JPS63500128A - comparison device - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention] ■■ le [river] This application is filed under U.S. Pat. The application is a continuation in part of Application No. 762.118, the latter of which is No. 762.1. No. 18 is titled “Comparative Device” and is published by the United States on April 2, 1985. The present invention generally relates to searching a subject or search record (hereinafter referred to as a "search"). A set of points in a reference file record (hereinafter referred to as "file") for one set of points. Concerning automatic computer control methods that contrast files and and each of the search points has a value related to a predetermined parameter. Special The present invention relates to a fingerprint comparison device in which a set of minutiae of a file is The search pairs are compared to the position parameters of a set of search minutiae. It determines whether sufficient agreement has been obtained to verify the phenomenon.

The mechanism of the fingerprint comparison device is known. Such a comparison device is the Wegstei n, J, H, + The M-40Fin er rint Matcher.

National Bureau of 5 standards Technic al Note 87EL LLSGoverno+ent Printing Office, Washington, DC 1975.

Wegstein's comparison mechanism allows the details of each file and the search The minutiae position is expressed by X, Y coordinates and angle. Described in Wegstein As shown, each file point gives a displacement value of the difference table (difference table). Compare each of the search points in each of the above X, Y and angle parameters so that compared. The difference table allows each entry to compare the search point to the file point. It is a giving table.

This table includes the displacement between two points in the X direction and the displacement between two points in the Y direction, and also It also includes the angular displacement between the two minutiae points being compared. Plot the X and Y displacement values. The difference plane is obtained.

The extent to which the search fingerprint is displaced using the cluster of points that this difference table contains , or the search fingerprint is rotated to the extent that the set of file points is rotated relative to the obtained fingerprint. I'm trying to give an indication. A performance index or comparison score is obtained. this A search fingerprint is matched against a file fingerprint only if the contrast score exceeds a threshold of It is thought that this was the case.

Another improved comparison mechanism is provided by Egstein, Joseph H. Omated Fingerprint Identification Sy stem, National Bureau of 5 standards Pu Publication 500-89+ February 1982.

The effectiveness of a comparison device is given as a function of its operating speed and error rate. Type I The error is in validating the match when the target finger is actually the same as the finger from which the file was obtained. This is a mistake in case of failure. This is due to the fact that the verification is incorrectly removed in the access system. This indicates that the Errors of type ■ are based on the fingerprint and actual file where the file was created. This error occurs when target fingerprints that are not the same are verified. This is an access system This indicates that the system improperly accepted the fingerprint.

Furthermore, the value of the contrast mechanism is that it increases the operating speed of the verification system. Depending on the extent to which it contributes to It is given according to the degree of contribution to

These three operating characteristics and the amount of deformation and misalignment that can be tolerated are mutually exclusive. They are in a bad relationship. Key and non-key points when comparing data obtained from fingerprints However, problems arise due to the fact that fingers are plastic. The imprint of a finger Since it changes over time, the obtained image will also change.

In light of the above, it is an object of the present invention to provide an improved comparison device, which Accommodates errors such as degrees, Type I errors, Type H errors, as well as deformations and misalignment. It is designed to improve the trade-off between basic characteristics, such as the ability to

Here, terminology will be explained for the following explanation. The following terms are specifically used throughout this specification. It is used repeatedly throughout the body.

The contrasting device of the present invention consists of a series of main components, which can be termed contrasting. It implements all judgment decisions in proper identification. In particular, this comparison device , four individual scale comparators integrated to provide variable scale contrast functionality. It is equipped with a network segment. This comparison device also has a scoring mechanism. . However, in order to operate this comparison device to make a judgment, additional file details are required. and a storage device for the search minutiae. comparison device is obtained by combining all of the above elements.

This is the surface on which the and Y difference values are plotted. These difference tables (hereinafter referred to as T) (abbreviated) and the difference plane are different from the image plane in which minutiae points appear. in this way Then, the entry point on the above difference plane is (a) given file point and given search and (b) this file point, representing the difference along the Y axis of the image plane between the points. given as a Y-coordinate point representing the difference along the Y-axis of the image plane between and the search point .

The term rQJ is used in different contexts with different subscripts to describe how the points relate to each other. This generally means whether the bond is securely bonded. Roughly speaking, this means Spatial peaking that refers to the space on the difference plane rather than on the plane eaking). The term Qc refers to how two points on the difference plane relate to each other. Gives a measure of closeness. The term Qs means that a particular point on a difference surface gives a measure of how close the points are to each other. Furthermore, the measured value Qt is gives an indication of how close the points are to each other. However, on this difference surface in terms of −i When expressing the proximity of , we will call it Q. The term QT is used here particularly for drawings. used interchangeably with Qt, which means the same thing. Please note this point.

As used herein, minutiae refers to basic microfeatures, i.e., the endpoints and branches of lines. used. These are well-known methods that have traditionally been used to identify fingerprints. . The position of each minutiae on the image plane is indicated by x, represented by The angle of the minutiae in the image plane is also used to identify the minutiae; can also be used as part of a butting mechanism. In an embodiment of the invention, The angles above are used only with respect to the initial qualification criteria for entries on the difference table. It will be done. Although this angle is not subsequently layered in this example, the butt mechanism has a Q factor of To generalize the difference table to include angular difference values as a contribution to Can be done.

tile! star escape The present invention will be briefly explained based on an embodiment. In the system, a set of search identification points obtained from the optical scanning of the finger A set of file identification points for the same finger of the object is compared. The distinguishing point for these is X.

Discrete points in the Y-plane, representing stable minutiae.

Minimizes the effects of custom distortions and changes that occur when the same finger scans different time intervals In order to do this, the matching system of the present invention divides the finger information into multiple segments, Additionally, each segment is sent to a separate comparator and compensation routine for processing. I have to. When such a finger undergoes changes over time, one of the above segments A correction factor is given based on the analysis of the difference table to compensate for the distortions and changes. The method used is Within each segment, search points and file points are compared. , thereby determining the initial (first) difference based on the distance between the points being compared in the X, Y axes. minute table is given. Proximity value (pro ximity value) Q is calculated. Based on these proximity values, each The above initial difference table is filled in for all but one entry representing the search point. will be paged. Points with large Q values are retained points.

Next, the basis for recalculating the proximity values for each entry on this purged table is As a foundation, the sufficiently reduced and purged (second) difference table described above is used. It will be done. The entry points on the above purge table with the maximum proximity value Q are DX and D It is processed as the origin on the difference plane where the Y value becomes zero. Other points on the difference plane A similar transformation is performed for . As a result, the two points, search point and file point, The search point/flash that constitutes the highest Q value on the difference table as if it were a match. File points are processed. This shift shifts the correction factor to the purged delta table. The calculation is based on the average residual difference plane developed between the search point and the file point on the file. calculated. Appropriate to the image plane position of the search point given on the purged difference table appropriate shifts and correction factors are applied.

A third difference table is established based on the corrected search point positions. this The third difference table is a difference table entry point representing search and file points. screened against. However, these points are further separated from a predetermined number of pixels. based on the one with the highest proximity value in the purged delta table above. It is limited to a predetermined number of entries. In this way, currently obtained, A more restricted (fourth) difference table is added to each entry point of this difference table. These recalculated proximity values are used as the basis for recalculating the proximity values for Used as the basis for scoring comparisons and determining discrimination.

This scoring method further calculates the number of entry points on the constrained difference table, the purge Number of entry points on purged delta table, overall proximity of purged delta table The score value and the number of lines in the limited delta table above and the purged delta text. including the total segment score value based on the ratio of the number of lines in the table. It uses parameters.

plate - plate Figure 1 shows the comparison mechanism of the present invention, the file minutiae and the services compared by this comparison device. FIG. The positioning provides a Q value and other figures of merit, and an indication of validation.

FIG. 2 is a block diagram illustrating the process in the comparison apparatus of the present invention.

Figures 3A to 3F show the comparator and correction process illustrated in Figure 2 in more detail. 1 is a diagram that is a flowchart. The explanation of the transition of processing between the above components. To facilitate the solution, Figure 3G shows the relationship between the processing targets shown in each of Figures 3A to 3F. Show relationships.

Figures 3H-3I are implementations of Figures 3A-3F where correction factors are provided on the Y and X axes. A modification of the example is shown. Figures 3H and 31 correspond to Figures 3E and 3F, respectively.

Figure 3J shows four points in the image plane. These four points are.

The pair S' and F' of file points and search points that give the highest Q in the difference table are . The other two points illustrated are the difference table pair Sk and Fk, which are the third Calculate the scale factor components SXT and WYT as shown in operation box 38a of diagram H. The pairs S' and F' are anti-compared to calculate the calculations, respectively.

FIG. 4A is a block diagram illustrating the four segment configuration of the present invention; The set of files and search points is divided into four segments, and the comparator and correction processing Four individual parallel stages of the process are given.

FIG. 4B is a schematic diagram showing the 16 segment configuration of the present invention; 3 illustrates dimensional segmentation and shows the degree of overlap of file segments;

In this case, there is no overlap of search segments.

Figures 5A and 5B are used in conjunction with the comparator of Figure 2 to verify ID verification. 2 is a flowchart showing a scoring method for giving.

Figure 5C shows the Q at perfect match as a function of line entries in the difference table. This is a table that provides values.

Figures 6A-60 illustrate variations of Figure 3 in which two major changes are made. being done. The first change changes the purge of the initial difference table in Figure 3D. service that appeared in the initial delta table is added to the purged delta table. It is guaranteed that there is one line entry for each point Is. Therefore, the sixth Comparing Figure A with Figure 3D will help understand the changes that have occurred.

