扩展工作条件下基于核免疫集成的SAR目标识别关键技术研究

项目名称： 扩展工作条件下基于核免疫集成的SAR目标识别关键技术研究

项目编号： No.61573375

项目类型： 面上项目

立项/批准年度： 2016

项目学科： 自动化技术、计算机技术

项目作者： 黄文龙

作者单位： 中国人民解放军空军工程大学

项目金额： 16万元

中文摘要： 在扩展工作条件(EOC)下，难以获得SAR目标的全面信息，再加上斑点噪声的影响，使得SAR图像分割和目标识别成为极具挑战性的任务。本项目借鉴免疫网络的结构自适应性和免疫全局优化特点，结合支持向量域描述理论，拟从仿生学角度给出解决问题的途径。首先研究有效的SAR图像不变特征的提取方法，然后利用支持向量域描述理论来改造抗体的识别邻域，构建一种完全非监督的免疫网络聚类算法，以结构自适应免疫抗体网络的构建为核心，结合分水岭和最小生成树算法来设计一种完全非监督的SAR图像层次分割模型。良好的分割结果将为后续SAR自动目标识别奠定很好的基础。这些研究成果将为SAR自动目标识别提供理论依据和技术支撑，具有重要的研究价值和广阔的应用前景。

中文关键词： SAR图像分割；人工免疫网络；结构自适应；聚类；半监督

英文摘要： Under the extended operating condition (EOC), it is difficult to obtain comprehensive information on the synthetic aperture radar (SAR) target, and coupled with the impact of speckle noise it becomes a challenging task for the SAR image segmentation and target recognition. Borrowing artificial immune network structure adaptive characteristic and immune multi-objective global optimization performance, the project attempts to make thorough research on correlative techniques for SAR image segmentation and recognition with a strong focus on bionics based on the theories of support vector domain description (SVDD). The project firstly attempts to make thorough research on effective SAR image feature extraction method, then to research a novel completely unsupervised immune clustering network based on the reform of antibody recognition neighborhood by SVDD. Taking the design of a structure adaptive immune antibody network as the core, and associating with watershed and minimum spanning tree algorithms, we want to structure a fully unsupervised SAR image segmentation layered model. Good segmentation results will establish the good foundation for the follow-up SAR automatic target recognition(ATR). These study achievements will be efficiently applied to the complex SAR ATR system, and have important research values and

英文关键词： SAR image segmentation；Artificial immune networks；Structural adaptation；Clustering；Semi-supervised