A method and system for processing face recognition data before payment

Abstract

The present invention relates to the field of face recognition technology, and in particular to a method and system for processing face recognition data before payment. The present invention analyzes each face structure area in a face image to be recognized of a payment user photographed by a face recognition terminal, and obtains the degree of illumination influence of each face structure area through the structural direction angle difference and overall grayscale difference between the face structure areas, thereby screening out abnormal corner areas according to the illumination influence degree, reducing the influence of feature corner point dislocation caused by illumination on the accuracy of face recognition, so that the accuracy of matching the face image to be recognized after removing the abnormal corner points with the face image stored in a face image database is higher, and the corresponding accuracy of face recognition before payment is also higher.

Description

Technical Field

[0001] The present invention relates to the field of face recognition technology, and in particular to a method and system for processing face recognition data before payment. Background Art

[0002] With the popularization of digital payment and the advancement of biometric technology, facial recognition payment is becoming more and more widely used, and it is more convenient than traditional digital payment methods. The principle of facial recognition payment is to collect the face image of the payment user to be recognized through the facial recognition terminal before the facial recognition payment is confirmed, and match the face image to be recognized with the image stored in the database, and perform face recognition before payment based on the matching results.

[0003] The prior art method of matching the collected face image to be identified with the image stored in the database usually adopts corner point detection, that is, the face image to be identified of the payment user is subjected to feature extraction by the corner point detection method, and the face image in the face database is matched according to the extracted corner points representing the key feature points. However, in the actual payment scenario, the facial features in the face image to be identified of the payment user may be affected by the ambient light, resulting in dislocation or unclear display; and the light sources around the payment user may come from different directions, resulting in different areas on the face of the payment user being affected by light differently; for example, the protruding positions of the face such as the nose and cheekbones are easily affected by light, resulting in some features of the area being dislocated or unclear, resulting in unclear or dislocated corner points of the corresponding area after corner point detection, thereby affecting the accuracy of matching the face image in the face database according to the corner points representing the key feature points detected by the corner points; that is, when the prior art does not consider the influence of light in the payment environment, the accuracy of matching the collected face image to be identified with the face image stored in the face image database by the corner point detection method is poor. Summary of the invention

[0004] In order to solve the technical problem that the existing technology has poor accuracy in matching the collected face image to be recognized with the face image stored in the face image database by the corner point detection method when the lighting effect in the payment environment is not considered, the purpose of the present invention is to provide a method and system for processing face recognition data before payment, and the technical solution adopted is as follows:

[0005] The present invention proposes a method for processing face recognition data before payment, the method comprising:

[0006] Before face recognition payment, a face recognition terminal is used to capture a face image of the paying user to be recognized; the face image to be recognized is divided into at least two face structure areas;

[0007] According to the grayscale gradient distribution trend in each facial structure region, the structural direction angle representing the structural influence direction of each facial structure region is obtained; according to the structural direction angle difference and the overall grayscale difference between each facial structure region and the adjacent facial structure region, the illumination influence degree of each facial structure region is obtained;

[0008] According to the degree of illumination influence, the abnormal corner point areas in all facial structure areas are screened out; after the abnormal corner points in the abnormal corner point areas in the face image to be identified are screened out, the face recognition before payment is performed through the face database.

[0009] Furthermore, the method for obtaining the structural direction angle includes:

[0010] For any face structure area:

[0011] The pixel with the largest gray value in the face structure area is used as the light source pixel; the pixel at the boundary position in the face structure area is used as the boundary pixel;

[0012] The line segment connecting the light source pixel point and each boundary pixel point is used as a reference line segment for each boundary pixel point; the pixels on each reference line segment are arranged in the order from the light source pixel point to the boundary pixel point, so as to obtain a pixel reference sequence corresponding to each reference line segment;

[0013] The grayscale variation disorder degree of each pixel on each reference line segment is obtained according to the grayscale difference between adjacent pixels in the pixel reference sequence relative to the grayscale difference between the light source pixel and the corresponding boundary pixel; the structural shadow influence degree of each reference line segment in the face structure area is obtained according to the overall size of the grayscale variation disorder degree of all pixels on each reference line segment;

[0014] The direction from the light source pixel point in each reference line segment to the corresponding boundary pixel point is used as the reference direction of each reference line segment. The degree of influence of the structural shadow is used as a weight, and combined with the reference direction, a structural direction angle representing the structural influence direction of the face structure area is obtained.

