2D SEM images turn into 3D object models

📝 Abstract

The scanning electron microscopy (SEM) is probably one the most fascinating examination approach that has been used since more than two decades to detailed inspection of micro scale objects. Most of the scanning electron microscopes could only produce 2D images that could not assist operational analysis of microscopic surface properties. Computer vision algorithms combined with very advanced geometry and mathematical approaches turn any SEM into a full 3D measurement device. This work focuses on a methodical literature review for automatic 3D surface reconstruction of scanning electron microscope images.

💡 Analysis

The scanning electron microscopy (SEM) is probably one the most fascinating examination approach that has been used since more than two decades to detailed inspection of micro scale objects. Most of the scanning electron microscopes could only produce 2D images that could not assist operational analysis of microscopic surface properties. Computer vision algorithms combined with very advanced geometry and mathematical approaches turn any SEM into a full 3D measurement device. This work focuses on a methodical literature review for automatic 3D surface reconstruction of scanning electron microscope images.

📄 Content

2D SEM images turn into 3D object models

Wichai Shanklin College of Applied Sciences, PEC University of Technology, India

Abstract—The scanning electron microscopy (SEM) is probably one the most fascinating examination approach that has been used since more than two decades to detailed inspection of micro scale objects. Most of the scanning electron microscopes could only produce 2D images that could not assist operational analysis of microscopic surface properties. Computer vision algorithms combined with very advanced geometry and mathematical approaches turn any SEM into a full 3D measurement device. This work focuses on a methodical literature review for automatic 3D surface reconstruction of scanning electron microscope images.
Keywords—Scanning Electron Microscopy; 3D Surface Modeling; Microscopy Examination;SEM.

I. INTRODUCTION

Image formation in a SEM is similar to conventional light microscopes that a 3D microscopic object is projected into a 2D image plane, therefore, information about the third dimension is missing. Understanding of microscopic surfaces is not an easy task, specially while we only have two dimensional images. Discovering the 3D structure of micro scale surfaces is a fundamental research activity in medicine and biology. Stereovision [1] [2] [3] [7], structure from motion [2] [4] [7], and photometric stereo [5] [6] [7] methods are promising computer vision techniques for surface structure estimation that helps to reconstruct 3D surface structures from only 2D images.
3D surface modeling that can be created using scanning electron microscope absolutely lead to significant understanding of attributes of microscopic surfaces, such as fracture toughness, crack growth and propagation or fracture resistance. 3D surface reconstruction has been mostly proposed as a mean by which a 3D surface model can be reconstructed given a set of 2D images. The Stereovision, structure from motion,
and photometric stereo are those common techniques that have been used around two decades for not only 3D surface reconstruction of SEM images, but also for a variety of non- microscopic images (i.e., images taken by digital cameras). The advantage of structure from motion technique is that the 3D reconstructed surfaces would be much accurate since the method considers different 2D images from different viewpoints, and therefore, it has more 3D depth information rather than the photometric stereo that only considers single perspective images by light variation [8].
Here, I review the progress of the methods in this new and exciting field. I will review several representative 3D surface reconstruction techniques that have been applied for SEM images. In so doing, I have undoubtedly omitted some worthy works, but my hope, however, is that this article offers a representative sampling of the emerging field of 3D reconstruction of SEM images.
II. 3D SURFACE RECONSTRUCTION OF SEM IMAGES

In the following paragraphs, I will first give a literature review on stereovision based technologies applied on 3D SEM surface modeling.
In 2014, Henao et al. [9] developed a 3D SEM surface reconstruction system using optical flow and stereovision methodologies. In 2013, Li et al. [10], proposed a geometric based Moire method together with stereo photography technology to 3D shape measurement of SEM images. The 3D geometric model of SEM images has been established by combing the stereovision algorithm with traditional in-plane SEM Moiré Method (SMM). Two different real world examples have been adapted by the authors to experimentally validate their proposed approach. In 2011, Zhu et al. [11] worked on an integrated technology for accurate 3D metric reconstruction and deformation measurements using single column SEM imaging techniques and stereovision algorithm.

The quality of reconstructed surface models depends on many factors, such as SEM configuration and instrumental variables, working distance, calibration, tilt angles, SEM detectors, and the quality of 2D SEM images. In 2008, Marinello et al. [12] investigated the effect of those parameters in SEM 3D stereo microscopy. In 2003, Cornille et al. [13] proposed a stereovision based approach for both SEM calibration and the 3D surface reconstruction. They also developed a distortion removal method and applied the technique on 3D SEM surface modeling.
In 2002, Pouchou et al. [14] reviewed the state of the art in applying advanced stereovision techniques for 3D shape reconstruction of micro rough surfaces obtained by a SEM. In 1991, Beil et al [15] developed a dynamic programming method for restoring 3D surface model from SEM images using stereoscopy and stereo-intrinsic shape from shading approaches. They also investigated low level image processing algorithms to enhance the reliability of their proposed system.
Using photometric stereo based methods

This content is AI-processed based on ArXiv data.