TW200412544A - Fingerprint identification using slice-matching method - Google Patents

Abstract

The method described in this article identifies fingerprints by using the reference fingerprint slices stored in the database to match with the input fingerprint image. If the matching score is higher than a custom threshold value, meaning the slice of input fingerprint image is very similar to one of the reference slices, the slice will then be chosen as a candidate slice. After all the slices of input image has been handled by the matching process, a candidate list is obtained. Stage 2 is to analyse the list. If there exists one set of candidate slices that fits the property of reference slices, the input fingerprint is matched. Mismatch otherwise.

Description

200412544

[Technology to which the invention belongs The invention is a device such as a mobile phone, which has a high comparison efficiency. Technical field] The method provides' fingerprint identification number assistants that are more efficient than traditional identification methods used in low-efficiency, etc. [Previous technology] Fingerprint identification is generally divided into feature point ratios and compared using complex calculations due to pattern comparison. For the two types, the former requires a larger memory space to collect the fingerprint images of the wealthy. [Summary of the Invention] The fragment comparison fingerprint recognition method can be more powerful and short comparison; the comparison method 'additionally space' is proposed in this article; the fragment comparison is proposed in this paper; several samples are taken from the complete memory image used by Er Chuyu Film ^ 1 mother-fingerprint When only fingerprint images are input from the named user, the H Beibu library 'when an unknown segment is taken out for rough comparison in the first stage? Then, the candidate fragments with a higher degree of template in the poor material library are added to +. Find the characteristics of similar sample fragments in the round-robin fingerprint and analyze the candidate's first-stage screening work: use ′ to determine whether the comparison is successful. Sample fragment "Refer to Figure 1 for comparison. Cell size 4, size and ^-fingerprint images are used as a custom 'to increase the area of the sample fragment or provide a higher accuracy rate according to user needs, as a price: The number of template fragments can be changed to longer processing time.

Page 5 V. Description of the invention (2) After the fragment is taken out, the processing program to filter the noise generated during the fingerprint capture will reduce the noise to the minimum. Second: Use the low-fingerprint fragment. In the end, · == "picture" is within the range of fingerprints after wave and binarization, and the fragment image is enough to contain the user input and the central area is used as a template. Discard not used, the comparison template is multiple images with a height of 1 4 0 ~ 7 between each sample (the index calculation formula ~ + 1 is selected as the pair stage, and the input fingerprint pair 'if it is the comparison score Candidate segments for higher segments; Candidates for identical segments. Height of identical segments + 10 pixels, template segments, 1 to 8, ..., 1 3 All plate segments can be customized from the center. I zed Match As shown in the formula on the second page of the illustration, if you want 80% to be a threshold value of 0.6 candidate fragments. Each piece is input on a certain input)) 8 pieces are all and all The fragment pattern of the fingerprint with a custom threshold can be divided into (Segment, for example, input pixels, then input 2 ~ 1 3 9 etc. 1 3, in Shui Wai) enter Index ~, the above pixel's fractional shift is 2 lines of comparison, (Lee) to calculate the range of results that match the segments that exceed 0.6 It ’s possible to select the board and 'I. It is possible to enter the fingerprint at the same time. The height of the fingerprint and the fingerprint are 3 fragments. 0 ~ + 2 0 The score will be compared with regular comparison. The margin will be selected at ~ 1. Meeting 200412544 V. Description of the invention (3) It is the 4th when the database is registered. The candidate list of the 5 template fragments is 16, 23, 33, 49, 5 into the film I, and the candidate of the second template fragment The list of candidates for 44, 59, 68, 73, and 8 board fragments is 1, 5, 67, 88, and gg. The analysis program first selects 1 as the candidate for the board fragment section 2 that is larger than 1 in the list of template fragment section 2. Then, according to the distance between template segment one and template segment two from template segment three, if the 'third candidate should appear in the middle of the candidate list into template segment three is not a candidate, if you think about it, find See if 5 0 or 5 2 is the candidate list where the candidate fragment and the model fragment are merged into the second candidate list. Continue to find the M candidate fragment that matches the model fragment. If you find a set of 58, 73, 88, then the same level Shift, if the three candidates match the sample fragment, the system will find 9, 5 9, 9 9 and 1, 0, 6 user rounds 8, 6 candidates clear 3, 8, 7, 7 from the first candidate 0 In turn, the first input fingerprint is 1, 9, 9, ^, 7 9, 9, 1 The input order is 3, 4, 2, 6, 5, 8 9 and the third sample is 1, 9, 3, 4, 5 7 A list of sample fragments to find a candidate 'Then, from the concept of sample slices, such as selecting 2 6 as a sample, it can be known that this is equal to sample fragment two and one and candidate two are the correct input fingerprint fragments. It is found that the error caused by the 51st input piece is not, if this is not the case, then the analysis program finds that there is no homogeneity at the end of the sample piece, then the system determines the certification candidate segment. In this case, whether the three candidates are The same is true with the transfer, and this group of candidates is successful. In this example, we can also wait for two sets of candidate fragments.

