CA1167166A

CA1167166A - Method for analyzing stored image details

Info

Publication number: CA1167166A
Application number: CA000379718A
Authority: CA
Inventors: Clifford H. Moulton
Original assignee: Individual
Current assignee: Individual
Priority date: 1980-07-03
Filing date: 1981-06-15
Publication date: 1984-05-08
Also published as: EP0043988B1; DE3172892D1; JPS5752972A; US4371865A; EP0043988A1

Abstract

Abstract A method for analyzing stored image details for identification purposes is disclosed in which slopes are abstracted from an image to provide three-dimensional recognition information. Data representing light levels of an image are stored in a picture memory device, which is analyzed in a predetermined manner to select absolute illumination magnitudes between fixed locations of the image. This information is directly related to the slope between the locations. Steeper slopes and their correspond-ing locations are stored as recognition data in a learn mode. In an access mode, the previously-obtained information is utilized to locate new data, and depending upon the degree of correlation therebetween, an indication of recognition is either verified or rejected.

Description

I J 6 7 7r. 6 B
rl~
METIIOD FOR ANALYZING STOR~ IMAGE DETAILS

Background of the Invention Electronic identification systems are utilized in a number Gf applications in which verification of personal identity lS required, sucn as to facilitate banking trans-actions and to permit access to restricted areas. Some ofthese systems merely read coded information magnetically stored on a plastic wallet-sized card, while more sophisti-cated systems are designed to provide a positive identifica-tion by reading an actual physical recognition pattern which is unique to an individual and then comparing the data derived therefrom with previously~stored data derived from the same pattern source.
U.S. Patent No. 4,186,378 teach~s an electronic identification system in which the palm of an individual's hand is scanned in a predetermined manner by an 1mage-sensing camera, and an image pattern corresponding to the palm print is stored. Recognition data to recognize the palm on sub-sequent presentations of the palm are abstracted from the image pattern in accordance with the most prominent details.
Thus, analysis of the stored picture deta1ls is based on t~o-dimensional pattern aspects of the image.

Summary of the Invention The present invention is related to methods for analyzing stored image details in identification systems, and in particular to a method for analyzing stored image details in accordance with the three-dimensional aspects thereo~.
The image details are obtained and ~tored in the conventional manner, as taught in U.S. Patent No. 4,186,378.
That is, the palm of an individual's hand is scanned in a 13~;7~6~) predetermined manner by a camera such as a vidicon or a solid-state charge-coupled image sensor to produce an analog signai which is proportional to the light levels received from the palm. The analog signal, which may be difEerentiated to enhance the definition of ridges and valleys of the palm print, is quantized by a conventi~nal ana og to~digital ~
verter to provide raw numerical digital data corresponding to the various light levels of the pattern, and then this raw data is stored in a 100-element by lOO-element memory array to represent a complete three-dimensional picture of a palm identity pattern.
The picture memory may be represented by a Cartesian coordinate system, wherein the picture~element (~ixel) array defines the X-Y plane, and the amplitude of stored data lS representing illumination levels defines the Z axis~ The 10,000 pixels of the 100-element by 100-element memcry array represent coordinate points of a picture and are subdivided into smàll groups each of which has a predetermined number of vectors in fixed locations in the X~Y plane. The groups are analyzed in a predetermîned manner to abstract from each group the vector havin~ the largest absolute ma~nitude Z
component difference f~om origin to end point. Depending upon vector length, the Z component provides information as to the slope of the vector with reference to the X~Y plane.
The steepness of the slope is assigned a quality value, with steeper slopes having a higher quality. Total information a~stracted from each group of pixels, therefore, is vector location, direction, polarity, and quality, The abstracted vectors are then ranked in their order from hi~hest quality ~0 to lowest quality, and then stored in a separate ar~ay. From this information, prominent recognition ddta representative of three-dimensional aspects of stored image details, such as slopes and flexions (indicative of valleys and peaks), 1 1~7~
~3--may be selected and encoded in cornpact form for permanent storage.
For verification purposes, the palm pattern is read and stored in the same manner as the original raw data was obtained so that a correlation process can take place to atten,~t to match the compacted recognition data with newly-obtained data. sased on this correlation, a decision is made as to whether an identification is verified.
It is therefore one object of the present invention to provide a method of recognizing an image to provide identification thereof by comparing selected slopes with previously-stored slopes.
It is another object of the present invention to provide in an identification system a novel method of analyz-ing stored image details in accordance with three-dimension-al aspects thereof.
It is a further object of the present invention to provide in an automatic electronic identification system recognition data by comparing slopes between fixed locations of an image.
Other~objects and advantages will become apparent to those having ordinary skill in the art upon a reading of the following description when taken in conjunction with the accompanying drawings, Brief Description of _he Drawings FIG. 1 is a block diagram of an identification system which employs the analysis method of the present inven-tion;
FIG. ~ shows a fractional portion of a picturememory lying the X-Y plane of a Cartesian coordinate system;
FIG. 3 shows a fractional portion of a picture .

