Automatic Fingerprint Identification System Using Robust Distance

Fingerprint recognition has attracted many researchers in order to conduct studies in this field. The need for finding an AFIS (Automatic Fingerprints Identification System) has been an interesting subject by scientists, bureaus and companies, because the increasing use of AFIS leads to solve many problems of identification, because there are no tools or machines in the region except classical methods. The fingerprint department at the Criminal Identification Bureau in Erbil uses habitually very classical methods to identify the fingerprints. There are (8000) fingerprints in that department related to criminals. To create an AFIS, we depended on the stages of image analysis using Visual Basic. These stages are very essential to obtain an excellent system for fingerprints recognition


Introduction
In the past decade, the interest in fingerprint-based biometrics systems has been increased significantly [14].Fingerprints are considered as an important device to the personal identification.Although, there are many other unique characteristics of an individual that can be used such as voiceprints, dental impressions, DNA, Retinal Patterns, Iris Patterns, Face Patterns, and even the shape of the Ear Lobes [3].They can by each of this Biometrics( * ) identifications recognize the individuals, but each one has its own merits and uses depending on the place and time of occurring.
In 1915, the idea of fingerprint had entered Iraq then some Indian and English experts constructed the first office of fingerprints in Basra.They applied Henry's method for the identification process.After that, this office was transferred to Baghdad City and named "Criminal Identification Bureau" under the supervision of Englishmen and Indians.In 1934, some officers were sent to India in order to study the science of fingerprints, and then they came back to Iraq [29].
The aim of this study is to make and find an AFIS using computer programs to recognize fingerprints, because the Criminal Identification Bureau in Erbil is using a classical method to do this work although there are (8000) fingerprints in that department related to criminals.

Fingerprint Patterns and their Features 2.1 Fingerprint Patterns
Each fingerprint has one of three basic patterns: Loop, Whorl and Arch [27], [5].
( * ) Biometrics: It is the science of measuring different aspects of the human body and provides near foolproof security because it relies on a person's unique makeup [24].
According to Edward Henry's classification, the fingerprint is classified into the following eight-subclasses [5]

Fingerprint Properties
Fingerprints have three important properties, as follows [5]: 1. Unchangeability: The ridges and details of FP's are constant through life until the skin disintegrates after death.2. Uniqueness: There are no two fingerprints, which have the same pattern.3. Classification: The variations of FP's are within a limit, which allow a systematic classification of the patterns.

Fingerprint Features
There are two main types of features in fingerprints used as an in identification of individuals [13]: A. Global Features B. They are those characteristics seen with the naked eye.The classifications of fingerprints are based on the global features, which include:

Pattern area
It is the part of the fingerprint that contains all the global features.While, the pattern area of loops and whorls can be easily defined, it is very difficult to define the area of arch patterns.

Core Point
It is located at the approximate center of the finger impression.

Delta
It is the point on the first bifurcation, abrupt ending ridge meeting of two ridges, dot, fragmentary ridge, or any point upon a ridge at the nearest center of divergence of two type lines.

Type lines
They are the two innermost ridges that start parallel, diverge, and surround or tend to surround the pattern area.They are often breaking the ridge immediately outside that line is considered to be its continuation.

Ridge count
It is the number of ridges in the Pattern Area.To establish the ridge count, an imaginary line is drawn from the Delta and the Core and each ridge that touches this line is counted.

B. Local Features
They are Minutia * points.It is possible for two or more individuals to have identical global features, but they still have different and unique fingerprints, because they have local features (minutia points): There are five different characteristics of minutia points in fingerprints:

Computer Imaging
Computer imaging is one of the important subjects nowadays; it deals with digital images including the Acquisition, Enhancement, and Recognition of images.Computer imaging can be defined as the acquisition and processing of visual information by computers [27].
There are two major kinds of computer imaging [27]: 1.Computer Vision: In computer vision the computer uses applications of the processed images.2. Image Processing: In image processing applications of the output images are for human use.
Computer imaging has many applications in medicine, industry, printing and publishing, cosmetics and personal grooming, astronomy, mineral analysis, fields of law enforcement and security and so on [18].4. Expert systems [9] Expert systems have been used successfully in various applications, ranging from image processing and industrial inspection to the medical diagnoses and process control.Expert systems are human-machine systems with specialized problem solving expertise.Thus, the name expert system reflects the fact that these systems are generally based on knowledge obtained from people who are experts in a specified discipline.

Classification
Sometimes the recognition process involves much time to do the operation of recognition.But another important procedure is to reduce the time of recognition, by partitioning the entire data to some classes by classification process.
The pattern classification, as related to image analysis, involves taking the features extracted from the image and using them to classify image objects automatically [23].
For a two-class pattern classification problem, there are the hypotheses: H 0: The pattern belongs to class one.H 1 : The pattern belongs to class two.If there are more than two classes, H k is the hypothesis, which assumes that the pattern belongs to class k.A typical example of multiple hypothesis testing is the classification of fingerprints type, where types of fingerprints are equal to five [25].

Data Collection Surces
We prepared AFIS, software in Visual Basic, to distinguish between fingerprints.In order to implement AFIS, we need fingerprint data; the required data were not available as mentioned before in "Erbil Criminal Identification Bureau", data was acquired from Bologna University in Italy.Those fingerprint images were taken from left thumb finger and corresponding to (21) persons (8 items per person).The type of all those fingerprints is (JPEG) and (256×256) pixels.In addition, (50) fingerprints were taken from left thumb finger corresponding to (25) persons (2 items per person) by a manual method, using a scanner of type (48) bit.These two fingerprints are selected from four impressions (items) of the same person in order to obtain good images in all directions.
For fingerprint purchasing refer to (www.nist.gov).