The second major difference in the embodiment of FIG. The reason is that it is carried out only along the X-axis). However, the size of the X-axis correction is In addition to this, distortion in the Y axis is also taken into consideration. This can be seen by comparing with Figure 3E. 6B shows the additional correction factor SIV.

As can be seen by comparing with Figure 3H, the correction factors SIW and COW are reduced in Figure 6B. has been excluded. Figure 6C shows correction of search point lSOX value by SIV and COv. It shows. No Y correction except for the Y shift indicated in function box 40a. Not possible.

FIGS. 7 and 8 illustrate the records used to generate file points maintained in storage device 18. It is a flowchart explaining a recording method.

Figures 9 and 10 generate fingerprint images from which files and search points are extracted. 1 illustrates one known optical scanning method for.

Good! row view The comparison mechanism determines how the set of details of the fingerprint image (search point) is compared to the set of details of the reference fingerprint image (search point). provides a score indicating whether the

The target fingerprint image is different from the reference fingerprint image even if both were taken from the same finger of the same person. However, they are different. Displacement and rotation of the finger can cause some of these differences. Ru. The above differences are also caused by the plasticity of the fingers, which means that a person can A function of the change in magnitude and direction of the force exerted by a person when placed on a tens or other support This results in a change in the configuration of minutiae. The condition of your fingers changes every day, and Given as a function of recent environment and usage conditions. As a result, the search from the image The set of minutiae points is different from the set of file points in the standard file if two fingerprints are from the same finger. Even what you get will be different.

Therefore, a comparison mechanism is used to provide a figure of merit for the verification decision. each The minutiae has X, Y coordinates on the image plane. The purpose of the comparison device according to the invention is mainly to The objective is to establish a difference table regarding the positional parameters on the X and Y image planes. .

Two sets of minutiae to be compared, namely the file cent JS and the search set Is However, United 5tates Patent Number 4+322+1 631ssued March 30.1982 and entitled Optical scanning of fingerprints as shown in FINGERIDENTIFICATION extracted from the image generated by The fingerprint image given in this way consists of many It is a binary image composed of image elements (pixels), and this image has standard details, especially lines. This makes it possible to identify the end point of a line and the branching of a line.

The details extracted from such optical images can be obtained by many known methods. Ru. Such a method is 5tock, R,M.

Automatic Finer rint Readin Proc by eedin's of the1972 carnahan Conference ce on Electronic Crime Counter-+neas ures, IJuniversity of Xentucky, Lexin gton Xentucky+1972, p16-28, and Banner , C, S, and 5tock, R, M, FINDERTHE FBI'S A approach to Automatic Fin er rintIden tification Procedure of a Conference ce on the Science of Finerrints, Ho meoffice, London, England.

Described in 1974.

The extracted details are stored in a reference record (file) in the memory unit of the automatic classification/identification device. ) or stored as a subject record (search point). Each point is x, y It has position coordinates and angular coordinates. Therefore, each point is represented by three parameters. It will be done. The comparison device automatically compares the file point parameters with the search point parameters. Provides a method to compare.

The contrast mechanism described here treats each such minutiae as a point and No discrimination is made between minutiae points and branch minutiae points. The set of minutiae in question is a set of mechanisms. Extracted from the image by either In practice, the contrast mechanism of the present invention is problematic. It is not required that the set of points be minutiae points. All that is required is that the set of points is It is said that the fingerprints of the person in question are unique and are therefore given a unique identity. That's true. Accordingly, with respect to the description and claims herein, the points of identification It should be understood that reference to It is.

System (1) FIG. 1 generally illustrates the relationship of the comparison device 10 of the present invention to the devices with which it is associated. This is what is illustrated. The optical scanning mechanism 12 is described in the already cited patent 4.322. ．． 163 and shown in Figures 9 and 10, which gives a binarized image. It works like that. The identification point extraction mechanism 14 extracts points that identify the fingerprint from the above image. Extract the set of This extraction mechanism 14 is a mechanism described in the above-mentioned literature regarding detailed item extraction. It is one of the The set of subject points is held in a search point memory device 16, which They will then be compared with the set of points in the file point memory device 18. Tsuki The combining device 10 stores the contents of the two memory units 16 and 18 as explained below. Compare and give a score indicating validation. The set of points in the file memory device 18 is arbitrary. determined by the following method. However, in order to minimize the type I and ■ errors already described, For this purpose, the set of points in file memory device 18 gives the set of points in search memory device 16. It is desirable that it be derived by the same mechanism as the one obtained.

File point JS and search point Is are based on three parameters: X, Y values and angle on the image plane. The degree values are stored in memory units 16 and 18. Favorable to facilitate comparison In the embodiment, JS and IS points are stored in order of Y value. In this way, J S, has the lowest Y value for any file point Q, JS, has the next Y value, and so on. Ru. By ordering by Y value in this way, we can give a loop routine and It becomes possible to facilitate point-to-point comparison.

An important feature of the system of the present invention is that it segments the fingerprint image into many individual zones. , applying comparison and contrast techniques to each zone or segment, respectively. It is in. This multi-segment configuration is explained in connection with FIGS. 4A and 4B. . The scoring method shown in Figures 5A and 5B is based on the quality of each segment of the contraster. It uses scores. Based on the above, the explanation of Fig. 2 and Figs. 3A to 3J is It should also be noted that the above routine is repeated for each image segment to be analyzed. Should be.

Combare (2s) By understanding the configuration diagram and functional diagram in Figure 2, the more detailed information shown in Figures 3A to 3J can be This will make it easier to examine and understand flowcharts.

As shown in FIG. 2, the points in the enclosures 16 and 18 are basically Compare and screen at 20 using known comparison methods to give a difference table 22. will be turned on. The Qs value for each entry on this difference table is calculated in 24, This difference table is then reordered into a sequence by Qs at 26. Placed. This means that the first entry in the difference table has the highest Qs value. It means something.

The routine performed in step 28 except for one entry representing the search point (all Used to purge from a table. This routine This is done by selecting the entry with the highest Qs value. Then a similar The routine is executed at 30 and the entry with the highest Qs representing the included file points Purge the delta table of multiple file point entries by keeping only the Ru. In this way, the purged difference table is given as shown at 32. It will be done. The important thing to say about this purged delta table is that the search fingerprint Assuming that the points actually correspond to file fingerprint points, this difference table is If we represent only the most possible search point-to-file point correspondences based on the Qs value of That's what I mean. Purging this difference table is the basis for almost all improvements to type I errors. It is thought that it gives

As shown in Figure 34, this purged difference table now shows the differences to the system. Qs(i is the entry of this purged delta table) will be recalculated based only on This purged delta table has the highest Qs value It is reconfigured to the beginning as shown at 36.

In this purged and reorganized delta table, the server with the highest Qs It is assumed that each entry represents a corresponding search and file point. These The points are here named S' and F', respectively. This search point P' is A shift procedure is performed to relocate the point on the image plane so that it has the same coordinates as point F'. child With the currently matched search and file point as the origin, as shown in 38, A scale factor is calculated for each of the other search points Sk. Each of these searches The scale factor for a point indicates that this search point is paired with an entry on the difference table. is calculated with respect to the file points that become greater than the threshold (in one embodiment) (20%) the individual search point scale factors are excluded from the calculation. Calculate like this All calculated individual search point scale factors are averaged and Gives scale factor correction. These sets of search points are added to the purged difference table. Only what is shown above.

゛ As shown in 40, the X and Y values of the image plane with respect to the search point are shifted, Search and file for highest Qs entry on purged difference table Match the point pair S', F'. This shift is the reordered purged difference By zeroing the DX and DY values for the first entry on the minute table It will be done. The same shift is done for all other DX and DY values so that the All of the search points are effectively shifted by the same amount.

As shown at 42, the search points are further rearranged based on the scale factor. Special After the above position shift, the reordered purged difference table DX and and DY values are corrected by the scale factor. A simple example of Figures 3E and 3F It should be noted that in , only the DX values are corrected by the scale factor. . The larger the value of this scale factor for DX or DY, the greater the relocation correction. It works to increase the size. This value for DX and DY is the stage 40 and is set to zero for the difference table entry with the highest Qs. Therefore, the search point for this highest Qs entry is determined by the above scale factor correction. It will not be affected.

The highest 15 entries are then rewritten above along with their new DX and DY values. Selected from the ordered purged difference table, the restricted difference shown in 46 Give a minute table. 48, where the new value for Qs is the highest Based only on up to 15 differential table entries and their new DX/DY values is calculated. for each of up to 15 entries on the new delta table. Once the above new Qs value is obtained, the total Q value QT is calculated. As shown in 50 5A and 5B, the QT and also the score value SQ are determined by the score routine shown in FIGS. 5A and 5B. The figure of merit used to determine whether a search point matches a file point by It is given as follows.

As shown in the flowchart, re-search point by shift and scale correction. Rather than affecting the value of the image plane position in the storage device 18, the placement This is done by changing X and DY. Data on the limited difference table above The number of entries is limited to 15, and these are the purged delta tables with the highest Qs values. There are 15 entries on the bull. However, this These entries are different from the corresponding entries in the purged delta table. Ru. This is because the DX and DYO values are corrected and reflect the relocated search point position. This is because it has been changed so as to reflect. Limited difference table DX and D These new values for Y represent up to 15 points in the restricted difference table. is used as the basis for calculating the Qc and Qs values for each of the .

