最近幾十年來，由於電腦硬體設備與電腦影像技術的演進，使得電腦影像辨識產生了蓬勃的發展，且被廣泛應用在現實生活上，例如:交通號誌影像辨識、人類動作影像辨識、人臉辨識系統等。
本論文對於現實生活交通號誌影像辨識做研究探討，因為它擁有多類別性、影像出現頻率不均、資料變異性高、影像背景複雜度高。這些問題在電腦影像辨識的技術開發上，都是非常具有挑戰性的。
本論文使用了在IJCNN 2011上所提供的一個取自於現實生活的交通號誌影像資料庫，它擁多類別影像、類別資料量分布不均、影像背景複雜度高、氣候變化與光照變化的影響、且資料量達50000筆以上。對這個資料庫，使用了影像特徵值擷取演算法Hog演算法及Haar-like演算法，對這些交通號誌做影像特徵值擷取，然後利用這些影像特徵值資料庫，運用機器學習資料探勘軟體WEKA做演算法分析，當中分析Decision Tree演算法、AdabootM1演算法與SVM演算法後，自行改進演算法發展出Boot-SVM演算法，並對建立多階層的演算架構模型，使Boot-SVM演算法結合多階層的演算架構模型作分析，發現可使準確度會被優化到94.42%。未來可以對較容易被影像分類模型所誤判的類別，做局部影像特徵值擷取優化的改進，或是改進機器學習演算法使其準確度更好。

Advanced computer vision and image processing technologies can greatly improve human living environment, and they have been widely used in various applications such as traffic sign image recognition, human body recognition, and face recognition.
It is difficult to develop automatic recognition systems for real life traffic signs as it involves multiple categories, contains subsets of classes appear very similar to each other, and tends to have large variations within class in visual appearances because of illumination changes, partial occlusions, rotations and weather conditions.
In the study, a real life Germany traffic sign image dataset is adopted for investigating suitable machine learning methods to such a complex problem. There are more than 50,000 traffic sign images collected. The data reflects strong variations in visual appearance due to changes of filming distance, illumination and weather conditions. In addition, partial occlusion and rotation effects make the task even harder. In our study, we built two kinds of dataset by different feature extraction algorithms, including Hog and Haar-like feature descriptors. WEKA (an open source machine learning software in Java) is utilized to test three existing machine learning algorithms, including AdabootM1, Decision tree and Support Vector Machine (SVM). We improved these machine algorithms to produce a new machine learning algorithm Boost-SVM. Moreover, we built a hierarchical learning model. Combining the two strategies mentioned above, we successfully built a more accurate image recognition model (94.42% accuracy for a 2-layer hierarchical learning Boost-SVM).
In future, we plan to enhance local image feature extraction methods for the image classes with high misclassifications and to modify machine learning algorithms for further improvements.

參考文獻
[1] ER Davies ,"Machine Vision: Theory, Algorithms, Practicalities",PP.2-11, 2004
[2] H Farid and EP Simoncelli, “Differentiation of discrete multi-dimensional signals”, IEEE Trans Image Processing, Vol.13(4), PP. 496--508, Apr 2004.
[3] J Canny, J,” A computational approach to edge detection”, IEEE Trans. Pattern Analysis and Machine Intelligence, 8(6):679–698, 1986.
[4] R Deriche, “Using Canny's criteria to derive a recursively implemented optimal edge detector”, Int. J. Computer Vision, Vol. 1, PP. 167–187, April 1987.
