Computer-Aided Intelligent Recognition Techniques and Applications

by Sarfraz, Muhammad

• ISBN13: 9780470094143
• ISBN10: 0470094141
• Edition: 1st
• Format: Hardcover
• Copyright: 2005-06-24
• Publisher: WILEY

Summary

Intelligent recognition methods have recently proven to be indispensable in a variety of modern industries, including computer vision, robotics, medical imaging, visualization and the media. Furthermore, they play a critical role in the traditional fields such as character recognition, natural language processing and personal identification. This cutting-edge book draws together the latest findings of industry experts and researchers from around the globe. It is a timely guide for all those require comprehensive, state-of-the-art advice on the present status and future potential of intelligent recognition technology. Computer-Aided Intelligent Recognition Techniques and Applications: Provides the user community with systems and tools for application in a very wide range of areas, including: IT, education, security, banking, police, postal services, manufacturing, mining, medicine, multimedia, entertainment, communications, data visualization, knowledge extraction, pattern classification and virtual reality. Disseminates information in a plethora of disciplines, for example pattern recognition, AI, image processing, computer vision and graphics, neural networks, cryptography, fuzzy logic, databases, evolutionary algorithms, shape and numerical analysis. Illustrates all theory with real-world examples and case studies. This valuable resource is essential reading for computer scientists, engineers, and consultants requiring up-to-date comprehensive guidance on the latest developments in computer-aided intelligent recognition techniques and applications. Its detailed, practical approach will be of interest to senior undergraduate and graduate students as well as researchers and industry experts in the field of intelligent recognition.

Author Biography

Dr. Sarfraz received his B.S.c.in mathematics and statistics from Govt. College, Lahore, Pakistan in 1977, an M.S.c in mathematics and an M.S.c in Numerical Analysis from the University of Brunel. He then went on to complete his P.h.D in Computer Aided Geometric Design also at The University of Brunel.

Dr. Sarfraz is currently Associate Professor at King Fahd University of Petroleum and Minerals, Saudi Arabia. He is a member of many professional bodies including the IEEE and the IEEE Computer Society. He has published three books and written numerous journal papers and reports.

Table of Contents

Preface	xvii
List of Contributors	xix
On Offline Arabic Character Recognition	1	(18)
Muhammad Sarfraz
Abdulmalek Zidouri
Syed Nazim Nawaz
Introduction	1	(3)
Structure of the Proposed OCR System	4	(2)
Preprocessing	6	(1)
Segmentation	7	(3)
Line Segmentation and Zoning	8	(1)
Word Segmentation	8	(1)
Segmentation of Words into Individual Characters	9	(1)
Feature Extraction	10	(1)
Recognition Strategy	11	(4)
Recognition Using the Syntactic Approach	12	(1)
Recognition Using the Neural Network Approach	13	(2)
Experimental Results and Analysis	15	(2)
System Training	15	(1)
Experimental Set-up	15	(1)
Results Achieved	15	(2)
Conclusion	17	(2)
Acknowledgement	17	(1)
References	17	(2)
License Plate Recognition System: Saudi Arabian Case	19	(14)
Muhammad Sarfraz