4. Description of the invention (4) This kind of unreasonable interval can be directly skipped by adding some restrictions to the analysis program, to avoid the situation where the two groups of candidates have the same horizontal displacement and cause misjudgment. The correct comparison result is shown in Figure 5. [Embodiment] The implementation is divided into three steps. Step 1: Take out N sample p-force databases from the fingerprint image during the login phase. The method is as follows: • The size and number of fragments can be customized according to user needs.

2. Use a low-pass filtering process to remove the fragment image. T, ^ ^ "Y, noise. 3" Binarize the weight image of the image for comparison. 4. Remove the central area of the fragment image as a template. Fragment. Step 2 · Use the model fragment and the input fingerprint to find the candidate list of the parent fragment in the authentication phase. The method is as follows: 1. The input fingerprint can be divided into (the input fingerprint image is louder 2: height + 1) X sample template fragment 2 · Each fragment of the input fingerprint is paired with all sample tablets, and the regular comparison index method is used to calculate the comparison. Eight = comparison 3. If the comparison score is higher than a custom threshold , Candidate fragment of the template fragment. &Amp; For the sample 4 · Generate a candidate list for each template fragment Step 3: Use the relative position characteristics of the template fragment = segment to judge the comparison result. Candidate film. Based on the relative distance between the sample fragments under the special material mf

Candidate list 200412544 5. In the description of the invention (5), the candidate fragment group with the same characteristics is found. If it cannot find it, the authentication fails. 2. If more than one set of candidate fragment groups are found, compare whether each of the candidate fragment groups in the group has the same horizontal displacement characteristics. If not, the authentication fails. 3. If there is a set of candidate segments that meet the above characteristics, the authentication is successful. [Simplified description of the icons] The first picture: (1) template fragment one (2) template fragment two (3) template piece, segment three second image: (1) original image (2) low-pass filtered image third Figure: (1) Low-pass filtered image (2) Low-pass filtered and binarized image Fourth image: Range of template segments Fifth image: (1) Original fingerprint segment (2) After candidate segment analysis After comparison

200412544

Page ίο

Claims (1)

200412544 • Fragment matching fingerprint identification method a. • Material database from fingerprint images during the login phase. ‘This identification method includes taking out N sample fragments and saving them: b. Capitalization b c Use of samples in the authentication phase Use the characteristics of the sample fragments to break the comparison results. The fragments are compared with the turn-in fingerprint. To analyze candidate fragments, to judge 2 aaaabb as in package a in application 1 · from paragraph 2 · will be 3 · will be 4. Take as package 1 in application b · enter 0 2 2 ● input patent scope: Login Refers to the size of the fragment of the movie, the range of patents for the fragment includes the following: Fingerprint + 1 pixel, ~ Ί, fragment. The fingerprint is used, and the identification method described in item 1 is used, in which the f-segment image in the pattern is used as a comparison and the number can be used according to the user's needs. The procedure is like binarization for profit comparison. The central image field of the image is used as the template area in the first item. The identification methods described in + and B can be divided into (round-in) m η s-?, ,, and ~ image degrees, such as input fingerprints. High] ~ 8 is 8 pixels, then the basis of the step is lost. Remove noise. segment. Among them, the steps are as follows: the plate fragmentation is 1 4 0. The fingerprints are 1 3 9 etc. 1 3 3 T i§ ^ are compared with all the samples in a regular comparison. The method is to calculate each piece of the comparison. Score. 200412544 6. Scope of patent application b 3. If the comparison score is higher than a certain custom threshold, it will be selected as a candidate fragment of the template fragment. One fingerprint input fragment may be a candidate for multiple template fragments at the same time. b 4 · Each template fragment generates a candidate list. 3. The identification method as described in item 1 of the scope of patent application, wherein step c includes: c 1 · Analyze the candidate list generated in step b. c 2 · According to the characteristics of the relative distance between the sample fragments, find the candidate fragment group with the same characteristics from the candidate list. If it cannot find it, the authentication fails. c 3. If it finds more than one set of candidate fragment groups that meet the characteristics, compare whether each fragment in the candidate fragment group has the same horizontal displacement characteristics. If not, the authentication fails. c 4 · If there is a set of candidate fragments that meet the above characteristics, the authentication is successful. 0 Page 12