基於 H-infinity優化控制器參數與神經網路加速還原影像動態模糊

簡易檢索 / 詳目顯示

回結果列表

研究生：	蔣暉岳 Hui-Yueh Chiang
論文名稱：	基於 H-infinity優化控制器參數與神經網路加速還原影像動態模糊 Dynamic Image Deblurring with H-infinity Optimized Controller Parameters and Neural Network Acceleration
指導教授：	顏家鈺 Jia-Yush Yen
口試委員:	陳亮嘉 Liang-Chia Chen 林紀穎 Chi-Ying Lin 王富正 Fu-Cheng Wang 陳永耀 Yung-Yaw Chen
學位類別：	碩士 Master
系所名稱：	工程學院 - 機械工程系 Department of Mechanical Engineering
論文出版年：	2023
畢業學年度：	111
語文別：	中文
論文頁數：	141
中文關鍵詞：	強健控制、粒子群最佳演算法、運動模糊、影像還原
外文關鍵詞：	Robust control, Particle swarm optimization, Motion blur, Image restoration
相關次數：	點閱：259 下載：9
分享至:	分享至facebook 分享至twitter

查詢本校圖書館目錄查詢臺灣博碩士論文知識加值系統勘誤回報

隨著科技的不斷進步，晶圓的尺寸逐漸擴大，並且自動化光學檢測應用也蓬勃發展，這使得晶圓檢測平台的運動控制性面臨更多挑戰。晶圓檢測平台是一套精密複雜的系統，其主要功能是在檢測物體快速移動時，透過拍攝影像並進行動態模糊還原。
在控制精密平台方面，本研究採用了現有的永磁同步線性馬達模型，並使用粒子群最佳化演算法以鑑別出運動平台各軸的受控場之傳遞函數。為了實現控制器的設計，我們採用了常數增益控制策略，我們採用了Tsai提出的鏈散射描述法（Chain scattering description）去設計常數增益的控制器，以達到系統的響應具有良好的穩定性和準確性。
　　影像還原方面，由於光學探頭共用給二維和三維的影像，因此二維影像將遇到色散的問題以及拍攝物體在移動下的動態模糊，本論文對二維影像進行運動影像演算法還原，使用Pan所提出的影像還原方法，將模糊的影像還原成清晰的影像。Pan的影像還原方法，將模糊影像的灰階值梯度分佈，透過影像先驗資訊(Image Prior)來成為影像還原模型的參數，以影像金字塔迭代的方式，找出影像的結構體再透過梯度灰階值分佈來得到模糊核(Blur Kernel)，最後得到清晰影像。傳統的迭代法通常將會耗時許久，所以為了提升還原效率，透過監督式神經網路來提高還原效率。
　　平台在移動下所拍攝的模糊影像，經由影像還原模型處理後，將對結果使用Anish Mittal所提出的無參考圖項指標(Natural image quality evaluator)去做影像還原的客觀評分。

With the increasing size of wafers and the widespread use of automated optical inspection, motion control performance of wafer inspection platforms faces new challenges. These platforms are complex systems that capture images and restore motion blur while objects are moving rapidly.
To achieve precise platform control, this study utilizes particle swarm optimization algorithm to identify the transfer function of each axis of the motion platform based on the existing model of permanent magnet synchronous linear motors. A constant gain control strategy is then employed, using Tsai's chain scattering description method to design the constant gain controller, ensuring stable and accurate system response.
In terms of image restoration, the shared optical probe for 2D and 3D imaging introduces challenges such as chromatic dispersion and dynamic blur caused by object motion. This paper focuses on restoring motion-blurred 2D images using Pan's proposed method. Pan's method utilizes the gray value gradient distribution and image prior information as parameters for the restoration model. By iteratively processing image pyramids, the structural elements of the image are identified, and the blur kernel is obtained based on the gradient gray value distribution, resulting in clear images. To improve restoration efficiency, a supervised neural network is employed.
The image restoration model processes the blurry images captured during platform movement, and the results are objectively evaluated using Anish Mittal's no-reference image quality evaluator for image restoration.