[0015] Furthermore, the method for calculating the grayscale change disorder degree includes:

[0016] ;

[0017] in, For the The first The pixel point in the reference sequence corresponding to the reference line segment The degree of disorder of grayscale changes of pixels; For the The first The pixel point in the reference sequence corresponding to the reference line segment The grayscale difference between a pixel and the previous pixel; For the Grayscale value of light source pixel in the individual face structure area; For the The first The gray value of the boundary pixel corresponding to the reference line segment; For the The first The number of pixels in the pixel reference sequence corresponding to the reference line segment.

[0018] Furthermore, the method for obtaining the degree of influence of the structural shadow includes:

[0019] The normalized value of the mean value of the chaotic degree of grayscale changes of all pixels on each reference line segment in each face structure region is used as the influence degree of the structural shadow on each reference line segment in each face structure region.

[0020] Furthermore, the method of obtaining a structural direction angle representing a structural influence direction of a face structure region by taking the structural shadow influence degree as a weight and combining it with the reference direction includes:

[0021] The angle difference between the reference direction of each reference line segment and the horizontal rightward direction is used as the reference angle of each reference line segment; the product of the reference angle and the degree of influence of the structural shadow is used as the weighted angle of each reference line segment;

[0022] The reference line segment whose corresponding structural shadow influence degree is greater than the preset structural influence threshold is taken as the structural influence line segment; the average of the weighted angles of all structural influence line segments is taken as the structural direction angle of the face structure area.

[0023] Furthermore, the method for obtaining the degree of illumination influence includes:

[0024] Each face structure region is taken as the target region in turn; the face structure region adjacent to the target region is taken as the adjacent region of the target region;

[0025] The negative correlation mapping value of the difference between the structural direction angle of each adjacent region and the structural direction angle of the target region is used as the adjacent gradient correlation of each adjacent region; the average of the adjacent gradient correlations of all adjacent regions corresponding to the target region is used as the regional gradient correlation of the target region;

[0026] According to the overall difference in grayscale values ​​between the target area and the corresponding adjacent areas, the significance of the grayscale value of the target area is obtained;

[0027] The normalized value of the product of the negative correlation mapping value of the regional gradient correlation and the gray value significance level is used as the illumination influence level of the target area.

[0028] Furthermore, the method for obtaining the gray value significance degree includes:

[0029] The difference between the mean grayscale value of all pixels in the target area and the mean grayscale value of all pixels in each adjacent area is taken as the reference grayscale deviation of each adjacent area; the mean of the reference grayscale deviations of all adjacent areas corresponding to the target area is taken as the grayscale significance of the target area.

[0030] Furthermore, the method for obtaining the corner point abnormal area includes:

[0031] The facial structure area where the illumination influence is greater than the preset influence threshold is regarded as the corner abnormal area.

[0032] Furthermore, the method of performing pre-payment face recognition through a face database after filtering out abnormal corner points in the abnormal corner point area in the face image to be recognized includes:

[0033] After filtering out the corner points in the corner point abnormal area in the face image to be identified that are located at non-edge positions, the remaining corner points of the face image to be identified are obtained; and the face image data in the face database are matched with the corner points according to the remaining corner points to obtain the face recognition result.

[0034] The present invention also proposes a pre-payment facial recognition data processing system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any one of the steps of a pre-payment facial recognition data processing method.

[0035] The present invention has the following beneficial effects:

[0036] In the actual payment process, the closer the distance between the paying user and the light source, the greater the illumination intensity corresponding to the face structure area directly illuminated by the light source, the greater the light reflected by the corresponding face structure area, and the more likely the feature corner points of the corresponding face structure area are to be misplaced or the features are not obvious, which is manifested as a relatively larger grayscale value of the corresponding face structure area on the face image to be identified; therefore, the degree of illumination influence can be reflected by the overall grayscale value of the face structure area. And according to the principle that light propagates in a straight line, when the light source is closer to the face structure area, the mapping direction corresponding to the shadows generated by the structures of each face structure area will be more different; and when the light source is farther away, the mapping direction corresponding to the shadows generated by the structures of each face structure area will be relatively smaller; therefore, the degree of proximity of the light source can be reflected according to the difference in the structural directions between the face structure areas, thereby reflecting the degree of illumination influence.

