A Fast Algorithm for Automated Quality Control in Surface Engineering

Ehsan Sheybani, Singli Garcia-Otero, Fedra Adnani, Giti Javidi

DSP Lab, Virginia State University, Petersburg, USA..

DOI: 10.4236/jsemat.2012.22019                

Abstract

In this article an approach to surface image quality assessment for surface pattern and object recognition, classification, and identification has been described. The surface quality assessment finds many industrial applications such as auto-mated, advanced, and autonomous manufacturing processes. Given that in most industrial applications the target surface is an unknown variable, having a tool to measure the quality of the surface in real time has a significant value. To add to the complication, in most industrial applications, the surface (and therefore its image) suffers from several physical phenomena such as noise (of several different kinds), time, phase, and frequency shifts, and other clutter caused by interference and speckles. The proposed tool should also be able to measure the level of deterioration of the surface due to these environmental effects. Therefore, evaluation of quality of a surface is not an easy task. It requires a good understanding of the processing methods used and the types of environmental processes affecting the surface. On the other hand, for a meaningful comparative analysis, some effective parameters have to be chosen and qualitatively and quantitatively measured across different settings and processes affecting the surface. Finally, any algorithm capable of handling these tasks has to be efficient, fast, and simple to qualify for the “real-time” applications.

Keywords

Surface Characterization; Advanced Techniques

1. Introduction

This research aims at providing image processing tools for comparison and assessment of a surface processed under different grades of a manufacturing process all the way up to optimal processing. Ability to measure the surface quality in real-time has many applications in manufacturing automation and product optimization, especially in processes in which the surface qualities such as roughness, grain size, thickness of coding, impurities size and distribution, hardness, and other mechanical properties are of importance. Surface analysis in manufacturing environments requires specialized filtering techniques. Due to the immense effect of rough environment and corruptive parameters, it is often impossible to evaluate the quality of a surface that has undergone various grades of processing. The algorithm presented here is capable of performing this comparison analytically and quantitatively at a low computational cost (real-time) and high efficiency. The parameters used for comparison are the degree of blurriness and the amount of various types of noise associated with the surface image. Based on a heuristic analysis of these parameters the algorithm assesses the surface image and quantifies the quality of the image by characterizing important aspects of human visual quality. Extensive effort has been set forth to obtain real-world noise and blur conditions so that the various test cases presented here could justify the validity of this approach well. The tests performed on the database of images produced valid results for the proposed algorithm consistently. This paper presents the description and validation (along with test results) of the proposed algorithm for surface image quality assessment.

This effort starts by assigning a value to the visual quality of several images from the same object whose surface has undergone different grades of the same process. Ideally, this algorithm is capable of identifying when an object is optimally processed. In doing so, we have identified some of the important parameters that affect the quality of a surface image and ways in which they can be measured quantitatively. The remainder of this paper is organized as follows: Section 2 gives the reader a little background on the requirements and components for this research and some of the challenges in surface image quality assessment; Section 3 discusses the methodology and the innovative techniques used in this research to overcome some of the challenges in surface image quality assessment; Section 4 describes the results obtained from this research; and Section 5 is reserved for conclusive remarks for this research and the direction of the future work in this field.

2. Background

In many advanced and automated industrial and manufacturing processes image processing algorithms are employed to analyze object surfaces and use the information obtained to improve the quality of the product such as finish, texture, color, placement, temperature, cracks, etc. [1-3]. One major disadvantage of these techniques is that collective environmental noise, speckles, and other artifacts from different sensors degrade the surface image quality in tasks such as surface pattern restoration, detection, recognition, and classification [4,5]. While many techniques have been developed to limit the adverse effects of these parameters on image data, many of these methods suffer from a range of issues such as computational involvement of algorithms to suppression of useful information [6,7]. Therefore, there is a great demand for a tool that could perform an accurate surface quality assessment. Since most surface images in industrial environments suffer from clutter, noise, and phase/pixel shifts, we have based this surface quality assessment algorithm on these parameters [8,9].

