A New Slice-based Concept For 3d Paper

Preparing to load PDF file. please wait...

0 of 0
100%
A New Slice-based Concept For 3d Paper

Transcript Of A New Slice-based Concept For 3d Paper

Preferred citation: M. Wiltsche, M. Donoser, W. Bauer and H. Bischof. A new slice-based concept for 3D paper structure analysis applied to spatial coating layer formation. In Advances in Paper Science and Technology, Trans. of the XIIIth Fund. Res. Symp. Cambridge, 2005, (S.J. I’Anson, ed.), pp 853–899, FRC, Manchester, 2018. DOI: 10.15376/frc.2005.2.853.
A NEW SLICE-BASED CONCEPT FOR 3D PAPER STRUCTURE
ANALYSIS APPLIED TO SPATIAL COATING LAYER FORMATION
Mario Wiltsche1, Michael Donoser1,2, Wolfgang Bauer1 and Horst Bischof 2
1Institute for Paper, Pulp and Fiber Technology, Graz University of Technology, Kopernikusgasse 24, 8010 Graz, Austria
2Institute for Computer Graphics and Vision, Graz University of Technology, Inffeldgasse 16, 8010 Graz, Austria

ABSTRACT
This paper introduces a new concept for digitizing the three dimensional paper structure, based on a fully automated microtomy process and light microscopy. The microscope can be moved in all three directions of space with high accuracy in order to be able to digitize large samples with high spatial resolution. All components are controlled by a PC interface which enables an automated digitization process.
The literature concerning 3D analysis of paper structure is reviewed. Non destructive and destructive techniques are compared.
Image analysis algorithms for creation of a detailed digital representation are described. This digital data set is analyzed, to derive characteristics of the paper structure.
As a first example of possible applications the analysis of the 3D coating layer formation is presented. The coating layer is detected by means of image analysis based on a 3D color

13th Fundamental Research Symposium, Cambridge, September 2005

853

M. Wiltsche, M. Donoser, W. Bauer and H. Bischof
segmentation concept. Initial experiments on analyzing coated paper samples prove the applicability of the concept.
The correctness of the implemented sample digitization process and following image analysis was validated.

854

Session 5: Structure and Imaging

A New Slice-based Concept for 3D Paper Structure Analysis

CONTENTS

1 INTRODUCTION

856

2 RELATED RESEARCH

856

2.1 Non destructive methods

857

2.1.1 Confocal laser scanning microscopy – CLSM

857

2.1.2 X-ray microtomography

858

2.1.3 Magnetic resonance imaging – MRI

860

2.1.4 Ultrasonic microscopy

860

2.2 Destructive methods

861

2.2.1 Sheet splitting

861

2.2.2 Serial sectioning

862

3 NOVEL APPROACH

863

3.1 Key aspects

863

3.2 Sample preparation

864

3.3 Design

864

3.4 Operating sequence

866

3.5 Stitching and aligning

867

3.5.1 Stitching

868

3.5.2 Aligning

870

4 IMAGE ANALYSIS

870

4.1 3D coating layer detection

871

4.1.1 3D color segmentation algorithm

872

4.1.2 Multivariate Gaussian distribution

875

4.1.3 Multivariate outlier detection

877

4.1.4 Bayes classification and Bhattacharyya distance

878

4.2 Calculation of coating thickness

881

4.2.1 Center line determination

882

4.2.2 Measurement of coating thickness

883

4.3 3D Visualization

883

5 RESULTS

884

5.1 Validation

884

5.2 Practical application

889

6 CONCLUSION

892

7 OUTLOOK

892

13th Fundamental Research Symposium, Cambridge, September 2005

855

M. Wiltsche, M. Donoser, W. Bauer and H. Bischof
1 INTRODUCTION
Many physical paper properties are strongly influenced by the spatial distribution of the raw materials in the sheet. Paper consists of a network of fibers, fiber fragments, filler particles and sometimes a coating layer.
Each paper grade has to meet particular requirements depending on its intended use. Therefore specialized paper structures should be designed and hence the analysis of the three dimensional paper structure with high spatial resolution is of high industrial interest.
Due to the constantly growing demands for higher and more uniform paper quality, papermakers have to gain more knowledge of the interrelations between paper structure and paper quality characteristics. For instance printability, strength properties and porosity and their relationship to paper structure hold high optimization potential for the papermaking process.
Two major challenges had to be mastered for the intended purpose. Three dimensional digitization at high resolution and extraction of spatial information by means of advanced image analytical methods.
In section 2 the literature concerning 3D analysis of paper structure is briefly reviewed. Non destructive and destructive techniques are compared.
Section 3 introduces a new concept for digitizing the three dimensional paper structure, based on a fully automated microtomy process and light microscopy.
The analysis of the 3D coating layer formation was developed as a first example of a possible application, it is described in section 4. The three dimensional coating layer formation should allow a better insight into several types of mottling problems, i. e. microgloss and backtrap mottle for coated paper grades. These nonuniformities are related to inhomogeneities in the paper and coating structure. Furthermore the modification of the base paper topography due to the coating process can be studied in future.
Validation results, which prove the usefulness of the novel concept, and first applications on coated industrial paper samples are presented in section 5.