[0037] Therefore, the present invention first calculates the structural direction angle that characterizes the structural influence direction of each face structure area; combined with the relationship between the overall grayscale value of the face structure area and the degree of illumination influence, the illumination influence degree of each face structure area is obtained according to the structural direction angle difference and the overall grayscale difference between each face structure area and the adjacent face structure area, and the abnormal situation of the corner points on each face structure area is indirectly reflected by the illumination influence degree, so as to screen out the corner point abnormal areas with blurred corner point features or missing corner point features in all face structure areas according to the illumination influence degree. Therefore, after further screening out the abnormal corner points in the corner point abnormal area in the face image to be identified, the face recognition before payment is performed through the face database, so as to reduce the influence of the feature corner point dislocation caused by illumination on the accuracy of face recognition, so that the accuracy of matching the face image to be identified after removing the abnormal corner points with the face image stored in the face image database is higher, and the corresponding accuracy of face recognition before payment is also higher. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0039] Figure 1 A flow chart of a method for processing facial recognition data before payment provided by one embodiment of the present invention;

[0040] Figure 2 A schematic diagram of a facial structure region division standard provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0041] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the method and system for processing face recognition data before payment proposed by the present invention, its specific implementation, structure, features and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0042] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0043] The following is a detailed description of a method and system for processing facial recognition data before payment provided by the present invention in conjunction with the accompanying drawings.

[0044] See also Figure 1 , which shows a flow chart of a method for processing face recognition data before payment provided by an embodiment of the present invention, the method comprising:

[0045] Step S1: Before face recognition payment, a face recognition terminal is used to capture a face image of the paying user to be recognized; and the face image to be recognized is divided into at least two face structure areas.

[0046] The embodiment of the present invention aims to provide a method for processing face recognition data before payment, which is used to analyze each face structure area in the face image to be recognized of a payment user photographed by a face recognition terminal, and obtain the degree of illumination influence of each face structure area through the structural direction angle difference and overall grayscale difference between the face structure areas, so as to screen out the abnormal corner area according to the illumination influence degree, reduce the influence of the feature corner point dislocation caused by illumination on the accuracy of face recognition, so that the accuracy of matching the face image to be recognized after removing the abnormal corner points with the face image stored in the face image database is higher, and the corresponding accuracy of face recognition before payment is also higher.

[0047] Before face recognition payment, the embodiment of the present invention uses a face recognition terminal to capture the face image of the paying user to be recognized. Specifically, the face recognition terminal faces the paying user's face and captures the face image video of the paying user, and selects one frame from the face image video and grayscales it to obtain the face image to be recognized required by the embodiment of the present invention. The reason for grayscale is to facilitate subsequent grayscale analysis.

[0048] Since the embodiment of the present invention needs to screen the corner point abnormal area in combination with the degree of illumination influence corresponding to the facial structure in different areas, it is necessary to first divide the face image to be identified into at least two face structure areas. In the embodiment of the present invention, the face key point detector in the Dlib library is used to detect the key feature points in the face image, and then the face area is divided into at least two face structure areas according to the position of the key feature points and the face structure area division standard. Please refer to Figure 2 , which shows a schematic diagram of a facial structure area division standard provided by an embodiment of the present invention, and divides the face image to be identified into multiple face structure areas through the lines between the key feature points. It should be noted that the implementer may also adopt other facial structure area division standards according to the specific implementation environment, such as dividing the face evenly, which will not be further elaborated here; in addition, it should be noted that all the facial images to be identified obtained by the embodiment of the present invention are data that the paying user has authorized to use.

[0049] Step S2: According to the grayscale gradient distribution trend in each facial structure region, a structural direction angle representing the structural influence direction of each facial structure region is obtained; according to the structural direction angle difference and the overall grayscale difference between each facial structure region and the adjacent facial structure region, the illumination influence degree of each facial structure region is obtained.

