Safety identification method and system based on neural network

Abstract

The present invention discloses a security identification method and system based on a neural network, the method comprising: obtaining a user image of a target identification user; determining a face image and a body image in the user image based on a classification algorithm; determining the posture information corresponding to the body image based on a posture recognition algorithm; and determining the malicious identification probability corresponding to the target identification user based on the face image and the posture information. It can be seen that the present invention can more accurately and efficiently identify identification users with malicious verification, improve the security of face recognition, and reduce the occurrence of security accidents.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a neural network-based security identification method and system. Background Art

[0002] With the development of face recognition technology and the increasing demand for smart access control, more and more access control systems use face recognition technology to allow users to verify their access. However, this has brought with it a test of the security of face recognition. Specific people have begun to use fake face images to try to break through access control. Therefore, how to verify faces more safely has become an important technical issue. When improving the security of face recognition, the existing technology often only considers the background of the area near the face and the characteristics of the face image itself to improve the verification security, but does not fully consider the user's body posture to improve the verification security. Therefore, when applied in a specific access control scenario, there is no way to fully utilize the user's overall image to improve the verification security. It can be seen that the existing technology has defects that need to be solved urgently. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide a neural network-based security identification method and system, which can more accurately and efficiently identify maliciously verified users, improve the security of face recognition, and reduce the occurrence of security accidents.

[0004] In order to solve the above technical problems, the first aspect of the present invention discloses a security identification method based on a neural network, the method comprising:

[0005] Obtain a user image of a target identification user;

[0006] Determine a face image and a body image in the user image based on a classification algorithm;

[0007] Determining posture information corresponding to the body image based on a posture recognition algorithm;

[0008] The malicious recognition probability corresponding to the target recognition user is determined according to the face image and the posture information.

[0009] As an optional implementation, in the first aspect of the present invention, determining the face image and body image in the user image based on a classification algorithm includes:

[0010] Inputting the user image into a trained first probability prediction algorithm model to obtain a first probability that each image part in the user image belongs to a face image;

[0011] Inputting the user image into a trained second probability prediction algorithm model to obtain a second probability that each image part in the user image belongs to a body image;

[0012] A face image and a body image in the user image are determined according to the first probability and the second probability.

[0013] As an optional embodiment, in the first aspect of the present invention, the first probability prediction algorithm model is obtained by training a training data set including multiple training user images and corresponding facial image annotations; the second probability prediction algorithm model is obtained by training a training data set including multiple training user images and corresponding body image annotations.

[0014] As an optional implementation, in the first aspect of the present invention, determining the face image and the body image in the user image according to the first probability and the second probability includes:

[0015] For each of the image parts, calculating a probability difference between the first probability and the second probability corresponding to the image part;

[0016] Determine whether the probability difference corresponding to the image portion is greater than a preset difference threshold, and obtain a determination result;

[0017] If the judgment result is no, determining that the image type corresponding to the image portion is other images;

[0018] If the judgment result is yes, determining the image type corresponding to the larger one of the first probability and the second probability as the image type corresponding to the image part; the image type is a face image or a body image;

[0019] Based on a dynamic programming algorithm, a face image and a body image in the user image are determined according to the image type corresponding to each of the image parts.

[0020] As an optional implementation, in the first aspect of the present invention, the determining the face image and the body image in the user image based on the dynamic programming algorithm according to the image type corresponding to each image part includes:

[0021] Setting the objective function to minimize the number of image parts in the face image group and the body image group;

[0022] Setting restrictions includes:

[0023] The image types corresponding to each of the image parts in the face image group are all face images;

[0024] The image type corresponding to each image part in the body image group is a body image;

[0025] The difference between the probability difference values ​​corresponding to any two image parts in the face image group or the body image group is less than a preset parameter threshold;

[0026] The distance between the positions of any two image parts in the face image group or the body image group is less than a preset distance threshold;

[0027] Based on a dynamic programming algorithm, according to the objective function and the constraint conditions, iterative grouping calculation is performed on all the image parts to obtain a face image group and a body image group that meet the conditions;

[0028] The face image group and body image group that meet the conditions are respectively determined as the face images and body images in the user image.

[0029] As an optional implementation, in the first aspect of the present invention, the posture information is a malicious posture or a normal posture; the malicious posture includes at least one of a holding object state, a hand covering the face state, a mask wearing state and a back-to-back state.

