项目名称: 基于人的视觉认知机制的目标识别与分类研究
项目编号: No.61305035
项目类型: 青年科学基金项目
立项/批准年度: 2014
项目学科: 自动化技术、计算机技术
项目作者: 王迪
作者单位: 温州大学
项目金额: 25万元
中文摘要: 近年来借助人类视觉的信息处理机制来增强机器视觉的识别能力已成为计算机视觉领域的研究热点之一。然而,如何模拟大脑视觉系统的典型功能或信息处理机制,使计算机拥有人类所具备的识别能力,却是我们面临的一大挑战。本项目拟借鉴人类视觉系统的信息处理机制以及视觉心理认知准则,结合现有的机器学习方法,构建具有人类某些视觉功能特性的目标识别和分类系统,具体包括:1)在视觉认知模型中引入人的选择性注意机制和学习机制,改善模型的学习能力,用于"学习"目标的原型特征向量,提高编码的可分性,2)模拟人的较高层次视觉系统的编码机制,建立基于流形学习理论的多项式双向映射模型,用于目标本质特征的快速提取,3)将视觉认知系统中的若干关键性机制引入到支持向量机中,用于提高分类器的分类速度和精度。所获结果将不仅为目标识别和分类问题提供新的解决方案,也有望对生物学、心理学的研究有启示作用,因而具有重要的理论意义与广阔的应用前景。
中文关键词: 图像恢复;字典学习;稀疏编码;特征提取;高效分类器
英文摘要: How to improve the visual recognition ability of machines by applying information processing system of human visions has become a hot research topic in computer vision field in recent years. However, the great challenge is how to imitate the typical function or information processing mechanism of the brain vision system to equip the computers with the recognition ability comparable to the mankind. In this project, combined with existing machine learning methods, we plan to establish a system of objects recognition and classification with characteristics of human vision based on the information processing mechanism and principles of human visual cortex. The concrete research of this project includes: 1) introduce the human visual selective attention mechanism and learning mechanism, and improve the learning ability in visual cognition models, then use for "learning" prototypical features of the objects and enhancing the separability of the coding, 2) simulate the coding mechanism of human advanced vision system, and establish a polynomial bidirectional mapping based on the manifold learning theory, then used for effectively extracting the essential features of objects, 3) introduce some key mechanisms of vision system in support vector machines (SVMs), and improve the classification speed and accuracy. The result
英文关键词: image restoration;dictionary learning;sparse coding;feature extraction;efficient classifier