Method for authenticating an identity document, and device for implementing the method
The method measures distances and ratios between facial features at different angles to validate 3D portraits in identity documents, addressing the lack of automated verification tools and improving security by distinguishing between 3D and 2D images.
Patent Information
- Application Number
- FR2021013085
- Authority / Receiving Office
- FR · FR
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2021-12-07
- Publication Date
- 2025-01-10
- Estimated Expiration
- 2041-12-07
AI Technical Summary
Existing methods lack an effective, automatic tool to verify the authenticity of 3D portraits in identity documents, making it difficult to quantify and control the quality of 3D features, which is crucial for preventing counterfeiting.
A method involving the measurement of distances between characteristic points on a face in a portrait at various angles, calculating ratios, and determining a trend curve to validate whether the portrait is 3D by comparing the slope of the curve to a reference value, without requiring registration in a database or specific indexing.
This method allows for reliable verification of 3D portraits by quantifying the variation in ratios with observation angles, distinguishing between 3D and 2D portraits, thus enhancing security and reducing reliance on human expertise.
Smart Images
Figure 00000021_0000 
Figure 00000021_0001 
Figure 00000022_0000
Abstract
Description
Title of the invention: Method for authenticating an identity document, and device for implementing the method
[0001] The invention relates to a method for authenticating an identity document.
[0002] It also relates to a device configured to implement the method.
[0003] To secure an identity document, it is becoming more common for an identity document to include a 3D portrait of an individual, for example the holder of the document, possibly coupled with their 2D portrait.
[0004] To limit the risks of counterfeiting, or better detect a counterfeit, it is interesting to be able to verify that a portrait is actually in 3D.
[0005] However, controlling a 3D rotation of a face is difficult to quantify.
[0006] Few, if any, tools for automatically controlling the quality of a 3D portrait after personalization of an identity document have been developed. However, there is a need for such an automatic control tool in the field, for example, to limit the risks of poor verification due to a control agent.
[0007] Incidentally, a need has therefore arisen to have a control method making it possible to verify that a portrait, theoretically 3D, is actually in 3D.
[0008] To this end, a method for authenticating an identity document is then proposed, according to a first aspect of the invention, the method comprising at least: - A step of providing an identity document comprising at least a first portrait representing a face; - A step of determining a direction of observation of the first portrait, for example horizontal (for example from left to right) or vertical (for example from top to bottom) relative to the portrait, or any combination of the two; - A step of identifying at least three characteristic points of the face of the first portrait, including a first point, a second point and a so-called “median” point;
[0009] and for different observation angles along the observation direction: • A step of measuring a first distance between the first point and the midpoint of the first portrait according to an observation angle considered among the different observation angles; • A step of measuring a second distance between the second point and the midpoint of the first portrait according to the observation angle considered; and • A step of calculating a ratio between the first distance and the second distance, for the observation angle considered, for the first portrait; • A step of processing the ratios calculated for each of the different observation angles for the first portrait; and • A validation step of the first portrait as a 3D portrait based on a result of the processing step.
[0010] For example, the result of the processing step may be a result of comparison of a parameter representative of an evolution of the ratios as a function of the observation angle compared to a reference value.
[0011] In other words, the processing step may include a step of determining a parameter representative of an evolution of the ratios as a function of the observation angle.
[0012] For example, the processing step may include a step of comparing the representative parameter with a reference value.
[0013] For example, if the representative parameter is greater than the reference value, then the first portrait is validated as a 3D portrait, otherwise the first portrait is not validated.
[0014] If it is validated, the step of validating the first portrait as a 3D portrait returns a positive result, i.e. confirming that the first portrait is indeed a 3D portrait.
[0015] If it is not validated, the validation step returns a negative result.
[0016] In an exemplary implementation, the processing step, or for example the step of determining the representative parameter, comprises: - A step of determining a trend curve based on the ratios calculated for each of the observation angles, for the first portrait; and - A step of calculating at least one steering coefficient of the trend curve for the first portrait.
[0017] The trend curve thus represents a mathematical function modeling an evolution of the ratios calculated as a function of the observation angle.
[0018] In an exemplary implementation, the trend line is a linear regression line, and the slope coefficient then corresponds to the slope of the line.
[0019] For example, the representative parameter is an absolute value of the slope coefficient.
[0020] Thus, in an interesting implementation example, the comparison step comprises a comparison of an absolute value of the direction coefficient with respect to a reference value, and if the absolute value of the direction coefficient is greater than the reference value, the step of validating the first portrait as a 3D portrait returns a positive result, i.e. confirming that the first portrait is indeed a 3D portrait.
[0021] A result of the processing step is then, for example, a result of comparing the absolute value of the at least one calculated steering coefficient with respect to the reference value.
