CN114582031A - Living body detection auxiliary method and device for face recognition system - Google Patents

Abstract

A living body detection auxiliary method and device for a face recognition system comprises establishing an attack suspect list and initializing the list; extracting a face region by using a face detection algorithm; calculating the living body probability of the single-frame face by using a neural network model; calculating a face characteristic value; according to the face characteristic value, matching 'attack suspects'; according to the comparison between the living body probability of the suspect and a set threshold value, the value of the confidence coefficient is adjusted, so that the current suspect is judged to be true or false; the confidence of the suspect of attack decays over time: traversing all 'attack suspects', confidence-attenuation step length every interval of time T; when the confidence is 0, the "attack suspect" is deleted. The method establishes an 'attack suspect' list based on face characteristic value matching, and can filter most of long-time continuous attacks; and the dynamic adjustment is carried out in the matching process, so that errors caused by occasional misjudgment are avoided, and the judgment result is more stable.

Description

A kind of living body detection auxiliary method and device for face recognition system

technical field

The present invention relates to the field of computer technology, and in particular, to an auxiliary method and device for living body detection used in a face recognition system.

Background technique

In the process of real-time face recognition, it is necessary to be careful not to use fake faces for face recognition, so it is necessary to perform liveness detection before face recognition.

Many existing live detection algorithms are based on single-frame images. The specific method is to first perform face detection on the picture and then crop the face area, and then input the face area into the living body detection model to obtain the result. Because the results of single-frame image recognition often cannot have very high accuracy, when an attacker continuously conducts live attacks on the scene for a long time, there is a certain probability that the detection will break through at a certain time. The longer the time, the higher the probability of breakthrough. .

Therefore, it is urgent to propose a new method to solve the risk of long-term attacks in response to the long-term attack by attackers.

SUMMARY OF THE INVENTION

The technical problem to be solved by the present invention is to provide an auxiliary method and device for living body detection used in a face recognition system, which can greatly reduce the risk of long-term attacks.

In a first aspect, the present invention provides an auxiliary method for living body detection for a face recognition system, which specifically includes the following steps:

Step S1: establish a list of "attack suspects", each attack suspect contains two fields of confidence level P _attack and face feature, and initializes it, which is initially empty;

Step S2: extracting the face area with a face detection algorithm;

Step S3: use the neural network model to calculate the living probability of a single frame of face, and set it as P _live ;

Step S4: Calculate the facial feature value;

Step S5: match the "attack suspect" according to the facial feature value;

Step S6: Update the confidence level P _{attack of the "current attack suspect", and adjust the value of the confidence level P attack according} to the comparison between the probability P _live of the suspect's living body and the set threshold P _{live_threshold} , so as to judge that the current suspect is true or false;

Step S7: Attack suspect confidence P _attack decays with time: traverse all "attack suspects" every time interval T, confidence P _attack - decay step size Δd2, until the end of 0.0; when the confidence P _attack is 0, delete The "attack suspect"; go to step S2.

Further, the step S5 specifically includes:

Step S51 : traverse the list of "attack suspects", compare the similarity of facial features, if it is higher than the threshold P _{live_threshold} , lock it as "current attack suspect";

Step S52: If it does not exist, add “attack suspect”, the confidence level P _attack is 0, and set as “current attack suspect”.

Further, the step S6 specifically includes:

Step S61: If the probability P _live of the suspect is less than the threshold P _{live_threshold} , the suspect confidence P _attack + update step Δd1, until the confidence P _attack reaches 1.0, and the deadline is reached; if the current suspect confidence P _attack is greater than the threshold P _{attack_threshold} , the return result is false;

Step S62: If the probability P _live of the suspect alive is greater than or equal to the threshold P _{live_threshold} , the suspect confidence P _attack - update step Δd1, until the confidence P _attack reaches 0.0; if the current suspect confidence P _attack is less than the threshold P _{attack_threshold} , returns true, otherwise false.

In a second aspect, the present invention provides a living body detection auxiliary device for a face recognition system, comprising:

The "attack suspect" list initialization module is used to establish a "attack suspect" list. Each attack suspect contains two fields, the confidence level P _attack and the face feature, and is initialized. The initial value is empty;

The face area extraction module is used to extract the face area by using the face detection algorithm;

The living body probability calculation module of a single-frame face is used to calculate the living body probability of a single-frame face by using a neural network model, and is set as P _live ;

The face feature value calculation module is used to calculate the face feature value;

The "attack suspect" matching module is used to match the "attack suspect" according to the facial feature value,

The "current attack suspect" confidence level P _attack update module is used to adjust the value of the confidence level P _attack according to the comparison between the suspect's living probability P _live and the set threshold P _{live_threshold} , so as to judge the current suspect as true or false;

Attack suspect confidence level P _attack decays with time module, used to traverse all "attack suspects" every time interval T, confidence level P _attack -0.1 until 0.0 cut-off; when confidence level P _attack confidence level is 0, delete The "attack suspect".

Further, the "attack suspect" matching module specifically includes:

Traverse the list of "attack suspects", compare the similarity of facial features, if it is higher than the threshold P _{live_threshold} , lock it as "current attack suspect";

If it does not exist, add "attack suspect", the confidence level P _attack is 0, and set it to "current attack suspect".

Further, the "current attack suspect" confidence level P _attack update module specifically includes:

If the suspect's living probability P _live is less than the threshold P _{live_threshold} , the suspect's confidence level P _attack +0.05, until the confidence level P _attack reaches 1.0; if the current suspect's confidence level P _attack is greater than the threshold P _{attack_threshold} , the return result is false;

If the suspect's living probability P _live is greater than or equal to the threshold P _{live_threshold} , the suspect's confidence level P _attack -0.05, until the confidence level P _attack reaches 0.0; if the current suspect's confidence level P _attack is less than the threshold P _{attack_threshold} , the return result is true , otherwise false.

