Center-environment feature models for materials image segmentation based on machine learning

Abstract

Materials properties depend not only on their compositions but also their microstructures under various processing conditions. So far, the analyses of complex microstructure images rely mostly on human experience, lack of automatic quantitative characterization methods. Machine learning provides an emerging vital tool to identify various complex materials phases in an intelligent manner. In this work, we propose a “center-environment segmentation” (CES) feature model for image segmentation based on machine learning method with environment features and the annotation input of domain knowledge. The CES model introduces the information of neighbourhood as the features of a given pixel, reflecting the relationships between the studied pixel and its surrounding environment. Then, an iterative integrated machine learning method is adopted to train and correct the image segmentation model. The CES model was successfully applied to segment seven different material images with complex texture ranging from steels to woods. The overall performance of the CES method in determining boundary contours is better than many conventional methods in the case study of the segmentation of steel image. This work shows that the iterative introduction of domain knowledge and environment features improve the accuracy of machine learning based image segmentation for various complex materials microstructures.

Semi-supervised learning of images with strong rotational disorder: assembling nanoparticle libraries

“…An oen-encountered scenario is the one in which the individual objects are separable, corresponding to the strong dilution of original solution, [23][24][25][26] rare defects, [27][28][29][30] or easily identiable borders of the objects. [31][32][33][34][35] In these cases, the compound images containing multiple objects can be separated into the patches containing individual objects of interest, albeit at arbitrary orientation, with positional jitter relative to the center of the patch due to the variability of object shapes. Correspondingly, analysis of such data via supervised or unsupervised machine learning methods needs to account for these factors of variability.…”

Correlative, ML based and non destructive 3D analysis of intergranular fatigue cracking in SAC305 Bi solder balls

“…Convolutional neural networks (CNNs) have proven advantageous over manual image analysis, as they are able to build highlevel features from lowlevel ones, providing accurate and e cient image recognition, object detection and image segmentation 24,25 . CNNs have been increasingly applied to medical and biological image analysis 25,27 and more recently, their use for image segmentation in materials science has been on the rise [28][29][30][31][32][33][34][35][36] . In microelectronics failure and reliability analysis, some work has been previously done on Xray tomography data 37,38 .…”

Semi-supervised machine learning workflow for analysis of nanowire morphologies from transmission electron microscopy images

“…[4][5][6][7][8] Recent advances in deep learning have led to a surge of applications in electron microscopy image analysis for a diverse set of tasks in two main categories: discriminative and generative. Discriminative tasks are tasks like morphology/phase classication, [9][10][11][12] particle/defect detection, [13][14][15][16] image quality assessment, [17][18][19] and segmentation [20][21][22][23][24][25] where the objective is quantied by how well the model can distinguish (1) between images or (2) between objects and their background. Generative tasks include microstructure reconstruction, [26][27][28] super resolution, [29][30][31] autofocus 32 and denoising 33,34 where the objective is generation of images with certain desired traits.…”