近年來人機互動以及工業自動化的興起，影像分析的相關演算法顯得日益重要，這些領域的應用大多透過影像當中的資訊與實際物體進行互動，並達到特定的目的。如何精確並快速分析必要的資訊，已成為致力發展的目標。傳統模板匹配Normalized Cross Correlation（NCC）常用來找出兩個影像區域的相似程度，但計算繁複執行時間長，儘管在Fast NCC已大幅改善執行時間，在目標物影像的旋轉與尺寸會與模板不同時，必須對每個可能的變化都逐個比對，執行時間又會大幅上升，RST即是利用旋轉不變的特徵以及尺寸縮放不變的特徵解決此情況，但是執行時間仍然不夠快，本論文提出的模板匹配以RST為基礎進行改良，利用OpenCL及OpenMP進行平行化處理以及演算法優化加速，執行時間可減少97.3%，並且使用嵌入式多核心平台提升成本與設備空間的競爭力，比個人電腦便宜55%且空間節省95%，讓本論文系統在工業自動化上的系統整合具有優勢。

In recent years, the interaction between human-machine interaction and the rise of industrial automation has become increasingly important in image analysis. The application of these fields mostly interacts with the actual objects through images and achieves specific purposes. Traditional template matching Normalized Cross Correlation (NCC) is often used to find the similarity of two image areas. The computational complexity of execution time is long, although Fast NCC has significantly improved execution time. Rotation and size of the target image RST is the use of rotating invariant features and size scaling of the same characteristics to solve this situation. But the implementation of the time is still not fast enough. The template matching proposed in this paper is improved on the basis of RST, and the execution time can be reduced by 97.3% by OpenCL and OpenMP parallel processing, algorithm optimization, and the embedded multi-core platform is used to improve the cost and equipment space. Competitive, 55% cheaper than the personal computer and space savings of 95%, so that this paper system in the industrial automation system integration has advantages.

[1] J.E. Bresenham, "Algorithm for Computer Control of a Digital Plotter," IBM Systems Journal, pp. 25–30, 1965.
[2] J.E. Bresenham, "A Linear Algorithm for Incremental Digital Display of Circular Arcs," Communications of the ACM, pp. 100–106, 1977.
[3] M.S. Choi and W.Y. Kim, "A Novel Two Stage Template Matching Method for Rotation and Illumination Invariance," Pattern Recognition, pp. 119–129, 2002.
[4] H.Y. Kim and S.A. Araújo, "Grayscale Template-Matching Invariant to Rotation, Scale, Translation, Brightness and Contrast," IEEE Pacific-Rim Symposium on Image and Video Technology, Lecture Notes in Computer Science, Vol. 4872, pp. 100–113, 2007.
[5] B. Kisacanin, "Integral Image Optimization for Embedded Vision Applications," IEEE Southwest Symposium Image Analysis and Interpretation, pp. 181-184, 2008.
[6] B.D. Lucas and T. Kanade, "An Iterative Image Registration Technique with an Application to Stereo Vision," International Journal of Computer Vision, pp. 674–679, 1981.
[7] J.P. Lewis, "Fast Normalized Cross-Correlation," Vision Interface, pp. 120–123, 1995.
[8] D.G. Lowe, "Distinctive Image Features from Scale-Invariant Keypoints," International Journal of Computer Vision, pp. 91–110, 2004.
[9] A. Nayyar and V. Puri, "A Review of BeagleBone Smart Boards-A Linux/Android Powered Low Cost Development Platform based on ARM Technology," International Conference on Future Generation Communication and Networking, pp. 55-63, 2015.
[10] P. Poudel and M. Shirvaikar, "Optimization of Computer Vision Algorithms for Real Time Platforms," 42nd IEEE Proc. on South Eastern Symposium on System Theory, pp. 51-55, 2010.
[11] TI Processor SDK Libraries: http://processors.wiki.ti.com/index.php/Processor_SDK_Libraries
[12] TI OpenMP Accelerator Model: http://downloads.ti.com/mctools/esd/docs/openmpacc/index.html
[13] TI Sitara™ ARM® Processors Boot Camp Training Series: https://training.ti.com/sitara%E2%84%A2-arm%C2%AE-processors-boot-camp-training-series
[14] AM57x Sitara™ Processors Training Series: https://training.ti.com/am57x-sitara-processors-training-series
[15] Processor SDK Training Series: https://training.ti.com/processor-sdk-training-series
[16] TI OpenCL: http://downloads.ti.com/mctools/esd/docs/opencl/index.html
[17] TI OpenMP DSP: http://downloads.ti.com/mctools/esd/docs/openmp-dsp/index.html