摘要    II
Abstract    III
致謝    IV
目錄    V
圖目錄    IX
表目錄    XIV
第一章    導論    1
一、    研究背景與動機    1
二、    文獻回顧    2
（一）    精密運動平台    2
（二）    動態模糊還原數學模型    5
（三）    粒子群最佳化    7
（四）    影像還原分析指標    10
（五）    監督式神經網路    11
三、    論文架構    12
第二章    系統架構介紹    13
一、    硬體架構    13
（一）    線性馬達    15
（二）    驅動器    16
（三）    伺服控制架構    17
（四）    位置量測系統    21
第三章    平台數學模型建立與識別    23
一、    永磁同步線性馬達數學模型    23
二、    系統識別    29
（一）    X軸識別結果    33
（二）    Y軸識別結果    34
（三）    Z軸識別結果    35
三、    前饋控制增益設計    36
第四章    控制器設計    38
一、    強健控制介紹    38
二、    線性分式變換    39
三、    鏈散射描述    42
四、    頻譜分解    50
（一）    左右互質分解    50
（二）    J頻譜分解    54
五、    H∞控制器設計步驟    56
（一）    控制架構    61
（二）    使用粒子群最佳化尋找最佳參數    66
（三）    積分誤差函數    66
（四）    懲罰函數    68
第五章    影像還原演算法    70
一、    演算法概述    70
二、    動態模糊還原數學模型    71
（一）    影像先驗資訊    71
（二）    Lp正則項    72
（三）    迭代重新加權最小二乘法(IRLS)    72
（四）    衝擊濾波器(Shock Filter)    73
（五）    影像清晰度分析    74
三、    模糊影像還原模型    76
（一）    影像金字塔    77
（二）    還原影像演算法    78
四、    監督式神經網路    80
（一）    卷積層    80
（二）    激勵函數    80
（三）    殘差塊    82
（四）    損失函數    83
（五）    優化器    84
（六）    卷積神經網路    85
五、    小結    87
第六章    實驗結果    88
一、    實驗背景    88
二、    實驗流程    89
三、    控制器參數    90
（一）    X軸調諧結果    90
（二）    Y軸調諧結果    92
（三）    Z軸調諧結果    94
四、    定位控制實驗    95
（一）    X軸    95
（二）    Y軸    98
（三）    Z軸    102
（四）    小結    105
五、    影像演算法    106
（一）    一維相對運動對影像還原結果之影響    106
（二）    色散模糊還原結果    110
（三）    迭代演算法及神經網路    113
（四）    小結    116
第七章    結論與未來展望    118
一、    結論    118
二、    未來展望    119
參考資料    121


                                