2 RELATED RESEARCH
Numerous research activities have been focused on the development of a technology that extracts spatial structures of fibers, filler pigments and coating layer to build a three dimensional model of a paper sheet.
These technologies have to fulfill two conflicting requirements. First a high spatial resolution is required. The diameter of the fibers is about 10 to 30 μm,

856

Session 5: Structure and Imaging

A New Slice-based Concept for 3D Paper Structure Analysis
their length is in the magnitude of millimeters, the particle size of fillers is below 1 μm. Therefore a resolution of at least 1 μm is mandatory to detect all fiber components and important details like fiber morphology parameters accurately. Second a sufficient sample size is needed to get reliable and statistical meaningful results. This requirement originates from the heterogeneity of the paper itself, therefore at least one square centimeter should be analyzed. For every square centimeter of a typical sheet of paper the total number of fibers is 10 000 to 100 000, see Retulainen et al. [1]. Thus such a large number of fibers is required to obtain statistically meaningful results.
In principle two main approaches to analyze three dimensional paper structures are applied: non destructive and destructive methods.

2.1 Non destructive methods
Non destructive methods generally do not cause any changes in the sample structure during the inspection process, there are miscellaneous techniques available for three dimensional, non destructive analysis of materials.
In the following sections the applicability of CLSM – confocal laser scanning microscopy, X-ray microtomography, MRI – magnetic resonance imaging and ultrasonic microscopy for 3D analysis of paper materials is discussed.

2.1.1 Confocal laser scanning microscopy – CLSM
A survey of several techniques for multidimensional microscopy in different applications areas such as biology, medical and material science is given in [2].
CLSM – confocal laser scanning microscopy – has been commonly used in qualitative and quantitative analysis of paper structure in recent years. The major feature of the CLSM is its ability of optical sectioning through a sample and obtaining its 3D structure non destructively without any pretreatment.
Nanko et al. [3] were one of the first who described an application of CLSM on paper materials. The structure of interfiber bonding in hand sheets was investigated. Structures of the bonded zone and bonded fibers were characterized. The influences of beating, couching, pressing and drying were explored.
Moss et al. [4] reported studies of fiber morphology, distribution of fines and surface topography based on CLSM. Three dimensional reconstruction of the paper surface was shown.
Auran et al. [5] used the CLSM to measure the three dimensional surface

13th Fundamental Research Symposium, Cambridge, September 2005

857

M. Wiltsche, M. Donoser, W. Bauer and H. Bischof
structures of wood containing paper samples. In addition the pore size distribution was estimated.
Xu et al. [6, 7] describe a method for analyzing fiber orientation in paper samples based on CLSM. The variation of fiber orientation distribution through the thickness of a paper sample was studied.
The limitation of CLSM is that the signal intensity diminishes rapidly with increasing depth in the sample. The light beam is refracted by the outer fiber layers and particularly by filler pigments. Therefore it is not possible to acquire three dimensional data through the entire thickness of a paper sheet. Another disadvantage is the small sized field of view, which is usually below one square millimeter.