Mohammed Jameel Ahmed
Introduction	19	(1)
Structure of a Typical LPR System	20	(1)
Image Acquisition	21	(1)
License Plate Extraction	21	(5)
Vertical Edge Detection	23	(1)
Filtering	23	(1)
Vertical Edge Matching	24	(2)
Black to White Ratio and Plate Extraction	26	(1)
License Plate Segmentation	26	(1)
Character Recognition	26	(1)
Normalization	26	(1)
Template Matching	27	(1)
Experimental Analysis and Results	27	(5)
Conclusion	32	(1)
References	32	(1)
Algorithms for Extracting Textual Characters in Color Video	33	(18)
Edward K. Wong
Minya Chen
Introduction	33	(1)
Prior and Related Work	34	(1)
Our New Text Extraction Algorithm	35	(5)
Step 1: Identify Potential Text Line Segments	36	(2)
Step 2: Text Block Detection	38	(1)
Step 3: Text Block Filtering	38	(1)
Step 4: Boundary Adjustments	38	(1)
Step 5: Bicolor Clustering	38	(1)
Step 6: Artifact Filtering	39	(1)
Step 7: Contour Smoothing	39	(1)
Experimental Results and Performance	40	(7)
Using Multiframe Edge Information to Improve Precision	47	(1)
Step 3(b): Text Block Filtering Based on Multiframe Edge Strength	47	(1)
Discussion and Concluding Remarks	47	(4)
References	48	(3)
Separation of Handwritten Touching Digits: A Multiagents Approach	51	(16)
Ashraf Elnagar
Reda Al-Hajj
Introduction	51	(1)
Previous Work	52	(4)
Digitizing and Processing	56	(1)
Segmentation Algorithm	56	(5)
Extraction of Feature Points	56	(1)
The Employed Agents	57	(4)
Experimental Results	61	(4)
Conclusions and Future Work	65	(2)
References	65	(2)
Prototype-based Handwriting Recognition Using Shape and Execution Prototypes	67	(22)
Miguel L. Bote-Lorenzo
Eduardo Gomez-Sanchez
Yannis A. Dimitriadis
Introduction	67	(1)
A Handwriting Generation Process Model	68	(2)
The First Stages of the Handwriting Recognition System	70	(3)
Character Segmentation	70	(1)
Feature Extraction	71	(2)
The Execution of the Prototype Extraction Method	73	(9)
Grouping Training Samples	74	(1)
Refinement of the Prototypes	75	(1)
Experimental Evaluation of the Prototype Extraction Method	76	(6)
Prototype-based Classification	82	(5)
The Prototype-based Classifier Architecture	82	(1)
Experimental Evaluation of the Prototype Initialization	83	(1)
Prototype Pruning to Increase Knowledge Condensation	84	(1)
Discussion and Comparison to Related Work	85	(2)
Conclusions	87	(2)
Acknowledgement	87	(1)
References	87	(2)
Logo Detection in Document Images with Complex Backgrounds	89	(10)
Tuan D. Pham
Jinsong Yang
Introduction	89	(1)
Detection of Potential Logos	90	(1)
Verification of Potential Logos	91	(2)
Feature Extraction by Geostatistics	91	(2)
Neural Network-based Classifier	93	(1)
Experimental Results	93	(4)
Conclusions	97	(2)
References	97	(2)
An Intelligent Online Signature Verification System	99	(20)
Bin Li
David Zhang
Introduction	99	(3)
Process and System	100	(1)
The Evaluation of an Online Signature Verification System	101	(1)
Literature Overview	102	(5)
Conventional Mathematical Approaches	102	(2)
Dynamic Programming Approach	104	(1)
Hidden Markov Model-Based Methods	105	(1)
The Artificial Neural Networks Approach	106	(1)
Signature Verification Product Market Survey	106	(1)
A Typical Online Signature Verification System	107	(6)
Data Acquisition	107	(3)
Feature Extraction	110	(1)
Feature Matching	111	(1)
Verification	112	(1)
Proposed Online Signature Verification Applications	113	(3)
System Password Authentication	113	(1)