[0030] As an optional implementation, in the first aspect of the present invention, determining the posture information corresponding to the body image based on a posture recognition algorithm includes:

[0031] Inputting the body image into a preset posture recognition image model to obtain a predicted probability that the body image belongs to each of the malicious postures;

[0032] For each of the malicious gestures, determining whether the predicted probability corresponding to the malicious gesture is greater than a preset probability threshold;

[0033] If so, it is determined that the posture information corresponding to the body image includes the malicious posture.

[0034] As an optional implementation, in the first aspect of the present invention, determining the malicious recognition probability corresponding to the target recognition user according to the face image and the posture information includes:

[0035] Inputting the face image into a preset plane probability prediction algorithm model to obtain the plane probability that the face image belongs to a plane image;

[0036] Inputting the face image into a preset user probability prediction algorithm model to obtain a user probability that the face image belongs to a user who can pass the current access control;

[0037] Based on a preset rule of malicious scores corresponding to different postures, calculating a score corresponding to each malicious posture in the posture information;

[0038] Calculate the product of the score corresponding to each malicious posture and the corresponding probability weight to obtain a posture score corresponding to each malicious posture; the probability weight is proportional to the predicted probability;

[0039] Calculate the combination of the posture scores corresponding to all the malicious postures in the posture information to obtain a malicious score;

[0040] The weighted sum of the user probability, the plane probability and the malicious score is calculated to obtain the malicious identification probability corresponding to the target identified user; when the malicious identification probability is greater than a preset probability threshold, it is used to determine that the target identified user is a malicious identification user.

[0041] A second aspect of an embodiment of the present invention discloses a security identification system based on a neural network, the system comprising:

[0042] An acquisition module, used to acquire a user image of a target identification user;

[0043] A classification module, used for determining a face image and a body image in the user image based on a classification algorithm;

[0044] A recognition module, used to determine the posture information corresponding to the body image based on a posture recognition algorithm;

[0045] A determination module is used to determine the malicious recognition probability corresponding to the target recognition user based on the face image and the posture information.

[0046] As an optional implementation, in the second aspect of the present invention, the classification module determines the specific manner of the face image and the body image in the user image based on the classification algorithm, including:

[0047] Inputting the user image into a trained first probability prediction algorithm model to obtain a first probability that each image part in the user image belongs to a face image;

[0048] Inputting the user image into a trained second probability prediction algorithm model to obtain a second probability that each image part in the user image belongs to a body image;

[0049] A face image and a body image in the user image are determined according to the first probability and the second probability.

[0050] As an optional embodiment, in the second aspect of the present invention, the first probability prediction algorithm model is obtained by training a training data set including multiple training user images and corresponding facial image annotations; the second probability prediction algorithm model is obtained by training a training data set including multiple training user images and corresponding body image annotations.

[0051] As an optional implementation, in the second aspect of the present invention, the classification module determines the specific manner of the face image and the body image in the user image according to the first probability and the second probability, including:

[0052] For each of the image parts, calculating a probability difference between the first probability and the second probability corresponding to the image part;

[0053] Determine whether the probability difference corresponding to the image portion is greater than a preset difference threshold, and obtain a determination result;

[0054] If the judgment result is no, determining that the image type corresponding to the image portion is other images;

[0055] If the judgment result is yes, determining the image type corresponding to the larger one of the first probability and the second probability as the image type corresponding to the image part; the image type is a face image or a body image;

[0056] Based on a dynamic programming algorithm, a face image and a body image in the user image are determined according to the image type corresponding to each of the image parts.

[0057] As an optional implementation, in the second aspect of the present invention, the classification module determines the specific manner of the face image and the body image in the user image based on the dynamic programming algorithm according to the image type corresponding to each image part, including:

[0058] Setting the objective function to minimize the number of image parts in the face image group and the body image group;

[0059] Setting restrictions includes:

[0060] The image types corresponding to each of the image parts in the face image group are all face images;

[0061] The image type corresponding to each image part in the body image group is a body image;

[0062] The difference between the probability difference values ​​corresponding to any two image parts in the face image group or the body image group is less than a preset parameter threshold;

[0063] The distance between the positions of any two image parts in the face image group or the body image group is less than a preset distance threshold;

[0064] Based on a dynamic programming algorithm, according to the objective function and the constraint conditions, iterative grouping calculation is performed on all the image parts to obtain a face image group and a body image group that meet the conditions;

[0065] The face image group and body image group that meet the conditions are respectively determined as the face images and body images in the user image.

