项目名称: 差分演化算法中种群多样性的自主增强技术研究及其在高光谱遥感图像分类中的应用
项目编号: No.61305086
项目类型: 青年科学基金项目
立项/批准年度: 2014
项目学科: 自动化技术、计算机技术
项目作者: 杨鸣
作者单位: 中国地质大学(武汉)
项目金额: 25万元
中文摘要: 种群多样性是差分演化算法能够搜索到全局最优解的决定性因素,对于高维多峰优化问题,其影响更加显著。针对高光谱遥感图像分类这类高维多峰优化问题,现有算法的分类精度有待提高,而高精度的分类结果是实现遥感信息后续实际应用的前提。本项目对差分演化算法中的种群多样性进行研究,设计一种基于种群多样性自主增强技术的差分演化算法。这种算法通过分析种群的进化过程和分布状态来判断种群是否多样性缺失或进化停滞,自主确定种群多样性增强的时机及新种群分布的位置和范围,进而生成新种群来增强种群的多样性,消除种群进化停滞现象,提高其对于高维多峰优化问题的全局寻优能力;将此算法应用到高光谱遥感图像分类问题中,提高分类结果的精度,促进遥感信息处理技术的发展及其在国民经济中的应用。除了差分演化算法,种群多样性的自主增强技术还可以广泛地应用到其它演化算法中。因而本项目的研究具有重要的理论意义和应用价值。
中文关键词: 差分演化算法;高光谱遥感图像分类;种群多样性;自主增强;
英文摘要: The population diversity is the decisive factor in the differential evolution (DE) algorithm to search the global optimal solution, especially for DE on the high-dimension and multimodal problems. For the hyperspectral remote sensing image classification, which is a high-dimension and multimodal optimization problem, it needs to improve the accuracy of classification results, because high-precision remote sensing image classification is an important prerequisite of a variety of practical remote sensing applications. To enhance the performance of DE algorithms, in this project, we will study the self-enhanced population diversity and apply it into the DE algorithm. The improved DE algorithm can automatically identify the moment, which is the time that the population diversity needs to be enhanced, when the population diversity is poor or the population stagnates, by analyzing the evolutionary process and the distribution of population. When the moment is identified, the population will be regenerated to enhance its diversity or to eliminate the stagnation issue. The location and scope of the new population distribution can be automatically determined. The self-enhanced population diversity can improve the DE algorithms' performance of global optimization, especially for the optimization of high-dimension and mult
英文关键词: Differential evolution;Hyperspectral remote sensing image classification;Population diversity;Self-enhancement;