[5] OJ Tobias and R Seara,”Image segmentation by histogram thresholding using fuzzy sets”, Image Processing, IEEE Transactions on Date of Publication,Vo1.11(12),PP1457-1465, Dec 2002
[6] PJ Phillips and Hyeonjoon Moon ,et al. , “The FERET evaluation methodology for face-recognition algorithms”, Computer Vision and Pattern Recognition,PP.137-143,Juny1997
[7] R Brunelli and T Poggio ,”Face recognition: features versus templates”, Pattern Analysis and Machine Intelligence, IEEE Transactions on,Vol.15(10),PP.1042-1052, Oct 1993
[8] M Guillaumin and T Mensink , et al. ,” Face recognition from caption-based supervision”, International Journal of Computer Vision,Vol.96,PP 64-82, Number 2012
[9] T Christie and V Slaughter, “Movement contributes to infants’ recognition of the human form”, Cognition,vol 114(3),PP 329-337, March 2010,
[10] F Xie and G Xu , et al. ,”An improved thinning algorithm for human body recognition” Imaging Systems and Techniques, 2009. IST '09. IEEE International Workshop on, PP 416-420, May 2009
[11]. S Ali and M Shah, “Human action recognition in videos using kinematic features and multiple instance learning”, Pattern Analysis and Machine Intelligence, IEEE Transactions on, Vol.32(2), Feb. 2010
[12] X Baro and S Escalera , et al. ,”Traffic sign recognition using evolutionary adaboost detection and forest-ECOC classification” , Intelligent Transportation Systems, IEEE Transactions on , Vol.10(1) ,PP.113-126 , 2009
[13] R Andrzej and L Yongmin ,et al. ,”Real-time traffic sign recognition from video by class-specific discriminative features”, Pattern Recognition ,Vol.43(1) ,PP.416-430, January, 2010
[14] CG Kiran,. and LV. Prabhu , et al. , “Traffic sign detection and pattern recognition using support vector machine” , Advances in Pattern Recognition, 2009. ICAPR '09. Seventh International Conference on ,PP.87-90, 4-6 Feb. 2009
[15] J Cao and L Lou ,et al. ,”Traffic sign recognition using visual feature toward driver assistance system” , Communications in Computer and Information , Vol. 268,PP.15-20 ,2012
[16]. T Zhang and S Liu , et al. ,”Boosted multi-class semi-supervised learning for human action recognition” , Pattern Recognition ,Vol.44(10-11),
PP.2334—2342, October–November 2011
[17] T Mitchell,”Machine Learning, McGraw Hill.”, P.2.,1997,
[18] G James, “Variance and Bias for General Loss Functions”, Machine Learning,Vol.51(2), PP.115-135,May 2003
[19] R Timofte and K. Zimmermann , et al. , “Multi-view traffic sign detection, recognition, and 3D localization” ,Applications of Computer Vision (WACV), 2009 Workshop on ,PP.1-8, 7-8 Dec. 2009
[20] L Cai and C Zhang , et al. , “Traffic sign detection using surround suppression of texture edges and cascade detector”, Intelligent Transportation Systems (ITSC), 2010 13th International IEEE Conference on,PP.1369-1374,19-22 Sept. 2010
[21] A Ruta and F Porikli , et al. ,”In-vehicle camera traffic sign detection and recognition”,Machine Vision and Applications ,Vol.22,PP.359-375 , 2011
[22] S Kumari and SK Mitra ,“Human action recognition using DFT” ,Computer Vision, Pattern Recognition, Image Processing and Graphics (NCVPRIPG), 2011 Third National Conference , PP.239-242 ,15-17 Dec. 2011
[23] L Saldivar-Pinon and MI Chacon-Murguia , et al. , “Human sign recognition for robot manipulation” ,Pattern Recognition Lecture Notes in Computer Science, Volume 7329/2012,PP.107-116, 2012
[24] CC. Tsenga and JC Chen , et al. , “Human action recognition based on graph-embedded spatio-temporal subspace”,Pattern Recognition ,Vol.45(10), PP.3611-3624 ,October 2012
[25] A G Goldstein and ,EJ Mackenber ,et al. , “Recognition of human faces from isolated facial features: A developmental study” Psychonomic Science, Vol. 6(4), PP. 149-150, 1966.
