Chapter Six: Minutiae Extraction // done There is four
Chapter Six: Minutiae Extraction // done There is four type to represent the fingerprint, and the minutiae one of them, Due to its distinctiveness, compatible with feature that used by fingerprint human experts. Figure (6.1.1):Using minutiae.6.1 Ridge ThinningThe goal of Ridge Thinning is eliminate the redundant pixels of ridges to get one pixel wide, and filter the image by using MATLAB. For each scanning to all fingerprint image; Thinning algorithm is marked all redundant pixels, move all this redundant to (3X3) image window, Then it removed all marked pixel after several scans.The output of Morphological operation step is Thinned ridge map; this step goal is to remove the breaks, spike and isolated point; at this step any single point breaks or ridges are considered to eliminate processing noise.The figure(5.1.2)below is shown Thinned-ridge map in the left side, in the other hand the right side will show how the image look like after remove all spikes and breaks Figure (6.1.2):Thinned-ridge map6.2 Minutia MarkingAfter finger Ridge Thinning step, The Marking of Minutiae step became easily; since it’s considered by Crossing Number(CN) to extract the minutiae. In general this step used the output of fingerprint Ridge Thinning to analyze; to 3X3 window block.If the center pixel is one and it’s has three one value’s as neighboring then the center pixel is called “Ridge Branch or Bifurcation “; Look at figure “5.2.1”. Figure(6.2.1):Ridge Branches.Else if the center pixel value is one and it has exactly a one value as a neighbor, then the center pixel is called ” Ridge Ending or Termination ” look at figure 5.2.2 below Figure(6.2.2):Ridge Ending.The main difference between ridge end and ridge burifurcation point is the number of one-value neighbor; the ridge as we show has exactly one-value as a neighbor, in another hand the ridge burifurcation has three one-value neighbor. Figure(6.2.3): Ridge Burifurcation.Suppose that; if the most-upper and the most-right pixel’s with one-values have other neighbor’s out of the 3X3 block window, then two 1’s pixels will mark as a branch ridge , but one of these branch is locate in a small region. Therefore the routine check are required when none of branch neighbors added.Also, the distance computed between two neighbors ridge are called “Average inter ridge width” and it’s denoted by “D”, However the way to get the approximation value of D value is have only so simple step , such that: Step 1: Take the thin image ridges you need to scan the row of this image , then sum all one’s pixel value in one row.Step 2: By Dividing the length of the row by the sum that you retrieve in step 1 to get width of the InterRidge.Step 3: All InterRidge width are the average value to get “D” value.So, the way to compute Inter ridge distance is only the sum all pixels with value one. All thinned ridges of fingerprint images with the minutiae marking step, are Labeled by uniqueness ID used in next operations. Figure(6.2.4): minutiae Marking.Chapter Seven: post-processing //done7.1: False Minutia Removal Recognition:In fact all of previous stages are introduce only the artifacts which may leads to “Spurious Minutia”. As an example, Let the false breaks are obtained from insufficient amounted ink, so not all of over ink are eliminated. That’s mean all of The Pre-processing stage doesn’t fix the total fingerprint image. Because the false Minutia affected matching accuracy, we need some mechanism to remove all of false minutia to keep the fingerprint system effective.The figure(6.1.1) below will show you seven type of false minutiae structure. Figure(6.1.1): False minutia Types.M1 the valley ridge is a pierce spike ,M2 the connection between ridge are false spike ridge.M3 when same broke ridge are having 2 near bifurcation.M4 when a twice ridge are broke a point(having the same orientation angle and short distance). M5 similar to case four exception when a single part of an broken ridge is short when the terminate minutiae generated.M6 are extended from m4 case , but it’s include an extra property which is an third ridge is founded in middle of two parts broken ridge.M7 is denoted when a short ridge founded in threshold window.Those are all steps that you needed for removal false minutia: Compared the distance between terminations and bifurcation ridges, if it less than average inter-ridge( D),” Distance between two parallel neighboring ridge “, and case 1 of structure(two minutia in the same ridge)is checked; Then you need to Delete the both ridge. If distance between 2 bifurcation is less “D” value (case 2 and 3)Then, Removing the 2 bifurcation ridge only. If distance “D” is equal to the two termination, The direction are coincident with small variation, case 4,5,6 are satisfy, and it’s satisfy a condition where no any terminated ridge is locating between the 2 termination.Any minutiae derived from broke ridge (as false termination minutiae )it will be remove. The last step is denoted if the case 7 checks, the two termination locating in a short ridges, and the length is less than “D”, then we removed two terminations ridge .This stage has two mainly advantage; one of them, the ID ridge is used when seven type