object detection model was established to provide location priors. The model was mainly based on three image cues of a object

and they are saliency detection

color contrast and superpixel straddling. To determine the likelihood of image object contained in a window

the saliency scores of the three cues were calculated

and the saliency cues of the three objects were fused in a simple Bayesian framework by a machine learning center-surrounding proportion parameter. In experiments on the challenging PASCAL VOC 07 dataset

it shows that the detection rate is 28.94 %

the hit rate is 96.99% and the combined measuring result is better than any cue alone. In experiments on MSRC dataset

it shows that the proposed model is generic and efficient

the detection rate is 80.64 %

the hit rate is 99.10% and the average processing time is 40% less than that of Bogdan's model.These results from extensive field tests suggest that proposed model can provide better location priors to the object recognition and image segmentation where the location of object is unknown.

关键词

Keywords

references

LEIBE B, LEONARDIS A, SCHIELE B. Combined object categorization and segmentation with an implicit shape model [C]. Europeon Conference on Computer Vision, ECCV, 2004: 17-32.

FELZENSZWALB P F, GIRSHICK R B, MCALLESTER R B D, et al.. Object detection with discriminatively trained part based models [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, TPAMI, 2010: 1627-1645.

LI L J, WANG G, LI F F . OPTIMOL: automatic online picture collection via incremental model learning [C]. IEEE Conference on Computer Vision and Pattern Recognition, CVPR, 2007: 1-8.

GU C, LIM J J, ARBELAEZ P, et al..Recognition using regions [C]. IEEE Conference on Computer Vision and Pattern Recognition, CVPR, 2009:1030-1037.

KUMAR M P, TORR P H S, ZISSERMAN A. OBJCUT:Efficient segmentation using top-down and bottom-up cues[C]. IEEE Conference on Computer Vision and Pattern Recognition, CVPR, 2005:18-25.

LEIBE B, SCHIELE B. Interleaved object categorization and segmentation[C]. British Machine Vision Conference, BMVC, 2003:264-271.

FULKERSON B, VEDALDI A, SOATTO S. Class segmentation and object localization with superpixel neighborhoods [C]. IEEE Conference on Computer Vision and Pattern Recognition, CVPR, 2009: 670-677.

ALEXE B, DESELARES T, FERRARI V. What is an object?[C]. IEEE Conference on Computer Vision and Pattern Recognition, CVPR, 2010:73-80.

ACHANTA R, HEMAMI S, ESTRADA F, et al..Frequency-tuned salient region detection[C]. IEEE Conference on Computer Vision and Pattern Recognition, CVPR, 2009:1597-1604.

HOU X, ZHANG L. Saliency detection: a spectral residual approach[C]. IEEE Conference on Computer Vision and Pattern Recognition, CVPR, 2007:1-8.

EVERINGHAM M, Van Gool L, WILLIAMS C K I, et al.. The PASCAL Visual Object Classes Challenge 2006[R].2006.

SHOTTON J, JOHNSON M, CIPOLLA R.Semantic texton forests for image categorization and segmentation [C] . IEEE Conference on Computer Vision and Pattern Recognition, CVPR, 2009:1-8.

FELZENSZWALB P F, GIRSHICK R B, MCALLESTER D, et al.. Object detection with discriminatively trained part based models[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, TPAMI, 2010:1627-1645.

SHOTTON J, WINN J, ROTHER C, CRIMINISI A. Textonboost for im-age understanding: Multi-class object recognition and segmentation by jointly modeling texture[J]. International Journal of Computer Vision, IJCV, 2007:2-23.

VERBEEK J, TRIGGS B. Region classification with markov field aspect models [C]. IEEE Conference on Computer Vision and Pattern Recognition, CVPR, 2007.

EVERINGHAM M, VAN GOOL L, WILLIAMS C, et al.. Zisser-man. The PASCAL Visual Object Classes Challenge 2007[R].2007.

浏览量

679

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

集成视觉显著性和群决策的航空零件孔特征检测

结合分数阶显著性检测及量子烟花算法的NSST域图像融合

结合Retinex校正和显著性的主动轮廓图像分割

光学遥感图像中复杂海背景下的舰船检测

基于元胞自动机的动态背景运动目标检测