基于遗传算法的离散数据特征点识别与提取  

Feature Points Recognition and Extraction from Scattered Data Based on Genetic Algorithm

作　　者：吕秋娟[1,2] 方素平[1] 张镇 

机构地区：[1]西安交通大学机械制造系统工程国家重点实验室,陕西西安710049 [2]第二炮兵工程大学,陕西西安710025

出　　处：《东华大学学报（自然科学版）》2012年第5期597-600,共4页Journal of Donghua University(Natural Science)

摘　　要：针对ICT(industrial computed tomography)图像处理后零件轮廓的离散数据点,采用改进遗传算法的特征点自适应识别与提取方法对轮廓数据进行精简,以线段和圆弧为逼近基元,以较小的逼近误差(ISE)和较少的特征点为优化目标;对种群采取分类初始化,大大缩小了种群规模;变异概率和交叉概率自适应生成,加快了收敛速度.实例表明改进的遗传算法有更高的优化速度和全局搜索能力,特征点的提取效果较好.It is a key problem that the feature points recognition and extraction from the scattered data ot the ICT （industrial computed tomography） image. The line and circular arc are regarded as the fitted basic cell, an improved genetic algorthm is then proposed for feature points recognition and extraction using the least error and least feature points as the optimized aim. First, the contour points are condensed to reduce the population scale. Second, mutation probabilities and crossover probabilities are caculated adaptively. In this method, the reduced population and expanded search space improve the fitness function, which cause the fitness more effective. Some practical examples are then presented which verified the effectiveness of the method.

关 键 词：特征点 遗传算法 适应度函数 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]