图像物体分类与检测算法综述.PDFVIP

第36 卷 计 算 机 学 报 Vol.36 2013 年 论文在线发布号 No.12 CHINESE JOURNAL OF COMPUTERS 2013 Article Online No.12 图像物体分类与检测算法综述 黄凯奇,任伟强,谭铁牛 ( 中国科学院自动化研究所，模式识别国家重点实验室智能感知与计算研究中心, 北京市 中国 100190) 摘 要 图像物体分类与检测是计算机视觉研究中的两个重要的基本问题，也是图像分割、物体跟踪、行为分析等其他 高层视觉任务的基础。本文从物体分类与检测问题的基本定义出发，首先从实例、类别、语义三个层次对物体分类与检 测研究中存在的困难与挑战进行了阐述。接下来，本文以物体检测和分类方面的典型数据库和国际视觉竞赛PASCAL VOC 竞赛为主线对近年来物体分类与检测的发展脉络进行了梳理与总结，指出表达学习和结构学习分别对于物体分类与检测 的作用。最后本文对物体分类与检测的发展方向进行了思考和讨论，探讨了这一领域下一步研究的方向。 关键词 物体分类；物体检测；计算机视觉；特征表达；结构学习； 中图法分类号 TP391.4 A Review on image object classification and detection HUANG Kai-Qi, REN Wei-Qiang, TAN Tie-Niu ( National Laboratory of Pattern Recognition, Center for Research on Intelligent Perception and Computing, Institute of Automation, Chinese Academy of Sciences, Beijing 100190, China) Abstract Image object classification and detection are two of the most essential problems in computer vision. They are the basis of many other complex vision problems, such as segmentation, tracking, and action analysis. In this paper, we try to give an analysis of object classification and detection based on PASCAL VOC challenge, which is generally acknowledged as a public evaluation for object recognition. We first discuss the importance of object classification and detection; next we summarize the difficulties and challenges in the development of basic object recognition. Then we review the yearly achievements in the study of object classification and detection. Finally we discuss the development directions of object classification and detection, from the view of representations learning and structure learning. Key words Object Classification; Object Detection; Computer Vision; Feature Representations; Struct