[0050] In the actual payment process, the closer the distance between the paying user and the light source, the greater the illumination intensity corresponding to the face structure area directly illuminated by the light source, the greater the light reflected by the corresponding face structure area, and the more likely the feature corner points of the corresponding face structure area will be misplaced or the features will not be obvious. According to the principle that light propagates in a straight line, when the light source is closer to the face structure area, the mapping direction corresponding to the shadows generated by the structures of each face structure area will be greatly different; when the light source is farther away, the mapping direction corresponding to the shadows generated by the structures of each face structure area is relatively small; therefore, the degree of proximity of the light source can be reflected according to the difference in the structural direction between the face structure areas, thereby reflecting the degree of illumination influence; therefore, it is necessary to first analyze the structural direction angle of each face structure area. First of all, it is necessary to consider that when light is irradiated on the face, the illumination intensity at different positions will show a uniform change, so when only affected by light, the corresponding grayscale value will also show a uniform change trend. However, due to the existence of the face structure, the grayscale gradient law will be destroyed at the position affected by the face structure, so the direction of the face structure influence can be determined according to the destroyed grayscale gradient law, that is, the structural direction angle of each face structure area can be obtained. The embodiment of the present invention obtains a structural direction angle representing a structural influence direction of each facial structure region according to a grayscale gradient distribution trend in each facial structure region.

[0051] Preferably, the method for obtaining the structural direction angle includes:

[0052] For any face structure area:

[0053] The pixel with the largest grayscale value in the face structure area is used as the light source pixel; the pixel at the boundary position in the face structure area is used as the boundary pixel; the line segment connecting the light source pixel and each boundary pixel is used as the reference line segment for each boundary pixel; the pixels on each reference line segment are arranged in the order from the light source pixel to the boundary pixel to obtain a pixel reference sequence corresponding to each reference line segment; according to the grayscale difference between adjacent pixels in the pixel reference sequence, relative to the grayscale difference between the light source pixel and the corresponding boundary pixel, the degree of grayscale change disorder of each pixel on each reference line segment is obtained.

[0054] Preferably, the method for calculating the grayscale change disorder degree includes:

[0055] ;

[0056] in, For the The first The pixel point in the reference sequence corresponding to the reference line segment The degree of disorder of grayscale changes of pixels; For the The first The pixel point in the reference sequence corresponding to the reference line segment The grayscale difference between a pixel and the previous pixel; For the Grayscale value of light source pixel in the individual face structure area; For the The first The gray value of the boundary pixel corresponding to the reference line segment; For the The first The number of pixels in the pixel reference sequence corresponding to the reference line segment. Greater than 1, because the first pixel in all pixel reference sequences is a light source pixel, there is no previous pixel, and there is no need to calculate the degree of grayscale change disorder.

[0057] For each facial structure area, if the facial structure area is greatly affected by light, the corresponding light source pixel point is the position in the facial structure area that is most affected by light. Therefore, a straight line is extended to the edge with the light source pixel point as the center, and the direction corresponding to the grayscale change on each corresponding reference line segment that conforms to the grayscale gradient feature is the direction corresponding to the influence of light; therefore, the structural direction corresponding to the facial structure area can be obtained based on the direction corresponding to the reference line segment that does not conform or conforms poorly to the grayscale gradient feature.

[0058] For the pixel reference sequence of each reference line segment, if the grayscale change on the reference line segment conforms to the grayscale gradient feature more, then in the order from the light source pixel point to the boundary pixel point, the grayscale value of the pixel point should conform to the monotonically decreasing uniform change more. Therefore, for each pixel point on the reference line segment, if the grayscale value difference between the pixel point and the previous pixel point is closer to the ratio of the grayscale value difference at both ends of the pixel reference sequence corresponding to the corresponding reference line segment to the number of pixels, it means that the grayscale change at the local position of the pixel point conforms to the gradient law more, and the corresponding grayscale change is less chaotic. And when the overall grayscale change chaos of all pixels on the corresponding reference line segment is smaller, it means that the grayscale change of the pixel point on the reference line segment conforms to the grayscale gradient feature more, and the direction corresponding to the reference line segment is less likely to represent the structural direction. On the contrary, when the overall grayscale change chaos of all pixels on the corresponding reference line segment is greater, it means that the grayscale change of the pixel point on the reference line segment does not conform to the grayscale gradient feature, and the direction corresponding to the reference line segment is more likely to represent the structural direction.