[0022] For example, if an absolute value of the slope coefficient of the trend curve is greater than a reference value, then the first portrait is validated as a 3D portrait.
[0023] In an exemplary implementation, the trend line is a linear regression line, and the slope coefficient then corresponds to the slope of the line.
[0024] In another example of implementation, the representative parameter is a deviation of the calculated ratios.
[0025] The processing step, or even the step of determining a representative parameter, then includes, for example, a step of determining a deviation of the calculated ratios, for all the observation angles considered, for the first portrait.
[0026] The deviation may for example be a difference between a maximum ratio and a minimum ratio among all the ratios calculated for the different observation angles along the observation direction.
[0027] Or for example, the deviation can be a standard deviation of a distribution of all ratios.
[0028] As before, the determined deviation can be compared to a value of reference.
[0029] For example, the comparison step includes a comparison of the deviation with respect to a reference value, and if the deviation is greater than the reference value, the step of validating the first portrait as a 3D portrait returns a positive result, confirming that the first portrait is indeed a 3D portrait.
[0030] Such a method thus makes it possible to verify that the identity document includes a 3D portrait by measuring a variation in the inclination of the face as a function of an angle of observation of the identity document.
[0031] In the case of a 2D portrait, this ratio varies little depending on the observation angle. A direction coefficient or a corresponding deviation are therefore very low, even close to zero.
[0032] On the contrary, for a 3D portrait, this ratio, or the gap, varies much more strongly depending on the observation angle.
[0033] Such a variation then depends on the portrait and / or the angles of observation, for example.
[0034] Depending on the respective values of the first distance and the second distance, the ratio may be greater than or less than 1 (or equal).
[0035] Furthermore, depending on the face and / or the variation of the viewing angle according to the viewing direction (e.g. from left to right or from right to left, or from top to bottom or from bottom to top, or any combination of the two, so that the viewing direction can be any direction, which simplifies use), the trend curve obtained can be increasing or decreasing according to the angles observation. The trend curve is, for example, monotonous.
[0036] An absolute value of the slope coefficient greater than the reference value then means that the trend curve is steeper than a portion of a straight line whose slope is equal to the reference value.
[0037] The reference value may depend on each case.
[0038] It can then, for example, be prefixed by the provider of the identity document.
[0039] According to an exemplary implementation, the reference value is a parameter represented representative obtained in the same way for a 2D portrait, as described later.
[0040] Such a method thus allows 3D security parameter authentication.
[0041] Furthermore, it does not require registration of the identity document in a database.
[0042] Such a method also does not require a specific indexing marking in the identity document.
[0043] The characteristic points of the face are for example chosen from usual characteristic points for the analysis of a face; for example the first and second points can correspond to the outer corners of the eyes, while the midpoint can correspond to the tip of the nose.
[0044] The authentication method does not require prior knowledge of the characteristic parameters of the identity document, i.e. it does not require knowledge of the characteristic points of a visual of the document. The method is interoperable, and thus works on any identity document presenting at least one portrait, or even, for example, a 2D portrait and a 3D portrait (in theory).
[0045] For example, the method may use a face detection algorithm, then include an analysis step, for example of the trend curve, or even a comparison step, for example of the trend curve with a reference curve, or with a curve obtained for the same portrait in 2D.
[0046] According to an interesting option when the identity document also includes a 2D portrait, theoretically of the same face, the method then includes, for example, steps similar to those described previously, applied to a second portrait, which is a 2D portrait.
[0047] For example, the method, or even the processing step more specifically, comprises: - A step of identifying at least three characteristic points of the face of the second portrait, including a first point, a second point and a so-called “median” point;
[0048] and for different observation angles along the observation direction, for example possibly the same as those for the first portrait, the method comprises for example: • A step of measuring a first distance between the first point and the point median of the second portrait according to the observation angle considered among the different observation angles; • A step of measuring a second distance between the second point and the midpoint of the second portrait according to the observation angle considered; and • A step of calculating a ratio between the first distance and the second distance, for the observation angle considered, for the second portrait.
[0049] Although the observation angles may differ from those used to observe the first portrait, the observation direction is for example the same, to observe both the first portrait and the second portrait.
[0050] For example, the processing step may then include a step of determining a parameter representative of an evolution of the ratios as a function of the observation angle for the second portrait.
[0051] In an exemplary implementation, the processing step may further comprise a step of recording the representative parameter for the second portrait as a reference value.
[0052] In an exemplary implementation, the processing step further comprises: - A step of determining a trend curve based on the ratios calculated for each of the observation angles, for the second portrait; - A step of calculating at least one steering coefficient of the trend curve for the second portrait.
