景象适配性分析与模板制备方法研究
文献类型：学位论文
| 作者 | 郭桂旭 |
| 学位类别 | 硕士 |
| 答辩日期 | 2014-05-28 |
| 授予单位 | 中国科学院沈阳自动化研究所 |
| 导师 | 惠斌 |
| 关键词 | 图像匹配 景象匹配辅助导航 适配性 基准图 |
| 其他题名 | Study on Scene Matching-area Suitability Analysis and Template Preparation |
| 学位专业 | 机械电子工程 |
| 中文摘要 | 景象匹配辅助导航是精确制导武器的一种重要制导方式，由于其自主性强，打击精度高等特点得到了广泛的应用。在景象匹配中，基准图的质量对制导精度的影响很大，目前的基准图制备是专家根据自己的经验完成，同样的条件下不同的专家选择的基准图也不同，存在主观性强、标准不统一的问题。基准图制备技术作为景象匹配导航的一项关键技术，是提高匹配概率的重要途径，因此景象适配性分析与模板制备方法研究具有重要的理论意义和实用价值。本文基于模板匹配分析了景象匹配中下视导航阶段和前下视自动目标识别阶段的成像特点，结合对图像匹配系统深入的分析，提出了一套相对完整的适配性分析与基准图制备方法。基准图的适配性问题是整个景象匹配系统的核心问题，涉及到景象匹配系统的不同方面。本文首先讨论了景象匹配涉及的各关键技术及其关系，指出了适配性分析是评价基准图的景象信息与景象结构的抗干扰能力，适配性分析是提高匹配概率的重要途径而不是决定匹配概率，合理的划定了适配性分析的研究范围。景象匹配导航系统中，基准图的工作范围包括下视导航阶段和前下视自动目标识别阶段。下视景象匹配导航阶段需先从景象区域中提取一定数目的候选适配区进行后续的适配性分析，前下视自动目标识别阶段要求制备的基准图在透视变换下仍能保持性能。针对这些要求，本文首先提出了一种融合独特性的候选适配区初选算法，该方法采用信息熵、图像方差、边缘密度、边缘方向离散度、改进的独立像元数作为特征指标，对每个特征指标以特征的独特性进行加权，采用融合的策略形成了候选适配区的选择方法；其次，从景象匹配曲面出发，介绍了景象匹配曲面的性质和匹配曲面特征对应的物理意义，提出了合理的景象适配性度量指标；最后针对自动目标识别阶段的前下视成像特点，分析了传统的模板在前下视情况下适配性能退化严重的原因，提出了一种符合前下视适配性能的基准图制备方法。 |
| 索取号 | TJ765/G94/2014 |
| 英文摘要 | Scene matching aided navigation is an important integrated navigation method for precision-guided weapons and has been widely used on account of its strong independence and high accuracy. In the scene matching , the quality of the reference image make a great influence on the final guidance result ,and preparation for reference is completed manually based on their experiences, while the reference images, selected by different people, are often different under the same conditions. Preparation for reference images is an essential part of scene matching navigation, and an important way to improve the matching probability, so the study on scene suitability analysis and template preparation method has important significance.In this paper, we propose a set of relatively complete suitability analysis and preparation of the reference chart method, based on the analysis of visual imaging features in visual navigation and automatic target recognition stages, and image matching system. The matching-area suitability is the core issue of the whole scene matching system, involves the scene matching system different parts. In this paper, after discussing the scene matching key technologies and its connections, we pointed that reference suitability analysis is a technique for study the ability to against the real-time image diversity with the reference image by analyzing the reference construction, the range of reasonable delineation study suitability analysis.In scene matching navigation system, the work scope of reference image includes navigation stage and automatic target recognition stage. In scene matching navigation stage, we need to extract a certain number of candidates for subsequent matching suitability analysis; in automatic target recognition stage, we need to guarantee the reference image performance in perspective transformation. To deal with these problems, firstly ,we proposes a candidate reference images selection algorithm combining uniqueness, in which we uses information entropy, image variance, edge density, edge direction dispersion and improved independent pixel numbers as the characteristic indexes and fuse these indexes weighted by its uniqueness; secondly, we explain the characteristics of the matching surface and its corresponding physical meaning and propose a reasonable matching suitability measure method; finally, after considering the imaging features in automatic target recognition stage, we analysis the reason of traditional template performance degradation and propose a preparation method of reference image reasonable matching suitability measure method presented before the next depending on the and the reference image area adaptation process for preparing. |
| 语种 | 中文 |
| 产权排序 | 1 |
| 页码 | 63页 |
| 分类号 | TJ765 |
| 源URL | [http://ir.sia.ac.cn/handle/173321/14795]  |
| 专题 | 沈阳自动化研究所_光电信息技术研究室 |
| 推荐引用方式 GB/T 7714 | 郭桂旭. 景象适配性分析与模板制备方法研究[D]. 中国科学院沈阳自动化研究所. 2014. |
入库方式： OAI收割
来源：沈阳自动化研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。