Internet E-commerce Application	114	(2)
Conclusions	116	(3)
References	116	(3)
Hybrid Fingerprint Recognition using Minutiae and Shape	119	(12)
Asker Bazen
Raymond Veldhuis
Sabih Gerez
Introduction	119	(1)
Elastic Deformations	120	(2)
Elastic Minutiae Matching	122	(4)
Local Minutiae Matching	122	(1)
Global Minutiae Matching	123	(3)
Shape Matching	126	(1)
Results	126	(3)
Conclusions	129	(2)
Acknowledgement	129	(1)
References	129	(2)
Personal Authentication Using the Fusion of Multiple Palm-print Features	131	(14)
Chin-Chuan Han
Introduction	131	(2)
Preprocessing	133	(2)
Step 1: Image Thresholding	134	(1)
Step 2: Border Tracing	134	(1)
Step 3: Wavelet-based Segmentation	135	(1)
Step 4: Region of Interest (ROI) Generation	135	(1)
Feature Extraction	135	(1)
Enrollment and Verification Processes	136	(4)
Multitemplate Matching Approach	136	(1)
Multimodal Authentication with PBF-based Fusion	137	(2)
Adaptive Thresholding	139	(1)
Experimental Results	140	(2)
Experimental Environment	140	(1)
Verification Using a Template Matching Algorithm	140	(1)
Verification Using PBF-based Fusion	141	(1)
Conclusions	142	(3)
References	142	(3)
Intelligent Iris Recognition Using Neural Networks	145	(24)
Muhammad Sarfraz
Mohamed Deriche
Muhammad Moinuddin
Syed Saad Azhar Ali
Introduction	145	(2)
Literature Review	147	(1)
Some Groundbreaking Techniques	148	(6)
Daugman's Method	149	(1)
Boles's Method	150	(1)
Method of Dyadic Wavelet Transform Zero Crossing	151	(3)
Neural Networks	154	(4)
Multilayer Feed-forward Neural Networks (MFNNs)	154	(2)
Radial Basis Function Neural Networks (RBFNNs)	156	(2)
Proposed Method	158	(4)
Localizing the Iris	158	(1)
Finding the Contour	159	(1)
Feature Extraction	159	(3)
Iris Pattern Recognition	162	(1)
Experimental Results	162	(2)
Results for an MFNN	162	(1)
Results for an RBFNN	162	(2)
Graphic User Interface (GUI)	164	(2)
Concluding Remarks	166	(3)
References	166	(3)
Pose-invariant Face Recognition Using Subspace Techniques	169	(32)
Mohamed Deriche
Mohammed Aleemuddin
Introduction	169	(3)
Background	170	(1)
The Problem of Pose	171	(1)
Review of Biometric Systems	172	(6)
Summary of the Performance of Different Biometrics	173	(4)
Selecting the Right Biometric Technology	177	(1)
Multimodal Biometric Systems	177	(1)
Face Recognition Algorithms	178	(5)
Template-based Face Recognition	180	(1)
Appearance-based Face Recognition	180	(1)
Model-based Face Recognition	181	(2)
Linear Subspace Techniques	183	(8)
Principal Component Analysis	184	(1)
Linear Discriminant Analysis	185	(4)
Independent Component Analysis	189	(2)
A Pose-invariant System for Face Recognition	191	(7)
The Proposed Algorithm	192	(1)
Pose Estimation using LDA	192	(2)
Experimental Results for Pose Estimation using LDA and PCA	194	(1)
View-specific Subspace Decomposition	194	(1)
Experiments on the Pose-invariant Face Recognition System	195	(3)
Concluding Remarks	198	(3)
References	198	(3)
Developmental Vision: Adaptive Recognition of Human Faces by Humanoid Robots	201	(40)
Hon-fai Chia
Ming Xie
Introduction	201	(1)
Adaptive Recognition Based on Developmental Learning	202	(3)
Human Psycho-physical Development	202	(1)
Machine (Robot) Psycho-physical Development	203	(1)
Developmental Learning	204	(1)
Questions to Ponder	204	(1)
Developmental Learning of Facial Image Detection	205	(17)
Current Face Detection Techniques	205	(1)