[0053] The trend curve for the ratios of the second portrait is for example obtained in the same way as for the first portrait.
[0054] This is for example also a linear interpolation.
[0055] In an exemplary implementation, the characteristic points of the face of the second portrait correspond to the characteristic points of the face of the first portrait.
[0056] In an exemplary implementation, the processing step may comprise a step of recording an absolute value of the direction coefficient of the trend curve for the second portrait as a reference value.
[0057] Thus, for example, the absolute value of the direction coefficient of the trend curve for the first portrait compared to a reference value is then compared to the absolute value of the direction coefficient of the trend curve for the second portrait.
[0058] If the first portrait is indeed a 3D portrait, then the absolute value of the slope of the trend curve for the first portrait is greater than the absolute value of the slope of the trend curve for the second portrait (which is the 2D portrait).
[0059] In an exemplary implementation, the processing step comprises a step of calculation of a ratio between the slope of the trend curve for the first portrait and the slope of the trend curve for the second portrait, in absolute value.
[0060] The representative parameter includes, for example, the ratio between the direction coefficient of the trend curve for the first portrait and the direction coefficient of the trend curve for the second portrait, in absolute value.
[0061] And for example, the step of comparing a parameter representative of an evolution of the ratios as a function of the observation angle with respect to a reference value comprises a comparison of the ratio with a corresponding reference value.
[0062] For example, if the ratio is greater than a predetermined threshold, then the method comprises the step of validating the first portrait as a 3D portrait.
[0063] The threshold is for example greater than 1, for example at least equal to 1.1, or even 1.5, or even preferably at least equal to 2.
[0064] According to another example of implementation, the processing step comprises a step of calculating a deviation of the ratios, representing a spread of the ratios obtained for all the observation angles for the first portrait.
[0065] According to an interesting option, the processing step includes a step of calculating a deviation of the ratios, representing a spread of the ratios obtained for all the observation angles for the second portrait.
[0066] For example, the reference value includes the difference in ratios obtained for the second portrait.
[0067] For example, the step of comparing a parameter representative of an evolution of the ratios as a function of the observation angle with respect to a reference value comprises a comparison of the deviation of the ratios obtained for the first portrait with respect to the deviation of the ratios obtained for the second portrait.
[0068] For each portrait, the difference represents, for example, a maximum difference between the ratios calculated for each of the observation angles, i.e., for example, a difference between a maximum ratio and a minimum ratio obtained for the portrait, over all the observation angles.
[0069] This is for example a simplified technique in which a trend curve of the ratios for each of the portraits is optional.
[0070] The spread of 2D ratio and 3D ratio values taken at different angles are compared.
[0071] This comparison is possibly translated graphically by a curve (for example a straight line) with a direction coefficient comparable or even identical to the ratio obtained previously between the direction coefficients of the curves for the two portraits.
[0072] An advantage of this approach is that the observation angle does not need to be determined, or the ratios do not need to be sorted according to the viewing angle. It is just better to have several viewing angles to see that the ratios of a 3D portrait vary more than for a 2D portrait.
[0073] In a case where the first portrait is not 3D, the comparison curve is shortened, or even assimilated to a point, because in this case the ratios of the two portraits would be very similar or even identical.
[0074] According to another interesting option, the method may further comprise a step of verifying a concordance of the first portrait and the second portrait, for example according to an observation angle perpendicular to each of the first portrait and the second portrait (for example noted a = 0°).
[0075] A perpendicular observation of the portrait generally corresponds to an observation of the face from the front. Thus, whether the portrait is in 3D or 2D, the two images must then correspond when the portraits are authentic.
[0076] Such a comparison of the first portrait and the second portrait helps to verify the identity of the holder.
[0077] Such a step can, for example, take place before checking whether the first portrait is indeed a 3D portrait.
[0078] Also provided, according to another aspect, is a device for authenticating an identity document, which is configured to implement the steps of the method as described previously.
[0079] For example, the device comprises: - An optical system configured to view a first portrait and identify at least three characteristic points of a face of the first portrait, including a first point, a second point and a so-called “median” point; - A calculator configured to measure a first distance between the first point and the midpoint, measure a second distance between the second point and the midpoint and calculate a ratio between the first distance and the second distance, for each of different observation angles considered, and to process ratios calculated for each of the different observation angles, for the first portrait; - A display configured to communicate validation information for the first portrait as a 3D portrait.
[0080] Such a device thus allows a security agent or other user to know whether the portrait being checked is actually a 3D portrait.
[0081] The validation information may include a value display, and / or an indicator light, for example green (if validation) or red (if the portrait is not validated as a 3D portrait for example).
