Real-time FPGA-based template matching module for visual inspection application

Template matching enables localization of objects under inspection, but suffers from long computation due to high computation complexity. In this work, a real-time FPGA-based template matching module which accelerates time consuming normalized cross-correlation (NCC) template matching was presented. To improve NCC computation speed, we simplified the original NCC algorithm and designed a pipelined parallel processing circuit architecture. Experimental results showed that our FPGA module accelerates NCC speed up to 80 times faster than PC performance. The real-time template matching module has been integrated into an LED die inspection system to localize LED dies. This real-time template matching module can be applied to LED, PV, and semiconductor inspection and manufacturing applications. Furthermore, this module can also be applied to vision applications for either service or industrial robots.

[1]  Xiaotao Wang,et al.  FPGA Based Parallel Architectures for Normalized Cross-Correlation , 2009, 2009 First International Conference on Information Science and Engineering.

[2]  Hyunsu Cho,et al.  Pattern classification of solder joint images using a correlation neural network , 1996 .

[3]  Du-Ming Tsai,et al.  Rotation-invariant pattern matching using wavelet decomposition , 2002, Pattern Recognit. Lett..

[4]  Akira Tomono,et al.  Affine-Invariant Recognition of Gray-Scale Characters Using Global Affine Transformation Correlation , 2001, IEEE Trans. Pattern Anal. Mach. Intell..

[5]  F Kvasnik,et al.  Wafer fault measurement by coherent optical processor. , 1994, Applied optics.

[6]  N. Gupta.  A VLSI Architecture for Image Registration in Real Time , 2007, IEEE Transactions on Very Large Scale Integration (VLSI) Systems.

[7]  Luigi di Stefano,et al.  An efficient algorithm for exhaustive template matching based on normalized cross correlation , 2003, 12th International Conference on Image Analysis and Processing, 2003.Proceedings..