[1]李新德,张晓,朱博.基于立体视觉的一般物体识别方法[J].东南大学学报(自然科学版),2013,43(4):711-716.[doi:10.3969/j.issn.1001-0505.2013.04.008]
 Li Xinde,Zhang Xiao,Zhu Bo.Generic object recognition method based on stereo vision[J].Journal of Southeast University (Natural Science Edition),2013,43(4):711-716.[doi:10.3969/j.issn.1001-0505.2013.04.008]
点击复制

基于立体视觉的一般物体识别方法()
分享到:

《东南大学学报(自然科学版)》[ISSN:1001-0505/CN:32-1178/N]

卷:
43
期数:
2013年第4期
页码:
711-716
栏目:
自动化
出版日期:
2013-07-20

文章信息/Info

Title:
Generic object recognition method based on stereo vision
作者:
李新德张晓朱博
东南大学复杂工程系统测量与控制教育部重点实验室, 南京210096; 东南大学自动化学院, 南京210096
Author(s):
Li Xinde Zhang Xiao Zhu Bo
Key Laboratory of Measurement and Control of Complex Systems of Engineering of Ministry of Education, Southeast University, Nanjing 210096, China
School of Automation, Southeast University, Nanjing 210096, China
关键词:
一般物体识别 立体视觉 图像分割 GPU加速
Keywords:
generic object recognition stereo vision image segment GPU(graphic processing unit)acceleration
分类号:
TP24
DOI:
10.3969/j.issn.1001-0505.2013.04.008
摘要:
为了使计算机具有与人类相似的在复杂背景下识别一般物体的视觉处理能力,提出了一种基于立体视觉的一般物体识别方法.该方法的核心在于融合二维图像信息和双目相机获取的深度信息,对视野中的环境进行物体定位、图像分割、特征描述以及物体识别.通过双目相机获取环境的三维点云信息,并利用mean-shift算法进行聚类,剔除干扰点,从而实现物体在二维图像上的定位与分割.利用含有空间关系的BoW模型对分割后独立区域内的物体进行识别,得出判别结果.此外,在利用sift算法进行特征点提取以及利用mean-shift算法进行聚类的环节中,采用CUDA环境下的GPU进行加速处理,提高了处理速度.实验结果表明,所提方法具有较好的识别效果和鲁棒性.
Abstract:
In order to make computer have visual processing capability similar to humans for generic object recognition in complex background, a generic object recognition method based on stereo vision is proposed. The kernel of this method lies in fusing 2D image information and depth information from binocular camera for object localization, image segment, feature description and object recognition in field of vision. 3D point cloud information of environment is acquired from stereo camera, and the mean-shift algorithm is used to cluster those points. Object localization and image segment is achieved on 2D image after some interference points are eliminated. In terms of the model of BoW(bag of words)with spatial relationship, objects in independent area after segmentation are recognized and the discriminated results are obtained. In addition, GPU(graphic processing unit)acceleration technology under CUDA(compute unified device architecture)environment is applied to feature extraction by the sift algorithm and clustering by the mean-shift algorithm in order to greatly improve the processing speed. The experimental results show that this method is robust and the recognition effect is good.

参考文献/References:

[1] 田渊栋. 基于特征组合的一般物体识别相关算法研究[D].上海:上海交通大学计算机科学与工程系,2007.
[2] 王利明. 机器视觉中物体识别方法的研究与探讨[D]: 上海:复旦大学计算机科学技术学院,2009.
[3] Felzenszwalb P F, Huttenlocher D P. Efficient matching of pictorial structures[C]//Proceedings of 2000 IEEE Conference on Computer Vision and Pattern Recognition. Hilton Head, South Carolina,USA, 2000:66-73.
[4] Kwon J,Lee K M. Highly non-rigid object tracking via patch-based dynamic appearance modeling[J]. IEEE Transaction on Pattern Analysis and Machine Intelligence, 2013.(to appear)
[5] Niu Z X, Hua G, Gao X B, et al. Spatial-DiscLDA for visual recognition[C]//Proceedings of 2011 IEEE Conference on Computer Vision and Pattern Recognition. Colorado Springs, CO,USA, 2011:1769-1776.
[6] Wang B T,Xiong H K, Jiang X Q, et al. Semi-supervised object recognition using structure kernel[C]//Proceedings of 2012 IEEE International Conference on Image Processing.Orlando, Florida, USA, 2012: 2157-2160.
[7] Ren Z, Yuan J S, Meng J J, et al. Robust part-based hand gesture recognition based on finger-earth movers distance[J]. IEEE Transaction on Multimedia, 2013.(to appear)
[8] Felzenszwalb P F, Girshick R B, McAllester D, et al. Object detection with discriminatively trained part-based models[J]. IEEE Transaction on Pattern Analysis and Machine Intelligence, 2010, 32(9):1627-1645.
[9] Helmer S, Lowe D.Using stereo for object recognition[C]//Proceedings of 2012 IEEE International Conference on Robotics and Automation. Anchorage, Alaska, USA, 2010:3121-3127.
[10] Fukunaga K, Hostetler L D. The estimation of the gradient of a density functions with applications in pattern recognition[J]. IEEE Transactions on Information Theory, 1975, 21(1):32-40.
[11] Csurka G, Dance C R, Fan L, et al. Visual categorization with bags of keypoints[C]///Proceedings of ECCV04 Workshop on Statistical Learning in Computer Vision. Prague, Czech Republic, 2004:59-74.
[12] 李新德,张晓,朱博,等.一种基于新奇GOR+GPU算法的机器人视觉导航方法[J].机器人,2012,34(4):466-475.
  Li Xinde, Zhang Xiao, Zhu Bo, et al. A visual navigation method for robot based on a GOR and GPU algorithm[J]. Robot, 2012,34(4):466-475.(in Chinese)
[13] Lowe D G. Distinctive image features from scale-invariant key pints [J]. International Journal of Computer Vision,2004, 60(2):91-110.
[14] 张舒,褚艳利.GPU高性能运算之CUDA[M].北京:中国水利水电出版社,2009.

相似文献/References:

[1]罗翔,席文明,颜景平.一种双目主动立体视觉系统的目标定位算法[J].东南大学学报(自然科学版),2002,32(1):59.[doi:10.3969/j.issn.1001-0505.2002.01.014]
 Luo Xiang,Xi Wenming,Yan Jingping.Scheme of target 3D point positioning for binocular active stereo visual system[J].Journal of Southeast University (Natural Science Edition),2002,32(4):59.[doi:10.3969/j.issn.1001-0505.2002.01.014]

备注/Memo

备注/Memo:
作者简介: 李新德(1975—),男,博士,副教授,博士生导师,xindeli@seu.edu.cn.
基金项目: 国家自然科学基金资助项目(61175091)、江苏省自然科学基金资助项目(BK2010403)、江苏省“青蓝工程”优秀青年骨干教师计划资助项目、东南大学优秀青年教师教学、科研资助计划资助项目(3208001203).
引文格式: 李新德,张晓,朱博.基于立体视觉的一般物体识别方法[J].东南大学学报:自然科学版,2013,43(4):711-716. [doi:10.3969/j.issn.1001-0505.2013.04.008]
更新日期/Last Update: 2013-07-20