Mathematics and the information science research
1. Main Research field of our team
1.1. The Mathematical method and System Implementation of Fingerprint and Palmprint Images Compression.
1.2. The Mathematical Method and System Implementation of Fingerprint and Palmprint Recognition and Machine Vision.
1.3. The Mathematical method and System implementation of Mobile Communication Network.
1.4. The Mathematical method and Application Research of A new generation of Computer Network
2. The main research team members
Professor： Tian-de GUO; Sui-xiang GAO
Associate Professor: Tong ZHAO; Wen-guo YANG
Lectuer: Ge CHEN, Cong-ying HAN
Postdoc: Zhi-peng JIANG
together with 7 PhD students and 11 Master students.
25 in total
3. Research Team Working Philosophy
Problem solving oriented, gradually established the working mode which combines academic research with the production practice, conducts academic research based on practical problems, then resolve the practical problem by academic research.
4. The introduction to the research results
4.1. The police fingerprint and palm print image data compression and reproduction
4.1.1. The principle of the algorithm
We have proposed the data compression and reproduction algorithm for fingerprint and palm print image using matrix optimize by examine the characteristic of its texture.
4.1.2. The evaluation of the research results
Chinese security technology certification center was commissioned by Ministry of Public Security to open call the public (include university, Scientific research departments and Fingerprint company) for the fingerprint image data compression software and test them comprehensively in June, 2008. The “fingerprint image data compression algorithm and system implementation” which was developed by professor Tian-de GUO from the graduate School of Chinese Academy of Sciences ranked first in all the indicators (PSNR, rate of change of pattern, rate of change of feature, degree of change of feature).
The experts consider that the indicators of the compression algorithm developed by School of Mathematical Sciences of graduate School of Chinese Academy of Sciences can meet the demand of second automatic feature extraction in reproduction image and recommended the national unification use. Ministry of Public Security ask for the opinion of twenty-second bureau, fifth bureau, sixth bureau, the criminal investigation department and the fingerprint company and they all agreed to adopt uniform compression software, and the fingerprint image data compression software developed by school of mathematical sciences of graduate school of Chinese academy was in good condition in nearly a year of actual application testing.
In April 2009, Ministry of public security Released document to determine that we adopt the fingerprint image data compression software developed by graduate school of Chinese academy as the nation's uniform finished product in long-term use nationwide. The public safety net industry standard of People's Republic of China “fingerprint image data compression DLL interface” (GA 784-2008) and “the fingerprint image data compression ratio” (GA 788-2008) which were drafted mainly by Graduate school of Chinese academy of sciences has released by Ministry of public security. Since the compression techniques is the research results of technology support plan for the eleventh five-year plan, when existing funds support, Ministry of public security advice the graduate school of Chinese academy of sciences to agree to use this compression technology for free in the national scope, and promises to guarantee do technical support.
At the vice President of the Chinese academy, academician Hejuan yin’s request, the achievements has reported to the high technology bureau of Chinese academy.
4.2. Mathematical Methods and System Implementation of Fingerprint Identification, Palmprint Identification and Machine Vision
4.2.1. Mathematical Methods and System Implementation of Fingerprint Identification
Fingerprint as the primary evidence has been accepted due to its uniqueness and invariance. In China, the automated fingerprint identification system has been used since 1990. So far, the Provinces (cities) have all established the automated fingerprint identification system
More than 30 million stamps fingerprint exist in our country. The number is increased by 20% every year. The expansion of the fingerprint database brings some questions, such as the instability of the system and the relative decline in precision and speed .By increasing the number of the devices; we can only solve the problem about large-capacity storage of the fingerprint database, while other questions have not been effectively solved. This restricted the further development of the automated fingerprint identification system, so the automated fingerprint identification system needs to be completely updated.
At present, the foundation of processing and matching algorithms of fingerprint images at the level of millions has been built in china. However, some aspects are in the research stage at level of ten million, for example, the decline of the matching accuracy and speed and inquiry among different fingerprint identification systems. The research of the Automatic Fingerprint Identification System with independent intellectual property rights at the level of ten millions will promote the development of the national automated fingerprint identification system and the rapid detection. There are Immeasurable social significance for the social stability, state property safety and the people's safety.