[0066] As an optional implementation, in the second aspect of the present invention, the posture information is a malicious posture or a normal posture; the malicious posture includes at least one of a holding object state, a hand covering the face state, a mask wearing state and a back-to-back state.

[0067] As an optional implementation, in the second aspect of the present invention, the recognition module determines the specific manner of the posture information corresponding to the body image based on the posture recognition algorithm, including:

[0068] Inputting the body image into a preset posture recognition image model to obtain a predicted probability that the body image belongs to each of the malicious postures;

[0069] For each of the malicious gestures, determining whether the predicted probability corresponding to the malicious gesture is greater than a preset probability threshold;

[0070] If so, it is determined that the posture information corresponding to the body image includes the malicious posture.

[0071] As an optional implementation, in the second aspect of the present invention, the specific manner in which the determination module determines the malicious recognition probability corresponding to the target recognition user according to the face image and the posture information includes:

[0072] Inputting the face image into a preset plane probability prediction algorithm model to obtain the plane probability that the face image belongs to a plane image;

[0073] Inputting the face image into a preset user probability prediction algorithm model to obtain a user probability that the face image belongs to a user who can pass the current access control;

[0074] Based on a preset rule of malicious scores corresponding to different postures, calculating a score corresponding to each malicious posture in the posture information;

[0075] Calculate the product of the score corresponding to each malicious posture and the corresponding probability weight to obtain a posture score corresponding to each malicious posture; the probability weight is proportional to the predicted probability;

[0076] Calculate the combination of the posture scores corresponding to all the malicious postures in the posture information to obtain a malicious score;

[0077] The weighted sum of the user probability, the plane probability and the malicious score is calculated to obtain the malicious identification probability corresponding to the target identified user; when the malicious identification probability is greater than a preset probability threshold, it is used to determine that the target identified user is a malicious identification user.

[0078] The third aspect of the present invention discloses another security identification system based on a neural network, the system comprising:

[0079] A memory storing executable program code;

[0080] a processor coupled to the memory;

[0081] The processor calls the executable program code stored in the memory to execute part or all of the steps in the neural network-based security identification method disclosed in the first aspect of the present invention.

[0082] The fourth aspect of the present invention discloses a computer storage medium, which stores computer instructions. When the computer instructions are called, they are used to execute some or all of the steps in the neural network-based security identification method disclosed in the first aspect of the present invention.

[0083] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0084] The present invention can determine the face image and body image in the user image based on the classification algorithm, and then determine the posture information corresponding to the body image based on the posture recognition algorithm, so as to comprehensively and accurately determine the malicious recognition probability corresponding to the target identified user, thereby being able to more accurately and efficiently identify the identified users with malicious verification, improve the security of face recognition, and reduce the occurrence of safety accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. 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 creative work.

[0086] Figure 1 It is a flowchart of a neural network-based security identification method disclosed in an embodiment of the present invention.

[0087] Figure 2 It is a structural schematic diagram of a neural network-based security identification system disclosed in an embodiment of the present invention.

[0088] Figure 3 It is a schematic diagram of the structure of another neural network-based security identification system disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0089] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0090] The terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, device, product or equipment that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units that are inherent to these processes, methods, products or equipment.

[0091] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0092] The present invention discloses a security identification method and system based on a neural network, which can determine the face image and body image in the user image based on a classification algorithm, and then determine the posture information corresponding to the body image based on a posture recognition algorithm, so as to comprehensively and accurately determine the malicious identification probability corresponding to the target identification user, thereby being able to more accurately and efficiently identify the identification user with malicious verification, improve the security of face recognition, and reduce the occurrence of security accidents. The following are detailed descriptions.

[0093] Embodiment 1

[0094] See also Figure 1 , Figure 1 : is a flowchart of a neural network-based security identification method disclosed in an embodiment of the present invention. Figure 1The described neural network-based security identification method can be applied to a data processing system / data processing device / data processing server (wherein the server includes a local processing server or a cloud processing server). Figure 1 As shown, the neural network-based security identification method may include the following operations:

[0095] 101. Obtain a user image of a target identification user.

[0096] 102. Based on the classification algorithm, determine the face image and body image in the user image.

[0097] 103. Based on a posture recognition algorithm, determine the posture information corresponding to the body image.

[0098] 104. Determine the malicious recognition probability corresponding to the target recognition user based on the face image and posture information.

