Selecting an Automated Fingerprint System(AFIS)

Selecting an Automated Fingerprint System (AFIS)
Shri Sharada Prasad, IPS

Fingerprint record keeping is in vogue in police department for more than a hundred years. However the efforts to computerize them with a view to facilitate their storage and processing is a recent phenomenon. These efforts were first made, by utilizing the principles followed by the fingerprint experts in fingerprint classification. However this line of computerization was soon abandoned for many reasons including that this type of classification is possible only if all the Ten Prints are available and therefore the partial prints could not be stored in a classification system for identification and matching.

It was noted that the fingerprint ridges were having some faults like break, bifurcation, islands etc. These could be identified (feature extraction) and if there are two such points, they could be linked together by counting the ridges between them as well as by noting their orientation with respect to other such faults. If there are more than two such faults, the computer can be programmed to prepare a map linking all the faults and preparing a unique map for each fingerprint. This map would appear similar to the air-route map of any country showing different criss-crossing straight lines linking different airports. This mapping facilitates matching of even partial fingerprint, provided the number of faults in the partial prints is above an acceptable level. This can be compared to the scenario that if in any partial map the air-links of south India tally fully with some complete map out of a set of country air-route maps, we are sure that the partial map must be the part of air-route map of India. However inside the computer the information about the finger print is kept in the form of a mathematical template as it occupies less space (bytes) and is easier to match.

Besides processing the fingerprints, the computer programs have been developed to handle palm prints with equal ease, deft and accuracy.

Computerization of fingerprint permits very fast data matching unthinkable in manual system. It also provides for fingerprint data to be stored in a classification free system, thereby eliminating the errors due to wrong classification. It permits automatic comparison of the fresh print with that stored from earlier records and gives option to fingerprint expert to replace the earlier poor print with the better print now available.

Besides above tasks, which were earlier performed by the experts though very slowly, the automation provides for certain activities that were almost impossible in the manual system. This is in the area of matching the partial print of one Digit with the digits stored in the entire database. Modern algorithms provide for such comparison to be made even if no Core or Delta is present in the partial print. Reconstruction of print, mage enhancement, analysis of super- imposed prints and removing the background of the prints for analysis are some other features, which the computerization facilitates.

When any agency decides to computerize the fingerprint records, it must crystallize its expectations from the investment in the computerization. These maybe
1. To facilitate investigations of criminal cases by timely advice to the investigative officer.
2. To identify a person for his previous conviction of for solving cases of unidentified dead bodies.
3. Identifying the re-entrants in any immigration control system.
4. Issue of driving licenses, credit cards, social security cards, national ID systems and cards for obtaining subsidized items (like ration card).

Every agency would better carry out a study, decide upon and freeze the investment expectations, as there are different automated systems to achieve different objectives.

Secondly the agency must decide the optimum database size for which the system should be configured and also the maximum the which would be tolerable for processing a query. The other important feature to decide is about the number of queries that would be expected to be processed simultaneously. For many agencies the existing data base size is of no relevance because due to many reasons the fingerprint work might be lying un attended and due to obsolescence of the database the results are poor which in turn discourage the users in sending fresh records for search or record-keeping. A better parameter to look for would be the number of property crimes in the area as the largest number of fingerprints get generated in these crimes. These cases also throw up maximum numbers of chance prints.

In case it decided to use the automated system to keep records of Re-entrants on immigration front, or of excise cases then that relevant database should be consulted (say the number of immigrants handled of the number of persons arrested for excise violations). Since the larger database would mean larger investment, the database size is important. Growth of database size at a reasonable rate should also be kept in view. At the same time too pessimistic estimate of growth of database size is not advised as at least in the field of crime investigation, the computerization is known to fuel the use of finger print technology in crime solving and there by quickly adding mass to the existing database. A 10% annual growth rate in database size after the computerization should be a fair expectation and the system should be designed for optimum performance for at least next 7 to 10 years. Similarly the frequency of use of fingerprint system before the computerization and after the computerization is likely to undergo a sea change. More and more satisfied users are likely to send fresh queries and also the success stories may prompt the non-users to go for it. Thus for example where the fingerprint bureau was receiving 5 queries a day before computerization, it may get anywhere between 20 to 30 queries per day by the end of the first year of computerization. The system should be capable in handling the increased workload without needing an up gradation within the warranty period!