A correction routine is used to shift and correct the search point, thereby When the correct finger is submitted for identification, the search points will match the file points more closely. When a further incorrect finger is submitted, the search point is Points are shifted further away from the file point than they would have been without the correction routine It becomes like this. Therefore, the correction routine according to the present invention eliminates errors of type I and ■. It has the effect of reducing both.

Combare flowchart (-3A~3G).

If you remember the system shown in Figure 2, the detailed system will be explained in Figures 3A to 3G. will be easier to understand.

An initial difference table is created from the set of file points in the storage device 18 and the target image in the storage device 16. is formed based on a comparison with a set of search points. In forming this table, Each file point JS is compared with each search point within the limits of . end of line and minute There is no equivalence distinction between the branch point and the branch point. A portion of such a difference table is shown in Figure 3B. vinegar.

Various entries are created in the table for each comparison between file point JS and search point ISO. formed into a circle. The entry DX is equal to the displacement between these two points along the Y axis of the image plane. be destroyed. The other entry DY is made equal to the displacement along the Y axis. Between these two points A third entry is made indicating the angular difference. DX and DY entries are pixels between two points It is a number. Plotting all DX and DY entries forms a difference plane.

The number of entries in the initial difference table (D, T) is the number of entries for a given file point. It is limited in that it does not include a search point. One example is shown in Figure 3A. , the search point IS excluded from the difference table CD,T,) is a given reference point , where (i) the search point is centered at the file point <61 is outside the 61-pixel box, or (ii) the angle of the search point is outside the band of plus or minus 22.5 degrees around the angle to the point.

In the illustrated embodiment, the angle value for each point is It is only used as Otherwise they would not be used, but the device of the present invention It is extensible to use angle information as another indicator of alignment. In this way , the number of entries in the initial difference table will be limited.

The difference table shown in Figure 3B represents a portion of a normal difference table. Ru. Each entry is given by a line number. Each file point JS is a file point Center on! ＜　61　Compared with each search point IS in the x61 pixel box Ru. The Y-axis and Y-axis displacements DX and DY are respectively entered in the difference table. . The difference table is ordered by the Y value of the reference point. J in memories 16 and 18 Because we have ordered the S and IS points, we can include a loop routine to speed up the comparison. It becomes possible.

Any search point with Y value less than or equal to that of the window assembled on a given JS is It does not need to be compared with point S. In this way, it is determined that the Y value of IS is below the window. This means that the following 13 points can be reviewed immediately. 3rd Diagram B is used to calculate the Q value for each entry on the difference table. It shows a routine. en) This Q(+!! is the value Qs for IJ) Ru. Each individual value Qc of a given difference table entry is equal to one other difference table. Given based on comparison with entry. A high Qc means that two entries are This indicates that they are relatively close to each other on the difference plane. In particular, two differences The concave point is determined by taking the difference between the two DX values, as shown by equation 24a in Figure 3B. , is compared by adding it to the difference between the two DY values to obtain the TR value. next , this TR value is subtracted from the constant KR, which in this example has the value 15, and Qc give value. 24b, when (KR-TR) is negative, the contribution of Qc is No grant will be accepted in any way. All Qc values for a given entry are added to give the Qs value. The basic purging and control given in action 46 of FIG. In a limited difference table, the overall value QT for the set of points being compared is also calculated. calculated. In particular, Qs−ΣQc Figure 3C is a difference table with Qs values assigned to each entry in the difference table. This shows a routine for ordering. This ordering is routine and fairly standard. It is a typical routine. One item and two functions worth noting in relation to Figure 3C 50 a and 50b. This function 50a simply provides an output value SQ for the scoring system. is zero when there is only one line in the difference table. Immediately In other words, if ST is equal to or less than 1, SQ has a value of zero. and be immediately dismissed. This gives the scoring system an indication of the discrepancy and the identification It will be denied. The value KSW is determined by blocks 38-44 of FIG. Increments to rlJ after the function (see 44a in Figure 3F) is completed. K.S.W. 1'', the value SQ is calculated as shown by the formula shown at 50b. Ru. However, this is done after the system processing through Pron 44 in Figure 2 is completed. occurs only.

The output A of Figure 3c, which is actually an ordered difference table, is shown in Figure 3D. purged by the routine that was created. This purge will remove all multiple search delta table entries. and all multiple file difference table entries. like this Where there are multiple entries, the entry with the highest Qs value is retained. This pa The image routine is a very important routine that is used for search printing and file printing. The noise level is significantly lower, which reduces type errors considerably. given a reduced difference table. If these prints are not compatible, The algorithm reduces the number of random correlations and further improves type H errors. purged The number of entries 5old in the difference table is the value used to calculate the output factor SQ. and retained.

The portion of FIG. 3D surrounded by a dotted line with the reference numeral 31 is a specific purge routine; The routine may be repeated at other stages of processing in various embodiments described herein. It is. The purge routine 31 portion of the purge method in FIG. A routine where P has a single slot for each line entry on the difference table It is. This purge routine is initialized to zero. "Zero" status means difference indicates that the corresponding line entry in the minutes table should be retained It is something. When the status of the P buffer slot is changed to “1”, the page routine 31 operates to delete the line entry. this purge In routine 31, three indicators are used. Index "i" is on the difference table Indicates that the line is being considered. Indicator “C” is how many lines are on par. Indicates what should be recorded. The index rSJ is determined after one or more other lines have been purged. , indicates where the retained line entry is on the difference table. Do it like this As shown in 31m, if the line entry has a “1” status in the P buffer, If the count C of line entries to be eliminated is 31n is incremented by 1 as indicated by , and the line entry is indicated by 31P. It is located at line 256 as shown in FIG. Place the line entry on line 256 , this effectively excludes the line from the difference table. on the difference table The line STO number is always less than 256, so the line is a functional box 322 is excluded. If the line entry being considered is in function box 31m As shown, the corresponding slot Pi of the P buffer is then in the “zero” state. If the line entry is given a However, this line S is deleted by the purge function. There are as many lines above line "i" as there are lines "C" to be processed.

Purge routines 31b, 31c, and 31d shown in other figures.

31e and 31f are functionally equivalent to the routine shown at 31 in Figure 5D. be. Other purge routines are used to delete differential tables that are deleted for other reasons. It runs on a digital table.

Next, the purged difference table is subjected to the Qs calculation procedure of FIG. 3B, and further subjected to the Qs calculation procedure of FIG. The new Qs values are ordered as shown in the figure.

The above purged difference table is shown in the Qs calculation and in Figures 3B and 30. As the ordination is recirculated through the stages, the gamma value increments by 1 as shown in Figure 3C. 50c), as well as the calculation of the scale factor COv as shown in Figure 3E. A difference table entry is given to D as a basis. As shown here, The search point represented by the highest Qs entry on the paged difference table is the difference matches the corresponding file point represented by the same entry in the table It seems so. That is, they are considered to coincide on the image plane. difference table For each other entry on the bull, all other search points are is considered to be displaced from the corresponding reference point. The correction factor should now reflect this distortion. Calculated. As shown at 38a in FIG. 3E, the individual scale factor components SXT is calculated for each search point, and a separate search point is Exists for each entry.

Of course, the scale factor component for the highest Qs entry on this difference table is It is b.

According to the embodiment shown in Figure i3E, scale factor correction is only done in the X direction. ・It should be noted that this is not done in the Y direction. ・Figure 3H shows ・Scale factor correction An example is shown in which this is done in both directions. As you can see from this diagram, the species consisting of ° In the case of fingerprint scanning, the distortion in the Y direction is relatively small compared to that in the X direction; In that case, the embodiment of FIG. 3E is sufficient.

The scale factor correction SXT for each detail Sk,i is based on the displacement value XIJ Given. XIJ is search details Sk.

It is the distance along the X-axis between i and the compared file details Fk,j. That is, 38e As shown in , XIJ=XI−XJ. (Figure 3J shows these relationships. This is a visual representation. ) Individual scale factor correction SXT gives the highest Q. The search point S' that is searched is is given based on the assumption that the file point F' is shifted to coincide with the file point F'. Ru. Each of the other search points Sk is shifted by the same amount on the image plane. The search image is When actually corresponding to the image, this shift usually occurs when the other search points Sk are The corresponding file point Fk that is paired as an entry in the difference table is It acts to move the other search points Sk mentioned above. Other searches above Unless the point Sk coincides with the corresponding file point Fk (corresponding search and The lack of agreement (with respect to the image) is thought to be due to plastic distortion of the fingerprint image. .

Furthermore, the farther the paired points Sk and Fk are from the points S' and F', the greater the absolute distortion. Therefore, the component SXT for the correction factor CO■ is must be normalized to the distance away from . In this way, the correction factor The child component SXT is (a) the distance XIJ between the two points Sk and FkO to be compared, and (b) the sa It is the ratio between the distance XIO that the search point Sk is away from the reference point S'. This is a function box It is indicated by 38a.

The actual correction factor co■, shown in box 38b, has the value "1" (which means no correction 5XTO) and the individual correction factor component 5XTO. Individual correction factor The average of is simply the individual correction factor SXT (this sum is denoted by the symbol SX in Figure 3E). ) and the number of included search points l5XO. If the two points being compared Sk When Fk and Fk match, the XIJ value becomes zero, and the component S for the correction factor XT becomes zero. The difference between the average correction factor and the value "1" is the difference between the search point on the image plane. This is the actual correction factor CO■ used for the X value.

However, as shown in 38c, the component under the individual scale factor SX exceeds 20%. This component is not taken into account when

This is important to prevent abnormal components from entering the overall correction factor Co■. It has been found that Skureen.

This prevents excessive correction. After all, if the correction exceeds 20%, this indicates that the compared points do not match and are probably not from the same person.