[0099] It can be seen that the above-mentioned embodiments of the invention can determine the face image and body image in the user image based on the classification algorithm, and then determine the posture information corresponding to the body image based on the posture recognition algorithm, so as to comprehensively and accurately determine the malicious recognition probability corresponding to the target identified user, thereby being able to more accurately and efficiently identify maliciously verified identified users, improve the security of face recognition, and reduce the occurrence of safety accidents.

[0100] As an optional embodiment, in the above step, determining the face image and body image in the user image based on the classification algorithm includes:

[0101] Inputting the user image into the trained first probability prediction algorithm model to obtain a first probability that each image part in the user image belongs to a face image;

[0102] Inputting the user image into the trained second probability prediction algorithm model to obtain a second probability that each image part in the user image belongs to a body image;

[0103] A face image and a body image in the user image are determined according to the first probability and the second probability.

[0104] It can be seen that through the above-mentioned optional embodiments, the probability of each image part in the user image belonging to a face image or a body image can be determined respectively based on the two prediction algorithm models, so as to comprehensively determine the face image and body image in the user image, facilitate the subsequent malicious recognition probability determination, and assist in more accurately and efficiently identifying maliciously verified users, thereby improving the security of face recognition and reducing the occurrence of safety accidents.

[0105] As an optional embodiment, in the above steps, the first probability prediction algorithm model is obtained by training a training data set including multiple training user images and corresponding facial image annotations; the second probability prediction algorithm model is obtained by training a training data set including multiple training user images and corresponding body image annotations.

[0106] It can be seen that through the above optional embodiments, the training details of the two prediction algorithm models are clarified, so as to accurately determine the probability that each image part in the user image belongs to a face image or a body image, so as to comprehensively determine the face image and body image in the user image, so as to facilitate the subsequent malicious recognition probability determination, and assist in more accurately and efficiently identifying maliciously verified users, thereby improving the security of face recognition and reducing the occurrence of safety accidents.

[0107] As an optional embodiment, in the above step, determining the face image and the body image in the user image according to the first probability and the second probability includes:

[0108] For each image part, calculating a probability difference between a first probability and a second probability corresponding to the image part;

[0109] Determine whether the probability difference corresponding to the image portion is greater than a preset difference threshold, and obtain a determination result;

[0110] If the judgment result is no, determining that the image type corresponding to the image portion is other images;

[0111] If the judgment result is yes, the image type corresponding to the larger one of the first probability and the second probability is determined as the image type corresponding to the image part; optionally, the image type is a face image or a body image;

[0112] Based on a dynamic programming algorithm, the face image and body image in the user image are determined according to the image type corresponding to each image part.

[0113] It can be seen that through the above-mentioned optional embodiments, the image type corresponding to each image part can be determined based on the calculation and judgment of the probability difference between the first probability and the second probability corresponding to each image part, and then the face image and body image in the user image can be determined based on the dynamic programming algorithm, so as to facilitate the subsequent malicious recognition probability determination, assist in more accurate and efficient identification of maliciously verified users, improve the security of face recognition, and reduce the occurrence of safety accidents.

[0114] As an optional embodiment, in the above steps, based on a dynamic programming algorithm, according to the image type corresponding to each image part, determining the face image and the body image in the user image includes:

[0115] The objective function is set to minimize the number of image parts in both the face image group and the body image group;

[0116] Setting restrictions includes:

[0117] The image type corresponding to each image part in the face image group is a face image;

[0118] The image type corresponding to each image part in the body image group is a body image;

[0119] The difference between the probability difference values ​​corresponding to any two image parts in the face image group or the body image group is less than a preset parameter threshold;

[0120] The distance between the positions of any two image parts in the face image group or the body image group is less than a preset distance threshold;

[0121] Based on the dynamic programming algorithm, according to the objective function and the constraints, all image parts are iteratively grouped to obtain the face image group and the body image group that meet the conditions;

[0122] The face image group and the body image group that meet the conditions are respectively determined as the face images and the body images in the user image.

[0123] It can be seen that through the above-mentioned optional embodiments, based on the set objective function and constraints, the facial image and body image in the user image can be accurately determined based on the dynamic programming algorithm, which is convenient for the subsequent determination of the probability of malicious recognition, and assists in more accurately and efficiently identifying maliciously verified users, thereby improving the security of facial recognition and reducing the occurrence of safety accidents.

[0124] As an optional embodiment, in the above steps, the posture information is a malicious posture or a normal posture; the malicious posture includes at least one of a holding object state, a hand covering the face state, a mask wearing state and a back-to-back state.