of false minutiae structure are strictly defined to distinguish minutiae; The second one is in the order of remove procedure that consider to reduce the complexity computation since that may utilize the relation between the type of false minutiae.Chapter Eight: Minutiae match // doneThe role of minutiae matching algorithm is to check if two fingerprint image are belongs to the same person or not, in another word it used to determining if the minutiae sets are perform to the same person’s finger or not.This stage have consecutive sub stage: Alignment stage. Matching stage.The next subsections will explain the main role for each these stage.8.1: Alignment StageGiving many minutiae images to compare and match step , when you choose any minutiae from each image, there is some information must be recorded, like:1-We need to calculate x and y coordinate for each points.2-The output of coordinate points are define the type of Minutiea ridge (end ridge , bifurcation , .. ).3-the direct link grade with the other minutiae in the neighborhood.After that, calculate the similarity of the two ridge, if the similarity between two images is larger than threshold value; we will transform set of minutiae to new system coordination, where the origin is on references point, the x axis represent the coincident by the references point direction. The series of x value are represent of the ridge point “called the associated ridge to each minutiae”. 1-Assuming I is denoted to average inter-ridge length, n value is set to 10: “the total length ridge .8) then we move to next step.Otherwise, complete the applying matching algorithm for next-pair ridges. 2- Respected with minutiae references, we translation and rotation other minutiae according to this formula: x , y and ? parameters for the minutiae reference will become Figure (7.1.1), Show the rotation of the image according to minutiae references orientation. Figure (7.1.1):Rotate the coordinate systemWhile the Distance -D- was calculating, the rotation is calculated implicitly and stored with each minutiae coordinate then its reduce the process time. According the references minutiae, rotation role make to transform minutiae and matching them with unified x and y coordinate.8.2: Matching Stage: In this stage it must verification from a parameters x, y and ? if are the same for two identical minutiae, and that is done by place the bound box value around each of minutia template, and compare it, if the difference between minutiae is tiny and the minutiae is matching with a rectangle shape then two minutiae will check as a Minutiea matched pair, every minutiae must has a matched minutiae pair or has no matched.The ratio of final matching defined by :Ratio =(the total of matched pairs)/(number of minutiae founded in the template )Score of fingerprint matching is ratio*100 , the range is from 0-100 ; if the founded score is larger than threshold then 80% of probability is the two compared template are from same finger. Chapter Nine: Conclusion and Future Work // done 9.1: Evaluation of the system:The reliability of the fingerprint image systems are precision relies that obtain in the minutiae extraction process, however the main factor that can damaging the corrected location of minutiae is poor quality for image; this factor was solved when we used Alignment- based match. There are many type for performance evolution indexes, which use for determining performance of fingerprint recognition systems such that:9.1.1:False Rejection Rate: As we say in previous sections; the biometric security system maybe rejected correct attempt access by authorized user; so the measure “FAR” is used in this type, the following formula is basically used to enhance the FFR.In general, FRR ” Incorrect Rejected ” is the measure of possible of biometric system to classify any authorized user as invalid user Therefore, FRR is basically state the ratio value between the number of False Rejection and number of identification attempt.9.1.2: False Acceptance Rate:FAR happened when the security of biometric systems are accepted the attempt of the unauthorized user. Systems FAR are indicate the ratio value between number of false acceptance and number of identification attempt. FAR(%) = (FA/n)*100 ;where FA= false accepted , n =sample number.Another Formula :FAR is basically define when the system obtain unauthorized access as a valid user. When we say the biometric system become more security that’s mean the system is provide low False Accepted Rate. 9.2: Conclusion:When we backed many years ago, we found many technique used to verification the fingerprint such as: Traditional technique: It the widest technique, it worked by taking impression of hand’s used ink then pressing covered finger on to the paper to obtain fingerprint. Those used technique based on minutiae matching technique ,ARPS and Pattern-based matching, which are relatively fast and less expensive, improves the matching speed, improve feature extraction and accuracy more than other algorithms. After system analysis, Euclidean matching templates algorithm is the best, since it’s computed matching rate in less time and easy way. All future work based on PCA recognition tried to achieve more accuracy and performance .