If an individual scale factor component SXT fails the CC test at 38c, the corresponding T, line entry is deleted. To facilitate understanding of Figure 3E, The purge routine is omitted. It is exactly the routine 31 shown in Figure 3D. The same is expected to occur at output M in FIG. 3E. purge louche like this The location of the button is shown at 3IC in the embodiment of FIG. 3H.

In this case, the screening factor DD of 38d is less important.

If the distance between the search point under consideration and the search point with the highest Q is 8 pixels or less, The positive component SXT is not calculated. This eliminates over-correction due to noise points. Therefore, a good search point is one that is close to the search point with the highest Q. I can't say anything.

Once the correction factor CO■ is determined, the purged difference, represented by the table (a) First, the highest Q search point is combined with its partner's file point. Transform each search point by an amount equal to the matching transformation, and then (b) scale factor ′ By relocating the search point based on CO■, it is possible to reposition it on the image plane. 6. Figure 3F shows the routine in which the repositioning of each search point Sk takes place. . The equations shown in function boxes 40a and 42a transform the search point and It gives results adjusted according to the Kehl factor correction CO■. above box 40a and 42a are understood to refer to the sample points on the image plane as shown in Figure 3J. becomes easier. As shown in block 40a and as shown in FIG. 3J, For a given search point Sk,i the distances X and Y are calculated (in Fig. 3F with the symbol X A represents the distance XI shown in Figures 3E and 3J). Assume no distortion (which means assuming that the correction factor C0V=1), the image plane The position of the upper point Sk,i is given by the distance XI and Yl from the file point F'. F ' is obtained from the difference table entry with the highest Q. The search point S' is point F'. In this way, set the X and Y values to point F'. The required shift is obtained by adding to the X and Y values. X value with scale factor Co■ By changing, an adjustment is made to the strain along the X-axis. Function box 42 It is noted that a indicates subtraction of the X value from the XXF value. This is simply a seed It simply arises from trends established elsewhere for the calculation of each factor. These Values added algebraically to points XXF and YYF are subtracted as long as the trend is maintained. Or expressed as addition.

In this example of FIGS. 3E and 3F, the reconstruction based on the scale factor CO The placement routine is done only in the X axis.

The function shown in 43a can be used to set the difference table entry by setting XYK=6. Establishes a window that excludes the bird from further processing, but this window The search points and file points represented by Six or more picture elements are spaced apart from each other along this axis. In this way, par The above difference table entry is removed from the indexed difference table. and, 44a, the above purged diff with such excluded points Tables are further limited to only 15 entries. These 15 entries corresponds to the entry on the purged delta table with the highest Qs value It is.

Next, these only 15 difference table entries contain only the relationships between these The routine shown in FIG. 3B is passed to recalculate the Qs value based on . However, the As shown at 44C in Figure 3F, the constant KR is centered at "7", so The difference table to be formed here has a TR distance equal to or greater than "7" ( (on the difference plane) to eliminate all Qc points.

44a, if the KSW value is now incremented to 1, the function of FIG. Their outputs are added to allow calculation of SQ values. Indicated by 50b in Figure 3C. As mentioned above, the above SQ value is essentially the difference table (Sold) that has been purged. The above limited difference divided by the number of entries and multiplied by the number of entries in the table (ST) is the sum of Qs values from the minute table (QT). This value KRR is Since QT makes a significant contribution and normalizes the result by multiplying QT by 0.5, QTO It is given to normalize the value. The value for ST should not be more than 15. However, the denominator 5oldO value is usually thicker than St (it should be noted that It is possible.

Importantly, the minutiae comparison mechanism (a) makes it easier to purge delta tables; (b) of the above purge to limit potential searches and file point comparisons ) before calculating the proximity value used to determine whether a match has been obtained. This includes a combination with a method of shifting and correcting the effective position of the search point on the image plane. Ru. This combination is superior to other matching devices in improving Type I errors. This gives very important results. This also improves type ■ errors. . Thus, a feature of the present comparison device is that the nature of the present mechanism corrects both forms of error. In order to improve It has the advantage of being moved to the bell.

2　　　　(-3H~3J~ When actually applying the present invention, it is necessary to scale not only along the Y-axis but also along the Y-axis. It is desirable to correct the factors. For example, a set of searches obtained from ink fingerprints If a contrastor is applied to contrast a point to a set of file points, then along both axes It may be desirable to provide a scale factor correction using

Figures 3H and 3■ show that these provide scale factor correction along two axes. Identical to Figures 3E and 3F, except that it shows the required steps. be. Therefore, where the function is essentially the same, those used in Figures 3E and 3F The same reference numbers are used in FIGS. 3H and 31.

However, in order to make it easier to understand this two-axis correction, some explanations will be repeated. .

As shown in the figure, for any given search point IS, the distortion is usually in one of the two axes. grows along the line. A set of search points can be obtained to determine which axis has the largest strain. given as a function of a particular finger impression. Which of the two axes, X and Y, is thicker? A judgment box 39a compares the absolute values of the displacements XI and Yl to determine whether the distortion is present. determined by This determination is made for each search point. One set of search points Given a given case, most points tend to give five similar directions of distortion. .

As shown in the box in Figure 3H, the distance XI is (a) the highest in the difference table. (b) the X value of the search point S′ that is a member of the pair with Q; This is the distance on the image plane between the X value of k and the X value. These two distances are each distance XX ■ and SXK. The Y axis distance MY■ can also be considered in a completely similar manner.

As shown in decision boxes 38d and 39d, the absolute value of either XI or Yl is less than a predetermined value DD (which is eight picture elements in the illustrated embodiment), then The search point does not contribute to the scale factor correction, and as shown in 39e, the search point does not contribute to the scale factor correction. The process is incremented by 1 and the next search point Is is determined by this scale factor correction. selected for. The reason for this is that if the search point is close to S', then The position of is thick (does not seem to be distorted) with respect to The reason is that it does not need to be incorporated into the calculation of the value of the factor correction.

The initial parameters for calculating the scale factor correction shown in Figure 3H are block 3. 9g and blocks upstream thereof are provided as shown. In particular, search Point S' and file point F' (points related to the highest Q entry on the difference table) The X and Y values on the image plane are the values XXV and YYV shown in 39g.

They are XXF and YYF. The value for K is that K=1 is the first value on the difference table. line (this line is obtained from the comparison of search point S' and file point F'). Therefore, it is initially cented as "2". The line on this difference table is It is not compared against itself. This line is that on the difference table This is the line to which other lines are compared. Next, add a difference table from =2 to the line For each search point on the difference table throughout, the rest of the routine shown in Figure 3H is executed. The section shows the output scale factor correction COV, COW, CIV, SIW. It will be done.

These values and relationships, as well as the possible composition of those shown in connection with Figure 3H. For ease of visualization, Figure 3J shows the services from the top line of the difference table. Example of point S' and point F' on the image plane and other lines on the difference table Another example search point Sk and file point FkO from is given.

If the search point Sk is outside the eight pixel proximity criteria of function boxes 38d and 39d, If so, the value shown in function box 39f is the value shown in function box 39m or 39 n is given to allow calculation of the scale factor as indicated in either . Which of these scale factor components are calculated is indicated in decision box 39a. Determined by comparison.

The scale factor components SXT and WYT are calculated as shown in function box 39n. Let's assume that. At this point, as shown in 39p and 39q, these If either component SXT or WYT exceeds 20%, both components should be considered further. In fact, a signal is sent to the purged buffer 58, which The corresponding search point is deleted from the difference table.

This is why purged routine 31c is included.

Decision box 39j is significant only during recording operations. This is a recording flag.

During the recording procedure described below, the deletion routine provided by buffer 45 is executed. However, for the present purpose, this record flag 39r can be omitted. can.

These individual scale factor components SXT and WYT are determined in decision boxes 39p and 39q. , they are added as shown by function box 39s, and then Using the sum of the averages, the correction values CO■ and SIW are calculated as shown in function box 39t. is calculated.

If the decision box 39a indicates that the absolute value of Yl is greater than the absolute value of XI The calculations shown in function boxes 39m and 39n are performed, which results in the function box Addition to the calculation of scale factor correction value COW and SIV as shown in step 39W. A value is given.

As indicated by the two decision boxes 39x, branch 39m of this procedure or When there is no scale factor component in any of 39n, the correction scale factor COV , SIW, COW, SIV as appropriate, and two function boxes 39y cents to values of 1 and 0 as shown in .

Figure 31 is completely similar to Figure 3F. In the operator box 40a The same goes for shifts. Relocation based on scale factor correction is performed in boxes 42b, 4 This is done in both the X and Y axes, as shown at 2C. Judgment box 42 d is the scale factor correction to be applied if the X-axis distortion is greater than or equal to the Y-axis distortion. Whether it is an equal COV/SIW pair or SIν where the Y-axis strain is greater than the X-axis strain /COW pair.

If not, the D-T control to 66 element windows in 43a and 15 entries in 44a. and the restriction of the TR distance to 7 (KR = 7 cents) as shown in 44c. ) is the same as that in FIG. 3F. The purge routine 31d is , a line edge outside the window 43a around the search point Si to which the file point Fi corresponds. The purpose is to eliminate entries.

Star i dish ■ Soukekikai Figures 6A, 6B, and 6C show variations of the embodiment shown in Figures 3A to 3F. Ru. This is the presently preferred embodiment and is superior to the embodiment of Figures 3A-3F. It has three advantages.

Figures 6A, 6B, and 6C correspond to Figures 3D, 3E, and 3F, respectively. . Therefore, the explanations for Figures 3A, 3B, 3C, and 3C are as follows: This is almost the same as the embodiment shown in FIG. 6 except for the explanation.