2.1.2 X-ray microtomography
For applications in the field of biology and material sciences X-ray microtomographs with high resolution and small sample dimensions have been developed. These tomographs have also been commonly used in paper research in recent years. In principle there are two different ways of depicting 3D structures based on microtomography – beam absorption or phase contrast.
The most straightforward way to provide the necessary contrast for imaging is by using beam absorption [8, 9]. This is the traditional use of Xray tomography in medical applications. The necessary contrast for imaging between high density bone regions and low density tissue regions based on absorption is sufficient. Paper fibers are made up of carbohydrates and lignin, which contain only the light elements carbon, oxygen and hydrogen. The weakly absorbing fibers make absorptive contrast unsuitable at higher resolutions of about 1 μm.
Goel et al. [10] have shown the application of absorptive microtomography on paper with a commercial CT scanner. The porosity and specific surface area of commercial liquid packaging board samples were analyzed. The results for porosity showed good agreement with results obtained with mercury intrusion porosimetry. The pixel size was about 2.3 μm.
Huang et al. [11] and Ramaswamy et al. [12] analyzed the pore structure of different handsheets with absorptive microtomography. These results help to explain differences in liquid and vapor transport through different directions of the paper structure. The pixel size in this case was about 2 μm.
Gupta et al. [13] analyzed numerous 3D structural characteristics – like porosity, specific surface area, tortuosity – of commercial paper and hand sheets based on absorptive microtomography. Results clearly show significant differences between the samples.

858

Session 5: Structure and Imaging

A New Slice-based Concept for 3D Paper Structure Analysis
Aaltosalmi et al. [14] compared numerical and experimental techniques for analyzing flow in fibrous porous materials. The results computed from the tomographic images obtained with a commercial CT scanner showed agreement with results from other methods.
For materials containing only light elements like paper, phase contrast based X-ray microtomography leads to much better image quality. Phase contrast arises in regions of sharp changes in the refractive index of the sample, such as borders between fibers and pores [15, 16].
Phase contrast can only be obtained when the beam is at least partially spatially coherent. This is the case for X-rays obtained from the third generation of synchrotron sources [15], like at the European Synchrotron Radiation Facility – ESRF – in Grenoble. First demonstrations of the feasibility of phase contrast tomography were reported by Raven et al. [17] and Snigirev et al. [18] in the mid-1990s.
At the ID 22 MICRO-FID beam line at Esrf a lot of research on paper materials has been done. The effective pixel size can be varied between 0.35 and 2.8 μm, with a field of view changing correspondingly between 0.7 and 5.6 mm [15, 19]. Samuelson et al. [8, 9] and Holmstad et al. [20] discuss the application of this technique on paper materials. In order to ensure robust analysis of the 3D fiber network based on this image data, extensive image processing is required to enhance contrast and reduce noise, see Antoine et al. [15, 16].
Holmstad et al. [21] reported successful application of absorption mode imaging with the ESRF synchrotron source. A cubic voxel size of 0.7 μm3 was achieved which allowed quantitative analysis of detailed structural properties.
Holmstad et al. [22] compared 3D data obtained with monochromatic synchrotron radiation in phase contrast mode with a high resolution of about 1 μm with a commercial CT scanner at low resolution of about 5 μm. The results show, that the low resolution images are only suitable for comparative studies. Fundamental research requires high resolution 3D data for accurate measurements. Knackstedt et al. [23] reported similar findings: one data set obtained at a synchrotron facility with high resolution (voxel size 0.35 μm) was compared with another obtained from a commercial CT scanner with low resolution (voxel size 2 μm). These results revealed the particular importance of high resolution images for reliable 3D analysis of paper structure.
The level of detail in the synchrotron images is very impressive. Single fibers can be distinguished easily. This is almost impossible in images which were obtained by absorptive microtomography with commercial CT scanners. In this case spatial resolution and contrast is too low for detailed analysis of the three dimensional fiber network structure.

13th Fundamental Research Symposium, Cambridge, September 2005

859

M. Wiltsche, M. Donoser, W. Bauer and H. Bischof
The challenge of the very high spatial resolution of the synchrotron images is the huge amount of volume data. Therefore just small sample sizes below one square millimeter can be handled by commercial personal computers [20]. A disadvantage of this technique is the requirement of a synchrotron source. Hence this method is far too expensive and therefore not feasible for routine analysis on commercial paper samples.

2.1.3 Magnetic resonance imaging – MRI
MRI – Magnetic resonance imaging – also known as NMR – nuclear magnetic resonance – is well established for visualizing spatial distributions of fluids in all kinds of materials. MRI has been applied to study paper drying and moisture transport in paper [24–28].
Lehmann et al. [29] describe a method based on MRI to analyze the internal structure in filter media. MRI of the original filter media yields inappropriate image quality. Therefore the pore structure in between the fiber network was filled up with water and a contrast agent to get sufficient image quality. The spatial fiber network was calculated from the negativ pattern of the pore structure. A voxel size of 59 μm3 was achieved, which is not suitable for 3D analysis of paper samples.
Due to the low maximum resolution of approximately 10 μm [30] MRI is from today’s point of view not suitably for three dimensional paper structure analysis.

