• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:
    Method for analyzing an imprint, comprising the step of acquiring an imprint image (20) as well as the steps of: - implementing a filtering treatment on the imprint image to estimate, for each pixel of the imprint image, a first peak frequency (21) in the imprint, and producing, thanks to the first frequencies associated with the pixels of the imprint image, a first frequency map (22) of the image imprint; - divide the imprint image into a plurality of windows each comprising a plurality of pixels, calculate a Fourier Transform of each window to estimate a second peak frequency for all the pixels of said window, and produce thanks to the second associated frequencies to the pixels of the windows a second frequency map of the imprint image; - merge the first frequency map and the second frequency map to obtain a consolidated frequency map of the footprint image.
    公开号:FR3084945A1
    申请号:FR1857460
    申请日:2018-08-10
    公开日:2020-02-14
    发明作者:Iana IATSUN；Laurent KAZDAGHLI
    申请人:Idemia Identity and Security France SAS；
    IPC主号:
    专利说明:

    The invention relates to the field of fingerprint analysis methods.
    BACKGROUND OF THE INVENTION
    The use of biometric recognition means is more and more frequent to protect access to secure areas: businesses, banks, airports, prisons, etc.
    In particular, it is possible to identify and authenticate a person thanks to their unique fingerprints.
    A fingerprint image, representative of a fingerprint and acquired by a sensor, for example an optical, thermal or ultrasonic sensor, can be seen as an alternation of peaks and valleys.
    A footprint image also includes a number of characteristic points, or minutiae, which correspond, for example, to the end of a crest or to the division in half of a crest (bifurcation). To compare efficiently and reliably two footprint images and therefore two sets of minutiae, it is first necessary to accurately estimate the local frequency of the peaks. Indeed, in the event of a bad estimate of the local frequency of peaks, a doubling of peaks may appear. False peaks produce nonexistent minutiae which tend to distort identification and authentication.
    However, a precise evaluation of the local frequency of peaks can sometimes be complex to carry out. In general, the main obstacle to accurately assess the local frequency of peaks is the poor quality of fingerprint images. The degraded structure of a footprint image complicates the analysis of the data obtained. Estimating the local frequency may become impossible if there is noise on the imprint image, or if there is a scar.
    Other phenomena can degrade the accuracy of the estimation of the local peak frequency: presence of pores, frequent change of direction of the peaks, significant difference in the distance between the peaks from one person to another, variation in the local frequency significantly within the same footprint, difference in period of the peaks and the 10 valleys, noise, etc.
    Thus, with reference to FIG. 1, a distinction is made on the ridges 1 of an impression image 2 of enlarged pores 3 of the skin, which have a large opening because of the heat. On the processed image 4, 15 obtained from the imprint image 2 using a first known method of analysis, it can be seen that a certain number of peaks are doubled due to the poor estimate of the local frequency. ridges. This processed image 4 can be compared to an expected image 20 5, processed effectively, which does not exhibit this doubling of the peaks. We can see that, compared to the expected image 5, the processed image 4 is degraded significantly.
    To accurately estimate the local frequency of the 25 peaks, it was envisaged to implement a second known analysis method, which is described with reference to FIGS. 2 and 3.
    The local frequency of the peaks 6 in a sample of a footprint image 8 is defined as being equal to the number of peaks 6 by the length of the sample along an axis orthogonal to an axis of local orientation of the peaks 6. The local frequency is different between two fingerprints, but it can also vary within the same fingerprint.
    The second known method of analysis of the local frequency uses information on the number of pixels between two peaks 9 of gray levels.
    The determination of these gray levels is carried out in two main steps which use sliding windows 10. First, the local orientation of the peaks 6 is estimated in each sliding window 10. It is calculated as being equal to the ratio between the responses of the intensity gradients obtained by a Sobel filter in vertical and horizontal direction. Then, a sliding window 10, oriented in the direction of the global orientation of the ridges 6, is placed around each pixel. For each column of the sliding window
    10, the values s of pixels are accumulated. So, the distance between of them peaks 9 represents the distance between the ridges 6. We name T 1, 12 and T3 the distances between three peaks 9 successive. These distances corresponding to of the intervals in which gray levels are less than one threshold predetermined 12.It is estimated the local frequency using the formula : 1 f ~ T ± + T 2 + T 3 Point weakness of this method is that the
    determination of the local frequency is imprecise when the imprint image is blurred or of poor quality.