[1] J. Pan, R. Liu, Z. Su, and X. Gu, "Kernel estimation from salient structure for robust motion deblurring," Signal Processing: Image Communication, vol. 28, no. 9, pp. 1156-1170, 2013/10/01/ 2013, doi: https://doi.org/10.1016/j.image.2013.05.001.
[2] M.-C. Tsai and D.-W. Gu, Robust and optimal control: a two-port framework approach. Springer Science & Business Media, 2014.
[3] C. Li-Dong, "Objective assessment to restoration of global motion-blurred images using traveling wave equations," in Third International Conference on Image and Graphics (ICIG'04), 18-20 Dec. 2004 2004, pp. 6-9, doi: 10.1109/ICIG.2004.108.
[4] S. Nah, T. Hyun Kim, and K. Mu Lee, "Deep multi-scale convolutional neural network for dynamic scene deblurring," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 3883-3891.
[5] Aerotech. "AGS15000 Direct-Drive Gantry." https://www.aerotech.com/product/gantries/ags15000-direct-drive-gantry/ (accessed.
[6] Aerotech. "PlanarHDX Two Axis, Silicon Carbide, Planar Air-Bearing Stage." https://www.aerotech.com/product/stages-actuators/planarhdx-two-axis-silicon-carbide-planar-air-bearing-stage/ (accessed.
[7] Aerotech. "QNP-XY Series Two-Axis, XY, Piezo Nanopositioning Stages." https://www.aerotech.com/product-catalog/piezo-nanopositioners/qnp-xy-series.aspx (accessed.
[8] R. Chen, Z. Jiao, and L. Yan, "Dual linear motors synchronous control for horizontal axis of far-field target motion simulators," in 2016 IEEE International Conference on Aircraft Utility Systems (AUS), 10-12 Oct. 2016 2016, pp. 810-813, doi: 10.1109/AUS.2016.7748164.
[9] T. Samad, "A Survey on Industry Impact and Challenges Thereof [Technical Activities]," IEEE Control Systems Magazine, vol. 37, no. 1, pp. 17-18, 2017, doi: 10.1109/MCS.2016.2621438.
[10] C. Lee and S. M. Salapaka, "Fast Robust Nanopositioning—A Linear-Matrix-Inequalities-Based Optimal Control Approach," IEEE/ASME Transactions on Mechatronics, vol. 14, no. 4, pp. 414-422, 2009, doi: 10.1109/TMECH.2009.2023903.
[11] Y. Bin, H. Chuxiong, and W. Qingfeng, "Adaptive robust precision motion control of high-speed linear motors with on-line cogging force compensations," in 2007 IEEE/ASME international conference on advanced intelligent mechatronics, 4-7 Sept. 2007 2007, pp. 1-6, doi: 10.1109/AIM.2007.4412558.
[12] M. Tsai and C. Tsai, "A chain scattering-matrix description approach to H/sup infinity / control," IEEE Transactions on Automatic Control, vol. 38, no. 9, pp. 1416-1421, 1993, doi: 10.1109/9.237659.
[13] M. Tanaka, K. Yoneji, and M. Okutomi, "Motion blur parameter identification from a linearly blurred image," in 2007 Digest of Technical Papers International Conference on Consumer Electronics, 2007: IEEE, pp. 1-2.
[14] A. N. Rajagopalan and R. Chellappa, Eds. Motion Deblurring: Algorithms and Systems. Cambridge: Cambridge University Press, 2014.
[15] J. Kennedy and R. Eberhart, "Particle swarm optimization," in Proceedings of ICNN'95-international conference on neural networks, 1995, vol. 4: IEEE, pp. 1942-1948.
[16] Y. Zhang, S. Wang, and G. Ji, "A Comprehensive Survey on Particle Swarm Optimization Algorithm and Its Applications," Mathematical Problems in Engineering, vol. 2015, p. 931256, 2015/10/07 2015, doi: 10.1155/2015/931256.
[17] 朱軒佑, "精密晶圓檢測平台的高度穩定軌跡與速度控制," 2020.
[18] M. Clerc and J. Kennedy, "The particle swarm - explosion, stability, and convergence in a multidimensional complex space," IEEE Transactions on Evolutionary Computation, vol. 6, no. 1, pp. 58-73, 2002, doi: 10.1109/4235.985692.
[19] A. Mittal, R. Soundararajan, and A. C. Bovik, "Making a “completely blind” image quality analyzer," IEEE Signal processing letters, vol. 20, no. 3, pp. 209-212, 2012.