[0125] It can be seen that through the above optional embodiments, the content of the posture information is clarified, which can more accurately characterize the user's current posture, facilitate the subsequent determination of the probability of malicious recognition, and assist in more accurately and efficiently identifying maliciously verified users, thereby improving the security of face recognition and reducing the occurrence of safety accidents.

[0126] As an optional embodiment, in the above step, determining the posture information corresponding to the body image based on the posture recognition algorithm includes:

[0127] Inputting the body image into a preset posture recognition image model to obtain a predicted probability that the body image belongs to each malicious posture;

[0128] For each malicious gesture, determine whether the predicted probability corresponding to the malicious gesture is greater than a preset probability threshold;

[0129] If so, it is determined that the posture information corresponding to the body image includes the malicious posture.

[0130] It can be seen that through the above-mentioned optional embodiments, the predicted probability that the body image belongs to each malicious posture can be determined based on the preset posture recognition image model, so as to determine the malicious posture type in the posture information corresponding to the body image, facilitate the subsequent malicious recognition probability determination, and assist in more accurately and efficiently identifying maliciously verified users, thereby improving the security of face recognition and reducing the occurrence of safety accidents.

[0131] As an optional embodiment, in the above step, determining the malicious recognition probability corresponding to the target recognition user according to the face image and posture information includes:

[0132] Inputting the face image into a preset plane probability prediction algorithm model to obtain the plane probability that the face image belongs to a plane image;

[0133] Input the face image into a preset user probability prediction algorithm model to obtain the user probability that the face image belongs to a user who can pass the current access control;

[0134] Based on the preset rules of malicious scores corresponding to different postures, calculate the score corresponding to each malicious posture in the posture information;

[0135] Calculate the product of the score corresponding to each malicious posture and the corresponding probability weight to obtain the posture score corresponding to each malicious posture; optionally, the probability weight is proportional to the predicted probability;

[0136] Calculate the combined posture scores corresponding to all malicious postures in the posture information to obtain a malicious score;

[0137] The weighted sum of the user probability, plane probability and malicious score is calculated to obtain the malicious identification probability corresponding to the target identification user; when the malicious identification probability is greater than a preset probability threshold, it is used to determine that the target identification user is a malicious identification user.

[0138] It can be seen that through the above optional embodiments, the scores corresponding to whether the facial image belongs to a plane image, the degree of fit between the facial image and the user, and the malicious posture can be calculated respectively, and the probability of malicious identification can be comprehensively determined. The behavior of maliciously identified users can be effectively characterized and quantified, so as to achieve more accurate and efficient identification of maliciously verified users, improve the security of facial recognition, and reduce the occurrence of safety accidents.

[0139] Embodiment 2

[0140] See also Figure 2, Figure 2 : is a schematic diagram of a neural network-based security identification system disclosed in an embodiment of the present invention. Figure 2 The described neural network-based security identification system can be applied to a data processing system / data processing device / data processing server (wherein the server includes a local processing server or a cloud processing server). Figure 2 As shown, the neural network-based security identification system may include:

[0141] The acquisition module 201 is used to acquire the user image of the target recognition user.

[0142] The classification module 202 is used to determine the face image and body image in the user image based on the classification algorithm.

[0143] The recognition module 203 is used to determine the posture information corresponding to the body image based on the posture recognition algorithm.

[0144] The determination module 204 is used to determine the malicious recognition probability corresponding to the target recognition user according to the face image and posture information.

[0145] It can be seen that the above-mentioned embodiments of the invention can determine the face image and body image in the user image based on the classification algorithm, and then determine the posture information corresponding to the body image based on the posture recognition algorithm, so as to comprehensively and accurately determine the malicious recognition probability corresponding to the target identified user, thereby being able to more accurately and efficiently identify maliciously verified identified users, improve the security of face recognition, and reduce the occurrence of safety accidents.

[0146] As an optional embodiment, the classification module determines the specific manner of the face image and the body image in the user image based on the classification algorithm, including:

[0147] Inputting the user image into the trained first probability prediction algorithm model to obtain a first probability that each image part in the user image belongs to a face image;

[0148] Inputting the user image into the trained second probability prediction algorithm model to obtain a second probability that each image part in the user image belongs to a body image;

[0149] A face image and a body image in the user image are determined according to the first probability and the second probability.

[0150] It can be seen that through the above-mentioned optional embodiments, the probability of each image part in the user image belonging to a face image or a body image can be determined respectively based on the two prediction algorithm models, so as to comprehensively determine the face image and body image in the user image, facilitate the subsequent malicious recognition probability determination, and assist in more accurately and efficiently identifying maliciously verified users, thereby improving the security of face recognition and reducing the occurrence of safety accidents.