1 ;1 ~7:L66 memory subdivided into 9-pixel blocks;
FIGS. 4A through 4F illustrate a 9-element group of pixels and the analysis sequence thereof;
FIG. 5 is a diagram of the analysis pattern of a 4-pixel by 4-pixel array;
FIG. 6 is a diagram of the analysis pattern o~ a 5-pixel by 5-pixel array; and FIG. 7 shows a block di.agram of the details of the data recognition analyzer portion of the system of FIGq l.
Detailed ~escri~tion of the Invention Referring now to FIG. 1, a block diagram of an identification system which employs the analysis ~ethod of the present invention i$ shown. Generally, the overall system comprises a recognition data acquisition unit lO, a picture memory device 20, a recognition data analyzer 25, a data storage unit 30, a test unit 32, a utilization device 34, a key~oard 36, and a process and control logic unit 40. The system is ~asically that shown and described 20. in U.S. Paten~ No. 4,186,378. A suitable jig device (not shown) may be provi~ed for the placement of a human hand to ensure proper registration of the palm print for the initial recording of a recognition pattern by the acquisition unit lO and eaah su~sequent presentation of the palm print for identity verification. The recognition data acquisiti.on ~nit lO comprises a camera 12, an enhance circuit 14, and an analog-to-digital converter (ADC) 16.
The camera, which may suitably be a television-type vidicon or a solid-state charge coupled image sensor, 3Q raster scans the palm print, outputting an analog voltage signal which is proportional to the light levels obtained from the print on each horizontal scan wherein a ~3 I lG7166 positive peak representing high illumination corresponds to a ridge in the palm print pattern and a negative peak cor responds to a valley in the palm prin-t pattern. The enhance circuit 14, which is not essential to the system, enhances the positive and negative peaks of the analog signal to provide a preater pronunciation of light and dark 7evels~
A conventional differentiating circuit such as a se:~les capacitor and a shunt resistor will provide desired enhance-ment in accordance with the compone!nt values selected. The enhanced analog signal is then quantized by the ADC 16 to provide numerical digital data which corresponds to the various voltage levels quantized. Many conventional analog-to-digital converters are commercially available for this purpose The quantized, or "digitized" signal is then stored line by line in a 10,000-element picture memory device 20 s~ch t~.at a 100-element by 100-element image of the palm pattern is stored. If this image were read out and viewed on an X-Y display device in the 100 by l00 ~ormat, it would be discerned that the vertically-oriented pattern components, or ridge and valley lines, are more prominent than the horiz~
ontally-oriented lines because of the enhancement process which takes place as each horizontal line is recorded. Thus an optimized image may be formed ~or the analysis and test procedure which will be described later.
The recognition data analyzer 25 includes a number bf read-only mernories (ROM's) containing specific lugic steps (program instructions burned in~, and operates in concert w;th and under the control of the process and control lo~ic unit 40, which may suitably be a microprocessor unit, for analysis of the pattern ima~e stored in the picture memory devi~e 20 !