The first explanation relates to Figure 3B. According to the embodiment of FIG. 6, KR is set to a value of 7. will be cut. If this is done, the decision box TR: KHR must be included. It has been found that there is no.

Figure 6A shows the currently preferred purging method for the difference table (D,T,'') It is something. This method differs from the method shown in Figure 3D mainly for the following reasons. Ru. That is, the purge in FIG. 6A removes all multiple files and multiple search points (Is and JS points) for the search points represented by the initial difference table. The implementation takes into account the fact that it results in a purged delta table with no entries. In other words, it incorporates the "at-back" J characteristic. be. Purged difference table for each IS point populated into the initial difference table It has been found to improve the accuracy of containing file entries. Therefore, search and After the file purge routines are pursued, the raw data of all multiple IS and JS points is Highest Q score difference table entry for any 13 points not represented by purge It is desirable to undo the file. As a result, there are more line edges than search points. There are no line entries, and there is a net entry such that there is one line entry for each search point. A delta table with purges is given. This is a search point for the initial certification routine. There are few line entries, so the search points that are not represented in the initial difference table are It is no longer represented in the purged delta table.

The method shown in Figure 6A requires the maintenance of three buffers. 1st The buffer MS is called a mark search buffer. This MS buffer is the initial difference Each has 13 points that are members of the minute table entries. MS buffer space The lot consists of 13 entries on the purged difference table after the conversion purge. Used to indicate whether or not to be maintained as a member.

At the beginning of the purge routine, this MS buffer contains a "zero" table for each of the 13 points. give an indication. This zero display is not shown in the difference table where 13 points were purged. It means that.

The purge routine proceeds and the line entries are to be kept in the delta table. When determined, the MSS buffer entry for the corresponding 13 points, the initial difference table 13 points are represented in the purged difference table by one of the entries from Bull. The display is changed to "1" indicating that it should be done.

The mark file buffer MF has exactly the same structure for each file point JS and is perform exactly the same function.

The purge routine is executed and the purged difference table contains IS and JS points. information that should be represented is formed in the MS and MF buffers. determine which IS points are unrepresented after the purge routine is completed. A sequence of information is given that sets the routine to be executed. Then each one of the purged difference tables Select the appropriate line entry from the original difference table to contain the 3-point display. can be done.

The third buffer P is called the purge buffer. This is the label on the initial difference table. Each in-entry has a slot. This P buffer is on the difference table. Whether the corresponding line entry should be kept or the purged delta table It merely contains an indication of whether or not the item should be excluded from the bull.

The indication that the P buffer slot is "zero" indicates that the initial ordered difference The corresponding line in the table should be kept in the purged delta table. means. The P buffer slot display “1” is in the initial ordering difference table. means that the corresponding line in the purge should be excluded from the purged delta table. Taste. In this way, the table of "zero" and "1" status in P-buffer The use shown initializes all slots in each of the three buffers to zero. It differs from the use of status indications in MS and MF buffers in that it has advantages. It is what it is.

Referring now to FIG. 6A, all of the operations at 53a, 53b and 53c initializes the three relevant buffers to zero, status in each slot of each buffer. It shows that As shown in Figure 53d, the number of slots in the P buffer is initially It is equal to the number of lines (St) of the difference table. As shown in 53e, the MSS buffer The number of face slots is equal to the number of search points Sct, and as shown in 51f, MF The number of buffer slots is equal to the number of file points Fct.

The routine shown generally at 54 in Figure 6A represents the raw purge routine. . Each of the rstJ line entries in the initial ordered difference table is stored in MSS buffer I. It is compared at 54a with the status indicated for point S. These line entries The lis are processed as such in order of Qs value. When this IS slot is zero, its The difference table entry is the step at the JS point in the MF buffer, as shown at 54b. compared with the data.

When this status is also zero, its The status of the 13 line entry points and the JS point is the MS and MF buffer, respectively. It is changed to "1" of Fufa. This routine executes the entry on the initial ordered difference table. Repeated for each bird. 13 points for any of these entries Or if any of the JS points are already in the MS or MF buffer with status rlJ. When given, it corresponds to the difference table line, as shown in 54d. The P buffer slot has its status changed from “0” to “1”. It becomes.

Thus, upon completion of the routine indicated at 54, the P-buffer will contain a set of differences. It maintains the display '0' for the minute table line entry, which contains the IS point and No duplicates of JS points are given. This P buffer is used by all other differential table The in-entry will be displayed as "1".

The routine indicated at 55 stores in the P buffer the routine indicated at 54. A delta table line entry with IS members that are not represented after a raw purge Add "0" status (retention indication) to the bird. As shown in 55a and the status "0" (this is not indicated by IS according to the raw purge of 54). It means something. ) is identified for each MSS buffer slot. Next, 55 As shown in b, each difference table entry in Qs order is compared and the member Identify the difference table entry with the missing IS from the MSS buffer.

55c, the P buffer corresponding to the identified difference table entry The slot has its status changed to "0" and its line entry The status of the MS buffer slot indicating the Is member of the resource is changed to "1". has been done. In this way, the P-buffer stores each IS member of the initial difference table. indicates at least one and only one line entry from the initial delta table with things are caused. The purge routine of FIG. 6A is ordered by proximity value Qs. The highest Qs for each Is were purged. given to the line on the difference table.

Purgil similar to the equivalent routines shown in Figures 3D, 3E, 3F, and 31. The routine 31e stores all line entries whose P buffer has status “1”. Perform the actual purge.

FIG. 6B shows the CIB correction factor as an addition to the routine of FIG. 3E, or As the COW and STW correction factor routines are removed from the configuration of Figure 3H. I can understand. Basically, the SXT value from each search point IS is shown in Figure 3E. The correction factor CO■ is calculated as shown in 56a, and the correction factor CO■ is 0. Calculated from the average SXT after excluding SXT values of 2 or more. In Figure 6B The criterion shown at 38d in Figure 3E is not used and therefore the relatively small XI Search points with values are used in the embodiment of FIG.

Similarly, as shown in Figure 3H, the WXT value and the corresponding SIV correction are Calculated in the illustrated example. Here again, the DD proximity criteria (38d and 3 in Figure 3H) 9d) is not used in Figure 6B. However, E　E　 The window generated by the criterion =0.2 is retained (see 56b).

The configuration in Figure 6B is as shown at 56c between Sk and FkO (see Figure 3J). Take the X-axis distance given as XIJ (i.e. XI-XJ) and calculate the X and Y distortions. It can be thought of as normalizing the assumed error for . in this way ``If error JXIJ is divided by the distance in the X-axis direction between S' and Sk, then XIJ is normalized to the distortion (see 56d). Also, this error XI, If L is divided by the distance in the Y-axis direction (Yl) between S' and Sk, then for the strain in the Y-axis direction, XIJ is standardized. As shown in 56e, the SXT value is added and further WX The T value is added. Using these average values, COv and Give SIV. As a result, both COV and SIV are in the X-axis direction between Sk and FkO. Gives a measure of "error".

The purge routine 31f in FIG. 6B is performed, and the correction factor component SXT or Or, if WXT is given, the in-entry is excluded from the difference table. 56a and 56 As shown in b, when any of these components is 0.2 or more, the purge bag is The file 58 corresponds to the line entry given as status "1". You will have lots. This purge routine 31f is the purge routine 31f shown in FIG. works exactly like Line 31 to eliminate that line entry, and Make all line entries below the specified line entry into one line. .

The shift correction shown at 40a in FIG. 6C is the same as in the embodiment of FIG. 3F. .

However, as shown in box 57a, the correction of the 5tOX value is based on the cov and SIV factors. This is taken into consideration.

As shown in Fig. 57b, an image is formed on the image plane around the shifted and corrected Si point. be done. This window extends 5 pixels in both directions along the X axis from point Si, and expands 20 pixels in both directions. Extend in both directions along the Y axis from point Si. The corresponding file point Fi is outside the window Once exited, the corresponding line entry is added to the difference table in the manner shown in Figure 3F. be excluded from

Segment system (4A's and 4IL! Fingers given over time) Distortions and changes in properties are partially achieved by the system described above with reference to Figures 2 and 3. be compensated. However, in one preferred application of the above system, fingerprint or The finger image is divided into four segments.

These are four vertical segments.

According to one example of scanning, most of the distortion occurs along the X-axis.

Therefore, in this example, the X-axis is divided into four segments, one segment The zone associated with the ment will extend along the entire Y axis. In Figure 4A As shown, this includes four separate file minutia stores 18a-18d and and four search point storage devices 162 to 16d. Corresponding units in Figure 4A As shown in the figure, various file details are stored in storage devices 18a-18d. The points overlap considerably. In particular, this overlap is given over 60 pixels, Therefore, the first file point storage device 18a stores file images from X=o to X=94. The one-way file point storage device 18b has X of X=34 to X=158. It will cover file points that have values. However, the search point storage values may overlap. There's nothing wrong with that. As shown in Figure 4A, each search segment is Data 17a-17d are compared against corresponding file segments. here This comparator is essentially the device shown in FIGS. 2 and 3. especially, From the comparison and screen unit 20 to the calculation unit 5o shown in FIG. All the units in are connected to each of the four comparators 17a to 17d as they are used.

As shown in FIG. 4A, each of the units 17a to 17d in FIG. The five items of specified output data shown in the figure are given. This output data is shown in Figure 5A. and processed by the scoring system in Figure 5B to provide verification of identification. .