Low Level Processing
The processing of algorithm used was as follows (figure, 7.

Intermediate Level Processing
In this stage, three processes were made; Segmentation, Representation and Description.The size of the image in the right picture box is reduced to (160 x 160) pixels.After that, these (160 x160) pixels are divided into (64) squares; the size of each square is equal to (20 x 20) pixels (400 pixels per square).This is the segmentation process, figure (7.2).
Directly after segmentation process, the representation process comes.The internal characteristics of the image are chosen, therefore, the present values in each pixel are the raw data that we depend on to represent each square inside the fingerprint image.
After that, the image was described by an approach, there are many approaches to describe the image, we have chosen statistical approaches.8 parameters were used in this step, they are as follows: 1. Mean (µ) 2. Median (Med).Where: a i : It is the ith parameter for the first square.b i : It is the ith parameter for the second square.After computing the (64) distances of the selected fingerprint to be recognized, all fingerprints would be stored in the storage unit.The minimum distances for each square from these (64) squares are computed, in order to know which fingerprint has a maximum number of squares and, at the same time, possess the minimum distances (i.e., the fingerprint that matches the unknown fingerprint must possess a great number of squares which have minimum distances).While, this operation is limited to the following: If the number of squares that have minimum distances is less than (10), the matching process must be rejected.The result of recognition process is failure and the AFIS gives a message that the name of that person is unknown (the fingerprint is unrecognized).If the number of squares has minimum distances greater than or equal to 10, it will lead to a message appearing in the message box telling the user of the name of the person, who possesses the fingerprint.Depending on the references [20] and [28], we have written this subroutine concerning the third current stage of image analysis.
In addition to robust distance, many measures are applied to find the distance between two images, such: f.Mahalanobis Distance.g.Similarity Measure.h.Correlation.In order to know which of the above measures the best is, we wrote a specific subroutine to compute them.Clicking the button (test) on the main form of AFIS leads to obtain the success rate of the particular measure used in the recognition process.Therefore, in each trial, we can change the measure and click the "test" button to know the best measure.The robust distance overcomes all other measures and obtains the percentage of succeeding equal to (84.5%).On the other hand, Mahalanobis Distance came in the second order with percentage of succeeding equal to (82%).

Conclusions
1. Using a specific scanner for acquiring fingerprint images leads to increasing the accuracy of FPR compared with the classical methods for acquisition the fingerprint images.
2. Segmentation of the region in the fingerprint image by the researcher's method led to simplify the describing and recognizing processes, in addition to increase the resolution (accuracy) of the process.3. The robust distance of calculating the distance between the two vectors has a good role in finding the minimum distance between two fingerprint images.The success rate of matching was reaching (84.5 %). 4. Expert systems are powerful methods in matching fingerprints, because they depend on simple procedures, and give excellent results. 5.The average time of completing the analysis process from acquisition step to a final step of intermediate level processing is equal to (0.33) seconds, and the average time of recognition and interpretation is equal to (0.28) seconds.
9. Recommendations 1.The use of (AFIS) in "Erbil Criminal Identification Bureau" 2. The conducting of another study on the classification of fingerprint classes to reduce the time of recognition 3. Producing or buying a specific scanner for acquiring fingerprint images to the display unit of the computer directly.

(
* ) Minutia points are the points at which ridges end, fork and change.

3 .
Spatial Frequency-Spatial frequency refers to how far apart the ridges are in the neighborhood of the minutia point.4. Curvature-refers to the rate of change of ridge orientation.5. Position-The position of the minutia point refers to its x, y location.Figure (2.2) Local Features:

Figure 2
Figure 2.2 Local Features: A-Ridge Ending.B-Ridge Bifurcation.C-Ridge Divergence D-Dot or island.E-Enclosure.F-Short Ridge.

1 ): 1 .
Choose a fingerprint image.2. IF the size of the image is equal to (200 x 200) pixels THEN GO TO the step 6 END IF. 3. Shrinking or stretching.4. Showing the image in the left picture box. 5. Copying the fingerprint image.6. Transforming the image's copy from color type to grayscale type.7. Deleting the first and last 20 rows and columns from the image's copy.8. Showing the image in the right picture box.9. Stopping.

Figure 7 . 1 :
Figure 7.1 : Main form of the program

: a. Arch: is sub-classified into: 1
. Arch (A): It is the pattern in which the ridges enter in one side, rise in the middle, and flow or tend to flow out from the other side as shown in figure (2.1a).2. Tented Arch (T): It is the pattern in which the ridge have the tendency to enter from one side and flow out from the other side, with the exception that the ridges from either an angle or an upthrust on the center.Figure (2.1b) shows a typical tented arch fingerprint.b.

Loop: Its is sub-classified into:
3. Right Loop (R) or Radial Loop: It is the pattern in which one or more of the ridges enter in either the side of the impression, recurve, touch or pass an imaginary line drawn from the delta to the core, and terminate on, or toward the same side of the impression from which such a ridge or ridges entered.

. Whorl (W): It is sub -classified into:
Enclosure: A ridge which is divided into two and then reunites again creating an enclosed area of ridge -less skin, see figure (2.2E).F. Short Ridge: An extremely short ridge, but not so short and appears as a Dot or Island, see figure (2.2F).2. Orientation-Each Minutia point faces in particular direction.