Furthermore, to achieve a comprehensive model and an algorithm that can handle a wide-range of surface imaging applications, we have proposed adaptive parameters and thresholds that can be adjusted to the type of object surface, manufacturing process, and optimal grade of operation. The noise in this case consists of Gaussian, salt and pepper, and shot noise. The blur consists of different levels of pixel displacements and angular rotations. We have used a variety of the most prevalent techniques recommended in the literature to include noise and blur in the images [10,11]. Wavelet transforms have been employed for analyzing noise in image data as suggested by relevant literature [12-15]. This research requires several hardware and software components that set up the framework for image processing and analysis. These include Matlab analysis and modeling software, a laptop equipped with at least 2 GB of memory to run computationally intensive calculations, and programming (C/C++) environments to run programs and extract data. The digital signal processing algorithms serve to manipulate data so that they would be a good fit for image processing and analysis. In these algorithms a wavelet based approach has been considered for de-noising the image datasets. A detailed description of the technique follows in the next section.

3. Methodology

Figures 1 and 2 below show the placement of the proposed algorithm in a given industrial image processing setup and its functional block diagram. Ideally, the proposed algorithm should be able to look at an image received from the processing sensor of the manufacturing cell (which is cluttered with various noise and blur effects of the environment), compare it to the outcome of various grades of the same cell (A1, A2, and A3) for the same surface (Image 1, Image 2, and Image 3), and decide which one is closer to the optimal threshold set for that particular process. As such, to validate the capability of the proposed algorithm in assessing and comparing the quality of surface images, it must be tested with known images as compared to their cluttered and processed versions.

Figure 1. The general path for an image from processing cell to manufacturing cell, the alternative paths for processing, and the proposed algorithm for surface quality assessment.

Figure 3 shows the block diagram of the validation approach combined with the details of the proposed algorithm shown in Figure 2. Consistency in quality measure figures is the key to the successful validation of this approach and its applicability to a wide range of surface images from different manufacturing processes. The objective is to have one algorithm that works with surface images from different set of surface processes. To show consistency in results, the tests have been repeated with the original image (O), original image plus noise (O + N), original image plus blur (O + B), and original image plus noise and blur (O + N + B) and the results have been shown for all cases in Table 1. As depicted in Figure 2, the proposed algorithm consists of several modules, each unique in its design and purpose, while applicable to a broad array of surface images. These modules are described below:

Conflicts of Interest

The authors declare no conflicts of interest.