Criteria of Developmental Learning for Facial Image Detection	206	(1)
Neural Networks	206	(1)
Color Space Transformation	207	(3)
RCE Adaptive Segmentation	210	(8)
Implementation	218	(1)
Questions to Ponder	218	(2)
Experimental Results	220	(2)
Developmental Learning of Facial Image Recognition	222	(14)
Wavelets	222	(1)
Wavelet Packet Analysis	223	(1)
Feature Extraction by Wavelet Packet Analysis	224	(3)
Hidden Markov Models	227	(7)
Feature Classification by Hidden Markov Models	234	(1)
Questions to Ponder	234	(1)
Experimental Results	235	(1)
Discussion	236	(5)
References	237	(4)
Empirical Study on Appearance-based Binary Age Classification	241	(16)
Mohammed Yeasin
Rahul Khare
Rajeev Sharma
Introduction	242	(1)
Related Works	243	(1)
Description of the Proposed Age Classification System	243	(4)
Database	244	(1)
Segmentation of the Facial Region	245	(1)
Preprocessing	246	(1)
Feature Extraction	246	(1)
Classifying People into Age Groups	246	(1)
Empirical Analysis	247	(5)
Performance of Data Projection Techniques	247	(1)
The Effect of Preprocessing and Image Resolution	248	(1)
The Effect of Pose Variation	248	(1)
The Effect of Lighting Conditions	249	(1)
The Effect of Occlusion	249	(1)
The Impact of Gender on Age Classification	250	(1)
Classifier Accuracies Across the Age Groups	251	(1)
Conclusions	252	(5)
Appendix A: Data Projection Techniques	253	(1)
Principal Component Analysis (PCA)	253	(1)
Non-Negative Matrix Factorization (NMF)	253	(1)
Appendix B: Fundamentals of Support Vector Machines	253	(1)
Acknowledgement	254	(1)
References	254	(3)
Intelligent Recignition in Medical Pattern Understanding and Cognitive Analysis	257	(18)
Marek R. Ogiela
Ryszard Tadeusiewicz
Introduction	257	(2)
Preliminary Transformation of Medical Images	259	(2)
Structural Descriptions of the Examined Structures	261	(2)
Coronary Vessel Cognitive Analysis	263	(2)
Understanding of Lesions in the Urinary Tract	265	(3)
Syntactic Methods Supporting Diagnosis of Pancreatitis and Pancreatic Neoplasm	268	(3)
Semantic Analysis of Spinal Cord NMR Images	271	(1)
Conclusions	272	(3)
References	273	(2)
The Roadmap for Recognizing Regions of Interest in Medical Images	275	(22)
Sabah M.A. Mohammed
Jinan A.W. Fiaidhi
Lei Yang
Introduction	275	(1)
Convolutional Primitive Segmentation	276	(4)
Thresholding Primitive Segmentation	280	(1)
Morphological Primitive Segmentation	280	(1)
Erosion	280	(1)
Dilation	281	(1)
Opening and Closing	281	(1)
Hybridizing the Primitive Segmentation Operators	281	(4)
Region Identification Based on Fuzzy Logic	285	(8)
Experimental Results	289	(4)
Conclusions	293	(4)
References	294	(3)
Feature Extraction and Compression with Discriminative and Nonlinear Classifiers and Applications in Speech Recognition	297	(22)
Xuechuan Wang
Introduction	298	(2)
Standard Feature Extraction Methods	300	(1)
Linear Discriminant Analysis	300	(1)
Principal Component Analysis	301	(1)
The Minimum Classification Error Training Algorithm	301	(3)
Derivation of the MCE Criterion	301	(2)
Using MCE Training Algorithms for Dimensionality Reduction	303	(1)
Support Vector Machines	304	(3)
Constructing an SVM	304	(2)
Multiclass SVM Classifiers	306	(1)
Feature Extraction and Compression with MCE and SVM	307	(1)
The Generalized MCE Training Algorithm	307	(1)
Reduced-dimensional SVM	307	(1)
Classification Experiments	308	(8)
Deterding Database Experiments	309	(2)