The four segments shown in Figure 4A are vertical segments, and these are the segments that represent the image. It is divided into four overlapping zones along the X axis. Each segment is Because of this, it covers over 173 statues.

Figure 4B shows a 16 subspace in which the image is divided into a set of four segments. FIG. 4A shows the configuration of FIG. 4A into the system. Above for the embodiment of FIG. 4A. The comment made in Figure 4B is based on the fact that 16 comparators 17 are used in the embodiment of Figure 4B. The same applies to the embodiment of FIG. 4B, except for the following. Therefore, There are 16 sets of output data, one set from each comparator, thus one set represents each of the 16 subspaces. This output data is shown in Figure 5A. and processed by the scoring system in Figure 5B to provide verification of identification.

Verification speed is important in access systems, and the described multiple contrast arrangement operate in parallel. Furthermore, in the access system, an array of search point sets is used. control is relatively difficult and therefore requires significant segment overlap. There is a point. On the other hand, in a fingerprint identification system, for example, a fingerprint card is a file point. gives a set of search points to be compared with the set of , but with slightly greater flexibility over time. can be used and also provide greater control over alignment. is possible. Therefore, in such a system, the The overlap is fairly small, or if the alignment is properly controlled, there will be no overlap. becomes unnecessary.

Important in overcoming errors introduced by unavoidable differences between search image and file image The important thing is the segmentation of the field and the segmentation of each field within this field. The use of scale factors specific to the Low Type 1 and Low Type H E It is the above combinations that give the power of this system that give the combination of ra. .

In particular, the linear scale factor calculated for each segment is This facilitates the identification of clusters. Distortion and noise present in fingerprint files are class make identification of data difficult. However, when the segments are smaller, there is a simple method to compensate for the distortion. The ability of simple scale factor correction reduces cluster masking. in this way Therefore, segmentation is particularly useful in facilitating the identification of fingerprints buried in noisy environments. It is useful for

An important feature of the system of the present invention is that it segments the fingerprint image into a number of individual zones; It consists in applying comparison and matching methods to each zone or segment separately. . This multi-segment device is described in connection with FIGS. 4A and 4B. 5th A.

The scoring method shown in Figures 5B and 50 is based on the quality scores from each segment of the matcher. This is a scoring method that uses cores. Thus, regarding the description of FIGS. 2 and 3A-3J, In turn, the described routine is iterated for each image segment analyzed. It should be noted that

The corrected 1, D, -5A and 5B in Figure 5A are 1 of the 4 score units. 3 is a flowchart showing two operations. Each of these scoring units has 4 are coupled to individual outputs from individual comparator segment units 17a-17d. It will be done. Thus, the following details of the identification method are given in the four-segment embodiment of FIG. 4A. given based on.

It is recognized when any of the four sets of output parameters satisfy the verification criteria shown in Figure 5A. This means that something else has been verified. This is a search in any of the 4 segments. There is a quick procedure when there is a high degree of correlation between points and file points, as shown in Figure 5B. This eliminates the need to use a

In particular, if the number of lines ST in the final restricted difference table is 4 or more, the QT ratio is 0. ．． If it is equal to or greater than 85, then 1,D,is considered verified. child The QT ratio of QT and the sample in the restricted set obtained from the process of Figure 3F is The ratio between the QT that would be obtained if there was a perfect match between the It is. For example, as shown in Figure 5C, 5T=4 (i.e., 4 in the difference table). QT=42 gives a perfect match. Figure 3F At the end of the process, = 7, so Qc cannot be greater than 7. I want to be noticed. In this way, if a perfect match is obtained, the raw QT-84 will be 0 ．． 5 x raw QT = r42J. Figure 56 is for difference tables of other sizes. Figure 5A shows the QT values for other perfect matches. , including the remaining of the five parameter outputs that allow identification verification. There are many other identification paths, but these paths are self-explanatory from FIG. 5A.

The important thing to realize here is that besides the identity being verified, there are two main outputs: It means that there is. One of these outputs is the correlation between the search point and the file point. Seki is 1.0. Although not large enough to provide verification of If the value is large, add the value "1" to the [Vote] J parameter. It includes that. However, if these criteria are not met, please refer to Figure 5A. The output actually gives a value ■ equal to "0".

In this way, when identity is not verified under one of the paths in FIG. 5A, The output from the unit in Figure 5B is (a value equal to ``1'' or ``0'') is applied to

The inputs for FIG. 5B are 1. for any of the four units in FIG. 5A. D , is not verified, the output from each of the four units in Figure 5A It is. As shown by the loop containing action box 52a, the value of ■ and the value of FIG. Other output values from the unit are for four fingerprint segments (n=1 to n=4). and then added.

5T (sum of f (total number of 4 combined comparator units) is "25" If they are equal to or greater than that, the identification is verified, and if the added ST value is 6 or more, In the following cases, identification will not be specifically verified. However, as shown in the figure, ST is between 6 and 15 and the sum of SQ is greater than 100, other considerations for verification path exists and the identity will therefore be verified. Also, as shown in Figure 5B, There are two other paths to identity verification.

As indicated by action 52b, the segment product core SQA of the individual comparison device but also due to the relationship between the four butting device segments taken two at a time. Final comparison considering the determined mutual contrast device segment scores SQ, , fi A combined value of the device's segment score SQ is calculated. Separate contrast device segmentation The calculation of the client quality score SQ,1 is shown in action box 50b of FIG. 3C. death However, as shown in operation box 52b of FIG. 5B, the various segments of the comparison device The combined value of quality scores SQ from segments SQ, , is the quality score for each segment SQ, . calculated to take into account the relationship between the quality scores of each segment as well as the quality scores of each segment. . These relationships SQ, (i.e., SQ+z, SQ+3.SQ14 , SQZ3, and 5Q34) is the QT ratio (see box 51a in Figure 5A). ) and the SQ value (see Figure 4A).

As can be seen by considering the restricted case, for each of the two segments being compared, The SQ value between two segments when added to the SQ value is the combined , equal to the SQ value for two segments considered as one segment.

According to the formula SQ, , the product of two ST values is gives a factor representing the additional pairs that are possible when considered as a ment.

For example, assuming there is a perfect match, therefore with particular reference to the table of Figure 5c, If segment rmJ has 5 lines in the difference table, i.e. 5T=5) , and assuming that segment rnJ has 6 lines in the difference table, the individual The possible pairs of these two segments taken separately are for a total of 25 pairs. It can be seen that they are 10 and 15, respectively. However, these two One segment is taken as a unit for a total of 11 lines in the difference table. If so, there are 55 possible pairs between these 11 lines. two STs When multiplied by the values, the result is 5×6=30. The value “30” is added to “25”. If it is, the result will be "55". In this way, for segment m The product of ST for segment n and ST for segment n is the segment when there is a perfect match. provides an appropriate factor to compensate exactly for what is lost due to perfect match If not, the ratio of the two proximal values (QT ratio) is an appropriate reduction of the contribution of SQ, , Give a grading.

In legally mandated work, the identification procedures in Figures 5A and 5B are not required. There is a possibility that A predetermined small number of items will be withdrawn with the highest priority for further consideration. The sequence of quality scores SQ (according to the calculation in box 52b) is It may be possible to just link it. In such cases, search For each of a predetermined number of files selected for the category of fingerprints that the code consists of. are compared, and the one with the largest SQ value is selected. For such applications In this case, the 1, D, and L thresholds shown in Figures 5A and 5B are no longer used.

Zeng Ji t, (''''''7 and 8) The value of a contrast device is that the file points are a reasonable number of repeatable traditional details. This can be greatly improved by having adequate and valid records to include:

Therefore, the recording method shown in Figures 7 and 8 is of considerable importance. Record-scanning methods usually give good, usable recording files, but If you want to create a recording file according to the recorder who visually selects the points that The manual recording method of Figure 8 can be used.

One useful advantage of this type of contrast device is that once the repeatable basic details are given, The reason for this is that it can be identified even if there are other points in the file. That is, , the purge and shift method eliminates noise-induced errors so that proper identification can be performed. Gives the final difference table excluding file entries.

This allows the use of a variety of recording and verification techniques, e.g. Ten points from each finger can be recorded in the same file. At this time , by specifying a search point with any one of the three fingers above, the person in question can be identified. become capable.

As another example, an individual for whom the number of relevant details to repeat is very small (e.g., 4) may This can record the same finger three times separately and provide 12 file points. will cause a Type I error for the individual even if there are only a few search points that provide validation. greatly reduces the risk of

As a third example, it is possible to record the fingers of two individuals in the same file.

If a high degree of security is required, separate fingers from two individuals at the same time may be used. Access can be permitted only if the search file contains the details shown in It is Noh.

As shown in the flowchart in Figure 7, the basic automatic recording procedure This provides the first or initial recording scan (N=1) for forming the file.

For each of these scans, the finger to be recorded is subjected to 6 consecutive scans (N=2-7). The initial (N=1) recording scan is performed using the extracted points. Compare with the points obtained when . Once a match is obtained, that point in the initial recording scan The number of scans verifying each point during the initial recording scan is marked as verified. Determine the rNJ value at that point. 60c, the reduced recording file An initial recording file is created and verified at least four times, i.e. N=4 or more. limited to the point of the file.

The comparison method in question is different from the one shown in Figure 2, except that no correction factor COv is included. The method shown in FIGS. In particular, as shown in FIG. As shown, when the value CC is set to "0", the C0vO value automatically becomes "1". Additionally, there may be windows or boxes that limit or screen the extent to which comparison points can be formed. The pixels are basically cents equal to 250 pixels by 25 pixels. This is shown in Figure 3. This is done with the XYK values shown equal to "12" cents. initial recording file The point is calculated as if this point appeared in the restricted difference table obtained from the output of Figure 3F. It is considered that only in this case, it was compared with one of the following recording scans. ID verification Since nothing is included, subsequent calculations of values such as Qt need not be performed. However, note The recording time processing is the verification time used up to the point where the ID verification step is required. processing is applied.

