Abstract
Agglomeration is a major challenge in the research of nanodielectrics. Recognition of agglomerates in scanning electron microscopy (SEM) images can effectively support tackle this issue. Motivated by the fast development of image recognition, we propose a new approach for agglomerates recognition in SEM images of nanodielectrics by semantic segmentation algorithm. On the basis of convolutional neural network, pixel blocks classification network and full convolutional segmentation network employed with data augmentation are investigated in this work. Both networks can preliminarily recognize the agglomerates of spherical silica-based blend polyethylene nanocomposites. The average intersection over union (mIoU) of the pixel blocks classification network is 0.837 and it takes 48 seconds to process an image, while the mIoU of the full convolutional segmentation network is 0.777 and it takes 0.059 seconds to process an image.