[0082] According to an interesting option, the device can further include a memory configured to record at least one reference value.
[0083] The memory can also record several reference values, depending on the representative parameter considered, for example including a threshold or any other useful value.
[0084] For example, the calculator can further be configured to determine a parameter representative of an evolution of the ratios as a function of the observation angle.
[0085] For example, the calculator may further be configured to determine a steering coefficient of a trend curve of the ratios calculated for the first portrait, and / or the deviation of the ratios.
[0086] For example, the calculator can further be configured to compare the parameter representative of a change in the ratios as a function of the observation angle with respect to a corresponding reference value.
[0087] For example, the calculator may be configured to compare an absolute value of the slope coefficient of the trend line for the first portrait to the reference value.
[0088] According to yet another interesting option, the device is further configured to apply the aforementioned steps to a second portrait, which is for example a 2D portrait of the same face, in theory.
[0089] Such a device makes it possible to verify the concordance between the 2D portrait and the 3D portrait, in particular in a more objective manner, for example by limiting a risk of not identifying a falsified document due to a lack of training and / or knowledge of a security agent, such as a police officer, who is required to check the document.
[0090] Such a device also makes it possible to carry out this verification digitally.
[0091] For example, the optical system is further configured to identify at least three characteristic points of a face of a second portrait, including a first point, a second point and a so-called “median” point.
[0092] For example, the calculator is further configured to measure a first distance between the first point and the midpoint of the second portrait, and measure a second distance between the second point and the midpoint of the second portrait.
[0093] For example, the calculator is further configured to calculate a ratio between the first distance and the second distance for the second portrait, for each observation angle considered.
[0094] In an exemplary embodiment, the calculator is configured to determine a parameter representative of an evolution of the ratios as a function of the observation angle for the second portrait.
[0095] For example, the calculator is configured to determine an absolute value of a direction coefficient of a trend curve of the ratios calculated for each of the angles observation, for the second portrait and / or calculate a deviation of the calculated ratios.
[0096] In an exemplary embodiment, the memory is configured to store the representative parameter for the second portrait as a reference value.
[0097] For example, the calculator may further be configured to calculate a ratio between the direction coefficient of the trend curve for the first portrait and the direction coefficient of the trend curve for the second portrait, in absolute value.
[0098] The invention, according to an exemplary embodiment, will be well understood and its advantages will appear better on reading the detailed description which follows, given for information purposes and in no way limiting, with reference to the appended drawings in which:
[0099] [Fig.l] represents an identity document comprising a 3D portrait and a 2D portrait of a holder of the identity document;
[0100] [Fig.2] schematically shows an example of measurement of two distances between a first point of the face and a midpoint, and a second point of the face and the midpoint, for a considered observation angle;
[0101] [Fig.3] illustrates different characteristic points of the face on the 3D portrait and the 2D portrait, according to an exemplary implementation of the invention;
[0102] [Fig.4] illustrates the straight line obtained for a 3D portrait and for a 2D portrait, in an example of implementation of the invention;
[0103] [Fig.5] illustrates the curve obtained for a comparison of the ratios of the 3D portrait with those of the 2D portrait, in another example of implementation of the invention; and
[0104] [Fig.6] schematically represents an example of embodiment of a device authentication method for implementing an authentication method according to an implementation mode of the invention.
[0105] [Fig.l] illustrates a step E1 of providing an identity document 10 comprising at least a first portrait 12 representing a face.
[0106] The identity document 10 here includes a main portrait 11 of the holder, located towards an upper left corner of the identity document.
[0107] It also includes here, according to an interesting option to reinforce the security of identification of the document, a secondary portrait 12 of the holder.
[0108] The secondary portrait 12 is here smaller in size than the main portrait 11, and is located in a different location, here more in the center of the identity document.
[0109] The identity document 10 also includes various inscriptions and information about the holder, and / or other security elements which are not described in detail here.
[0110] In the context of the present invention, it is interesting that the identity document includes at least one 3D portrait.
[0111] Using a 3D portrait in an identity document as a security feature is increasingly becoming a key technology for authenticating an identity document.
[0112] To create such a 3D portrait, technologies such as SLI, FLI, or others have been developed, and most often use lenticular networks with spherical or cylindrical lenses.
[0113] For example, a 3D portrait using an array of cylindrical lenses arranged vertically relative to the portrait allows, by varying the angle of observation of the portrait, to observe a rotational movement of the face of the portrait along a vertical axis, also called a “no-no” movement, while a 3D portrait using an array of cylindrical lenses arranged horizontally relative to the portrait allows, by varying the angle of observation of the portrait, to observe a rotational movement of the face of the portrait along a horizontal axis, also called a “yes-yes” movement.