While choosing an automated finger print identification system (AFIS) the following points may be considered.
1. Accuracy
2. Speed
3. Ease of use
4. Upgrades
5. Inter-connectivity between different system and other programs.

The speed of the system depends on many factors. These are related to the hardware performance as well as the power of the matching algorithm. The sped may be specified in terms of ten print records scanned by the computer per second. In a typical system this may range from 10,000 records to 500,000 records per second per matcher unit depending upon the system chosen. Bytes used for storage of a single typical record will decide how much space will be required by the database for a given database size Lesser the bytes per record, smaller is the disc storage space requirement. Typical requirement per record may go up to 10 KB.

As regards the matching accuracy is concerned, there are standard parameters to measure the same. Standard benchmarking procedures measure what is known as false acceptance rate and true rejection rate. In a good system both should be nearly zero. The system should give high accuracy rate in matching the partial and latent prints as well.

Besides above points it will be important to find out about the availability of after sales service maintenance and technical assistance arrangements with the vendor.

Some vendors insist upon proprietary hardware or the hardware from of a particular supplier. Most likely in such cases the hardware would be costlier than the equivalent system available in the open market. Secondly the maintenance for all time to come will have to be entrusted to the hardware supplier at a price dictated by him. On the other hand many AFIS vendors offer systems that can be run on any hardware provided it meets the minimum requirement Similarly some AFIS programs are platform independent (i.e. they can use any operating system like Unix, Lynix, Windows etc) at least for up to medium size databases (I million 10 print records).

The system should permit search with or without information about Core and Delta or about a particular digit. Some systems are unable to store all the ten prints and leave out little finger's fingerprint. In India we require to store all the ten print.

The system should have provision for accepting fingerprints through live scanners, digital cameras scanning the latent print or the live mug shot. It should also accept signatures and demographic data (Criminal Attribute Data Base). While printing the reports it should be able to fetch these records and prepare a combined report containing expert opinion, grounds for opinion, the questioned print, the matched print, the demographic details and the mug shot of the individual.

Many AFIS systems are sensitive to the orientation of fingerprint and cannot give accurate results if the fingerprints are not oriented properly (say they are placed upside down or side ways). Such systems will increase the turn around time for the finger print query.

Some vendors try to market the systems that are essentially Access Control Systems based on fingerprint as biometrics as AFIS for the police departments. However few queries regarding the finger print reports would distinguish them from AFIS, which is a forensic application.

The chosen AFIS should be able to prepare and print quality and quantity reports. It should be able to print individual workstation performance and the time taken in processing different queries. It should keep a log of all users of the AFIS so as to permit carrying of the audit trails. It should be able to provide the list of Records that are fit for weeding out.

                             AFIS should provide fingerprint matching in the following ways

                                           1. Ten print to ten print match
                                           2. Single digit print to ten print match
                                           3. Partial print to ten print match
                                           4. Partial print to partial print.

The last one facility is required to work out Serial Crimes that thought unsolved need to be looked into, from the angle if the identical criminal was involved in all of the crimes.

Existing systems
1.Andhra Pradesh has installed FACTS-4 from CMC in 1998 at a cost of Rs. 3.81 Crores. It has a central database (1.28 Lakhs in January 2001) in Hyderabad SCRB and local database at single workstation at the districts (28). Local units are on-line connected with central database at Hyderabad. In the year 2000 it could trace 19 chance prints and could successfully match 200 search slips.

2.Delhi installed FACTS -4 in 2000 at the cost of Rs. 73 Lakhs. It consists of a central database (86,500 Ten Print Records) at Malviya Nagar Fingerprint bureau and proposed to have workstation in each police district. Due to communication line problems the workstations are at present with State FP Bureau it self. 3 cases (3 chance prints) were worked out with its help in 2001 out of 242 chance print examined.