[26] J Wright, and AY Yang, et al. ,”Robust face recognition via sparse representation”,Pattern Analysis and Machine Intelligence, IEEE Transactions on ,Vol.31(2),PP.210-227,Feb. 2009
[27] R Chellappa and CL Wilson, et al. ,”Human and machine recognition of faces: a survey” ,Proceedings of the IEEE ,Vol.83(5),PP.705-741,May 1995
[28] J Alon and V Athitsos , et al. , “A unified framework for gesture recognition and spatiotemporal gesture segmentation “Pattern Analysis and Machine Intelligence, IEEE Transactions on, Vol.31(9),PP.1658-1699, Sept. 2009
[29] J Liu and Z Wang, et al. ,”uWave: Accelerometer-based personalized gesturerecognition and its applications”,Pervasive and Mobile ComputingVol.5(6), PP.657-675 , December 2009
[30] R Poppe ,“A survey on vision-based human action recognition”, Image and Vision Computing ,Vol.28(6), PP.976–990, June 2010,
[31] International Joint Conference on Neural Networks 2011
“http://ijcnn2011.cs.umass.edu/”
[32] NISYS GmbH, Software tools, ” Advanced development and analysis framework”,“http://www.nisys.de/index.php?folder=Software&page=adaf.php”
[33] F Moutarde and A Bargeton, et al. , “Robust onvehicle real-time visual detection of american and european speed limit signs with a modular traffic signs recognition system,” in Proceedings of the IEEE Intelligent Vehicles Symposium, PP. 1122–1126 ,2007
[34]A Broggi, and P Cerri , et al. , “Real time road signs recognition,” in Proceedings of the IEEE Intelligent Vehicles Symposium, 2007,PP. 981–986 ,2007
[35]C G Keller and C Sprunk , et al. ,”Real-time recognition of U.S. speed signs,” in Proceedings of the IEEE Intelligent Vehicles Symposium, 2008, PP. 518–523,2008
[36]S Maldonado Bascon and J. Acevedo Rodrıguez ,et al , “An optimization on pictogram identification for the road-sign recognition task using SVMs,” Computer Vision and Image Understanding, Vol. 114, no. 3, PP. 373–383, 2010.
[37] N Dalal and B Triggs,”Histograms of oriented gradients for human detection “,Computer Vision and Pattern Recognition, 2005. CVPR 2005. IEEE Computer Society Conference on ,Vol.1,PP.886-893,2005
[38]H Farid and E P Simoncelli,”Optimally Rotation-Equivariant Directional Derivative Kernels”, Computer Analysis of Images and Patterns, PP. 207--214, Sep 1997.
[39] P Viola and M Jones ,”Rapid Object Detection using a Boosted Cascade of Simple Features”, Computer Vision and Pattern Recognition, 2001. CVPR 2001. Proceedings of the 2001 IEEE Computer Society,Vol.1.1 ,PP. I-511 - I-518,2001
[40] F Suard and A Rakotomamonjy , et.al. ,“Pedestrian detection using infrared images and histograms of oriented gradients ”Intelligent Vehicles Symposium, 2006 IEEE,PP.206-212,2006
[41] M Bertozzi and A Broggi , et al. , “A pedestrian detector using histograms of oriented gradients and a Support Vector Machine Classifier”,Intelligent Transportation Systems Conference, 2007. ITSC 2007. IEEE,PP.143-148,2007
[42] SK Pavani and D Delgado , et al. ,“Haar-like features with optimally weighted rectangles for rapid object detection” ,Pattern Recognition,Vol.43(1),PP.160-172,2010
[43] X Yuan and X Shan , et al. , “A combined pedestrian detection method based on Haar-Like features and HOG features” ,Intelligent Systems and Applications (ISA), 2011 3rd International Workshop on,PP.1-4,2011
[44] JR Quinlan, “C4.5:Programs for machine learning”,1993
[45] YFreund and R E. Schapire , “Experiments with a new boosting algorithm.”,Proc International Conference on Machine Learning, PP.148-156, 1996
[46] Y Freund and RE Schapire ,"A decision-theoretic generalization of on-Line Learning and an Application to Boosting", 1995
[47] C Cortes and V Vapnik , "Support-Vector Networks", Machine Learning,Vol.20, 1995
[48] Weka 3: Data Mining Software in Java “http://www.cs.waikato.ac.nz/ml/weka/”