Certain selected recognition data, to be described later, obtained by the recognition data analyzer 25 is stored along with a user's identlty code, obtained from keyboard 36, l J ~

in the data storage unit 30.
For identity verification, the user places his hand in the iig device mentioned earlier and enters an identifi-cation humber into the keyboard 36. The process and control logic unit 40 turns on the camera L2 to read the palm pattern.
The enhanced palm print pattern is stored line by line into the 10,000-element picture memory 20 in the same ~anner described earlier. The user's identity code number ensures retrieval of the correct data from the storage unit 30.
The reco~nition data analyzer 25 then analyzes the newly-store~
image using the analysis pattern and data as originally' abstracted~ That is, since key information pre~iously has been abstracted from image and stored in learn mode, it is necessary in an access mode only to see if similar ~ey infor-mation exists in the newly-stored image; obviating the need to subject a palm pattern to a complete and perhaps lengthy analysis upon subsequent presentations of the palm ~or ident-ity purposes. The newly-abstracted information is sent to the test unit 32 along with the originally-obtained recog-ni`tion data to determine whether a reasonable correlationexists. The test unit 32 includes a number o ROM's con-taining specific logic steps (progr~ instructions burned in) and operates in concert with and under control the process and logic control unit 40 to determine the numerical agreement or degree of agreement be,tween the new and retrieved-from-storage recognitlon data. Added steps ma~ be incorporated to translate or skew or rotate the prior stored recognition data for a comparison of best fits to better match the new image details to correct or translational (X-Y displace-3Q ment) or rotational registration errors~ An identitydecision is made as to whether a reasonable match exists or does not exist between the stored reco~nition data and the new recognition data, and an output signal is applied I ~7~66 -- , to a utillæation device 3q indicating verification or rejection of the new recognition data.

ANALYSIS PROCESS
For this discussion, it w:ill be assumed that a complete image o~ a palm print is stored in the 100-element by 100-element picture memory 20 as described hereinabove;
that is, each picture element (pixel) has stored th~rein numerical digital data relating to a light level obtained from the palm print. The picture memory may be represented by a Cartesian coordinate system as shown in FIG. 2, wherein the pixel array defines the X-Y plane, and the numerical digital data values define the amplitude along the Z axis.
The 10,000 pixels of the memory array are subdivided into small groups, each of which is analyzed-in a predetenmined manner to pick form each group a vector having the largest absolute magnitude Z component difference from origin to end point. The Z component provides information as to the slope of the vector with reference to the X-Y plane. While a vector is not selected in its absolute (X2+Y2+Z2)~ sense in this example, it could be a possibility by proper scaling.
For example, FIG. 3 shows a portion of the memoxy array subdivided into 9-pixel blo¢ks, each comprising a 3-element by 3- element subarray. Starting at the bottom left corner of the memory array, each subaxray is selected for analysis in accordance with the X-Y location of a pre-determined home pixel in each subarray. Sequentially, this selection may be corrdinates X,Y = 1,1; 1,4; 1,7; 1,10;

4,1; 4,4; etc., as shown in FIG. 3, to cover a 99-element by 99-element portion of the 100 x 100 memory array~ Each 9-pixel block of t:he memory shown in FIG~ 3 is analyzed as follows. With reference to FIG. 4A, the outer pixels of ~he block are assigned addresses 1 through 8 in clockwise I :i 6 7 ~ 6~