[0114] A network of cylindrical lenses makes it possible, for example, to combine the two movements.
[0115] The 3D portrait can optionally be coupled with a 2D portrait.
[0116] This makes it possible, for example, to strengthen security and / or control devices to authenticate the identity document.
[0117] In the present exemplary embodiment, the main portrait 11 comprises the 2D portrait while the secondary portrait 12 comprises the 3D portrait.
[0118] However, according to an example not shown, the primary portrait could include the 3D portrait, and the secondary portrait could include the 2D portrait.
[0119] However, it is difficult to control a rotation of a face in a 3D portrait, especially quantitatively.
[0120] To remedy this, the invention here proposes a method in which distances between characteristic points of the face of a portrait are measured and monitored as a function of an angle of observation of the portrait.
[0121] [Fig.2] shows in more detail the face of one of the portraits 11, 12, 2D or 3D, of the identity document 10.
[0122] This figure first illustrates a step E2 of identifying at least three characteristic points of the face of the first portrait.
[0123] Indeed, it is possible to define on such a face different characteristic points 13. These are for example different points defined along a contour of the face, by the line of the eyebrows, the bridge of the nose, the contour of the nose or the lips, or others not illustrated.
[0124] It is considered here, by way of illustration, that it may be a 3D portrait observable according to different angles a (here schematized a;, ai, a2, a3, a4), along an observation direction D, which is here called horizontal (left-right observation). Such a portrait then comprises a lenticular network of cylindrical lenses extending vertically relative to the face of the portrait.
[0125] Observing the portrait from these different angles theoretically allows us to observe a “no-no” movement of the face, if it is indeed a 3D portrait.
[0126] According to another example not shown, to observe a “yes-yes” movement, different observation angles would be identified along an observation direction which would be vertical (up - down) in the example shown [Fig.2]. Such a portrait then comprises a lenticular network of cylindrical lenses extending horizontally relative to the face of the portrait.
[0127] Any combination of the two is also possible. Such a portrait then comprises a lenticular network of spherical lenses instead of cylindrical lenses. In such a case, the direction of observation can be any direction, which simplifies use.
[0128] Here, by way of example, three characteristic points of the face are identified, namely by definition: a first point 131 located on the contour of the face at the level of a temple of the face on the left in the image, a second point 132 located on the contour of the face at the level of a temple of the face on the right in the image and a so-called “median” point, here representing a tip of the nose.
[0129] According to a characteristic of the invention, for each of the observation angles (here for example ab a2, a3, a4), a first distance dO between the first point 131 and the midpoint 133 and a second distance dl between the second point 132 and the midpoint 133 are measured; which illustrates, for different observation angles (ai) according to the direction of observation, that the method comprises: - A step E3 of measuring a first distance dO between the first point and the midpoint of the first portrait 12 according to an observation angle considered among the different observation angles; and - A step E4 of measuring a second distance dl between the second point and the midpoint of the first portrait 12 according to the observation angle considered.
[0130] Then according to another interesting characteristic of the invention, a ratio “r” between the first distance dO and the second distance dl is calculated, for the observation angle a; considered. This illustrates a step E5 of calculating a ratio r between the first distance dO and the second distance dl, for the observation angle considered, for the first portrait 12.
[0131] In the case of a 2D portrait, this ratio varies little depending on the observation angle, whereas for a 3D portrait, this ratio varies much more strongly depending on the observation angle a;.
[0132] [Fig.3] represents the identity document 10 of [Fig.l] on which the characteristic points, for each of the main portrait and the secondary portrait, have been represented, for illustrative purposes.
[0133] In the context of authentication of an identity document, according to an interesting option, an authentication method comprises for example a prior step E8 of verifying a match between the main portrait and the secondary portrait.
[0134] For this, each of the portraits is for example observed perpendicularly (i.e. according to an observation angle a = 0 relative to the portrait).
[0135] A perpendicular observation of the portrait generally corresponds to an observation of the face from the front. Thus, whether the portrait is in 3D or 2D, the two images obtained must then correspond when the portraits are authentic.
[0136] In the event of a match, or independently of this comparison, the method of authenticating the identity document advantageously includes steps aimed at verifying whether one of the portraits, here the secondary portrait 12, is actually a 3D portrait.
[0137] A series of images of the identity document is taken from different observation angles; for example from at least three angles α, or even more, for example from -7° to +7°, every 0.5°.
[0138] Different characteristic points, as represented [Fig.2] are detected on the portrait, and at least one ratio r = dO / dl is calculated, at each angle.