3.Jammu & Kashmir purchased FACTS-3 in 1998 and dinstalled in Jammu at the office of IGP Crime & Railways but due to administrative difficulties the system is not in full use. As on January 2000, only 3992 ten print records could be entered and no chance print search was conducted. It is a stand-alone system with no networking facility.

4.Kerala installed FACTS-4 in 2000 with Database (1,10,178 in January 2002) only at State fingerprint bureau. It traced 19 chance prints in the year 2001 out of 918 chance print processed.

5.Tamil Nadu has installed FACTS in the year 1998 at the cost of Rs.1.81Crores. So far 1,55,953 ten prints records have been entered in it . Six districts have been connected with the main control system through Microwave communication system. In the year 2001 the system traced 91 chance prints out of 2530 chance print searches.

6.Karnataka has also acquired an AFIS recently and entered 61,671 records as on 31 December 2001. It has examined 1112 chance prints and successfully matched 72 chance prints. Haryana has also reportedly installed AFIS. However the details about AFIS installed there and its performance are not available.

Database sizes
1. Punjab has State Fingerprint Bureau at Philaur established in 1897. This bureau has more than 5.43 Lakhs ten print records and adds about 16,600 records annually.
2. Tamil Nadu State Fingerprint Bureau at Chennai was established in May1897. This bureau has more than 4.62 Lakhs ten print records and adds about 6,700 records annually.
3. Uttar Pradesh State Fingerprint Bureau is now at Luchnow. It was established first at Allahabad in 1899. This Bureau has more than 4.23 Lakhs ten print records and adds about 4,300 records annually.
4. Maharashtra State Fingerprint Bureau is at Pune. It was established in 1893, initially for Anthropometrical work. This bureau has more than 2.20 Lakhs ten print records and adds about 12,000 records annually.
5. Bihar fingerprint bureau started functioning at Ranchi in 1912 on separation of this work from Calcutta bureau. In 1915 it was shifted to Patna where it has now more than 2.08 Lakhs records. However annual addition to the records is only about 100 records.

Above information is furnished as obtained from the central finger print bureau. This will give a fair idea about the database sizes in older finger print bureaus with relatively larger database. While above is the position in some States which are leading in terms of Database size, there is no assurance that these records are all current. In fact, the routine weeding has not been carried out in most of the States and if this is done the actual holding may come down to about 60% of the present size. This fact is evident in very poor hit rate in case of Uttar Pradesh, West Bengal and Bihar. Further in many States the duplicates are not being routinely taken out and they add to the bulk. AFIS has facility to indicate and remove the duplicates. Thus while deciding about the database size of the proposed AFIS system, some cold analysis will be called for.

Players in the market
There are only a few manufacturers of AFIS in the world shoes products are good for forensic applications. Thus while there may be a number of vendors offering match or no match type of solutions, there will be only few whose product include report generation giving map of identical points and reasoning for the given result.

Addresses of some reputed Companies manufacturing Automated Fingerprint System suitable for forensic applications are given below.
1. CMC Ltd Hyderabad India: Postal address:- Manager Fingerprint system, CMC Center, Old Bombay Highway, Gachibowli, Hyderabad-500019, India. Tel:-91-40-3000501,3000401, Fax-91-40-3000509, e-mail: factsinfo @ cmcltd. Com

2. Printak,1250, North Tustin Avenue Anaheim California-92807 USA Ph-714238-2000

3. CIMSA (Now known as SAGEM-MORPHO), France: Postal address Sagem Morpho Inc 1145, Broadway plaza Suite 200 Tacoma, Washington USA 98402 Ph-8003462674

4. NEC, Japan

5. Zygox, Bangalore India Zygox Software (P) Ltd. # 575/A, 1" Main HAL 1st Stage, Indiranagar Bangalore- 560038 Ph-91-80-5286051/676/64, Fax-91-80-6286052

6. Identification systems, DERMALOG, Address: - Identification Systems DERMALOG Gmbh Mittelweg 120. 20148 Hamburg. Germany Tel- + 91-40-413227-0, fax- + 49- 40- 41 3227 44, e-mail: info @ dermalog. de Internet : http// www. Dermalog.de

