基于人工禁忌免疫原理的多源遥感图像自动配准研究

项目名称： 基于人工禁忌免疫原理的多源遥感图像自动配准研究

项目编号： No.41261091

项目类型： 地区科学基金项目

立项/批准年度： 2013

项目学科： 天文学、地球科学

项目作者： 叶发茂

作者单位： 南昌大学

项目金额： 48万元

中文摘要： 多源遥感图像配准由于待配准图像之间差异很大，所以相似测度难以寻找，搜索空间大而复杂，一般的搜索算法难以解决，然而图像配准是多源遥感信息融合的基础和必要条件，是遥感科学与技术中急需解决的前沿问题。本项目将免疫记忆机制和禁忌搜索思想与人工免疫原理相结合研究出禁忌人工免疫模型，并在此基础上针对遥感图像配准复杂的搜索问题，建立禁忌人工免疫原理搜索模型。该模型引入免疫记忆机制和禁忌搜索算法中的禁忌准则，避免一般人工免疫模型的迂回搜索问题，提高收敛速度；通过特赦准则保证细胞的多样性，避免过早收敛，提高全局搜索能力；并利用遥感图像已有信息和边缘提取的信息，建立禁忌人工免疫搜索模型初始化规则，减少算法随机性，提高初始化的有效性，增强配准的速度和鲁棒性。同时利用基于特征信息和局部灰度信息构造混合相似性测度进一步提高配准精度。通过本项目研究能够为实现多源遥感图像高精度配准提供更好的思路和方法。

中文关键词： 多尺度禁忌人工免疫模型；多源遥感图像配准；混合相似测度；边缘检测；

英文摘要： Registering two remote sensing images taken at different times and with different sensors that having variable spatial or temporal variations is difficult to find an effective similarity measure and a good search strategy. However, image registration is the precondition and foundation of the fusion of multi-source data and is a cutting edge issue in the remote sensing science and technology. In this research， A new search strategy combines artificial immune network algorithm and tabu search algorithm to solve the problem of great amount of computation of image registration. Through the introduction of immune memory mechanisms and tabu criterion, the model can reduce the repeated search and improve the convergence speed. The model uses aspiration criterion to keep cells diversity, so it can preventing premature convergence to local optima and improve the global search capability. At the same time, it builds the initialization rules according the existing information of remote sensing images and the information extracted from edges, that can reduce the randomness of algorithm and enhance the speed and robustness of image registration. It constructs hybrid similarity measure based on the feature information and local gray information to further improve the registration accuracy. This research will provide bette

英文关键词： multi-scale Tabu artificial immune network algorit；multi-source remote sensing image registration；mixed similarity measure；edge detection；