Title :
3D Reconstruction Model of Metal Fracture SEM Image and Realization
Author :
Kang, Ge-Wen ; Ren, Wen-Wei ; Chen, Heng-Li
Author_Institution :
Univ. of Electron. Sci. & Technol. of China, Chengdu
Abstract :
In this paper, a new neural-based reflectance model of the scanning electron microscope (SEM) image is proposed. The idea of this method is to adapt imaging mechanism of real secondary electron SEM. The conventional cost function is modified, which smoothness constraint is replaced by fractal constraint. The proposed method overcomes the disadvantage of tradition shape from shading algorithm. The detail features of metal fracture surface are reconstructed better. The contrastive experiments are performed by the conventional SFS model and our new model. The maximum brightness error of SEM image is decreased from 23.84% to 6.28% and the average brightness error is decreased from 4.46% to 1.07%. The experimental results show that the algorithm is very efficiency and accuracy for single metal fracture SEM image of the unknown light source direction.
Keywords :
fracture; image reconstruction; neural nets; scanning electron microscopy; structural engineering computing; 3D image reconstruction model; SEM; conventional cost function; fractal constraint; maximum brightness error; metal fracture; neural-based reflectance model; scanning electron microscope; smoothness constraint; unknown light source direction; Brightness; Cost function; Fractals; Image reconstruction; Light sources; Reflectivity; Scanning electron microscopy; Shape; Surface cracks; Surface reconstruction; 3D reconstruction; Fractal; Metal fracture; Neural network; SEM;
Conference_Titel :
Machine Learning and Cybernetics, 2007 International Conference on
Conference_Location :
Hong Kong
Print_ISBN :
978-1-4244-0973-0
Electronic_ISBN :
978-1-4244-0973-0
DOI :
10.1109/ICMLC.2007.4370431