TY - GEN
T1 - High-Performance Template Matching-Based Precision Measurement Using Googlenet
AU - Zhao, Chenyang
AU - Li, Bing
N1 - Publisher Copyright:
© 2019 IEEE.
PY - 2019/9
Y1 - 2019/9
N2 - This paper presents a method using GoogLeNet to increase the calculation speed of template matching-based precision measurement. An anisotropic surface pattern named polar microstructure is proposed to provide the global map for the matching. The previous matching is searched on the whole of the global map, this paper uses GoogLeNet to firstly identify the smaller area where the current position belongs, and then reduce the search area for the subsequent template matching so as to improve the calculation efficiency. To identify the unique pattern of polar microstructure, 25 different places from the microstructure surface and total of 1250 images are selected to train the network using GoogLeNet. After 6 epochs and 1404 iterations, the trained network has a 96.42% accuracy for the pattern identification of polar microstructure. The paper proves feasibility for artificial intelligence especially deep learning theory to be promisingly applied in the precision measurement area in the future.
AB - This paper presents a method using GoogLeNet to increase the calculation speed of template matching-based precision measurement. An anisotropic surface pattern named polar microstructure is proposed to provide the global map for the matching. The previous matching is searched on the whole of the global map, this paper uses GoogLeNet to firstly identify the smaller area where the current position belongs, and then reduce the search area for the subsequent template matching so as to improve the calculation efficiency. To identify the unique pattern of polar microstructure, 25 different places from the microstructure surface and total of 1250 images are selected to train the network using GoogLeNet. After 6 epochs and 1404 iterations, the trained network has a 96.42% accuracy for the pattern identification of polar microstructure. The paper proves feasibility for artificial intelligence especially deep learning theory to be promisingly applied in the precision measurement area in the future.
KW - GoogLeNet
KW - convolutional neural network
KW - precision measurement
KW - template matching
UR - https://www.scopus.com/pages/publications/85075713649
U2 - 10.1109/CCHI.2019.8901924
DO - 10.1109/CCHI.2019.8901924
M3 - 会议稿件
AN - SCOPUS:85075713649
T3 - Proceedings - 2nd China Symposium on Cognitive Computing and Hybrid Intelligence, CCHI 2019
SP - 241
EP - 245
BT - Proceedings - 2nd China Symposium on Cognitive Computing and Hybrid Intelligence, CCHI 2019
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2nd China Symposium on Cognitive Computing and Hybrid Intelligence, CCHI 2019
Y2 - 21 September 2019 through 22 September 2019
ER -