    OBJECT OF THE INVENTION
    The object of the invention is to improve the estimation of the frequency of the peaks of a footprint image representative of a footprint.
    SUMMARY OF THE INVENTION
    With a view to achieving this goal, a method for analyzing a fingerprint is proposed, comprising the step of acquiring, using at least one sensor, a fingerprint image representative of the fingerprint, as well as the steps, carried out by at least one electrical processing unit, of:
    implement a filtering treatment on the imprint image to estimate, for each pixel of the imprint image, a first frequency of peaks in the imprint, and produce thanks to the first frequencies associated with the pixels of the fingerprint image a first frequency map of the fingerprint image;
    dividing the footprint image into a plurality of windows each comprising a plurality of pixels, calculating a Fourier Transform of each window to estimate a second peak frequency for all the pixels in said window, and producing using the second frequencies associated with the window pixels a second frequency map of the footprint image;
    merge the first frequency map and the second frequency map to obtain a consolidated frequency map of the footprint image.
    The combined implementation of filter processing and spectrum analysis of the Fourier Transform, as well as the fusion of the first frequency map and the second frequency map, make it possible to very precisely estimate the frequency of the peaks of the footprint image.
    A system is also proposed comprising at least one sensor and at least one electrical processing unit, in which an analysis method such as that just described is implemented.
    The invention will be better understood in the light of the following description of a particular non-limiting embodiment of the invention.
    BRIEF DESCRIPTION OF THE DRAWINGS
    Reference will be made to the appended drawings, among which:
    FIG. 1 represents an impression image, a processed image and an expected image;
    Figure 2 shows a footprint image and a sliding window;
    Figure 3 is a graph showing gray levels associated with Figure 2;
    FIG. 4 represents a footprint image and a map of first frequencies;
    FIG. 5 represents a footprint image, windows of the footprint image, and Fourier Transforms of the windows calculated during the implementation of the analysis method according to the invention;
    FIG. 6 represents a window, the Fourrier Transform, the implementation of a radial distribution function, and a graph comprising a curve of the radial distribution function;
    FIG. 7 represents a footprint image and a map of second frequencies;
    FIG. 8 represents a footprint image, a map of first frequencies, a map of second frequencies and a map of consolidated frequencies;
    FIG. 9 represents an image of an imprint, a first image treated only with a filtering treatment, and a second image treated with the analysis method according to the invention.
    DETAILED DESCRIPTION OF THE INVENTION
    With reference to FIG. 4, the analysis method according to the invention firstly consists in acquiring an image of an imprint 20 representative of an imprint of an individual. The fingerprint here is a fingerprint, but it could be a different papillary fingerprint, for example a palm print.
    The imprint image 20 is acquired using a sensor which is for example an optical sensor, a thermal sensor, an ultrasonic sensor, etc.
    The analysis method then comprises a certain number of steps which are all carried out by an electrical processing unit. By “electrical processing unit” is meant any type of fixed or mobile electrical equipment, which comprises one or more processing components (microcontroller, processor, FPGA, DSP, etc.) adapted to execute instructions of a program for perform the tasks dedicated to it (and, in particular, to at least partially implement the analysis method according to the invention). We note that it is possible to carry out these steps not in one but in several electrical processing units, possibly connected to each other by any type of connection means (wired, wireless, network, etc.). Of course, the processing unit is capable of acquiring the imprint image 20 produced by the sensor.
    First of all, filter processing is implemented on the imprint image 20 to estimate, for each pixel of the imprint image 20, a first frequency of the peaks 21 in the imprint. Filtering treatment uses a filter bank here. The filter bank is in this case a Gabor filter bank.
    Filtering processing is carried out by an encoder which assigns to each pixel a first frequency value included in a finite list of predefined values. Here, the list includes three values.
    The first frequencies associated with the pixels of the imprint image 20 then produce a first frequency map 22 of the imprint image 20. On the first frequency map 22, the gray level of each pixel corresponds to the first frequency associated with it (that is, one of the three values in the list).