TIMIT Database Experiments	311	(5)
Conclusions	316	(3)
References	317	(2)
Improving Mine Recognition through Processing and Dempster--Shafer Fusion of Multisensor Data	319	(26)
Nada Milisavljevic
Isabelle Bloch
Introduction	319	(1)
Data Presentation and Preprocessing	320	(4)
IR Data	320	(1)
GPR Data	321	(2)
MD Data	323	(1)
Region Selection	324	(3)
IR Regions	324	(1)
GPR Regions	325	(1)
MD Regions	325	(2)
Choice of Measures and Their Extraction	327	(6)
IR Measures	327	(1)
GPR Measures	328	(5)
MD Measures	333	(1)
Modeling of Measures in Terms of Belief Functions and Their Discounting	333	(5)
IR Measures	334	(1)
GPR A-scan and Preprocessed C-scan Measures	335	(1)
GPR B-scan (Hyperbola) Measures	336	(1)
MD Measures	336	(1)
Discounting Factors	337	(1)
Region Association, Combination of Measures and Decision	338	(3)
Region Association	338	(1)
Combination of Masses	339	(1)
Decision	340	(1)
Results	341	(1)
Conclusion	341	(4)
Acknowledgement	342	(1)
References	342	(3)
Fast Object Recognition Using Dynamic Programming from a Combination of Salient Line Groups	345	(18)
Dong Joong Kang
Jong Eun Ha
In So Kweon
Introduction	345	(1)
Previous Research	346	(1)
Junction Extraction	347	(1)
Energy Model for the Junction Groups	348	(1)
Energy Minimization	349	(2)
Collinear Criterion of Lines	351	(2)
Parallelism	351	(1)
Normal Distance	352	(1)
Energy Model for the Junction Groups	353	(1)
Experiments	354	(6)
Line Group Extraction	355	(4)
Collinearity Tests for Random Lines	359	(1)
Conclusions	360	(3)
References	360	(3)
Holo-Extraction and Intelligent Recognition of Digital Curves Scanned from Paper Drawings	363	(26)
Ke-Zhang Chen
Xi-Wen Zhang
Zong-Ying Ou
Xin-An Feng
Introduction	363	(1)
Review of Current Vectorization Methods	364	(3)
The Hough Transform-based Method	365	(1)
Thinning-based Methods	365	(1)
The Contour-based Method	365	(1)
The Sparse Pixel-based Method	365	(1)
Mesh Pattern-based Methods	365	(1)
Black Pixel Region-based Methods	366	(1)
The Requirements for Holo-extraction of Information	366	(1)
Construction of the Networks of SCRs	367	(6)
Generating Adjacency Graphs of Runs	367	(1)
Constructing Single Closed Regions (SCRs)	368	(2)
Building Adjacency Graphs of SCRs	370	(1)
Constructing the Networks of SCRs	371	(2)
A Bridge from the Raster Image to Understanding and 3D Reconstruction	373	(6)
Separating the Annotations and the Outlines of Projections of Parts	373	(2)
Vectorization	375	(2)
3D Reconstruction	377	(2)
Classification of Digital Curves	379	(3)
Extracting the Representative Points of Digital Curves	379	(1)
Fitting a Straight line to the Set of Points	380	(1)
Fitting a Circular Arc to the Set of Points	381	(1)
Determining the Type	382	(1)
Decomposition of Combined Lines Using Genetic Algorithms	382	(4)
Initial Population	382	(2)
Fitness Function	384	(1)
Crossover	384	(1)
Mutation	384	(1)
Selection	385	(1)
Convergence and Control Parameters	385	(1)
Determination of the Relationships Between the Segments	385	(1)
Software Prototype	386	(1)
Conclusions	386	(3)
References	387	(2)
Topological Segmentation and Smoothing of Discrete Curve Skeletons	389	(22)
Wenjie Xie
Renato Perucchio
David Sedmera
Robert P. Thompson
Introduction	389	(1)
Basic Definitions	390	(2)
Topological Segmentation	392	(5)
Component Counting and Labeling	392	(1)
Classification of Skeleton Voxels	392	(1)
Local Junction Classification	393	(1)