7. Cogent System Incorporated, 209 Fair Oaks Ave, South Pasadena, California 91030 USA Tel- +1626 79 8090, Fax- + 1 626 799 8996 www cogentsystem.com, e-mail,:- info@ cogentsystems.com

Options, which must be considered for configurating the most suitable system

1. Entire system at State Finger Print Bureau with quarries received through post or manually.
2. Central database at State Fingerprint Bureau and digitization units at districts/ important locations say Range offices and exchange of data through reliable communication media.
3. Central database at State Fingerprint Bureau and distributed database at local station with capacity to process data at local level and also to forward and receive back information from central server at State Fingerprint Bureau through reliable communication media.

Capacity for the system

1. Database Size
a. Database location
(i) Above 300,000 FP records in central server
(ii) Between 100,000 and 300,000 FP records in central server.
(iii) Below 100,000 FP records in central server
(iv) District units may not require to store more than 10,000 FP records

b. In Assam the State FPB has 53,508 FB records in year 2000

c. Annual receipt in Assam is not exceeding 400 for record and 2500 for search. Only 5 search slips could be linked in year 2000.

d. Assam recorded 2404 cases of Burglary and 5138 cases of theft in the year 2001. Viewed in this context, the annual addition of record slip is very poor and with computerization may increase to at least 1500 per year, that is on average 6 per working day. Assam also does not process FP records of BD National to test if they are re-entrants. In case this is done the database size and volume of transactions would increase.

2. Processing speed
a. 10 FP slips for record per day
b. Between 10 to 20 FP slips for record per day
c. More than 20 FP slips for records per day
d. Output per day can be doubled by operating the system in three shifts (third shift for updating re-indexing and maintenance)
e. Another way to look at is the through put rate i.e. The time taken in giving the result of a query right from time of receipt in the bureau till printing the reply to the query. This may depend upon the time taken in
i. Scanning the query record
ii. Reading the handwritten text on the record
iii. Processing by the machine (may be in queue)
iv. Printing the reply
Total process may take any where between 15 minutes to few hours. Maharashtra receives on average 400 scarch slips on a typical working day at Pune. Tamil Nadu gets about 60 per day and West Bengal gets 20 per day. Assam gets only 7 slips per day for search. Thus we may decide the turn around time by finding out the same for the first query of the day and the last query of the day.

3. Importance of communication media
a. Media should have low noise rate
b. It should be able to carry digital data
c. Media should have fairly high up time and stable linkage
d. Data transfer rate should be pretty high say 64 KBPS Upwards

4. Availability of after sales support in India and time frame for attending the calls

5. Up gradation of the system and connectivity with higher versions

6. Annual maintenance charges and free warranty period

7. Capability of the supplier in making customized reports, summaries, - performance assessment work etc.

8. Price

Benchmarking by NCRB

NCRB is the supervising and administrative controlling authority of the Central fingerprint Bureau of India. In this capacity, NCRB took up the task of benchmarking the available AFIS. So far it has benchmarked two such systems namely CMC product FACTS- 4 and Zygox product AFIS NT-2000. These reports may help the State Police Departments in selecting the suitable product for their fingerprint bureau.

Preparing for the change over

Once it is decided to go for AFIS, the staff of finger print bureau must be trained and mentally prepared to handle the new task. As of now the non-experts are performing a large number of tasks in any finger print bureau. They also happen to be relatively younger in age. It has been found that they adapt the new technology better than the experts who may be less receptive to the new tools. Secondly the department must decide what arrangements would be made to transfer the data from the cards to the machine. Whether some company will be given a contract to complete the job as one time project or the existing hands in the bureau would be used to enter the data. In the latter case the benefits of computerization would be delayed. Thirdly a link will have to be established between Henry coded record slips and the computerized version of the same record for facilitating manual locating of the record. Fourthly the user like an investigating officer will have to be educated about the benefits of sending search slips, record slips and latent prints for facilitating the investigation. The court staff will have to be educated to prepare clean and legible record prints.

TOP