扫描电子显微镜图像的特征提取方法研究
2018-11-13刘科钟志成
刘科 钟志成



摘 要: 对扫描电子显微镜图像特征进行提取在提高显微镜图像利用率方面具有重要意义。传统提取方法主要通过模板匹配与遗传算法对特征进行提取,存在提取精度低、耗时长的问题。因此,提出基于阈值筛选的扫描电子显微镜图像的特征提取方法。先对影响扫描电子显微镜图像采集的扫描电路和采集部件进行分析,确定采集图像的准确性;根据特征向量相似性预提取图像特征,通过阈值筛选确定特征点位置实现对扫描电子显微镜图像特征的提取;以提取准确度和耗时为指标进行实验验证。实验结果表明,采用改进方法相比传统方法提取准确度提高了32.3%,耗时缩短了69.1 s。
关键词: 扫描电子显微镜; 图像采集; 遗传算法; 特征提取; 阈值筛选; 实验验证
中图分类号: TN911.73?34; TP391.41 文献标识码: A 文章编号: 1004?373X(2018)22?0068?04
Abstract: The feature extraction of scanning electron microscope (SEM) images is of great significance in the aspect of utilization rate improvement of microscope images. In the traditional extraction method, feature extraction is conducted mainly by means of template matching and genetic algorithm, which has the problems of low extraction precision and large time?consumption. Therefore, a feature extraction method based on threshold selection is proposed for SEM images. The scanning circuit and acquisition components that affect the acquisition of SEM images are analyzed to determine the accuracy of collected images. The image features are pre?extracted according to the similarities of feature vectors. The feature point position is determined by means of threshold selection to realize feature extraction of SEM images. A verification experiment was carried out, taking the extraction accuracy and time consumption as indexes. The experimental results show that, in comparison with the traditional method, the improved method can improve the extraction accuracy of 32.3%, and reduce the time consumption of 69.1 s.
Keywords: SEM; image acquisition; genetic algorithm; feature extraction; threshold selection; experiment verification
0 引 言
扫描电子显微图像是扫描电子显微镜依据高能电子及物质之间相互作用时不同信息形成机理,使用不一样的信息检测器收集信息,获得物质的微观形貌,以圖像的形式进行表现[1]。随着近几年计算机领域里图像处理技术的不断发展,显微镜随着图像采集卡和计算机接口技术的发展而被提出并且快速发展。传统显微镜多使用在教学及科研中,从一开始经过在三目头上加装模拟摄像机来达到最初显微镜像可视化的目的,到现在数字摄像机实现显微图像高清晰度输出,使得扫描电子显微镜图像特征提取成为现实。……