[0059] Therefore, the embodiment of the present invention further obtains the structural shadow influence degree of each reference line segment in the face structure area according to the overall magnitude of the grayscale change disorder of all pixels on each reference line segment; preferably, the method for obtaining the structural shadow influence degree includes:

[0060] The normalized value of the mean value of the grayscale change disorder of all pixels on each reference line segment in each face structure region is used as the structural shadow influence degree on each reference line segment in each face structure region. The greater the corresponding structural shadow influence degree, the less the grayscale change of the overall pixel points on the reference line segment conforms to the grayscale gradient feature, and the direction corresponding to the corresponding reference line segment is more likely to represent the structural direction. In the embodiment of the present invention, the normalization method adopts linear normalization, and the implementer can adjust the normalization method according to the specific implementation environment.

[0061] In the embodiment of the present invention, The first The method for obtaining the influence degree of the structural shadow of a reference line segment is expressed in the formula as follows:

[0062] ;

[0063] in, For the The first The influence degree of structural shadow of reference line segment, For the The first The number of pixels in the pixel reference sequence corresponding to the reference line segment; No. The first The pixel point in the reference sequence corresponding to the reference line segment The degree of grayscale change disorder of each pixel.

[0064] In another embodiment of the present invention, the normalized value of the standard deviation of the grayscale value difference of all pixels on the pixel reference sequence corresponding to each reference line segment is used as the degree of influence of the structural shadow on each reference line segment on each facial structure area; the larger the normalized value of the corresponding standard deviation of the difference value, the less the grayscale change in the corresponding pixel reference sequence conforms to the grayscale gradient feature, and the more likely the direction of the corresponding reference line segment is to represent the structural direction.

[0065] Further, the direction from the light source pixel point in each reference line segment to the corresponding boundary pixel point is used as the reference direction of each reference line segment; the degree of structural shadow influence is used as a weight, combined with the reference direction, to obtain a structural direction angle representing the structural influence direction of the face structure area. Preferably, the method of using the degree of structural shadow influence as a weight, combined with the reference direction, to obtain a structural direction angle representing the structural influence direction of the face structure area includes:

[0066] The angular difference between the reference direction of each reference line segment and the horizontal right direction is taken as the reference angle of each reference line segment; the product of the reference angle and the degree of structural shadow influence is taken as the weighted angle of each reference line segment; the reference line segment whose corresponding structural shadow influence degree is greater than the preset structural influence threshold is taken as the structural influence line segment; the average of the weighted angles of all structural influence line segments is taken as the structural direction angle of the face structure area.

[0067] Since the greater the degree of structural shadow influence, the greater the degree of influence of the structural shadow on the direction of the corresponding reference line segment, the greater the proportion of the reference angle of the corresponding reference direction, that is, the greater the weighted angle. Since the direction corresponding to the line segment with a smaller degree of structural shadow influence is the direction affected by illumination, the direction affected by illumination is screened out by the preset structural influence threshold, so that the structural direction angle obtained later based on the structural influence line segment is more accurate. In the embodiment of the present invention, the preset structural influence threshold is set to 0.5, and the implementer can adjust it according to the specific implementation environment.

[0068] In another embodiment of the present invention, the average value of the reference angles corresponding to the reference line segments whose corresponding structural shadow influence degree is greater than a preset structural influence threshold is directly used as the structural direction angle of the face structure region.

[0069] In the embodiment of the present invention, The method for obtaining the structural direction angle of the individual face structure area is expressed in the formula:

[0070] ;

[0071] in, For the The structural direction angle of the individual face structure area, For the The number of reference line segments in the individual face structure area whose structural shadow influence is greater than a preset structural influence threshold; For the The first The structural shadow influence degree of the reference line segment whose structural shadow influence degree is greater than the preset structural influence threshold; For the The first The reference angle of the reference line segment whose structure shadow influence degree is greater than the preset structure influence threshold.