[0151] As an optional embodiment, the first probability prediction algorithm model is trained by a training data set including multiple training user images and corresponding facial image annotations; the second probability prediction algorithm model is trained by a training data set including multiple training user images and corresponding body image annotations.

[0152] It can be seen that through the above optional embodiments, the training details of the two prediction algorithm models are clarified, so as to accurately determine the probability that each image part in the user image belongs to a face image or a body image, so as to comprehensively determine the face image and body image in the user image, so as to facilitate the subsequent malicious recognition probability determination, and assist in more accurately and efficiently identifying maliciously verified users, thereby improving the security of face recognition and reducing the occurrence of safety accidents.

[0153] As an optional embodiment, the classification module determines the specific manner of the face image and the body image in the user image according to the first probability and the second probability, including:

[0154] For each image part, calculating a probability difference between a first probability and a second probability corresponding to the image part;

[0155] Determine whether the probability difference corresponding to the image portion is greater than a preset difference threshold, and obtain a determination result;

[0156] If the judgment result is no, determining that the image type corresponding to the image portion is other images;

[0157] If the judgment result is yes, the image type corresponding to the larger one of the first probability and the second probability is determined as the image type corresponding to the image part; optionally, the image type is a face image or a body image;

[0158] Based on a dynamic programming algorithm, the face image and body image in the user image are determined according to the image type corresponding to each image part.

[0159] It can be seen that through the above-mentioned optional embodiments, the image type corresponding to each image part can be determined based on the calculation and judgment of the probability difference between the first probability and the second probability corresponding to each image part, and then the face image and body image in the user image can be determined based on the dynamic programming algorithm, so as to facilitate the subsequent malicious recognition probability determination, assist in more accurate and efficient identification of maliciously verified users, improve the security of face recognition, and reduce the occurrence of safety accidents.

[0160] As an optional embodiment, the classification module determines the specific manner of the face image and the body image in the user image based on the dynamic programming algorithm according to the image type corresponding to each image part, including:

[0161] The objective function is set to minimize the number of image parts in both the face image group and the body image group;

[0162] Setting restrictions includes:

[0163] The image type corresponding to each image part in the face image group is a face image;

[0164] The image type corresponding to each image part in the body image group is a body image;

[0165] The difference between the probability difference values ​​corresponding to any two image parts in the face image group or the body image group is less than a preset parameter threshold;

[0166] The distance between the positions of any two image parts in the face image group or the body image group is less than a preset distance threshold;

[0167] Based on the dynamic programming algorithm, according to the objective function and the constraints, all image parts are iteratively grouped to obtain the face image group and the body image group that meet the conditions;

[0168] The face image group and the body image group that meet the conditions are respectively determined as the face images and the body images in the user image.

[0169] It can be seen that through the above-mentioned optional embodiments, based on the set objective function and constraints, the facial image and body image in the user image can be accurately determined based on the dynamic programming algorithm, which is convenient for the subsequent determination of the probability of malicious recognition, and assists in more accurately and efficiently identifying maliciously verified users, thereby improving the security of facial recognition and reducing the occurrence of safety accidents.

[0170] As an optional embodiment, the posture information is a malicious posture or a normal posture; the malicious posture includes at least one of a state of holding an object, a state of covering the face with hands, a state of wearing a mask, and a state of facing away.

[0171] It can be seen that through the above optional embodiments, the content of the posture information is clarified, which can more accurately characterize the user's current posture, facilitate the subsequent determination of the probability of malicious recognition, and assist in more accurately and efficiently identifying maliciously verified users, thereby improving the security of face recognition and reducing the occurrence of safety accidents.

[0172] As an optional embodiment, the specific manner in which the recognition module determines the posture information corresponding to the body image based on the posture recognition algorithm includes:

[0173] Inputting the body image into a preset posture recognition image model to obtain a predicted probability that the body image belongs to each malicious posture;

[0174] For each malicious gesture, determine whether the predicted probability corresponding to the malicious gesture is greater than a preset probability threshold;

[0175] If so, it is determined that the posture information corresponding to the body image includes the malicious posture.

[0176] It can be seen that through the above-mentioned optional embodiments, the predicted probability that the body image belongs to each malicious posture can be determined based on the preset posture recognition image model, so as to determine the malicious posture type in the posture information corresponding to the body image, facilitate the subsequent malicious recognition probability determination, and assist in more accurately and efficiently identifying maliciously verified users, thereby improving the security of face recognition and reducing the occurrence of safety accidents.

