本文主要研究不同方位之入射光下的物件成像特徵於微組裝系統的相互關係，在應用影像回授進行微組裝時，由於工作空間有限，需考慮陰影在組裝過程中的影響，進而調整光源的入射角度。此外在組裝自動化的過程中，會因為系統的許多不確定變因造成微物件的定位出現誤差，導致組裝所需之影像特徵消失，而發生微組裝系統無法判斷物件位置造成自動化組裝失敗。因此本文利用類神經理論以及影像遮罩處理，讓微組裝系統學習如何的光場分佈，能使微組裝過程中的影像特徵得以清楚呈現，使系統在自動組裝物件的過程中，得以自行選擇最適當的光場來完成組裝的任務。
本研究微組裝系統的發展，以工業電腦並搭配LabVIEW 構築自動微組裝系統的人機介面，並合併MATLAB 進行類神經演算法發展軟體，由LabVIEW整合三個子系統（影像系統、微物件安置與定位系統及微組裝系統），且使用MATLAB進行類神經光源學習，學習影像與光場分佈間的關係，使系統得以自行調整適宜的光場進行組裝流程。
本研究成功的安置並定位微物件，並且成功以影像識別自動化組裝，物件直徑80μm，組合件孔徑100μm，其間隙比為0.2。

The present research investigates the relationship between image features and light source with different directions in microassembly system. Because of the workspace limitation, when a system uses image feedback to do the microassembly operation, the effect of the shadow of the object in the microassembly process needs to be considered in the adjustment of the direction of the light. Besides, in the automatic microassembly process, the system uncertainties will cause micro-object positioning error and results in disappearance of image features for microassembly operation. This situation will cause the system failure in obtaining the position information of micro object and automatic microassembly operation. By utilizing the neural network theory and image mask processing, we make the microassembly system to learn that how the light distribution will give clear image features. As a result, the system will choose the most suitable light distribution to achieve automatic microassembly operation.
The research of developing the microassembly system is constructed by utilizing the industrial computer with LabVIEW for man-machine interface software and MATLAB for implementing neural network algorithm. The system through LabVIEW integrates three subsystems including visual-servo system, micro object alignment system, and microassembly system and uses the neural network through MATLAB to learn the relationship between the images and light distributions. Therefore, the system will automatically adjust the light distribution to execute the microassembly process.
The research is achieved to implement alignment and positioning micro object and assembly operation through visual servo automatically. The diameter of peg is 80μm and the diameter of hole is 100μm, i.e., clearance ratio is 0.2.

[1] K.B. Yesin and B.J. Nelson, “A CAD Model Based Tracking System for Visually Guided Microassembly,” Journal of Robotica, Vol. 23, pp. 409-418, 2005.
[2] L. Wang, J.K. Mills, and W.L. Cleghorn, “Automatic Microassembly Using Visual Servo Control,” IEEE Transactions on Electronics Packaging Manufacturing, Vol. 31, pp. 316-325, 2008.
[3] A.N. Das, P. Zhang, W.H. Lee, D. Popa, and H. Stephanou, ”μ3:Multiscale, Deterministic Micro-Nano Assembly System for Construction of On-Wafer Microrobots,” IEEE International Conference on Robotics and Automation, pp. 461-466, 2007.
[4] T. Baidyk, E. Kussul, O.Makeyev, A. Caballero, L. Ruiz, G. Carrera, G. Velasco, “Flat Image Recognition in The Process of Microdevice Assembly,” Pattern Recognition Letters , Vol.25, pp. 107-118, 2004.
[5] 林敬祐, “ 影像伺服微物件姿態調整與定位組裝系統之發展,” 國立成功大學機械工程學系碩士論文, 2009.
[6] M. Probst, M. Fluckiger, S. Pane, O. Ergeneman, Z.n Nagy, B.J. Nelson, “Manufacturing of a Hybrid Acoustic Transmitter Using an Advanced Microassembly System,” IEEE Transactions on Industrial Electronics, Vol. 56, No. 7, pp. 2657-2666, 2009.
[7] H. Hashimoto, T. Kubota, M. Kudou, and F. Harashima, “Self-Organizing Visual Servo System Based on Neural Networks,” IEEE Control Systems, pp. 31-36, 1992.
[8] Y. Iwahori, J. Woodham, M. Ozaki, H. Tanaka, N. Ishii, “Neural Network Based Photometric Stereo with a Nearby Rotational Moving Light Source,” IEICE Trans. Inf. & Syst., Vol. E80-D, No. 9, pp. 948-957, 1997.
[9] Y. S. Hong, H. Jin, C. K. Park, “New Electrosensitive Traffic Light Using Fuzzy Neural Network,” IEEE Transactions on Fuzzy Systems, Vol. 7, No. 6, pp. 759-767, 1999.
[10] W.S. McCulloch and W.H. Pitts, “A Logical Calculus of the Ideas Immanent in Nervous Activity,” Bulletin of Mathematical Biophysics, Vol. 5, pp.115-133, 1943.
[11] E.B. Baum, and F. Wilczek, “Supervised Learning of Probability Distributions by Neural Networks,” New York: American Institute of Physics, pp. 52-61, 1988.
[12] S. Becker, “An Information-Theoretic Unsupervised Learning Algorithm for Neural Network,” University of Toronto, 1922.
[13] J. Alspector, J.W. Garnett, S. Haber, M.B. Porter, and R. Chu, “A VLSI Efficient Technique for Generating Multiple Uncorrelated Noise Sources and its Application to Stochastic Neural Networks,” IEEE Transactions on Circuits and Systems, pp. 109-123, 1991.
[14] D.E. Rumelhart, J.L. McClelland et al. (eds.), “Parallel Distributed Processing: Explorations in the Microstructure of Cognition,” MIT Press, Cambridge, MA, Vol. 1, pp. 318~362, 1986.
[15] 李士宗, “光散射法於自動微組裝系統之設計實現,” 國立成功大學機械工程學系碩士論文, 2011.
[16] P. Baldi, and K. Hornik, “Neural Networks and Principal Component Analysis: Learning from Examples without Local Minimum,” Neural Network 1, pp. 53-59, 1989.