direction around the 9-pixel block, with the home pixel being 1, and center pixel being unassigned. Beginning at the home pixel, pixels 1 and 3 are first tested by subtracting the digital number stored in pixel 1 from the digital number stored in pixel 3. The difEerence is the Z component, and this value, which may be either positive or negative, is stored in a random-access memory (RAM) along with the 1-3 vector location. Then pixels 1 ancL 4 are tested in the same manner, producing another Z component, which again i5 0 stored in the RAM along with the 1-4 vector location. Then the combinations 1-5, 1-6, and 1-7 are in turn examined in the same manner, as shown in FIG. 4B, with the values of the Z components being stored in the RAM along with the vector locations. The analysis continues as shown in FIGS. 4C, 4D, and 4E until fourteen æ component values along with their respective vector locations are stored in the random-access memory. Each ~rou~ of fourteen values is identified by the X,Y address location of the home pixel so that once the information is abstracted and stored, it may easily be retrieved with certainty as to the exact location from which it was taken. Suppose each pixel contains a 6-bit binary number indicative of the light level stored thereon.
Thus, each pixel has stored thereon a number between 0 and 63, depending upon the level of illumination the num~er represents. Suppose further that pixel 1 of group 1,1 has a value of 35, while pixels 3 and 7 ha~e values ~f ~6 and 28, respectively~ The information stored relating to each of these two vectors, then, would be groups of numerical data in the form of 1,1,1,3,11~1 and 1,1,1,7~7,0, respectively, with the first two numbers in each set indicating the X,Y
location of the home pixel, the second two numbers in each set indicating the vcctor locations, the fi~th number in the set indicating the value of the Z component of the vector, g and the last number of each set indicating the polarity of the slope, e.g., 1=+, 0=-, thereby giving complete information as to each vector. However, it can be appreciated that in this example, there are 33 X 33, or 1,089 blocks analyzed for the entire pic~e memory ~rray, each yielding fourteen vectors for a total of 15,246 vectors. Therefore, to save memory space, as each block is analyzed, only the one best of the fourteen vectors in each block is saved so that the number of vectors is 1,089. These vectors may then be ran~ed in a descending order from highest quality to lowest qllality, wherein quality may be equal to either the value of che Z component or a numerical value assigned to the steepness of the slope of the vector with respect to the X-Y plane.
Finally, one to two hundred vectors haviny the hiyhest quality are retained for storaqe in the data storage unit 30 for subsequent verification purposes. This analysis met~od therefore provides ~rominent recoqnition data relating to three-dimensional aspects of the stored image details, such as slopes and flexions tindicative of valleys and peaks).
While the ~ysis method here~ove descr~ ~ olves 3-pixel by 3-pixel subarrays each in a donut pattern yielding fourteen vectors, other analysis procedures ould be used without deviating from the general method. F~r example, FIG. 4 shows analysis of the 3-pixel by 3-pixel subarray using the center pixel as the home pixel and the analysis yieldin~ eight vectors. Also, the memory array 20 could be subdivided into 4-pixel by 4-pixel or 5-pixel by 5-pixel ( or more) subarrays as shown in FIGS. 5 and 6 respectively, which in the examples shown yield 22.and 12 vectors respectively. ~ther patterns are a matter of design choice. A further alternative is to divide the memory array into groups, such as the 3-pixel by 3-pixel subarrays discussed above, take the average o~
the data stored in the pixels of each group, and then locate fi ~) one or more vectors extending from one group to another throughout the memory arxay~ This alternative pro~ides~
region-to-region slope information, rather than point-to-point slope information as described above~ ~l FIG. 7 shows a block diagram of the details,o,f the data recognition analyzer 25. Included are a vertical- , scan read-only memory ~SCAN ROM) 50, and horizontal-scan read-only memory (HSC~N ROM~ 52, a test read~only memory ~TEST P~O~l 54, random-access mernory ~RAM~ 56, and a rank ,;
read-only memory (R~NK ROM) 58. Each of the ROM~s contains specific logic steps ~program instructions burned in~ and , operate in conjunction with process and control logic unit 40. The VSCAN and HSCAN ROM's locate the X~Y home pixels, and ~ector locations for each vector, the TEST ROM 54 tests for the slope, flexion, and polarity information o~ the vectors. Flexion is determined by comparing subsequent values of ~ectors selected as the image is scanned; it is , the x~te of change of slope~ These ROM's operate to ,,, abstract vector information from the picture memory ~0, storing intermediate and final results in the RAM 56. ,The RAN~ ROM 58 sorts the vectors from highest to lowest quality ~as defined abovel and selects a predetermined number, e.g., one ~o two hundredl of the hiyher quality vectors so that the.most prominent picture details may be encoded in compact form and stored in the data storage unit 30.
The reco~nition data analyzer 2S may be uti~ized in both the learn and access modes as discussed earlier~
In the learn mode, a complete analysis of the image details stored in picture memory 20 is made in accordance with the foregoing analysis description in order to abstract key information to store in the data storage unit 30~ In the access mode, a complete and lengthy ana'ysis upon sub,sequent presentations of the palm for identity purposes is not needed because the key information is already known and stored in the data storage unit 30. Upon recall o~ the previously-obtained information from storage, it is necessary only for data analyzer 25 to proceed at once to the picture locations from which the key information was abstracted in order; to abstract new information for comparison purposes. Thus, old information is utilized to aid in locating the new ir.formation to provide a quick and e~ficient lndentity verification.
The test unit 32 operates as described above.
Alternatively, since the previously-obtained information is utilized to locate new recognition information, a one-vector-at-a-time location and comparison could take place completely within in data analyzer without the need for a separate test unit. For example, upon the recall cf previously~stored recognition data, the address of the old vector or slope is utilized to locate a new oneO Then the ~uality value ma~ be compared immediately. The old :~ector may be translated or rotated on the new image to determine the best fit. The degree of correlation between the old and new data may be sent to a decision circuit, for which a minimum level of acceptance is predetermined to veriy or reject recognition of the new image.
It ~ill therefore be appreciated that the afore~
.. , . .~ .
mentioned and other objects have been achieved; however, it should be emphasized th~t t~e particular analysis method which is shown and described herein, is intended as merely illustrative and not restrictive of the invention.