[0177] As an optional embodiment, the specific manner in which the determination module determines the malicious recognition probability corresponding to the target recognition user according to the face image and posture information includes:

[0178] Inputting the face image into a preset plane probability prediction algorithm model to obtain the plane probability that the face image belongs to a plane image;

[0179] Input the face image into a preset user probability prediction algorithm model to obtain the user probability that the face image belongs to a user who can pass the current access control;

[0180] Based on the preset rules of malicious scores corresponding to different postures, calculate the score corresponding to each malicious posture in the posture information;

[0181] Calculate the product of the score corresponding to each malicious posture and the corresponding probability weight to obtain the posture score corresponding to each malicious posture; optionally, the probability weight is proportional to the predicted probability;

[0182] Calculate the combined posture scores corresponding to all malicious postures in the posture information to obtain a malicious score;

[0183] The weighted sum of the user probability, plane probability and malicious score is calculated to obtain the malicious identification probability corresponding to the target identification user; when the malicious identification probability is greater than a preset probability threshold, it is used to determine that the target identification user is a malicious identification user.

[0184] It can be seen that through the above optional embodiments, the scores corresponding to whether the facial image belongs to a plane image, the degree of fit between the facial image and the user, and the malicious posture can be calculated respectively, and the probability of malicious identification can be comprehensively determined. The behavior of maliciously identified users can be effectively characterized and quantified, so as to achieve more accurate and efficient identification of maliciously verified users, improve the security of facial recognition, and reduce the occurrence of safety accidents.

[0185] Embodiment 3

[0186] See also Figure 3 , Figure 3 It is another neural network-based security identification system disclosed in an embodiment of the present invention. Figure 3 The described neural network-based security identification system is applied to a data processing system / data processing device / data processing server (wherein the server includes a local processing server or a cloud processing server). Figure 3 As shown, the neural network-based security identification system may include:

[0187] A memory 301 storing executable program codes;

[0188] a processor 302 coupled to the memory 301;

[0189] The processor 302 calls the executable program code stored in the memory 301 to execute the steps of the neural network-based security identification method described in the first embodiment.

[0190] Embodiment 4

[0191] An embodiment of the present invention discloses a computer-readable storage medium storing a computer program for electronic data exchange, wherein the computer program enables a computer to execute the steps of the neural network-based security identification method described in the first embodiment.

[0192] Embodiment 5

[0193] An embodiment of the present invention discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to enable a computer to execute the steps of the neural network-based security identification method described in the first embodiment.

[0194] The above describes specific embodiments of the present specification, and other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily have to be performed in the specific order or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0195] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0196] For the convenience of description, the above device is described in various units according to their functions. Of course, when implementing this specification, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0197] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems, or computer program products. Therefore, the embodiments of this specification may be in the form of complete hardware embodiments, complete software embodiments, or embodiments in combination with software and hardware. Moreover, the embodiments of this specification may be in the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0198] This specification is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of this specification. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0199] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0200] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0201] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0202] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0203] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0204] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0205] This specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. This specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0206] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0207] Finally, it should be noted that the neural network-based security identification method and system disclosed in the embodiments of the present invention are only preferred embodiments of the present invention, and are only used to illustrate the technical solution of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, it should be understood by those skilled in the art that the technical solutions described in the aforementioned embodiments can still be modified, or some of the technical features therein can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A security identification method based on neural network, characterized in that: The method comprises: Obtain a user image of a target identification user; Determining a face image and a body image in the user image based on a classification algorithm includes: Inputting the user image into a trained first probability prediction algorithm model to obtain a first probability that each image part in the user image belongs to a face image; Inputting the user image into a trained second probability prediction algorithm model to obtain a second probability that each image part in the user image belongs to a body image; For each of the image parts, calculating a probability difference between the first probability and the second probability corresponding to the image part; Determine whether the probability difference corresponding to the image portion is greater than a preset difference threshold, and obtain a determination result; If the judgment result is no, determining that the image type corresponding to the image portion is other images; If the judgment result is yes, determining the image type corresponding to the larger one of the first probability and the second probability as the image type corresponding to the image part; the image type is a face image or a body image; Setting the objective function to minimize the number of image parts in the face image group and the body image group; Setting restrictions includes: The image types corresponding to each of the image parts in the face image group are all face images; The image type corresponding to each image part in the body image group is a body image; The difference between the probability difference values ​​corresponding to any two image parts in the face image group or the body image group is less than a preset parameter threshold; The distance between the positions of any two image parts in the face image group or the body image group is less than a preset distance threshold; Based on a dynamic programming algorithm, according to the objective function and the constraint conditions, iterative grouping calculation is performed on all the image parts to obtain a face image group and a body image group that meet the conditions; Determine the face image group and body image group that meet the conditions as the face images and body images in the user image respectively; Determining posture information corresponding to the body image based on a posture recognition algorithm; The malicious recognition probability corresponding to the target recognition user is determined according to the face image and the posture information.