    Referring to Figure 5, the footprint image 23 is further divided into a plurality of windows 25 which each include a plurality of pixels. Here, each window 25 has 128 × 128 pixels. At least two windows 25 overlap; here, in this case, all the adjacent windows 25 overlap.
    The imprint image 23 is also processed by applying a mask so as to eliminate the background.
    Thus, on the fingerprint image 26, the fingerprint is detached from its background. The following operations are implemented only on the fingerprint and not on the background.
    A Fourrier Transform 27 is applied to each window 25 of the imprint image 26.
    Then, with reference to FIG. 6, for each window 25, a spectrum of the Fourrier Transform 27 is analyzed. The analysis consists first of all in calculating a radial distribution function of the spectrum of the Fourier Transform 27.
    We then identify, for each window 25, the peak (s) 28 (or "energy peaks") of the curve 29 of the radial distribution function. The frequency corresponding to each peak 28 of window 25 is called "window frequency". Several window frequencies can therefore be associated with the same window 25.
    Then, a second peak frequency is associated with each pixel of each window 25 of the footprint image 30. When only one window frequency is associated with the same pixel, the second frequency of the pixel takes the value of the window frequency.
    However, it is possible that several window frequencies are associated with the same pixel. This results in particular from the fact that several window frequencies can be associated with the same window 25, but also due to the overlaps between the windows 25, which have the consequence that certain pixels belong to several windows 25.
    When several window frequencies are associated with the same pixel, the second frequency of said pixel is estimated by determining a median value of these window frequencies.
    We use the formula:
    F2 = MedianÇfi, f 2 , f. ,, fn), where F2 is the second frequency associated with a pixel where and the fl, f2, f ..., fn are the window frequencies associated with said pixel.
    With reference to FIG. 7, thanks to the second frequencies associated with the pixels of the windows of the imprint image 30, a second frequency map 31 is produced. The imprint image 30 is of reduced size compared to the image of footprint 20. The second frequency map 31 has the same size as the footprint image 30. It can be seen on the second frequency map 31 that the change in frequency is homogeneous. There is no abrupt change in frequency, which corresponds to reality.
    With reference to FIG. 8, the first frequency card 22 and the second frequency card 31 are then merged.
    The merging consists, for a given pixel, in associating with said given pixel an optimized frequency equal to a value from the predefined list of values closest to the second frequency. For pixels with no second frequency associated, the optimized frequency is equal to the first frequency.
    A consolidated frequency map 33 of the footprint image 20 is thus obtained.
    With reference to FIG. 9, a footprint image 40 acquired by the sensor can be compared, a first processed image 41 obtained by implementing only filter processing, and a second processed image 42 obtained with the analysis method. according to the invention.
    The local frequencies of the peaks of the second processed image 42 are much closer to those of the imprint image 40 than the local frequencies of the peaks of the first processed image 41. The second processed image 42 is more precise and cleaner. In particular, while the first processed image 41 includes areas 43 for doubling the peaks, this is not the case for the first processed image.
    The only implementation of filtering processing is a method that has many weaknesses. Since the number of filters is limited, as soon as the frequency of the peaks is higher or lower than the expected frequency, the calculations become erroneous. Duplication of the ridges can then occur and subsequently degrade the rest of the treatment. The filters are also sensitive to the degradations present on the imprint image. The presence of scars can change their responses. On the other hand, the filter computation times remain reasonable compared to the spectral approach. It is noted that the presence of an imperfection on the imprint image does not prevent the detection of energy peaks. The advantages of the two methods are therefore combined in the analysis method according to the invention, which makes it possible to obtain a precise, effective analysis method, with reduced treatment times.
    The invention is not limited to the particular mode of implementation which has just been described, but, on the contrary, covers any variant coming within the scope of the invention as defined by the claims.
    Note that it would be possible to implement the spectral approach only in certain areas of the footprint image. These zones are for example zones comprising sudden changes in the frequency of the peaks.
