项目名称: 协同Memetic计算模型及其在SAR图像变化检测中的应用
项目编号: No.61202176
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 计算机科学学科
项目作者: 马晶晶
作者单位: 西安电子科技大学
项目金额: 25万元
中文摘要: 本课题基于生物机体内神经系统与免疫系统的整合调节机理,构造协同Memetic计算模型并应用于SAR图像变化检测。研究内容包括:设计神经系统对免疫反应单向调节的拉马克学习策略;设计神经系统与免疫系统双向交互对免疫反应调节的班德文学习策略;模拟神经系统对免疫反应的多神经元网络调节机理,建立多策略学习和高阶学习模型;针对大规模NP-Hard优化和复杂数据聚类问题设计高性能求解算法,通过学习缓解搜索过程中因盲目性而导致的学习能力差、收敛速度慢的难题。基于协同Memetic计算模型,构造适合于SAR图像变化检测的差异图分析方法,克服对差异图概率统计模型的依赖。预期在本领域主流刊物和会议发表论文8~10篇;申报专利或软件著作权2~3项;联合培养博士、硕士3~5名。
中文关键词: Memetic计算;人工免疫系统;聚类;SAR 图像;变化检测
英文摘要: This proposal will build a cooperative memetic computational model based on the regulation between nervous system and immune system, and apply the new model to change detection in SAR images. The Lamarchian learning will be designed based on the unidirectional regulation of nervous system on the immune response. The Baldwinian learning will be designed based on the bidirectional reulation between nerous system and immune system. The cooperative learning and high-level learning models will be designed based on the network regulation of multiple neurons on immune response. High performance Memetic algorithms for NP-Hard optimization and complex data clustering will be proposed to overcome the main drawbacks in learning ability and convergence speed. Based on these models, the change detection algorithm in SAR image will be proposed to overcome the dependence on probability statistical model of difference image. We will publish 8-10 papers, apply 2-3 patents, and bring up 3-5 graduate students.
英文关键词: Memetic Computing;Artificial Immune System;Clustering;SAR image;Change Detection