Thick Junction Resolution	394	(1)
Branch Formation	395	(2)
Branch Filtering	397	(3)
Branch Classification	397	(1)
Noise Segment Removal	398	(2)
Branch Smoothing	400	(3)
Polynomial Branch Representation	400	(1)
Augmented Merit Functions	401	(2)
Results	403	(4)
Discussion	407	(4)
Acknowledgement	408	(1)
References	408	(3)
Applications of Clifford-valued Neural Networks to Pattern Classification and Pose Estimation	411	(28)
Eduardo Bayro-Corrochano
Nancy Arana-Daniel
Introduction	411	(1)
Geometric Algebra: An Outline	412	(4)
Basic Definitions	412	(1)
The Geometric Algebra of nD Space	413	(1)
The Geometric Algebra of 3D Space	414	(1)
Rotors	414	(1)
Conformal Geometric Algebra	415	(1)
Real-valued Neural Networks	416	(1)
Complex MLP and Quaternionic MLP	417	(1)
Clifford-valued Feed-forward Neural Networks	418	(3)
The Activation Function	418	(1)
The Geometric Neuron	418	(1)
Feed-forward Clifford-valued Neural Networks	419	(2)
Learning Rule	421	(1)
Multidimensional Back-propagation Training Rule	421	(1)
Geometric Learning Using Genetic Algorithms	422	(1)
Support Vector Machines in the Geometric Algebra Framework	422	(4)
Support Vector Machines	422	(1)
Support Multivector Machines	423	(1)
Generating SMVMs with Different Kernels	424	(1)
Design of Kernels Involving the Clifford Geometric Product for Nonlinear Support Multivector Machines	424	(1)
Design of Kernels Involving the Conformal Neuron	425	(1)
Clifford Moments for 2D Pattern Classification	426	(2)
Experimental Analysis	428	(8)
Test of the Clifford-valued MLP for the XOR Problem	428	(1)
Classification of 2D Patterns in Real Images	429	(2)
Estimation of 3D Pose	431	(1)
Performance of SMVMs Using Kernels Involving the Clifford Product	432	(2)
An SMVM Using Clustering Hyperspheres	434	(2)
Conclusions	436	(3)
References	436	(3)
Intelligent Recognition: Components of the Short-time Fourier Transform vs. Conventional Approaches	439	(14)
Leonid Gelman
Mike Sanderson
Chris Thompson
Paul Anuzis
Introduction	440	(1)
Theoretical Analysis	440	(7)
Application	447	(1)
Conclusions	448	(5)
Acknowledgement	450	(1)
References	450	(3)
Conceptual Data Classification: Application for Knowledge Extraction	453	(16)
Ahmed Hasnah
Ali Jaoua
Jihad Jaam
Introduction	453	(1)
Mathematical Foundations	454	(5)
Definition of a Binary Context	454	(1)
Definition of a Formal Concept	455	(1)
Galois Connection	456	(1)
Optimal Concept or Rectangle	457	(2)
An Approximate Algorithm for Minimal Coverage of a Binary Context	459	(3)
Conceptual Knowledge Extraction from Data	462	(3)
Supervised Learning by Associating Rules to Optimal Concepts	462	(1)
Automatic Entity Extraction from an Instance of a Relational Database	463	(2)
Software Architecture Development	465	(1)
Automatic User Classification in the Network	465	(1)
Conclusion	465	(4)
References	466	(3)
Cryptographic Communications With Chaotic Semiconductor Lasers	469	(16)
Andres Iglesias
Introduction	470	(2)
Semiconductor Lasers with Optical Feedback	472	(6)
Step 1: Choice of the Laser	472	(1)
Step 2: Determination of the Laser Equations and Parameters	473	(2)
Step 3: Choice of Some Accessible Parameter for Chaoticity	475	(1)
Step 4: Synchronization of the Chaotic Transmitter and Receiver Systems	476	(2)
Applications to Cryptographic Communications	478	(3)
Chaotic Masking	478	(1)
Chaotic Switching	479	(2)
Conclusions	481	(4)
References	482	(3)
Index	485