[0139] The method then comprises for example: - A step of determining a direction of observation of the secondary portrait 12, for example here according to the horizontal direction of observation D; - A step E2 of identifying at least three characteristic points 13 of the face of the secondary portrait 12, including a first point 131, a second point 132 and a so-called “median” point 133; - An optional step of determining several observation angles (00 according to the observation direction D;
[0140] and for each of the observation angles a;: • A step E3 of measuring a first distance dO between the first point 131 and the midpoint 133 of the secondary portrait 12 according to the observation angle considered; • A step E4 of measuring a second distance dl between the second point 132 and the midpoint 133 of the secondary portrait 12 according to the observation angle considered; • A step E5 of calculating the ratio r between the first distance dO and the second distance dl, for the observation angle considered, for the secondary portrait 12.
[0141] Then here, by way of illustration, the method comprises a processing step E6 of the ratios which comprises: - A step of determining a trend curve based on the ratios calculated for each of the observation angles, for the secondary portrait 12; and - A step of calculating a steering coefficient of the trend curve, for secondary portrait 12.
[0142] Here, the trend curve is a straight line which is represented in [Fig.4], by the straight line Cl.
[0143] Each of the points represented by “+” here represents a ratio (r = dO / dl for example, for the secondary portrait), on the ordinate, for an observation angle a; according to the observation direction D, on the abscissa.
[0144] Despite some dispersion of the points on the graph of [Fig.4], linked to a prototype 3D image not optimized in the context of this example and to measurement inaccuracies, the points nevertheless make it possible to determine the line Cl, for example by a least squares minimization or any other technique.
[0145] The line Cl is here a decreasing line.
[0146] We consider, for illustration purposes, that the line Cl, with the standard equation “y = ax + b”, where “a” represents the slope (direction coefficient), and “b” the ordinate at the origin, has the slope value: a = -0.014, or in absolute value lal = 0.014, representing here a representative parameter.
[0147] Such a slope value allows us to consider that the portrait in question is a 3D portrait. Indeed, the more the observation angle increases, the more the face looks to the side, here the right side for example, and consequently, the more the ratio r decreases.
[0148] Here, in parallel, since the identity document 10 also includes a main portrait 11, theoretically in 2D, the same steps are applied to the main portrait 11.
[0149] Thus, a series of images of the identity document is taken from different observation angles; for example from at least three angles α, or even more, for example from -7° to +7°, every 0.5°.
[0150] For each of the main portrait and the secondary portrait, different characteristic points, as represented [Fig.2] are detected.
[0151] At least one ratio is calculated for each portrait, at each angle.
[0152] The method, or even the processing step E6, then comprises, for example: - A step of identifying at least three characteristic points 13 of the face of the main portrait 11, including a first point, a second point and a so-called “median” point, which may for example correspond to those chosen for the secondary portrait 12; - Possibly a step of determining several observation angles (a;) according to the observation direction D; these are for example the same angles than those used for the analysis of the secondary portrait;
[0153] and for each of the observation angles a;: • A step of measuring a first distance dO between the first point and the midpoint of the main portrait 11 according to the observation angle considered; • A step of measuring a second distance dl between the second point and the midpoint of the main portrait 11 according to the observation angle considered; • A step of calculating a ratio between the first distance and the second distance, for the observation angle considered, for the main portrait 11; • A step of determining a trend curve based on the ratios calculated for each of the observation angles, for the main portrait 11; and • A step of calculating a steering coefficient of the trend curve, for the main portrait 11.
[0154] Such a trend curve is here a straight line represented on [Fig.4] also, by the straight line C2.
[0155] Each of the points represented by “o” here represents a ratio (r = dO / dl for example, for the main portrait 11) as a function of the observation angle a; considered.
[0156] Similarly, despite some dispersion of the points on the graph of [Fig.4], linked to a prototype 3D image not optimized in the context of this example and to measurement inaccuracies, the points nevertheless make it possible to determine the line C2, for example by a least squares minimization or any other technique.
[0157] The line C2 is also a decreasing line here.
[0158] We consider, for illustration purposes, that the line C2, with the standard equation "y = a'x+ b'", where "a'" represents the slope (direction coefficient), and "b'" the ordinate at the origin, has the slope value: a' = -0.003, or in absolute value la' I = 0.003, forming for example the representative parameter for the second portrait, or a reference value.
[0159] Such a slope value makes it possible to consider that the portrait in question is indeed a 2D portrait.
[0160] This also illustrates that for a 2D portrait, the ratio (r = dO / dl) varies relatively little depending on the observation angle.
[0161] In the present exemplary embodiment, it is possible to compare the slope of the line Cl (for the secondary portrait 12) with that of the line C2 (for the main portrait 11).
[0162] Comparing the line Cl and the line C2 makes it possible to compensate for variations possibly due to a change in perspective during the shot, for example.