References

[1]	J. D. Maxwell, Y. Qu and J. R. Howell, “Full Field Temperature Measurement of Specular Wafers during Rapid Thermal Processing,” IEEE Transactions on Semiconductor Manufacturing, Vol. 20, No. 2, 2007, pp. 137-142. doi:10.1109/TSM.2007.895204
[2]	K.-C. Huang; C.-L. Chang and W.-H. Wu, “Novel Image Polarization Method for Measurement of Lens Decentration,” IEEE Transactions on Instrumentation and Measurement, Vol. 60, No. 5, 2011, pp. 1845-1853. doi:10.1109/TIM.2011.2108070
[3]	Y. Cheng and M. A. Jafari, “Vision-Based Online Process Control in Manufacturing Applications,” IEEE Transactions on Automation Science and Engineering, Vol. 5, No. 1, 2008, pp. 140-153. doi:10.1109/TASE.2007.912058
[4]	D.-M. Tsai and J.-Y. Luo, “Mean Shift-Based Defect Detection in Multicrystalline Solar Wafer Surfaces,” IEEE Transactions on Industrial Informatics, Vol. 7, No. 1, 2011, pp. 125-135. doi:10.1109/TII.2010.2092783
[5]	F. Bernardini, I. M. Martin and H. Rushmeier, “High-Quality Texture Reconstruction from Multiple Scans,” IEEE Transactions on Visualization and Computer Graphics, Vol. 7, No. 4, 2001, pp. 318-332. doi:10.1109/2945.965346
[6]	J. P. W. Pluim, J. B. A. Maintz and M. A. Viergever, “Mutual-Information-Based Registration of Medical Images: A Survey,” IEEE Transactions on Medical Imaging, Vol. 22, No. 8, 2003, pp. 986-1004. doi:10.1109/TMI.2003.815867
[7]	S. Reed, I. T. Ruiz, C. Capus and Y. Petillot, “The Fusion of Large Scale Classified Side-Scan Sonar Image Mosaics,” IEEE Transactions on Image Processing, Vol. 15, No. 7, 2006, pp. 2049-2060. doi:10.1109/TIP.2006.873448
[8]	H. R. Sheikh and A. C. Bovik, “Image Information and Visual Quality,” IEEE Transactions on Image Processing, Vol. 15, No. 2, 2006, pp. 430-444. doi:10.1109/TIP.2005.859378
[9]	Z. Wang and A. C. Bovik, “A Universal Image Quality Index,” IEEE Signal Processing Letters, Vol. 9, No. 3, 2002, pp. 81-84. doi:10.1109/97.995823
[10]	R. Samadani, T. A. Mauer, D. M. Berfanger and J. H. Clark, “Image Thumbnails That Represent Blur and Noise,” IEEE Transactions on Image Processing, Vol. 19, No. 2, 2010, pp. 363-373. doi:10.1109/TIP.2009.2035847
[11]	M. Mignotte, “A Post-Processing Deconvolution Step for Wavelet-Based Image Denoising Methods,” IEEE Signal Processing Letters, Vol. 14, No. 9, 2007, pp. 621-624. doi:10.1109/LSP.2007.896183
[12]	V. Chappelier and C. Guillemot, “Oriented Wavelet Transform for Image Compression and Denoising,” IEEE Transactions on Image Processing, Vol. 15, No. 10, 2006, pp. 2892-2903. doi:10.1109/TIP.2006.877526
[13]	E. J. Balster, Y. F. Zheng and R. L. Ewing, “Feature- Based Wavelet Shrinkage Algorithm for Image Denoising,” IEEE Transactions on Image Processing, Vol. 14, No. 12, 2005, pp. 2024-2039. doi:10.1109/TIP.2005.859385
[14]	E. Sheybani, “Enhancement of Data in Integrated Communications, Navigation, and Surveillance Systems,” Proceedings of NASA/IEEE/AIAA ICNS 2011, Washington DC, 10-12 May 2011.
[15]	N. C. Pramod and G. V. Anand, “Nonlinear Wavelet Denoising for DOA Estimation by MUSIC,” 2004 International Conference on Signal Processing & Communications (SPCOM), Bangalore, 11-14 December 2004, pp. 388-392.
[16]	E. Davies, “Machine Vision: Theory, Algorithms and Practicalities,” Academic Press, New York, 1990, pp. 42- 44.
[17]	R. Gonzalez and R. Woods, “Digital Image Processing,” Addison-Wesley Publishing Com-pany, Reading, 1992, p. 191.
[18]	R. Haralick and L. Shapiro, “Computer and Robot Vision,” Addison-Wesley Publishing Company, Reading, 1992.
[19]	F. Jin, P. Fieguth, L. Winger and E. Jernigan, “Adaptive wiener filtering of noisy images and image sequences,” Proceedings of IEEE ICIP 2003, Barcelona, 14-17 September 2003.
[20]	LaserSoft Imaging, “SilverFast Unsharp Masking,” 2011. http://www.silverfast.com/highlights/
[21]	D. Ziou and S. Tabbone, “Edge Detection Techniques: An Overview,” International Journal of Pattern Recognition and Image Analysis, Vol. 8, No. 4, 1998, pp. 537- 559.
[22]	J. Canny, “A Compu-tational Approach to Edge Detection,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 8, 1986, pp. 679-714. doi:10.1109/TPAMI.1986.4767851
[23]	Wikipedia, “Signal-to-Noise Ratio,” 2011. http://en.wikipedia.org/
[24]	J. Stensby, “Notes on Matrix Norms,” 2011. http://www.ece.uah.edu