SIA OpenIR  > 光电信息技术研究室
几何与辐射联合李群及其图像匹配应用研究
Alternative TitleJoint geometric and photometric Lie group with applications on image registration
李晨曦1,2
Department光电信息技术研究室
Thesis Advisor史泽林
Keyword图像匹配 目标跟踪 大气辐射传输 亮度变化 李群
Pages121页
Degree Discipline模式识别与智能系统
Degree Name博士
2019-03-28
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract动态图像识别与精确跟踪是光电信息处理领域的重点研究方向。探测器与目标之间的相对姿态及相对距离的改变,不仅导致目标在像平面的几何变化,同时还会因大气辐射传输的影响使目标成像发生灰度变化,给目标实时跟踪和识别造成极大困难。建立在欧氏空间的传统处理算法难以解决问题,研究目标成像的几何和辐射变化的李群表征并提出图像匹配算法,是当前的一个热点命题,也是现代数学应用于光电信息处理的有益探索。论文以复杂条件下的图像匹配为研究主题,内容主要包括:第3章研究了基于李群聚类的形状识别方法。以形状为目标特征,在Grassmann流形上进行仿射不变形状表征。针对局部遮挡问题,提出了基于几何变换一致性约束的仿射不变形状匹配识别算法。首先根据形状的零曲率点进行形状的仿射不变分解,利用动态规划算法对得到的形状片段进行初始匹配,针对现有方法存在的误匹配问题,通过在李群空间上对部分形状匹配得到的几何变换集合进行聚类,有效去除了误匹配,提高了形状匹配识别的性能。第4章分析了大气辐射与传输对光电成像亮度变化的影响,发现并从数学上证明了红外成像灰度随距离变化符合李群特征,建立了李群模型(称之为视在辐射变换李群),为几何-辐射联合李群跟踪算法研究奠定了理论基础。第5章针对目标本征辐射亮度的变化,提出了几何-本征辐射联合变换李群及图像匹配跟踪算法。由于第4章建立的视在辐射变换李群是仿射光照模型的子群,本章在大气有关常数未知的情况下设计跟踪算法。将图像变换空间建立于射影几何变换和图像灰度变换的联合变换李群下,提出联合变换李群的参数化即李代数,采用高效二阶优化策略进行几何-灰度参数的联合求解。对光照剧烈变化的环境适应性实验表明算法具有很强的鲁棒性和很高的计算效率。算法不仅适用于单色可见光,同时也适应于红外成像跟踪。论文还进一步将其扩展到了多光谱彩色图像。 第6章针对距离变化情况下大气散射与传输影响的变化,提出了几何-视在辐射联合变换李群及目标跟踪算法。本章在大气有关常数合理估计的情况下直接根据第4章建立的视在辐射变换李群设计目标跟踪算法。将图像变换空间建立于射影几何变换和视在辐射变换李群的联合变换李群下,在联合变换李群的李代数空间上,对几何和视在辐射变换联合参数进行优化求解。实验表明本章直接利用视在辐射变换李群设计的算法对距离变化具有较强的适应性,且跟踪精度和计算效率都很高。
Other AbstractDynamic image recognition and accurate tracking are the key research directions in the field of photoelectric information processing. The change of the relative attitude and distance between the detector and the target not only leads to the geometric change of the target in the image plane, but also leads to the grayscale change of the target imaging due to the influence of atmospheric radiation transmission, which causes great difficulty in the real-time tracking and recognition of the target. Traditional processing algorithms based on Euclidean space are difficult to solve the problem. It is a hot topic to study the Lie group representation of geometric and radiative changes of target imaging and then propose image registration algorithms. This paper takes image registration under complex conditions as the research topic, and its contents mainly include: In chapter 3, the shape recognition method based on Lie group clustering is studied. Taking the shape as the target feature, the affine invariant shape characterization is carried out on the Grassmann manifold. Considering the local occlusion problem, an affine invariant shape matching and recognition algorithm based on the constraint of geometric transformation consistency is proposed. Firstly, affine invariant shape decomposition is carried out according to zero curvature point, then use dynamic programming algorithm to get the initial matching of the shape fragments. Against the false matching problem of existing methods, through clustering on Lie group space for the collection of geometric transformation obtained by the initial matching, in addition to the effective matching error, improve the performance of shape matching identification. Chapter 4 analyzes the influence of atmospheric radiation and transmission on the brightness change of photoelectric imaging, finds and proves mathematically that the change of infrared imaging gray with distance conforms to the Lie group characteristics, establishes the lie group model (called the apparent radiation transformation Lie group), and lays the theoretical foundation for the research of geometric and photometric joint Lie group tracking algorithm. In chapter 5, for the change of target eigenradiation brightness, the joint geometric and eigenradiation transformation Lie group and image matching-based tracking algorithm are proposed. Since the apparent radiation transform Lie group established in chapter 4 is a subgroup of the affine illumination model, the tracking algorithm is designed for the affine grayscale transformation model. The image transformation space is established in the joint Lie group of the projective geometry transformation and image gray-scale transformation, parameterized by the Lie algebra. The efficient second-order optimization strategy is employed to solve the joint geometrical and gray parameters. The experiments show that the algorithm has strong robustness and high computational efficiency under dramatic change of light environment. The algorithm is not only suitable for monochromatic visible light, but also for infrared imaging tracking. The paper also extends it to multi-spectral color images. In chapter 6, for the change of atmospheric scattering and transmission effects in the case of distance change, the joint geometric and apparent radiation transformation Lie group and image tracking algorithm are proposed. In this chapter, the image tracking algorithm is designed based on the apparent radiation transformation Lie group established in chapter 4. The image transformation space is established on the joint Lie group of projective geometric transformation and the apparent radiation transformation Lie group. Experiments show that the algorithm designed in this chapter by directly using the apparent radiation transformation Lie group has strong adaptability to the change of distance and high tracking accuracy and computational efficiency.
Language中文
Contribution Rank1
Document Type学位论文
Identifierhttp://ir.sia.cn/handle/173321/25160
Collection光电信息技术研究室
Affiliation1.中国科学院沈阳自动化研究所
2.中国科学院大学
Recommended Citation
GB/T 7714
李晨曦. 几何与辐射联合李群及其图像匹配应用研究[D]. 沈阳. 中国科学院沈阳自动化研究所,2019.
Files in This Item:
File Name/Size DocType Version Access License
几何与辐射联合李群及其图像匹配应用研究.(6089KB)学位论文 开放获取CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[李晨曦]'s Articles
Baidu academic
Similar articles in Baidu academic
[李晨曦]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[李晨曦]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.