(2) Main Research Results
(i) Singularity and core detection based on zero- pole model and generalized Hough transformation
(ii) The algorithm for ridge orientation based on grid interpolation model
(iii) The algorithm for fingerprint matching based on the compatibility of the minutia’s’ star structure
(iv) the algorithm for fingerprint matching based on convex hull of the local fingerprint
(v) Minutiae matching algorithm of the fingerprint based on bipartite matching and dynamic programming
(vi) The implement of a practical automated fingerprint identification system
As the continuous expansion of fingerprint storage capacity, the accuracy of the existing fingerprint recognition is seriously reduced, leading to results are not available. The current police fingerprint recognition system can only be used for millions of people. For Beijing, Shanghai and other large cities, it’s clearly unable to meet the identification requirements. The research group developed the ten million level (one hundred million fingerprints) Automated Fingerprint Identification System can be used for fingerprint database contains ten million people, and it has been tested on the ten million people fingerprint database of Beijing Public Security Bureau Criminal Investigation Corps. The indicators fully meet the requirements of the demand and the test results are very well. It is fully meet the needs of the provincial public security department. In fact, a variety of technical standards and API interface descriptions that are developed during research have become important reference standards and technical specifications for police fingerprint identification.
(4) The economic and social benefits of the results
The Automated Fingerprint Identification System at level of ten million is a technological breakthrough of the existing fingerprint identification system. It promotes the development of Automated Fingerprint Identification System in China and helps achieving rapid detection. It has immeasurable social significance to the stable of society and the protection of the safety of the state’s property and people's lives.
The obvious advantage of fingerprint identification makes it bound to be gradually applied in various fields such as banks, social security and national examinations. Large-scale, high-precision and efficient fingerprint recognition technology are the technical prerequisites and the protection of these applications.
(5) Research of civil fingerprint identification
The knowledge and experience accumulated in the police fingerprint feature extraction and matching have gradually opened up the opportunity in the civil areas of fingerprint identification. Currently the group is in cooperation with the Harbin Institute of Technology Software Engineering Company, and engaged in the fingerprint recognition algorithm design and software development of the social security system of Heilongjiang Province. The work is gradually carried out.
4.3.2. Palmprint identification system with mathematical methods
According to statistics, palmprint trace appears in about 30% criminal scenes with fingerprint marks (including fingerprints and palm prints), which indicates that palmprint occupies a large proportion of crime scene traces. Processing and recognition of palmprint traces provide effective clues for the investigation of cases and speed up the detection of criminal cases. As a result, there’s a growing need of automatic palmprint identification system.
(2)main research results
Algorithms are proposed in palmprint segmentation, palmprint splitting, feature extraction and palmprint matching, and identification problems between a full palm image and a full palm image or between a palm block and a full palm image are solved. In the step of palmprint segmentation, a new algorithm based on LBP and post-processing is presented; a reasonable sub-block processing method is experimented; the minutiae are used as feature which is the same in fingerprint processing; and effective algorithms are proposed in matching step. Moreover, palmprint system standerds are proposed, and a palmprint system with independent intellectual property rights is created.
Current international and domestic similar technologies and products are focus on online identification systems, and offline palmprint identification system can learn from that. In our implement, a series of offline key problems such as pre-processing, feature extraction and matching are solved.
At present, automatic palmprint identification systems used by local criminal investigation are introduced from France and USA, and is limited in small-scale investigation. It will fail on database with a number of 100000. For example, the stand-alone version of SPEX system introduced from France with a database capacity of only 10,000 and a speed of about 10 palmprints per second, has a high price of 200,000 Yuan. Our research group developed a palmprint identification system which reaches a capacity of million people level, and has an accuracy rate greater than 80% in feature extraction step. When indentify a scene palmprint with a least minutiae number of 24 in the database, the accuracy rate (lay in top ten) is higher than 65%.
(4)Economic and Social Benefits of Results
At present, though the introduction of palmprint automatic recognition software from France and the United States agency is limited to small application testing, we still cannot establish City-level (hundreds of thousands of people-level) palmprint database. For example: the capacity of introduced French SPEX system database (stand-alone version) designed only for 10000 people, matching speed is 10 entries per seconds, but the price is as high as 200,000 Yuan. Because foreign similar software product uses a complete set of sales methods, then system maintenance upgrade and price will be hold in the sellers, which led to the application of the system and the long-term development were controlled by others
At present, the FBI and the International Criminal Police Organization (InterPol) have not formulated the technical standards of palmprint automatic identification system. By contrast, domestic palmprint automatic identification technology in the criminal investigation application is still in the blank or just the beginning stage, there are no software products of the palmprint automatic recognition with independent intellectual property rights.
Because the principle of this project which regard research and development as the guide, application as the pillar is in accord with criminal investigation of the actual work needs, the Beijing public security built this database first which can support a millions level palmprint database for the current palmprint features extraction and matching work.