基于支持向量机目标检测算法综述_郭明玮.pdf

第29 卷 第2 期 控 制 与 决 策 2014 年 2 月 Vol. 29 No. 2 Control and Decision Feb. 2014 文章编号: 1001-0920 (2014) 02-0193-08 DOI: 10.13195/j.kzyjc.2013.0891 基于支持向量机的目标检测算法综述 郭明玮, 赵宇宙, 项俊平, 张陈斌, 陈宗海 (中国科学技术大学自动化系，合肥 230027) 摘 要: 目标检测的目的在于从静态图片或视频中检测并定位设定种类的目标物体, 已有研究大都将目标检测问题 简化为一个二分类问题. 鉴于支持向量机在模式识别领域尤其是解决二分类问题中所表现出来的优越性, 如何将其 应用于目标检测已成为当今计算机视觉领域关注的重点. 对此, 从支持向量机原理、目标特征模型构建、学习训练和 目标检测框确定等角度, 综述了基于支持向量机的目标检测算法的研究现状, 并就进一步的发展进行了展望. 关键词: 支持向量机；目标检测；局部可变模型；目标特征模型；学习训练 中图分类号: TP273 文献标志码: A Review of object detection methods based on SVM GUO Ming-wei, ZHAO Yu-zhou, XIANG Jun-ping , ZHANG Chen-bin, CHEN Zong-hai (Department of Automation，University of Science and Technology of China ，Hefei 230027 ，China ．Correspondent: CHEN Zong-hai，E-mail ：chenzh@) Abstract: The purpose of object detection is to detect and locate the object with a certain class from the static image or videos, and many studies simplify the object detection as a binary classification problem. For the reason that the support vector machine(SVM) can solve the pattern recognition problem well, especially the binary classification problem, how to use the SVM in computer vision becomes a hot point of many researchers. The status of object detection methods based on SVM is reviewed by introducing the concept and theory of SVM, the building of object feature model, training process and location of detection box. Finally, the future work of object detection methods based on SVM is discussed. Key words: SVM；object detection ；deformable part models ；object feature model ；training