面向目标检测的稀疏表示方法研究进展

Recent Advances of Sparse Representation for Object Detection

目标检测作为图像理解的一个基础而重要的课题深受国内外学者的重视,在军事和民用中具有广泛应用.应用背景的多样性和复杂性使得传统目标检测算法难以克服复杂背景、噪声干扰、光照变化以及非刚体形变、遮挡、弱特征、尺度、视角和姿态变化等因素的影响.近些年来发展起来的稀疏表示方法为图像处理及目标检测研究提供了新的思路,本文概述了稀疏表示基本概念和理论研究进展,综述了稀疏表示方法在目标特征学习、目标分类器和滤波器设计以及多源信息融合目标检测等目标检测领域中的国内外重要研究进展,并展望了稀疏表示方法在目标检测领域的发展方向.

Object detection is a basic and important subject in image understanding,which has attracted much attention from domestic and foreign scholars. Object detection has been widely used in military and civilian. The diversity and complexity of applications makes the traditional detection technique be affected by many factors such as complex background,noise,illumination variations,non-rigid deformation,occlusion,feeble features,scale,visual angle attitude and,etc.Recently,the developing method of sparse representation provides a novel research approach for image processing and objects detection. This paper overviews the basic concept of sparse representation and its recent progress in the theoretical study. The domestic and foreign research advances of sparse representation in object detection are summarized,especially in object feature learning,classifier and filter designing,multisource fusion detection. Meanwhile,some future directions of sparse representation in object detection are also addressed.