Next, as shown in Figure 7, the restricted file created in 60c is added to the basic file. The scan of the finger to be recorded is now performed three more times. These additional The three scans further enhance the verification of each point in the recorded file. At 60d As shown, the recorded file points are ranked in order by the number of verified rNJOs involved; Only points with at least 8 validation indications (N=8 or more) are retained. Other As a constraint, we set a maximum output file of 30 points based on the points with the highest number of verifications. file is established.

Since there is overlap between the file points of the four segments (file 18 in Figure 4A) 8 to 18d), it is possible to verify the initial scan point consisting of one or more times per scan. be. In this way, the criteria of 60c and 60d are obtained, which makes r It is contemplated that NJ is greater than or equal to "6" and "9", respectively.

King's death record A manual recording method can be used, as shown in Figure 8, in which case the recorder must 1) Formation of file points of Jl for one or two finger scans by a skilled person Assume that you are familiar with a selection method based on visual inspection that allows you to This method is automatic It is faster than recording and often gives better files. .

As shown in Figure 8, a single scan is made and the extracted points are given to the recorder. , the recorder performs the procedure of selecting the extracted points one by one. 40 recorders When you select the point point, the recorder pl is completed and these 40 or more points are saved in a file. Ru.

If the recorder selects fewer than 40 points, he will perform extra scans and review them. can be provided with additional point sets for

In particular, the above-mentioned extracted points R are automatically arranged by the recorder so that they have a one-to-one correspondence on the image plane. given to. First counter N (K represents the number of points reviewed by the recorder) The second counter Nytts is selected by the recorder as the content to be stored in the file. gives an indication of the number of points above. The selection routine in this regard has been reviewed. 0 judgment that continues as long as the number of points Nck is less than or equal to the number R of points extracted in the first scan As shown in box 62a, this selection routine allows the recorder to select 40 points. It ends when finished. If the number of selected points is 40 or less, the routine is 62b.

Continuing as shown in 62C and 62d, the recorder will record the number of points reviewed. Add points one by one until R or more or the number of selected points equals 40. Continue the selection operation.

All of the points R during scanning are considered and if the selected number does not equal 40, the decision box is As indicated by box 62e, the archivist decides to perform a second scan or Or terminate the process and set those points in the file. The second scan is performed Then, the counter Nck, which displays the number of points received, must be reset to "1". However, the counter Nyas is not reset. For this second scan In this case, the recorder will select a total of 40 points between the two scans according to the criteria. or the recorder decides to end the selection process and there are 4 Select whatever additional points are needed until you store a point less than or equal to 0. can be done.

Figures 9 and 10 illustrate the optical scanning method, which allows the basic image and data is given, search points are extracted from these images and data, and the storage 12.

Figures 9 and 10 show how a fingerprint is optically scanned and a modulated light beam is manually applied to a CCD array. 77 is a diagram illustrating a known mechanism for transmitting data to A 77. This is an already issued patent 4.322.163 and will not be discussed in detail here. Therefore I will stop here with the following explanation. i.e., a light beam such as that provided by laser 66 are suitably collimated by lenses 68 and 70 to provide an interrogation beam 71. Habo Toru A clear platen 62 is provided as a base on which the individual fingers F are placed. This plate The finger 72 is mounted in the movable carriage 74 so that the finger is guided along the interrogation beam 71. will be moved. As a result, a pattern of protrusions and recesses provided on the back surface of the platen 72 As a result, the reflection pattern is modulated with fingerprint information. By the condensing lens 76 The image carried by the reflected light beam is focused onto a linear array 77 of photodiodes. It will be done.

An encoder element 78 mounted on the carriage 74 detects movement of the carriage 76. and generates a synchronization signal every time the carriage moves a predetermined distance. This synchronization signal Scanning circuit 80 interrogates each of the photodiodes in array 77 sequentially.

In this way, one series of pulses is applied from the scanning circuit 80 to each scanning line.

Each pulse is representative of an image element or pixel.

The scanning mechanism of FIGS. 9 and 10 illustrates the finger scanning method in which the comparison device of the present invention is used. Please note that this is merely an example.

In this way, spatially located fingers can be scanned without fingerprint card scanning or platens. It becomes possible to input the modulated light beam obtained from the direct scanning of the array 77 into the array 77. .

to 1 difference eyebrow Be Figure 3F Figure 31 m, s used as single sunken wood. Figure 5A for each of n. Matchers used at the same time 2, so the maximum ST is 15 After correction, P, BKR=7 destroyed rm Figure 6A Fixed grade Figure 60 Automatic enrollment (scanning) Figure 8 6- Subject of correction Procedural amendment (formality) October 1986 Mr. Kunio Kogawa, Commissioner of the Patent Office 1.Display of the incident PCT/US86100635 2. Name of the invention comparison device 3. Person who makes corrections Relationship to the incident: Patent applicant Name: Finger matrix.

Incoholated 4. Agent Address: 8-10-5 Toranomon-chome, Minato-ku, Tokyo 105, Date of amendment order (11 Translations of the specification and claims (2) Translation of drawings 7. Contents of correction (11 Translations of the specification and claims (Contents December 1986, L date) Amended claims 1. (Correction) Use the difference table to identify one set of search identification points as one set of files. An improved comparison device that compares one set of search identification points to one set of two plants first memory means for storing as a number; a second memory means for storing a set of file identification points as a set of two plant numbers; Dan, 11 file divination and search divination 0 are interrelated... 1 divination above is applicable. each one is combined with a predetermined subcent of the search identification point, and multiple comparators coupled to corresponding predetermined subsets of the file identification points. comparator means, each of said multiple comparator means being a subset of said file points; giving a proximity value for each of said subset of search points with respect to the search points; between the search and file points within the corresponding subcent. are adjusted separately as a function of the displacement and strain of the previous point for a subset of file points. means for providing adjusted approximations for each of the search points; For each, a smaller number is determined based on the adjusted proximity value for that search subcent. means for providing at least one figure of merit; The figures of merit from each of the search subsets are combined to form the set of search points. and at least one overall figure of merit representing the degree of agreement between the set of file points. A comparison device comprising means for comparing.

25. (Correction) Use the difference table to convert IMi file identification points into a set of search identification points. A mechanical method that automatically contrasts another point, storing one set of search identification points as a set of two plant numbers in the first memory means; , storing one set of file identification points as a set of two plant numbers in the second memory means; And Saratei Otoriri Study J≧Vice] Shin 1. ! Those who perform fortune-telling on the Segment the set of search points and the set of file points into predetermined corresponding subsets. for each of the corresponding search and file subcents. compare individual searches and file points, and compare them for a subset of file points. providing a proximity value for each of the search points; a function of the search and file points in each of the proximal values and adjust for each of said search points with respect to the subcent of the file point. a step of giving an approximation value of Based on the adjusted proximity value for the search subset combining said figures of merit from each of said subsets; and at least one indicative of the degree of coincidence between the set of search points and the set of file points. a mechanical method comprising the steps of: providing an overall figure of merit;

international search report 1″“′″6MI　Ae″+=s+:a′N. Pm〒/+I<RA/nnAig

Claims (1)