    权利要求:
    Claims (8)
    [1" id="c-fr-0001]
    ! .. A method of analyzing an imprint, comprising the step of acquiring, using at least one sensor, an imprint image (20, 23, 40) representative of the imprint, as well as the steps, carried out by at least one electrical processing unit, of:
    implementing a filtering treatment on the imprint image to estimate, for each pixel of the imprint image, a first frequency of peaks (21) in the imprint, and producing thanks to the first frequencies associated with the pixels of the fingerprint image a first, frequency map (22) of the fingerprint image;
    dividing the footprint image into a plurality of windows (25) each comprising a plurality of pixels, calculating a Fourier Transform of each window to estimate a second peak frequency for all the pixels in said window, and producing using the second frequencies associated with the pixels of the windows a second frequency map (31) of the imprint image;
    merge the first frequency map and the second frequency map to obtain a consolidated frequency map (33) of the image d. ' imprint e.
    [2" id="c-fr-0002]
    2. Analysis method according to. claim 1, in which, for each window (25), the estimation of the second frequency of a pixel comprises the steps of
    this one function of distribution r; adiale c 1 'a. spectrum of the Transform! Fourier Island (27) of fire. and re, identified at least a peak (28) of the function of distribution radial and define 3U .Less a frequency of window horn • corresponding to the peak.
    [3" id="c-fr-0003]
    3. Analysis method according to claim 2, in which, when a single window frequency is associated with the same pixel, the second frequency of the pixel takes the value of the window frequency there, and when several window frequencies are associated with the same pixel, the second frequency of said pixel is estimated in
    5 determining a median value of these window frequencies,
    [4" id="c-fr-0004]
    4. Analysis method according to one of the preceding claims, in which at least two windows (25) overlap.
    10
    [5" id="c-fr-0005]
    5. Analysis method according to one of the preceding claims, in which, for each pixel, the value of the first frequency allocated to said pixel is included in a finite list of predefined values, and in which the step of merging the
    The first frequency map (22) and the second frequency map (31) consists, for a given pixel, in associating with said given pixel an optimized frequency equal to the predefined value of the finite list, which is the closest to the second pixel frequency.
    20 ’
    [6" id="c-fr-0006]
    6. Analysis method according to one of the preceding claims, in which the filtering treatment uses a filter bank.
    [7" id="c-fr-0007]
    7. Analysis method according to claim 6, wherein the bench
    25 filters is a Gabor filter bank.
    [8" id="c-fr-0008]
    8. System comprising at least one sensor and at least one electrical processing unit, in which one is used. analysis method according to one of the preceding claims.
    类似技术:
    公开号 | 公开日 | 专利标题
    EP3608834A1|2020-02-12|Method for analysing a fingerprint
    EP2902943B1|2016-10-12|Method for validating the use of a real finger as a support for a fingerprint
    FR3041423B1|2019-10-04|METHOD FOR EXTRACTING MORPHOLOGICAL CHARACTERISTICS FROM A SAMPLE OF BIOLOGICAL MATERIAL
    EP3214601B1|2020-10-07|Image enhancement method applicable to fingerprint images
    WO2014206887A1|2014-12-31|Method for detecting a real face
    EP3388976A1|2018-10-17|Method for detecting fraud
    FR2979727A1|2013-03-08|IDENTIFICATION BY RECOGNITION OF IRIS
    EP2917868B1|2020-05-13|Method of identification
    EP2652674B1|2015-01-21|Method of comparing images of irises by intelligent selection of textured zones
    FR3072199A1|2019-04-12|METHOD FOR DETECTING A PRESENCE OF A BODILY PORTION OF A DIGITAL FOOTPRINT ON A FINGERPRINT SENSOR
    EP3139307A1|2017-03-08|Method for characterising a material by speckle analysis
    EP3608835A1|2020-02-12|Method for processing fingerprint images
    FR3053500A1|2018-01-05|METHOD FOR DETECTING FRAUD OF AN IRIS RECOGNITION SYSTEM
    FR3035227A1|2016-10-21|METHOD FOR DETERMINING ACTIVITY OF AT LEAST ONE BODY PRESENTED IN EQUIPMENT METHOD FOR IDENTIFYING BODIES AND ASSOCIATED DEVICES
    FR3028376A1|2016-05-13|METHOD FOR DETECTING DEFECTIVE PIXELS.