.

Claims

The embodiments of the invention in which an exclusive property or privilege is claimed are defined as follows:

1. A method of analyzing an image to provide identification thereof, comprising the steps of:
storing said image in a picture storage device comprising an array of addressable picture elements each containing numerical data corresponding to levels of light;
electronically selecting one or more three-dimensional slopes between selected pairs of addressable fixed locations by subtracting numerical data stored in one or more picture elements at a first fixed location from numerical data stored in one or more picture elements at a second fixed location;
comparing said selected slopes with stored slopes previously selected from substantially the same locations of a prior image to determine a correlation therebetween;
and providing an identification of recognition based upon the degree of correlation between said selected slopes and said stored slopes.

2. A method in accordance with claim 1 wherein said stored slopes previously selected include location data identifying the locations in the image from which said slopes were selected, and said location data is utilized in said selecting step to locate said selected slopes for comparison with said previously selected slopes.

3. A method in accordance with claim 1 wherein said picture elements of said picture storage array are divided into groups of picture elements each arranged in a predetermined pattern having at least one first picture element representing a vector origin and at least one second picture element representing a vector end point in the X-Y

plane of Cartesian coordinate system, wherein said numerical data stored in said picture elements represents amplitudes along the Z axis of said coordinate system.

4. A method in accordance with claim 3 wherein said selected slope comprises the picture element addresses of said vector origin and end point, and the absolute Z-axis magnitude.

5. A method in accordance with claim 4 further comprising the step of abstracting a first set of slopes to provide said previously selected and stored slopes by addressing said groups of picture elements in a predetermined manner and selecting one slope from each group by calculating the absolute Z-axis magnitude of numerical data stored in preselected pairs of picture elements within said group, and storing the slope having the largest Z-axis magnitude.

6. A method of analyzing stored image details to provide recognition data therefrom, wherein the image is stored in a picture memory device comprising an array of picture elements defining the X-Y plane of a Cartesian coordinate system, comprising the steps of:
electronically addressing one or more picture elements in a first area of said picture memory device and obtaining the numerical value of data stored therein;
electronically addressing one or more picture elements in a second area of said picture memory device and obtaining the numerical value of data stored therein;
subtracting the numerical value obtained from said first area from the numerical value obtained from said second area to provide an absolute magnitude value representing the Z
axis of said coordinate system; and storing said addresses of said picture elements and said absolute magnitude value to provide recognition data thereby.

7. A method for analyzing the contents of a picture storage device to abstract prominent recognition data representative of three-dimensional aspects of stored image details, wherein the picture storage device comprises an array of picture elements arranged in groups each having a predetermined pattern, comprising the steps of:
addressing said groups in a predetermined manner;
addressing preselected pairs of picture elements at predetermined locations in each group and subtracting the numerical value stored in one of said pair of picture elements from the other to determine an absolute magnitude value;
temporarily storing the picture element locations and absolute magnitude value of the pair of picture elements in each group having the greater absolute magnitude value; and selecting a predetermined number of temporarily stored picture element locations and the respective corresponding absolute magnitude values thereof for permanent storage as recognition data.

8, A method in accordance with claim 7 further comprising the steps of recalling said recognition data from storage to analyze a subsequent image to provide identification thereof, comparing the absolute magnitude values of said recognition data with absolute magnitude values calculated at substantially the same picture locations of said subsequent image, and providing an indication of recognition based on the degree of correlation between said recognition data and data obtained from said subsequent image.

9. A method in accordance with claim 7 wherein said recognition data is numerical data indicating location, direction, and steepness of slopes.