[0163] Thus, here a ratio la / a'l = 4.667.
[0164] It is clear from the graph in [Fig.4] that the secondary portrait, represented by the line Cl, can therefore be considered as a 3D portrait.
[0165] Generally speaking, a portrait can be considered a 3D portrait when such a ratio becomes greater than a predetermined threshold (reference value for this case), for example when the ratio is greater than 1, or preferably 1.5, or even 2.
[0166] In the absence of a 2D portrait allowing a comparison of the direction coefficients, or when wishing to analyze a curve obtained intrinsically, for example here the curve Cl resulting from the secondary portrait, it is possible to compare the slope “a” (or generally the direction coefficient of the corresponding curve) to a reference value, noted “Vr”.
[0167] The reference value may depend on each case. It may, for example, be prefixed by the provider of the identity document or other.
[0168] Or in a case of taking into account the curve obtained for a 2D portrait, the absolute value of the slope coefficient can be taken as a reference value (for example here the reference value could be: Vr = la'1=0.003).
[0169] The absolute value of the direction coefficient obtained for a 2D portrait can then be recorded as a reference value, for example in a memory of an authentication device.
[0170] So, for example, if an absolute value of the director coefficient of the trend curve of the portrait considered is greater than the reference value (for example here if lal > Vr), then the portrait considered is validated as a 3D portrait.
[0171] In the present example, it appears that lal=0.017 is much greater than Vr = 0.003, which is consistent with the validation of the secondary portrait as a 3D portrait.
[0172] This illustrates an example of implementation of a step E7 of validation of the first portrait, here the secondary portrait 12, as a 3D portrait based on a result of the processing step E6.
[0173] [Fig.5] shows a comparison graph of the ratios represented [Fig.4], for every angle of observation.
[0174] On the graph of [Fig.5], the ratios of the secondary portrait 12 are represented on the abscissa, while the ratios of the main portrait 11, which is in 2D, are represented on the ordinate, for each observation angle.
[0175] For example, this involves carrying out a study of the dispersion for two lists of ratio values (for each of the two portraits considered), a simple mathematical tool is, for example, a calculation of standard deviation.
[0176] Indeed, if a sufficient number of images are acquired with varied angles (without needing to know the values of the angles), two lists of ratio values are obtained.
[0177] The processing step can then include a step of calculating the standard deviation for each list, i.e. for the ratios obtained for each portrait.
[0178] In the case of a 2D portrait and a 3D portrait, each of the lists can be easily classified because the standard deviation of the ratio values corresponding to the 3D portrait is significantly larger than that of the 2D portrait.
[0179] For example, we note o3D the standard deviation of the list of ratios presenting the highest standard deviation and o2D the other list.
[0180] It is then possible to validate the presence of a 3D portrait when the ratio R = o 3D / o2D is greater than a reference value V (with V greater than 1).
[0181] In the case of two 2D portraits, the standard deviations of the ratios would be substantially identical and then: R«l.
[0182] In a configuration where the relationship between ratio and angle is linear (case of a constant density law, as shown in figure 5), then we would possibly have R = o3D / o 2d = P3o / P2D with P3D the slope of the line between ratio and angle for the 3D portrait, and P2D the slope of the line between ratio and angle for the 2D portrait. Indeed, in this case we have, a priori, n , with — CLmax ^min- ^D~ 3^2
[0183] Here, for illustration purposes, the ratios of the secondary portrait 12 extend on the abscissa over a range of width RI, while the ratios of the main portrait 11 extend on the ordinate over a range of width R2.
[0184] By crossing, these areas RI, R2 form an RG zone.
[0185] In this zone, the points can also be represented by a curve (for example a straight line) whose direction coefficient is then comparable or even identical to the ratio obtained between the direction coefficients of the curves Cl, C2 of [Fig.4], i.e. R = P3D / P2D, as explained above.
[0186] This approach also makes it possible to visually observe that the ratios of a 3D portrait spread out more than for a 2D portrait.
[0187] In a case where the first portrait is not 3D, the RG zone would be reduced, or even comparable to a point.
[0188] To implement the method as described previously, the invention also proposes an authentication device 20, shown diagrammatically [Fig.6].
[0189] The authentication device 20 is then configured to implement the steps of the method as described previously.
[0190] For example, the authentication device 20 comprises: - An optical system 21 configured to view a portrait and identify at least three characteristic points of a face of the portrait, including a first point, a second point and a so-called “median” point; - A calculator 22 configured to measure a first distance between the first point and the midpoint, measure a second distance between the second point and the midpoint and calculate a ratio between the first distance and the second distance, for each observation angle considered, and to process at least ratios and / or, where appropriate, to determine a slope of an average straight line of the ratios calculated for each of the observation angles, for the portrait; - A display 23 configured to communicate information validating the portrait as a 3D portrait.