    FR3021783A1|2015-12-04|METHOD FOR VALIDATING THE AUTHENTICITY OF A HUMAN BODY ELEMENT
    WO2013135977A1|2013-09-19|System for determining the identification of a photographic apparatus on the basis of a photograph and method implemented in such a system
    EP2827282A1|2015-01-21|Method for verifying the veracity of a finger or a palm
    WO2014049149A1|2014-04-03|Method for detecting the reality of venous networks for the purposes of identifying individuals, and biometric recognition method
    EP3726423A1|2020-10-21|Device and method for biometric acquisition and processing
    FR2865057A1|2005-07-15|Finger print image improving process, involves constructing Gabor filters from parameters determined by average frequencies of finger print lines determined for image blocks to filter blocks, and finding image orientations
    FR3074295A1|2019-05-31|METHOD FOR DETECTING ROLL DEFECTS IN PRINTED GLASS
    FR3084943A1|2020-02-14|METHOD FOR OBTAINING A FINGERPRINT IMAGE
    FR3109457A1|2021-10-22|Presentation attack detection method for fingerprints
    FR3088755A1|2020-05-22|METHOD FOR UNFLOUTING AN IMAGE
    同族专利:
    公开号 | 公开日
    FR3084945B1|2020-08-07|
    US11055506B2|2021-07-06|
    US20200050820A1|2020-02-13|
    EP3608834A1|2020-02-12|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    US5799098A|1994-10-20|1998-08-25|Calspan Corporation|Fingerprint identification system|
    FR2865057A1|2004-01-09|2005-07-15|Zefyr Trechnologies|Finger print image improving process, involves constructing Gabor filters from parameters determined by average frequencies of finger print lines determined for image blocks to filter blocks, and finding image orientations|
    US7142699B2|2001-12-14|2006-11-28|Siemens Corporate Research, Inc.|Fingerprint matching using ridge feature maps|
    WO2006009319A1|2004-07-22|2006-01-26|Nec Corporation|Image processing system|US10562070B2|2016-05-10|2020-02-18|Invensense, Inc.|Receive operation of an ultrasonic sensor|
    US10539539B2|2016-05-10|2020-01-21|Invensense, Inc.|Operation of an ultrasonic sensor|
    US10984209B2|2017-12-01|2021-04-20|Invensense, Inc.|Darkfield modeling|
    US10997388B2|2017-12-01|2021-05-04|Invensense, Inc.|Darkfield contamination detection|
    US11151355B2|2018-01-24|2021-10-19|Invensense, Inc.|Generation of an estimated fingerprint|
    WO2020263875A1|2019-06-24|2020-12-30|Invensense, Inc.|Fake finger detection using ridge features|
    US11176345B2|2019-07-17|2021-11-16|Invensense, Inc.|Ultrasonic fingerprint sensor with a contact layer of non-uniform thickness|
    US11216632B2|2019-07-17|2022-01-04|Invensense, Inc.|Ultrasonic fingerprint sensor with a contact layer of non-uniform thickness|
    US11232549B2|2019-08-23|2022-01-25|Invensense, Inc.|Adapting a quality threshold for a fingerprint image|
    法律状态:
    2019-08-01| PLFP| Fee payment|Year of fee payment: 2 |
    2020-02-14| PLSC| Publication of the preliminary search report|Effective date: 20200214 |
    2020-07-21| PLFP| Fee payment|Year of fee payment: 3 |
    2021-07-22| PLFP| Fee payment|Year of fee payment: 4 |
    优先权:
    申请号 | 申请日 | 专利标题
    FR1857460A|FR3084945B1|2018-08-10|2018-08-10|PROCESS FOR ANALYSIS OF A FOOTPRINT|
    FR1857460|2018-08-10|FR1857460A| FR3084945B1|2018-08-10|2018-08-10|PROCESS FOR ANALYSIS OF A FOOTPRINT|
    EP19189490.6A| EP3608834A1|2018-08-10|2019-07-31|Method for analysing a fingerprint|
    US16/537,068| US11055506B2|2018-08-10|2019-08-09|Method of analyzing a fingerprint|
    [返回顶部]