基于机器的智能人脸识别

出版时间：2010-1  出版社：高等教育出版社  作者：牟登攀  页数：171  
Tag标签：无  

前言

　　We cant solve the problems by using the same kind of thinking we used when we created them.Albert Einstein （1879-1955）　　State-of-the-art machine-based face recognition technology， although boomingsince last decades， is still suffering a lot from critical research challenges， suchas the lack of fundamental intelligence， the difficulties of running completelyautomatically and unsupervisedly without separate training， and the typicalfailures of dealing with free face pose variations， etc. Those limitations greatlyhinder the wide applications it could have had. This book is the first to discussthe general engineering methods of imitating intelligent human brains forvideo-based face recognition. The advances and evidences from the cognitivescience research are introduced in this book， which further strengthen ourthoughts and proposals to achieve such a fundamental intelligence in machinevision.　　Regarding intelligence， we have defined two directions. The first effort isto simulate the ability of self-learning， self-matching and self-updating. Thisside of intelligence can be detailed into the following features： the whole rec-ognition procedure is running in an unsupervised， automatic， non-invasive，and self-updated way. It is important to note that， the fully automatic proce-dure is a generalized face recognition procedure， which includes the task ofenrollment （training） and updating as well. However， those steps are typicallyseparate arld supervised in machine learning， and therefore missing the essen-tials of intelligence.

内容概要

Machine, based Intelligent Face Recognition discusses the general engineering method of imitating intelligent human brains for video-based face recognition in a fundamental way, which is completely unsupervised,automatic, self-learning,self-updated and robust. It also overviews stateof-the-art researchon cognitive-based biometrics and machine-based biometrics, and especially the advances in face recognition.    This book is intended for scientists, researchers, engineers, and students in the field of computer vision, machine intelligence, and particularly of face recognition.

作者简介

　　Dr. Dengpan Mou，Dr.-Ing. and MSc from University of Ulm， Germany，is with Harman/Bedger Automotive Systems GmbH as technology expert，working on video processing， computer vision， machine learning and other research and development topics.

书籍目录

1  Introduction  1.1  Face Recognition--Machine Versus Human  1.2  Proposed Approach  1.3  Prospective Applications    1.3.1  Recognition in the Future Intelligent Home    1.3.2  Automotive    1.3.3  Mobile Phone for Children  1.4  Outline  References2  Fundamentals and Advances in Biometrics and Face Recognition  2.1  Generalized Biometric Recognition  2.2  Cognitive-based Biometric Recognition    2.2.1  Introduction    2.2.2  History of Cognitive Science    2.2.3  Human Brain Structure    2.2.4  Generic Methods in Cognitive Science    2.2.5  Visual Function in Human Brain    2.2.6  General Cognitive-based Object Recognition    2.2.7  Cognitive-based Face Recognition    2.2.8  Inspirations from Cognitive-based Face Recognition  2.3  Machine-based Biometric Recognition    2.3.1  Introduction    2.3.2  Biometric Recognition Tasks    2.3.3  Enrollment--a Special Biometric Procedure    2.3.4  Biometric Methods Overview    2.3.5  Fingerprint Recognition  2.4  Generalized Face Recognition Procedure  2.5  Machine-based Face Detection    2.5.1  Face Detection Categories  2.6  Machine-based Face Tracking  2.7  Machine-based Face Recognition    2.7.1  Overview    2.7.2  Benchmark Studies of Face Recognition    2.7.3  Some General Terms Used in Face Recognition    2.7.4  Recognition Procedures and Methods    2.7.5  Video-based Recognition    2.7.6  Unsupervised and Fully Automatic Approaches  2.8  Summary and Discussions  References3  Combined Face Detection and Tracking Methods  3.1  Introduction  3.2  Image-based Face Detection    3.2.1  Choice of the Detection Algorithm    3.2.2  Overview of the Detection Algorithm    3.2.3  Face Region Estimation    3.2.4  Face Detection Quality  3.3  Temporal-based Face Detection    3.3.1  Overview    3.3.2  Search Region Estimation    3.3.3  Analysis of Temporal Changes  3.4  Summary  3.5  Further Discussions  References4  Automatic Face Recognition  4.1  Overview  4.2  Feature Extraction and Encoding  4.3  Matching/Classification    4.3.1  Image-based Classifier    4.3.2  Adaptive Similarity Threshold    4.3.3  Temporal Filtering  4.4  Combined Same Face Decision Algorithms  4.5  Summary  References5  Unsupervised Face Database Construction  5.1  Introduction  5.2  Backgrounds for Constructing Face Databases    5.2.1  Supervised Learning    5.2.2  Unsupervised Learning    5.2.3  Clustering Analysis  5.3  Database Structure    5.3.1  A Fused Clustering Method    5.3.2  Parameters in the Proposed Structure  5.4  Features of an Optimum Database  References6  State Machine Based Automatic Procedure  6.1  Introduction  6.2  States Explorations7  System Implementation  7.1  Introduction  7.2  Typical Hardware Configuration  7.3  Software Implementation    7.3.1  Overview    7.3.2  Implementation Efforts  7.4  Technology Dependent Parameters  7.5  Summary  References8  Performance Analysis  8.1  Introduction  8.2  Performance of Face Detection  8.3  Performance of Face Recognition  8.4  Performance of Database Construction Algorithms  8.5  Overall Performance of the Whole System    8.5.1  Online Version    8.5.2  Offline Version    8.5.3  Critical Assumptions  8.6  Summary  References9  Conclusions and Future Directions  9.1  Conclusions  9.2  Future DirectionsIndex

章节摘录

　　Although the two parties who hold opposite opinions provide us much in-formation for the face recognition in cortex, further cognitive research ishighly demanding for ending the debates and providing us a clearer answer.However, we, although as researchers in a different field, can now still figureout that, each side has unfortunately one limitation in common: the importanceof frontal lobe is not taken into consideration at all. As mentioned earlier, thefrontal lobe contributes to the high-level analysis such as reasoning, planning,and problem-solving, etc. Frontal lobe is performing the most complicatedtask, being expected to be involved in all brain process, and hence demonstrating the fundamental intelligence. This region should be definitely explored forthe face recognition procedure. In early 1990s, Gross [29] suggests that theface processing cells are extended to the frontal lobe. In reality, this study focuses on finding the visual ability of the frontal lobe rather than the intelligence of it. More recently, Mechelli et al. [30] and Johnson et al. [31] foundout that, the face processing task, although mainly performed in posterior cortical regions such as FFA, OFA and fSTS, is modulated by top-down signalsoriginating in prefrontal cortex. The main purpose in [31] is to point out that,refreshing is a component of more complex modulatory operations such asworking memory and mental imagery. And the refresh related activity maythus be involved in the common activation patterns seen across different cognitive tasks. In summary, most researches are still concentrating on specificand different prospects. However, they convincingly support our fundamentalopinion: the high-level intelligence performed in frontal lobe is crucial for facerecognition.　　It is important to note that, there is a high level research on cognitivebased face recognition, published by P. Sinha et al. [32]. They reported nineteen basic results from the face recognition by humans. Those high-level findings provide us further remarkable insights in designing the correspondingcomputer vision systems. In the following, we reorganize and analyze thenineteen results for the readers for a better understanding.~　Results for the image-based face recognition algorithms. The findings　　from Result 3 to Result 12 include the following information that contribute and/or influence the face recognition performance: high-frequency information, facial features, holistic processing, face aspect ratios （width and height dimensions）, encoding, shape information, color cues, contrast variations, and illumination changes.