项目名称: 基于特征结构关系的目标分类研究
项目编号: No.61203252
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 自动化学科
项目作者: 黄永祯
作者单位: 中国科学院自动化研究所
项目金额: 24万元
中文摘要: 目标分类是计算机视觉与模式识别领域的基本问题之一。在目标分类研究中,当前的主流算法是基于局部特征的,如视觉词典算法。这类方法在很多目标分类数据库和竞赛中都取得了很好的结果。基于局部特征的方法大部分都只关注局部特征本身,忽略了局部特征在图像空间中的结构关系,从而导致缺乏对目标的全局描述和理解。本项目拟对目标局部特征在图像空间中的结构关系进行较系统地研究,不仅关注其认知机理、物理意义和数学内涵,还采用局部特征概率密度估计、图结构匹配以及视觉单词相关这三种方法来研究目标特征在图像空间结构关系的建模问题。本项目是对目标分类研究的一项重要技术改进,其研究方法和成果有望使我们在目标分类领域引领新的发展潮流,成为具有重要影响力的前沿性研究。
中文关键词: 计算机视觉;模式识别;目标分类;图像空间;结构关系
英文摘要: Object classification is one of the fundamental problems in computer vision and pattern recognition. The widely used approach in object classification is the local feature based model, e.g., bag-of-visual-words. It achieves the state-of-the-art performance in many popular databases and competitions of object classification. The local feature based model focuses on describing local features but ignores their structural relations in the image space, which is very important for understanding global meaning of objects. In this proposal, we are intended to study the relations of local features, including its cognitive mechanism and mathematical models. Specifically, we propose to model features' relations from three aspects: probability density estimation, graph matching and visual word specific methods. This project is crucial to bring technical innovation for object classification. We believe that this work has the potential to become an influential study and create a new direction of object classification.
英文关键词: computer vision;pattern recognition;object classification;image space;structural relations