www.winxp.fora.pl

[tomoyq60]

,[link widoczny dla zalogowanych]

long-term Since the development of high robustness of the human face and handwriting recognition algorithms and systems for authentication of the following difficulties:
the same person within the class of face and handwriting samples vary widely, ie the pose, illumination and expression, age and other changes, the same person's face image has changed dramatically, because of the time, place and other changes, the same person writing glyph will have a huge difference; the sample has a wide range of different people there are similarities, for example: the face of different people with similar eyes, nose and mouth and other facial structures, the same,[link widoczny dla zalogowanych], different people write the same word handwriting strokes tend to have similar structures. These difficulties make high-performance face and handwriting identification and authentication very difficult for the development of practical, automatic face recognition and handwriting recognition systems pose significant challenges.

Tsinghua University research team led by Professor Ding Xiaoqing for national security,[link widoczny dla zalogowanych], public safety and information security and other major national needs, through years of independent innovation,[link widoczny dla zalogowanych], research and development of high-performance, practical handwriting recognition of human face and authentication system. The system includes TH-FaceID face recognition authentication subsystem and TH-WriterID certified handwriting recognition subsystem, and subsystem integration of the two biometric identity authentication system, realized in the context of complex facial images and video automatic detection, identification and authentication, hundreds / thousands of orders of magnitude large database of face information, active vision face recognition automatic control,[link widoczny dla zalogowanych], and integrated handwriting recognition handwritten document retrieval and identification and so on. projects have been granted patents 4, software copyright 2 entry, and another two PCT international patent applications and two domestic patent applications.

The system's first real-time risk-sensitive AdaBoost algorithm for face detection and eye location, vision-based statistical model of high grade Identification of face recognition algorithms and high-performance sample-based anti-face authentication algorithm; home and abroad for the first time the practical handwriting identification and retrieval methods, the original character of the statistics based on a single framework for handwriting recognition, handwriting recognition fusion method for multiple characters and based on negative samples of handwriting authentication methods; effective biometric identification based on a single multi-mode generalized confidence universal biometric fusion. Xiao-Qing Ding team also successfully developed a variety of markets and practical high-performance image / video, and handwriting identification face authentication system .

In the face recognition industry's most authoritative FRVT2006 evaluation of face recognition technology, the 22 participating in the test Face the most famous international companies and academic research institutions, led by Xiao-Qing Ding Tsinghua University Research Center for Intelligent Information Processing Text is the only large-scale database testing of a complete academic institutions, but also the only institution from China. test results show that, TH-FaceID testing face recognition systems are all excellent Face Recognition in the performance of the human eye system, which all participants in the system has the lowest equal error rate.

in the field of pattern recognition by the international conference ICPR2004 the most important organizations, based on the BANCA people defined by the face database and its test protocol --- the most difficult race FAT2004 in face verification, TH-FaceID face recognition system evaluation indicators in all three lower than the second to about 50%% error Results The rate of the first, issued by the International Association for Pattern Recognition's \Department of Electronics, University of Intelligent Information Processing Laboratory, etc. Photo 5 units participated in the test. TH-FaceID identification face recognition system certification in all indicators of the accuracy of all the advantages of a larger first prize.