[20] C. Ledig et al., "Photo-realistic single image super-resolution using a generative adversarial network," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 4681-4690.
[21] Aerotech. "Ndrive HLe High-Performance Linear Digital Drive." https://www.aerotech.com/product/motion-control-platforms/ndrive-hle-high-performance-linear-digital-drive/ (accessed.
[22] Aerotech. "Ndrive ML High-Performance Linear Digital Drive." https://www.aerotech.com/product/motion-control-platforms/ndrive-ml-high-performance-linear-digital-drive/ (accessed.
[23] Aerotech. "A3200 help." https://www.aerotech.com/resources-manuals-help-files/ (accessed.
[24] W. C. contributors. "linear motor." https://commons.wikimedia.org/w/index.php?title=File:Linearmotorprinzip.png&oldid=471658136 (accessed.
[25] 鄒應嶼. "交流伺服驅動系統簡介." http://pemclab.cn.nctu.edu.tw/peclub/W3cnotes/cn03/ (accessed.
[26] 余睿洋, "大行程共平面晶圓檢測平台之精密二維定位控制," 機械工程學研究所, 國立臺灣大學, 2017年, 2017.
[27] Aerotech. "Advanced Motion Control Optimizes Mechanical Micro-Drilling." https://www.aerotech.com/advanced-motion-control-optimizes-mechanical-micro-drilling/ (accessed.
[28] L. Liu, S. Tian, D. Xue, T. Zhang, and Y. Chen, "Industrial feedforward control technology: a review," Journal of Intelligent Manufacturing, vol. 30, pp. 2819-2833, 2019.
[29] ControlEngTV. "How to Tune Servo Systems for Optimal Performance Part 2." https://www.youtube.com/watch?v=_mFsrmXVIp8&ab_channel=ControlEngTV (accessed.
[30] 黃偉倫, "精密單層雙軸氣浮運動平台之迫近與定位控制," 臺灣大學機械工程學研究所學位論文, 國立臺灣大學, 2016. [Online]. Available: http://dx.doi.org/10.6342/NTU201600641
[31] M. Shahrokhi and A. Zomorrodi, "1 Comparison of PID Controller Tuning Methods," 2012.
[32] R. K. Sahu, S. Panda, U. K. Rout, and D. K. Sahoo, "Teaching learning based optimization algorithm for automatic generation control of power system using 2-DOF PID controller," International Journal of Electrical Power & Energy Systems, vol. 77, pp. 287-301, 2016/05/01/ 2016, doi: https://doi.org/10.1016/j.ijepes.2015.11.082.
[33] Y. K. Soni and R. Bhatt, "BF-PSO optimized PID controller design using ISE, IAE, IATE and MSE error criteria," International Journal of Advanced Research in Computer Engineering & Technology (IJARCET), vol. 2, no. 7, pp. 2333-2336, 2013.
[34] R. A. Krohling and J. P. Rey, "Design of optimal disturbance rejection PID controllers using genetic algorithms," IEEE Transactions on evolutionary computation, vol. 5, no. 1, pp. 78-82, 2001.
[35] M. H. Marzaki, M. Tajjudin, M. H. F. Rahiman, and R. Adnan, "Performance of FOPI with error filter based on controllers performance criterion (ISE, IAE and ITAE)," in 2015 10th Asian Control Conference (ASCC), 31 May-3 June 2015 2015, pp. 1-6, doi: 10.1109/ASCC.2015.7244851.
[36] 陳潔安, "以人工智慧建立精密平台定位控制器," 機械工程學研究所, 國立臺灣大學, 2019年, 2019.
[37] D. Krishnan and R. Fergus, "Fast image deconvolution using hyper-Laplacian priors," Advances in neural information processing systems, vol. 22, 2009.
[38] Edmund. "USAF 1951 Glass Slide Resolution Targets." https://www.edmundoptics.com/f/1951-usaf-glass-slide-resolution-targets/12064/ (accessed.
[39] N. J. Gregory Hollows. "Resolution." https://www.edmundoptics.com.tw/knowledge-center/application-notes/imaging/resolution/ (accessed.
[40] N. J. Gregory Hollows. "Contrast." https://www.edmundoptics.com.tw/knowledge-center/application-notes/imaging/contrast/ (accessed.
[41] K. He, X. Zhang, S. Ren, and J. Sun, "Deep residual learning for image recognition," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770-778.
[42] IDS. "UI-2230SE-M-GL Rev.3." https://www.1stvision.com/cameras/IDS/dataman/UI-2230SE-M-GL%20Rev.3.pdf (accessed.
[43] IDS. "UI-3080CP Rev. 2." https://en.ids-imaging.com/store/ui-3080cp-rev-2.html (accessed.

簡易檢索 / 詳目顯示

相關論文