2.1.4 Ultrasonic microscopy
Another method for acquiring three dimensional data is ultrasonic microscopy which is commonly used in medical and industrial applications. Fenster and Downey [31] give a review of the state of technology in 3D ultrasonic imaging. Various applications of ultrasonic microscopy in medical diagnostics, industrial control sensors and nondestructive evaluation are discussed in [32].
Most industrial and medical imaging is done at ultrasonic frequencies from 1 to 10 MHz, sometimes frequencies up to 1000 MHz and more are employed. With these high frequencies spatial resolutions below 1 μm are attainable [32].
Figure 1 shows the achieved image quality for a SC paper sample and a woodfree coated sample. The measurements were performed at a frequency of 50 MHz and an acoustic velocity of 3000 m/s. The axial resolution towards the paper z direction (ZD) was 60 μm.
It showed that it is extremely difficult to reconstruct the three dimensional

860

Session 5: Structure and Imaging

A New Slice-based Concept for 3D Paper Structure Analysis

Figure 1 Images obtained with ultrasonic microscopy. Trials were performed at the Institute of Medical Physics and Biophysics, Medical Faculty of the Martin Luther
University Halle-Wittenberg in Germany.

paper structure from the data of sound reflections. The attenuation of the ultrasonic waves inside the fiber network of the analyzed samples prevents a total penetration through the entire thickness, this damping occurred in particular in the woodfree coated sample.
Further optimization of this technique is essential to make useful 3D analysis of paper materials, especially a higher resolution and a homogenous analysis of the entire sample thickness are required.

2.2 Destructive methods
The most relevant destructive methods with widespread acceptance for analyzing three dimensional paper structures are sheet splitting and serial sectioning.

2.2.1 Sheet splitting Knotzer et al. [33, 34] and Hirn et al. [35, 36] developed a technique in

13th Fundamental Research Symposium, Cambridge, September 2005

861

M. Wiltsche, M. Donoser, W. Bauer and H. Bischof
which samples are split by using a laminating device. With this method it is possible to split a paper sample into very thin layers. An 80 g/m2 sheet is split in 50 to 100 layers. ErkkilÄ, Pakarinen and Odell [37, 38] developed a splitting technique with duct tapes to split paper samples in about 10 layers. Thorpe [39] introduced a similar splitting technique to analyze fiber orientation in copy paper. In that case the sample was split into 13 layers.
In these methods the local fiber orientation is determined by means of image analysis in every single layer. The layers are digitally composed to a 3D model of fiber orientation. Using this method local structures in the fiber orientation are made visible, also the local twosidedness of the fiber orientation, which might cause distortions of the sheet, is determined.
The advantage of these splitting techniques is the sample size in the magnitude of several square centimeters. Hence macro effects like cockling or finger-ridging can be explored.
The disadvantages are first the only quasi three dimensional analysis, i.e. it is possible to analyze only the in-plane fiber orientation in every single layer but not the out-of-plane fiber orientation throughout the paper thickness of felted network structures [40]. Second it is impossible to analyze filler and coating layer distribution based on sheet splitting. Furthermore the influence of the splitting process itself on the result is probably considerable.

2.2.2 Serial sectioning
Methods which are based on serial sectioning by a microtome allow a true three dimensional analysis with sufficient resolution and potentially sufficient sample size. In these methods series of vertical cross sections are cut off an embedded paper sample and each section is digitized. Based on these data the three dimensional, internal structure of the sample can be analyzed.
Yang et al. [41] was one of the first who published a method on the basis of serial sectioning. Cut sections were transferred individually on a microscope slide. Single fibers were tracked through the slice images by hand. Hence, the manual work needed was tremendous for even tiny samples. The state of bonding, bonded area, aspect ratios and moments of inertia of fibers were measured.
Hasuike et al. [42] described a similar method. A 0.2 × 0.2 mm2 sized area of a laboratory sheet containing 138 fibers was analyzed. Out-of-plane fiber orientation as well as fiber-to-fiber bonding state were investigated and 3D visualizations of fiber segments were presented.
A combination of sectioning by a microtome and imaging with CLSM is an alternative technique, which is reported by He et al. [43]. Paper cross sections were imaged with a CLSM inside the embedding mold just below the

862

Session 5: Structure and Imaging
AnalysisResolutionPaper StructurePaperImaging