项目名称: 过完备字典自适应优化理论及应用研究
项目编号: No.61272050
项目类型: 面上项目
立项/批准年度: 2013
项目学科: 自动化技术、计算机技术
项目作者: 沈琳琳
作者单位: 深圳大学
项目金额: 60万元
中文摘要: 申请人主持的青年科学基金项目在过完备字典设计和稀疏表示求解算法等领域进行了深入研究,在国际权威期刊和会议上发表学术论文25篇(SCI论文9篇、EI论文19篇)。研究工作发现基于二维/三维Gabor函数的字典存在原子数量庞大、计算复杂以及缺乏自调节机制等问题。 针对上述问题,本项目重点研究过完备字典中的原子筛选和优化策略,探索将字典引入稀疏表示求解的目标函数来进行自适应调整,提高目前生物启发式优化算法的全局搜索能力,最终实现对噪声不敏感、计算量小、准确和高效的多光谱图像分类系统。项目最终将寻求把研究成果应用到高光谱遥感图像中的地物分类、多光谱乳腺切片图像中的癌细胞诊断以及高光谱人脸识别等问题。 项目涉及模式识别、智能优化、遥感和医学等不同学科领域,对于生命健康、疾病诊断和地球环境监测具有非常重要的研究意义。
中文关键词: Gabor特征;稀疏表示;字典学习;高光谱图像分类;
英文摘要: The applicant''s previous project mainly focuses on over-complete dictionary design using 2D/3D Gabor functions and the optimization algorithms for sparse representation. The research output has been presented in 25 papers published at international journals and conferences. Out of them, 9 are indexed by SCI and 19 are indexed by EI. The research works also found that current dictionary is computationally very expensive due to the large number of atoms. Following the research work, this proposal aims to study the methods to remove redundant atoms, the approaches to include dictionary learning for sparse representation, the abilities of optimization algorithms on global search and finally develop robust and efficient object recognition system. The research output of this project will be applied to remote sensing image classification, breast cancer diagnosis and hyperspectral face recognition. As an inter-discipline proposal covering pattern recognition, optimization, remote sensing and medical diagnosis, this project will be very useful for life health, disease diagnosis and earth environment protection.
英文关键词: Gabor Feature;Sparse Representation;Dictionary Learning;Hyperspectral Image Classification;