[0072] According to the principle that light propagates in a straight line, when the light source is closer to the face structure area, the mapping direction corresponding to the shadow generated by the structure of each face structure area will be more different; when the light source is farther away, the mapping direction corresponding to the shadow generated by the structure of each face structure area is relatively small; therefore, the proximity of the light source can be reflected according to the difference in the structural direction between the face structure areas, thereby reflecting the degree of illumination influence. And for each face structure area, the greater the illumination intensity it receives, the greater the light reflected by the face structure area, which is manifested in the face image to be identified as the corresponding face structure area is relatively brighter, and the corresponding grayscale value is larger overall. Therefore, the embodiment of the present invention combines the relationship between the overall grayscale value of the face structure area and the degree of illumination influence, and the relationship between the structural direction difference between the face structure areas and the degree of illumination influence, and obtains the degree of illumination influence of each face structure area according to the structural direction angle difference and the overall grayscale difference between each face structure area and the adjacent face structure area.

[0073] Preferably, the method for obtaining the degree of illumination influence includes:

[0074] Since it is necessary to compare each facial structure area with other areas to characterize the degree of illumination influence, the embodiment of the present invention takes each facial structure area as the target area in turn; takes the facial structure area adjacent to the target area as the adjacent area of ​​the target area; and takes the adjacent area as the comparison area of ​​the target area for subsequent analysis.

[0075] The negative correlation mapping value of the difference between the structural direction angle of each adjacent area and the structural direction angle of the target area is used as the adjacent gradient correlation of each adjacent area; because the greater the difference in structural direction between the face structure areas, the closer the light source distance, and the greater the corresponding degree of illumination influence. Therefore, the smaller the adjacent gradient correlation of each adjacent area corresponding to the target area, the greater the adjacent gradient correlation, and the smaller the corresponding degree of illumination influence. Therefore, the adjacent gradient correlation of all adjacent areas corresponding to the target area is further combined, and the mean of the adjacent gradient correlation of all adjacent areas corresponding to the target area is used as the regional gradient correlation of the target area, so that the greater the regional gradient correlation, the less the target area is affected by illumination.

[0076] In the embodiment of the present invention, The method for obtaining the regional gradient correlation of the individual face structure area is expressed in the formula as follows:

[0077] ;

[0078] in, For the The regional gradient correlation of individual facial structure regions, For the The number of adjacent regions corresponding to the individual facial structure region; For the The structural direction angle of the individual face structure area is related to the corresponding The angle between the structural orientation angles of adjacent regions; is a normalization function, and the normalization method in the embodiment of the present invention adopts linear normalization.

[0079] When the light intensity of the face structure area is greater, the corresponding overall gray value is greater than that of other areas. Therefore, the embodiment of the present invention obtains the gray value significance of the target area according to the overall difference of the gray value between the target area and the corresponding adjacent areas. Preferably, the method for obtaining the gray value significance includes:

[0080] The difference between the mean grayscale value of all pixels in the target area and the mean grayscale value of all pixels in each adjacent area is used as the reference grayscale deviation of each adjacent area; the mean of the reference grayscale deviation of all adjacent areas corresponding to the target area is used as the grayscale significance of the target area. The mean grayscale value corresponding to each facial structure area can characterize the overall grayscale value of the corresponding area; therefore, the larger the grayscale value of the target area relative to each adjacent area, that is, the larger the reference grayscale deviation of each adjacent area, the larger the overall grayscale value of the target area; that is, the greater the grayscale significance of the target area, the greater the impact of light on the target area.

[0081] In the embodiment of the present invention, The method for obtaining the gray value significance of the individual face structure area is expressed in the formula as follows:

[0082] ;

[0083] in, For the The gray value significance of the individual facial structure area; For the The number of adjacent regions corresponding to the individual facial structure region; For the The mean gray value of all pixels in the individual face structure area; For the The corresponding first The mean gray value of all pixels in the adjacent area; For the The corresponding first The reference grayscale deviation of adjacent regions.

[0084] Since the smaller the regional gradient correlation is and the greater the grayscale value significance is, the more the target area is affected by light, the embodiment of the present invention uses the normalized value of the product of the negative correlation mapping value of the regional gradient correlation and the grayscale value significance as the degree of light influence on the target area.

[0085] In the embodiment of the present invention, The method for obtaining the degree of illumination influence on the individual face structure area is expressed in the formula as follows:

[0086] ;

[0087] in, For the The degree of lighting influence on the structural area of ​​the individual's face; For the The gray value significance of the individual facial structure area; For the The regional gradient correlation of individual facial structural regions; is an exponential function with a natural constant as base; is a normalization function. All normalization functions in the embodiments of the present invention adopt linear normalization. The implementer may adopt other normalization methods according to the specific implementation environment, which will not be further elaborated here.

