Extract Nanoporous Gold Ligaments from SEM Images by Combining Fully Convolutional Network and Sobel Operator Edge Detection Algorithm

In order to quantitatively analyze the nanoporous gold (NPG) sponge-like structure, a new method is developed to extracted NPG ligament accurately by combining Fully Convolutional Network (FCN) and Sobel operator edge detection algorithm. The image datasets of NPG morphology are acquired by scanning electron microscope (SEM). The datasets are standardized and transformed into the unified TFRecord data format provide by TensorFlow. FCN was used to identify the ligaments and pores in SEM images preliminary. In addition, we further used Sobel operator edge detection algorithm to determine the ligament boundary precisely. The results show that the accuracy of NPG ligament recognition can reach 95%. This proposed method provides a new technique for analyzing the structural geometry and microstructure characteristics of nanoporous materials.

Keywords: Nanoporous gold, Image Analysis, Fully Convolutional Network, Sobel operator

Suggested Citation: Suggested Citation

Gao, Panpan and Song, Yu and Minhui, Song and Qian, Ping and Su, Yanjing, Extract Nanoporous Gold Ligaments from SEM Images by Combining Fully Convolutional Network and Sobel Operator Edge Detection Algorithm. Available at SSRN: https://ssrn.com/abstract=3979931 or http://dx.doi.org/10.2139/ssrn.3979931