2. The neural network-based security identification method according to claim 1, characterized in that: The first probability prediction algorithm model is trained by a training data set including multiple training user images and corresponding facial image annotations; the second probability prediction algorithm model is trained by a training data set including multiple training user images and corresponding body image annotations.

3. The neural network-based security identification method according to claim 1, characterized in that: The posture information is a malicious posture or a normal posture; the malicious posture includes at least one of a holding object state, a hand covering face state, a mask wearing state and a back-to-face state.

4. The neural network-based security identification method according to claim 3, characterized in that: The determining the posture information corresponding to the body image based on the posture recognition algorithm includes: Inputting the body image into a preset posture recognition image model to obtain a predicted probability that the body image belongs to each of the malicious postures; For each of the malicious gestures, determining whether the predicted probability corresponding to the malicious gesture is greater than a preset probability threshold; If so, it is determined that the posture information corresponding to the body image includes the malicious posture.

5. The neural network-based security identification method according to claim 4, characterized in that: The determining, based on the face image and the posture information, a malicious recognition probability corresponding to the target recognition user includes: Inputting the face image into a preset plane probability prediction algorithm model to obtain the plane probability that the face image belongs to a plane image; Inputting the face image into a preset user probability prediction algorithm model to obtain a user probability that the face image belongs to a user who can pass the current access control; Based on a preset rule of malicious scores corresponding to different postures, calculating a score corresponding to each malicious posture in the posture information; Calculate the product of the score corresponding to each malicious posture and the corresponding probability weight to obtain a posture score corresponding to each malicious posture; the probability weight is proportional to the predicted probability; Calculate the combination of the posture scores corresponding to all the malicious postures in the posture information to obtain a malicious score; The weighted sum of the user probability, the plane probability and the malicious score is calculated to obtain the malicious identification probability corresponding to the target identified user; when the malicious identification probability is greater than a preset probability threshold, it is used to determine that the target identified user is a malicious identification user.

6. A security identification system based on neural network, characterized in that: The system comprises: An acquisition module, used to acquire a user image of a target identification user; A classification module, used to determine the face image and body image in the user image based on a classification algorithm, including: Inputting the user image into a trained first probability prediction algorithm model to obtain a first probability that each image part in the user image belongs to a face image; Inputting the user image into a trained second probability prediction algorithm model to obtain a second probability that each image part in the user image belongs to a body image; For each of the image parts, calculating a probability difference between the first probability and the second probability corresponding to the image part; Determine whether the probability difference corresponding to the image portion is greater than a preset difference threshold, and obtain a determination result; If the judgment result is no, determining that the image type corresponding to the image portion is other images; If the judgment result is yes, determining the image type corresponding to the larger one of the first probability and the second probability as the image type corresponding to the image part; the image type is a face image or a body image; Setting the objective function to minimize the number of image parts in the face image group and the body image group; Setting restrictions includes: The image types corresponding to each of the image parts in the face image group are all face images; The image type corresponding to each image part in the body image group is a body image; The difference between the probability difference values ​​corresponding to any two image parts in the face image group or the body image group is less than a preset parameter threshold; The distance between the positions of any two image parts in the face image group or the body image group is less than a preset distance threshold; Based on a dynamic programming algorithm, according to the objective function and the constraint conditions, iterative grouping calculation is performed on all the image parts to obtain a face image group and a body image group that meet the conditions; Determine the face image group and body image group that meet the conditions as the face images and body images in the user image respectively; A recognition module, used to determine the posture information corresponding to the body image based on a posture recognition algorithm; A determination module is used to determine the malicious recognition probability corresponding to the target recognition user based on the face image and the posture information.

7. A security identification system based on neural network, characterized in that: The system comprises: A memory storing executable program code; a processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the neural network-based security identification method as described in any one of claims 1-5.

Images