基于近邻主特征匹配的微纳米尺度位移测量
发布时间：2019-01-09 08:38
【摘要】：提出了一种基于近邻主特征匹配的亚像素级位移测量方法.改进后的近邻主特征提取过程通过修正散度矩阵的构造,最大化相邻位移图像块投影距离,提高了算法的精度和稳定性.通过将训练过程离线化,提出了基于近邻主特征匹配的微纳米位移测量算法,并通过仿真实验验证了图像块在不同大小和位置情况下算法的精度.在高精度纳米平台、高倍显微镜及标准栅格构成的系统中进行了多角度的实验,验证了算法的有效性.算法的测量精度比传统的图像块匹配方法提高了近10倍,特别是算法对于图像块位置和大小的选择鲁棒性更高.
[Abstract]:A subpixel level displacement measurement method based on nearest neighbor principal feature matching is proposed. By modifying the construction of divergence matrix, the improved nearest neighbor main feature extraction process maximizes the projection distance of adjacent displacement image blocks, and improves the accuracy and stability of the algorithm. By de-linearizing the training process, a micro-nanometer displacement measurement algorithm based on nearest neighbor principal feature matching is proposed, and the accuracy of the algorithm is verified by simulation experiments. Experiments on high precision nanoplatform, high power microscope and standard grid system are carried out to verify the effectiveness of the algorithm. The measurement accuracy of the algorithm is nearly 10 times higher than that of the traditional image block matching method, especially the robustness of the algorithm to the selection of the location and size of the image block.
【作者单位】： 东北大学计算机科学与工程学院;常熟理工学院计算机科学与工程学院;
【基金】：国家自然科学基金资助项目(61305025) 江苏省高校自然科学基金资助项目(15KJB520001) 中央高校基本科研业务费专项资金资助项目(N120404008)
【分类号】：TP391.41
[Abstract]:A subpixel level displacement measurement method based on nearest neighbor principal feature matching is proposed. By modifying the construction of divergence matrix, the improved nearest neighbor main feature extraction process maximizes the projection distance of adjacent displacement image blocks, and improves the accuracy and stability of the algorithm. By de-linearizing the training process, a micro-nanometer displacement measurement algorithm based on nearest neighbor principal feature matching is proposed, and the accuracy of the algorithm is verified by simulation experiments. Experiments on high precision nanoplatform, high power microscope and standard grid system are carried out to verify the effectiveness of the algorithm. The measurement accuracy of the algorithm is nearly 10 times higher than that of the traditional image block matching method, especially the robustness of the algorithm to the selection of the location and size of the image block.
【作者单位】： 东北大学计算机科学与工程学院;常熟理工学院计算机科学与工程学院;
【基金】：国家自然科学基金资助项目(61305025) 江苏省高校自然科学基金资助项目(15KJB520001) 中央高校基本科研业务费专项资金资助项目(N120404008)
【分类号】：TP391.41
【相似文献】
相关期刊论文 前1条
1 王正直;辜萍;杨，
本文编号：2405389
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2405389.html