[0088] Step S3: Screen out abnormal corner points in all facial structure areas according to the degree of illumination influence; after screening out abnormal corner points in the abnormal corner points in the facial image to be identified, perform pre-payment face recognition through the face database.

[0089] As the degree of illumination influence on the face structure area is greater, that is, the greater the illumination intensity, the greater the light reflected by the corresponding skin surface, and the feature corner points of the corresponding face structure area are more likely to be misplaced or have unclear features; therefore, the embodiment of the present invention further screens out the corner point abnormal areas according to the illumination influence degree of each face structure area. The embodiment of the present invention screens out the corner point abnormal areas in all face structure areas according to the illumination influence degree; preferably, the method for obtaining the corner point abnormal areas includes:

[0090] The face structure area with the illumination influence degree greater than the preset influence threshold is regarded as the corner abnormal area. In the embodiment of the present invention, the preset influence threshold is set to 0.8, and the implementer can adjust it according to the specific implementation environment.

[0091] Further, the embodiment of the present invention removes abnormal corner points in the abnormal corner point region in the face image to be identified, and then performs face recognition before payment through the face database, thereby reducing the influence of feature corner point misalignment caused by illumination on the accuracy of face recognition. Preferably, after removing abnormal corner points in the abnormal corner point region in the face image to be identified, the method for performing face recognition before payment through the face database includes:

[0092] Taking into account that corner points at edge positions are not easily affected by light, while corner points at non-edge positions may become blurred or misaligned after being exposed to strong light, the embodiments of the present invention screen out corner points at non-edge positions in the corner point abnormal area in the face image to be identified, and obtain the remaining corner points of the face image to be identified; perform corner point matching on the face image data in the face database according to the remaining corner points to obtain a face recognition result, thereby reducing the influence of feature corner point misalignment caused by light on the accuracy of face recognition, so that the accuracy of matching the face image to be identified after removing the abnormal corner points with the face image stored in the face image database is higher, and the corresponding accuracy of face recognition before payment is also higher.

[0093] To summarize, the present invention analyzes each facial structure area in the facial image to be identified of the paying user captured by a facial recognition terminal, and obtains the degree of illumination influence of each facial structure area through the structural direction angle difference and the overall grayscale difference between the facial structure areas, thereby screening out abnormal corner areas according to the degree of illumination influence, reducing the influence of the misalignment of characteristic corner points caused by illumination on the accuracy of facial recognition, so that the accuracy of matching the facial image to be identified after removing the abnormal corner points with the facial images stored in the facial image database is higher, and the corresponding accuracy of facial recognition before payment is also higher.

[0094] The present invention also proposes a pre-payment facial recognition data processing system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any one of the steps of a pre-payment facial recognition data processing method.

[0095] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0096] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A method for processing facial recognition data before payment, characterized in that: The method comprises: Before face recognition payment, a face recognition terminal is used to capture a face image of the paying user to be recognized; the face image to be recognized is divided into at least two face structure areas; According to the grayscale gradient distribution trend in each facial structure region, the structural direction angle representing the structural influence direction of each facial structure region is obtained; according to the structural direction angle difference and the overall grayscale difference between each facial structure region and the adjacent facial structure region, the illumination influence degree of each facial structure region is obtained; Screen out abnormal corner points in all facial structure areas according to the degree of illumination influence; after screening out abnormal corner points in the abnormal corner points in the facial image to be identified, perform pre-payment facial recognition through the facial database; The method for obtaining the structural direction angle includes: For any face structure area: The pixel with the largest gray value in the face structure area is used as the light source pixel; the pixel at the boundary position in the face structure area is used as the boundary pixel; The line segment connecting the light source pixel point and each boundary pixel point is used as a reference line segment for each boundary pixel point; the pixels on each reference line segment are arranged in the order from the light source pixel point to the boundary pixel point, so as to obtain a pixel reference sequence corresponding to each reference line segment; The grayscale variation disorder degree of each pixel on each reference line segment is obtained according to the grayscale difference between adjacent pixels in the pixel reference sequence relative to the grayscale difference between the light source pixel and the corresponding boundary pixel; the structural shadow influence degree of each reference line segment in the face structure area is obtained according to the overall size of the grayscale variation disorder degree of all pixels on each reference line segment; The direction from the light source pixel point in each reference line segment to the corresponding boundary pixel point is used as the reference direction of each reference line segment. The degree of influence of the structural shadow is used as a weight, and combined with the reference direction, a structural direction angle representing the structural influence direction of the face structure area is obtained.