[0191] Such an authentication device 20 thus allows a security agent or other user to know whether the controlled portrait is actually a 3D portrait.
[0192] The display 23 illustrated here comprises, for illustration purposes, two diodes, for example a green one (if validation) and a red one (if the portrait is not validated as a 3D portrait for example).
[0193] According to an interesting option, the device further comprises here a memory 24 configured to record at least one reference value, for example a threshold, or any other useful parameter.
[0194] In the present example, the device is further configured to apply the aforementioned steps to both the secondary portrait and the primary portrait, as needed.
[0195] Such a device is further configured to verify the concordance between the main portrait and the secondary portrait.
Claims
Claims
1. Method for authenticating an identity document (10), the method comprising at least: - A step (El) of providing an identity document (10) comprising at least a first portrait (12) representing a face; - A step of determining a direction of observation of the first portrait (12); - A step (E2) of identifying at least three characteristic points (131, 132, 133) of the face of the first portrait (12), including a first point (131), a second point (132) and a so-called “median” point (133); and for different observation angles (00 depending on the observation direction: • A step (E3) of measuring a first distance (dO) between the first point and the midpoint of the first portrait (12) according to an observation angle considered among the different observation angles; • A step (E4) of measuring a second distance (dl) between the second point and the midpoint of the first portrait (12) according to the observation angle considered; and • A step (E5) of calculating a ratio (r) between the first distance (dO) and the second distance (dl), for the observation angle considered, for the first portrait (12); • A step (E6) of processing the ratios (r) calculated for each of the different observation angles for the first portrait (12); the processing step comprising a step of determining a parameter representative of an evolution of the ratios (r) as a function of the observation angle, and a step of comparing the representative parameter with a reference value; and • A step (E7) of validating the first portrait (12) as a 3D portrait based on a result of the processing step.
2. The method of claim 1, wherein the determining step of a representative parameter comprises: - A step of determining an average curve (Cl) as a function of the ratios calculated for each of the observation angles, for the first portrait (12); and - A step of calculating at least one direction coefficient of the average curve (Cl) for the first portrait (12); And in which the comparison step comprises a comparison of an absolute value of the direction coefficient with respect to a reference value, and if the absolute value of the direction coefficient is greater than the reference value, the step of validating the first portrait (12) as a 3D portrait returns a positive result, confirming that the first portrait (12) is indeed a 3D portrait.
3. Method according to claim 1, in which the step of determining a representative parameter comprises a step of determining a deviation of the calculated ratios, for all the observation angles considered, for the first portrait (12), and in which the comparison step comprises a comparison of the deviation with respect to a reference value, and if the deviation is greater than the reference value, the step of validating the first portrait (12) as a 3D portrait returns a positive result, confirming that the first portrait (12) is indeed a 3D portrait.
4. Method according to any one of claims 1 to 3, comprising a step of identifying at least three characteristic points of a face of a second portrait (11) of the identity document (10), including a first point, a second point and a so-called "median" point; and for different observation angles along the observation direction, the method comprising: • A step of measuring a first distance between the first point and the median point of the second portrait (11) according to the observation angle considered among the different observation angles; • A step of measuring a second distance between the second point and the median point of the second portrait (11) according to the observation angle considered; and • A step of calculating a ratio between the first distance and the second distance, for the observation angle considered, for the second portrait (11).
5. Method according to claim 4, in which the processing step further comprises a step of determining a parameter representative of a change in the ratios as a function of the observation angle for the second portrait (11), and in which the processing step further comprises a step of recording the parameter representative for the second portrait (11) as a reference value.
6. Method according to any one of claims 4 or 5, comprising a step (E8) of verifying a match between the first portrait (12) and the second portrait (11).
7. Authentication device (20) of an identity document (10) configured to implement a method according to any one of claims 1 to 6.
8. Authentication device (20) according to claim 7, characterized in that it comprises: - An optical system (21) configured to view a first portrait (12) and identify at least three characteristic points of a face of the first portrait (12), including a first point, a second point and a so-called "median" point; - A calculator (22) configured to measure a first distance between the first point and the median point, measure a second distance between the second point and the median point and calculate a ratio between the first distance and the second distance, for each of different observation angles considered, and to process ratios calculated for each of the different observation angles, for the first portrait (12); - A display (23) configured to communicate information for validating the first portrait (12) as a 3D portrait.
9. Authentication device (20) according to any one of claims 7 or 8, characterized in that it further comprises a memory (24) configured to record at least one reference value.