[Claims] 1. A set of search identification points is compared to a set of file identification points using a difference table. an improved comparison apparatus for storing a first memory means for storing a set of search discrimination points as a set of binary signals; and a second memory means for storing a set of file discrimination points as a set of binary signals. memory means; and multiple comparator means each coupled to a predetermined subset of said search identification points and a corresponding predetermined subset of said file identification points, each one of said multiple comparator means being coupled to a corresponding predetermined subset of said file points. Pair each of the said proximity values with the one that gives the proximity value for each of the subset of points. the displacement and distortion functions between said search and file points in the corresponding subset and means for individually adjusting and providing an adjusted proximity value for each of said search points with respect to a subset of file points; means for providing a figure of merit; and means for combining the figures of merit from each of the subsets to provide at least one overall figure of merit representative of the degree of agreement between the set of search points and the set of file points. A comparison device. 2. 2. The comparison device according to claim 1, wherein the file point and search point signals represent detailed identification points on the upstream side of a fingerprint image. 3. The subset of the file points is a superimposed subset on the image plane. Comparison device according to item 2. 4. each of said multiple comparator means is coupled to a respective one of said subsets of said search and file identification point signals to provide an initial difference table; and for each entry on said initial difference table, a first neighbor. The first subsection giving the value a first difference table, and a second difference means, which calculates a sample from the initial difference table extract a set of entries representing that there is no copy of the search point entry, and base the extracted set of entries on the entry in the initial difference table that has a higher proximity value. a second proximity value for each entry on the purged difference table; a second subset comparator means, wherein said second proximate value is a proximate value to be adjusted by said adjusting means. Comparison device according to item 1 above. 5. The adjusting means is means for performing difference plane transformation of an entry on the purged difference table having the highest proximity value to the origin of the difference plane represented by the purged difference table, and A move in which all entries are converted by the same amount a correction factor that differentially adjusts each of the entries on the purged difference table; means for providing a child of the purged difference text after the transformation adjustment; 5. A comparison device as claimed in claim 4, further comprising: means that is a function of the average of the difference values of the entries of the table. 6. Each of the means for providing a figure of merit calculates a third difference based on the adjusted difference value. third difference means for providing a partial table, wherein an entry on the third difference table corresponds to an entry in the purged difference table having the adjusted difference value; A third comparator giving a third proximity value for each entry. means for providing the at least one figure of merit based on the third proximity value; third subcomparator means for providing the third proximity value based on the adjusted difference value; and means for providing the at least one figure of merit based on the third proximity value. A comparison device according to claim 4. 7. Each of the means for providing a figure of merit calculates a third difference based on the adjusted difference value. third difference means for providing a partial table, wherein entries of the third difference table correspond to entries of the purged difference table together with the adjusted difference values; and each entry of the third difference table. a third subcomputer that gives a third proximity value for controller means, wherein the third proximity value is provided based on the adjusted difference value. and further comprising means for providing the at least one figure of merit based on the third proximity value. A comparison device according to claim 5. 8. The contrast device according to claim 5, wherein the difference value used to provide the average, which is the basis of the correction factor, is limited to a difference value that is less than or equal to a first predetermined magnitude. Place. 9. 8. A comparison device according to claim 7, wherein the difference value used to provide the average on which the correction factor is based is limited to a difference value that is less than or equal to a first predetermined magnitude. 10. 5. The comparison apparatus according to claim 4, wherein said second difference means extracts one initial difference table entry representing each of said search points. 11. 8. A comparison apparatus according to claim 7, wherein said second difference means extracts one initial difference table entry representing each of said search points. 12. a restriction hand that is combined with the third difference table to provide a restricted difference table; 7. The method according to claim 6, wherein the limited difference table further comprises a limiting means for limiting the limited number of entries to a predetermined number of entries based on the entry having the highest proximity value to the third difference table. Contrast device. 13. a restriction hand that is combined with the third difference table to provide a restricted difference table; 8. The method according to claim 7, wherein the limited difference table further comprises a limiting means for limiting the limited number of entries to a predetermined number of entries based on the entry having the highest proximity value to the third difference table. Contrast device. 14. a restriction hand that is combined with the third difference table to provide a restricted difference table; 12. The method according to claim 11, wherein the limited difference table further comprises a limiting means for limiting the limited difference table to a predetermined number of entries based on the entry having the highest proximity value to the third difference table. Contrast device. 15. The entries of the initial difference table are divided into pixels that are within a predetermined distance from each other on the image plane. 8. A comparison apparatus as claimed in claim 7, further comprising first window means for limiting the search and file points. 16. The entries of the third difference table are located within a predetermined distance from each other on the difference plane. The claimed invention further comprises second window means for restricting the comparison of the difference table entries. Comparison device according to item 12. 17. 14. The comparing device according to claim 13, further comprising a second window means for restricting entries of the limited difference table to difference table entry comparisons that are within a predetermined distance from each other on the difference plane. 18. Use a difference table to match one set of file identification points to one set of search identification points. An improved comparison device for combining a set of search discrimination points with a set of binary signals. first memory means for storing a set of file identification points as a set of binary signals; second memory means for storing a set of file identification points as a set of binary signals; a first difference means for providing a first proximity value for each entry on the initial difference table; and a set of entries representing that there is no copy of the search point entry from the initial difference table. a second difference means for extracting, wherein the set of extracted entries is the closest one; a second difference means for providing a purged difference table based on an entry in the initial difference table having a purge value, and further providing a purged difference table based on the extracted entry; 2nd comparator giving 2 proximity values adjustment means for adjusting the difference value of the purged difference table as a function of the second proximity value to partially correct for shifts and distortions between the search and file points; and the adjusted difference value. a third difference means for providing a third difference table based on the third difference table, wherein the entries on the third difference table correspond to the entries in the purged difference table together with the adjusted difference value; 3. A third comparator hand that provides a third proximity value for each entry in the 3-difference table. step, wherein the third proximity value is a third value given based on the adjusted difference value. and providing at least one figure of merit based on the third proximity value. A comparison device consisting of a means for comparing. 19. 19. The comparison device of claim 18, wherein the file point and search point signals provide minutiae identification points on the image plane of the fingerprint. 20. The adjustment means is configured to provide a difference plane transformation of the entry having the highest proximity value in the purged difference table to the origin of the difference plane represented by the purged difference table. means for providing a difference plane transformation in which all of the entries on the purged difference table are transformed by the same amount; Means for providing a correction factor that differentially corrects each of the entries, the correction factor The child is the average of the difference values of the entries on the purged difference table after the conversion adjustment. and a correction factor providing means provided based on the Comparison device included. 21. the difference values used to provide the average on which the correction factor is based are limited to difference values less than or equal to a first predetermined magnitude, and are combined with the third difference table to provide a restricted difference table. It's a means Therefore, the limited difference table has the highest proximity value to the third difference table. 20. The comparison device according to claim 19, further comprising a limiting means for limiting to a predetermined number of entries based on the entry. 22. The entries of the initial difference table are divided into pixels that are within a predetermined distance from each other on the image plane. and a second window means for restricting entries of the third difference table to difference table entry comparisons that are within a predetermined distance from each other on the difference plane. Comparison device according to item 19. 23. a first window means for limiting the entries of the initial difference table to search and file points that are within a predetermined distance from each other on the image plane; Te 21. The comparison device according to claim 20, further comprising second window means for restricting table entry comparison. 24. The entries of the initial difference table are arranged as a first window means for restricting the entries of the third difference table to points and file points; 25. The comparison device according to claim 21, further comprising a second window means for limiting the blue entry comparison. A set of file identification points is used as a set of samples using a difference table. A mechanical method for automatically comparing search discrimination points to search discrimination points, the method comprising: storing a set of search discrimination points as a set of binary signals in a first memory means; and storing a set of search discrimination points as a binary signal set. segmenting the search point set and the file point set into predetermined corresponding subsets; and storing a separate search for each of the corresponding search and file subsets in a second memory means. and file points, and compare the previous with respect to a subset of file points. providing a proximity value for each of the search points; and adjusting each of the proximity values as a function of the search and file points in the corresponding subset, and providing an adjusted proximity value for each of the search points with respect to the subset of file points. providing at least one figure of merit based on the adjusted proximity value for the search subset; and combining the figure of merit from each of the search subsets to determine the set of search points and the set of search points. indicates the degree of agreement between a pair of file points. and providing an overall figure of merit. 26. 26. The method of claim 25, wherein the file point and search point signals represent minutiae points on an image plane of a fingerprint. 27. 27. The method of claim 26, wherein the subsets of file points are superimposed subsets on the image plane. 28. The comparing step provides an initial difference table for the corresponding subset of search and file points. providing a first proximity value to each entry on the initial difference table; and extracting a purged difference table from the first difference table, wherein the purged difference table is searched. A set of entries that do not have a copy of the point entry. an entry in the purged difference table is given based on the entry in the initial difference table with the highest proximity value; and further determining a second proximity value for each entry on the purged difference table. providing a second proximity value, the second proximity value being the proximity value to be adjusted by the adjustment step; 28. The method of claim 27, comprising: step. 29. The adjusting step includes shifting the entry on the purged difference table with the highest proximity value to the origin of the difference plane represented by the purged difference table; shifting all other entries on the purged difference table by the same amount; and then shifting the purge based on the average of the difference values of the entries in the purged difference table. a step of correcting each of said entries on the stored difference table by a correction factor; and providing a third difference table based on the shifted and corrected entries. 29. The method of claim 28, further comprising: providing an adjusted proximity value based on the third difference table. 30. Using a difference table, one set of search identification points is matched to one set of file identification points. 1. A mechanical method for automatically comparing a set of search identification points as a set of binary signals in a first memory means; the step of storing in the second memory means as a signal; comparing the search and file identification point signals to provide a set of difference table entries for an initial difference table; and calculating a first proximity factor for each entry on the initial difference table. extracting a second difference table from the initial difference table, the second difference table being limited to a set of entries that do not have a copy of the search point entry; an entry in the second difference table is given based on an entry in the initial difference table having a larger proximity value; and a step of calculating a second proximity value for each entry on the second difference table. , adjusting the difference value of the second difference table as a function of the second proximity value to partially correcting shifts and distortions between sets of points and file points; and comparing the adjusted entries in said second difference table; providing a third proximate value for each entry of a bull; and providing at least one figure of merit based on the third proximate value. 31. 31. The method of claim 30, wherein the file point and search point signals represent minutiae points on an image plane of a fingerprint. 32. The adjusting step includes assigning the entry on the second difference table having the highest second proximity value to the second difference table. converting to the origin of a difference plane represented by a square, and converting all other entries on the second difference table by the same amount, and difference values of the entries in the second difference table after the conversion adjustment. and differentially adjusting each of the entries on the second difference table by a factor proportional to the mean of the second difference table. 33. 33. The method of claim 32, further comprising the step of limiting the difference values used to give the factor to difference values that are less than or equal to a predetermined magnitude. 34. The number of adjusted entries in the second difference table is set to the highest among the second proximal values. further comprising limiting to a predetermined number of entries based on those having a high value. 34. The method according to claim 33. 35. Each entry in the difference table is set as a sensor located within a predetermined distance from each other on the image plane. as claimed in claim 30, further comprising the step of limiting to How to put it on. 36. Each entry in the difference table is set as a sensor located within a predetermined distance from each other on the image plane. as claimed in claim 32, further comprising the step of limiting to How to put it on. 37. The entries of the difference table are arranged within a predetermined distance from each other on the image plane. 31. The method of claim 30, further comprising the step of restricting to entry comparisons. 38. The entries in the difference table are divided into difference tables that are within a predetermined distance from each other on the difference plane. 37. The method of claim 36, further comprising the step of limiting to table entries.