2. A method for processing facial recognition data before payment according to claim 1, characterized in that: The calculation method of the grayscale change disorder degree includes: Among them, R k,h,x Δg is the degree of grayscale change disorder of the x-th pixel in the reference sequence of pixels corresponding to the h-th reference line segment in the k-th face structure area; k,h,x is the grayscale difference between the xth pixel and the previous pixel in the reference sequence of pixels corresponding to the hth reference line segment in the kth face structure region; g ′ k is the gray value of the light source pixel in the kth face structure area; g ′ k ′ ,h is the gray value of the boundary pixel corresponding to the hth reference line segment in the kth face structure area; n k,h is the number of pixels in the reference sequence of pixels corresponding to the h-th reference line segment in the k-th face structure region.

3. A method for processing face recognition data before payment according to claim 1, characterized in that: The method for obtaining the degree of influence of the structural shadow includes: The normalized value of the mean value of the chaotic degree of grayscale changes of all pixels on each reference line segment in each face structure region is used as the influence degree of the structural shadow on each reference line segment in each face structure region.

4. A method for processing facial recognition data before payment according to claim 1, characterized in that: The method of obtaining a structural direction angle representing a structural influence direction of a face structure region by taking the structural shadow influence degree as a weight and combining it with the reference direction comprises: The angle difference between the reference direction of each reference line segment and the horizontal rightward direction is used as the reference angle of each reference line segment; the product of the reference angle and the degree of influence of the structural shadow is used as the weighted angle of each reference line segment; The reference line segment whose corresponding structural shadow influence degree is greater than the preset structural influence threshold is taken as the structural influence line segment; the average of the weighted angles of all structural influence line segments is taken as the structural direction angle of the face structure area.

5. A method for processing facial recognition data before payment according to claim 1, characterized in that: The method for obtaining the degree of illumination influence includes: Each face structure region is taken as the target region in turn; the face structure region adjacent to the target region is taken as the adjacent region of the target region; The negative correlation mapping value of the difference between the structural direction angle of each adjacent region and the structural direction angle of the target region is used as the adjacent gradient correlation of each adjacent region; the average of the adjacent gradient correlations of all adjacent regions corresponding to the target region is used as the regional gradient correlation of the target region; According to the overall difference in grayscale values ​​between the target area and the corresponding adjacent areas, the significance of the grayscale value of the target area is obtained; The normalized value of the product of the negative correlation mapping value of the regional gradient correlation and the gray value significance level is used as the illumination influence level of the target area.

6. A method for processing face recognition data before payment according to claim 5, characterized in that: The method for obtaining the gray value significance degree includes: The difference between the mean grayscale value of all pixels in the target area and the mean grayscale value of all pixels in each adjacent area is taken as the reference grayscale deviation of each adjacent area; the mean of the reference grayscale deviations of all adjacent areas corresponding to the target area is taken as the grayscale significance of the target area.

7. A method for processing facial recognition data before payment according to claim 1, characterized in that: The method for obtaining the corner point abnormal area includes: The facial structure area where the illumination influence is greater than the preset influence threshold is regarded as the corner abnormal area.

8. A method for processing facial recognition data before payment according to claim 1, characterized in that: The method for performing pre-payment face recognition through a face database after filtering out abnormal corner points in an abnormal corner point region in a face image to be recognized comprises: After filtering out the corner points in the corner point abnormal area in the face image to be identified that are located at non-edge positions, the remaining corner points of the face image to be identified are obtained; and the face image data in the face database are matched with the corner points according to the remaining corner points to obtain the face recognition result.

9. A pre-payment face recognition data processing system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method for processing facial recognition data before payment as described in any one of claims 1 to 8 are implemented.

Images