In a third aspect, the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, the processor implementing the program described in the first aspect when the processor executes the program method.

In a fourth aspect, the present invention provides a computer-readable storage medium on which a computer program is stored, which implements the method described in the first aspect when the program is executed by a processor.

The advantages of the present invention are:

1. The present invention establishes a list of "attack suspects" based on facial feature value matching, which can filter most of the long-term continuous attacks.

2. The present invention sets the confidence level for the "attack suspect", which is not black or white. Dynamic adjustment during the matching process avoids errors caused by occasional misjudgments, and the judgment results are more stable.

3. The present invention cleans up the list of "attack suspects" based on time decay, so as to avoid the extra cost of defense when there is no attack normally.

Description of drawings

The present invention will be further described below with reference to the accompanying drawings and embodiments.

FIG. 1 is a schematic flow chart of the method of the present invention.

Detailed ways

Example 1

As shown in Figure 1, an auxiliary method for living body detection for a face recognition system specifically includes the following steps:

Step S1: establish a list of "attack suspects", each attack suspect contains two fields of confidence level P _attack and face feature, and initializes it, which is initially empty;

Step S2: extract the face region with the MTCNN face detection algorithm;

Step S3: use the neural network model to calculate the living probability of a single frame of face, and set it as P _live ;

Step S4: Calculate the facial feature value;

Step S5: Match the "attack suspect" according to the facial feature value. The specific matching process includes:

Step S51 : traverse the list of "attack suspects", compare the similarity of facial features, if it is higher than the threshold P _{live_threshold} , lock it as "current attack suspect";

Step S52: if it does not exist, add “attack suspect”, the confidence level P _attack is 0, and set it as “current attack suspect”;

Step S6: Update the confidence level P _attack of the "current attack suspect", which specifically includes:

Step S61: If the probability P _live of the suspect alive is less than the threshold P _{live_threshold} , the suspect confidence P _attack +0.05 (update step Δd1 = 0.05, but not limited to this value), until the confidence P _attack reaches 1.0; if The current suspect confidence P _attack is greater than the threshold P _{attack_threshold} , and the returned result is false;

Step S62: If the probability P _live of the suspect is greater than or equal to the threshold P _{live_threshold} , the suspect’s confidence level P _attack is -0.05, until the confidence level P _attack reaches 0.0; if the current suspect’s confidence level P _attack is less than the threshold P _{attack_threshold} , return The result is true, otherwise false;

Step S7: The attack suspect's confidence P _attack decays with time: traverse all "attack suspects" at intervals (for example, one minute), the confidence P _attack -0.1 (the decay step size Δd2 = 0.1, but not limited to this value), until the cutoff of 0.0; when the confidence degree P _attack confidence degree is 0, delete the "attack suspect"; go to step S2.

Based on the same inventive concept, the present application also provides a device corresponding to the method in the first embodiment, and the details are in the second embodiment.

Embodiment 2

In this embodiment, a living body detection auxiliary device for a face recognition system is provided, including:

The "attack suspect" list initialization module is used to establish a "attack suspect" list. Each attack suspect contains two fields, the confidence level P _attack and the face feature, and is initialized. The initial value is empty;

The face area extraction module is used to extract the face area by using the face detection algorithm;

The living body probability calculation module of a single-frame face is used to calculate the living body probability of a single-frame face by using a neural network model, and is set as P _live ;

The face feature value calculation module is used to calculate the face feature value;

The "attack suspect" matching module is used to match the "attack suspect" according to the facial feature value. The specific matching process includes:

Traverse the list of "attack suspects", compare the similarity of facial features, if it is higher than the threshold P _{live_threshold} , lock it as "current attack suspect";

If it does not exist, add "attack suspect", the confidence level P _attack is 0, and set it to "current attack suspect";

"Current attack suspect" confidence level P _attack update module, including:

If the suspect's living probability P _live is less than the threshold P _{live_threshold} , the suspect's confidence level P _attack +0.05 (update step Δd1=0.05, but not limited to this value), until the confidence level P _attack reaches 1.0; if the current suspect The confidence level P _attack is greater than the threshold P _{attack_threshold} , and the returned result is false;

If the suspect's living probability P _live is greater than or equal to the threshold P _{live_threshold} , the suspect's confidence level P _attack -0.05, until the confidence level P _attack reaches 0.0; if the current suspect's confidence level P _attack is less than the threshold P _{attack_threshold} , the return result is true , otherwise false;

Attack suspect confidence degree P _attack decays with time module, used to traverse all "attack suspects" every time interval T, confidence degree P _attack -0.1 (decay step size Δd2 = 0.1, but not limited to this value), until 0.0 cutoff; when the confidence level P _attack confidence level is 0, delete the "attack suspect".

Since the device introduced in the second embodiment of the present invention is the device used to implement the method in the first embodiment of the present invention, based on the method introduced in the first embodiment of the present invention, those skilled in the art can understand the specific structure and deformation of the device. , so it is not repeated here. All devices used in the method of Embodiment 1 of the present invention belong to the scope of protection of the present invention.

Based on the same inventive concept, the present application provides an electronic device embodiment corresponding to the first embodiment. For details, refer to the third embodiment.

Embodiment 3

This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, any one of the embodiments in the first embodiment can be implemented.

Since the electronic device introduced in this embodiment is the device used to implement the method in the first embodiment of the present application, based on the method introduced in the first embodiment of the present application, those skilled in the art can understand the electronic device in this embodiment. The specific implementation manner and various modifications thereof, so how the electronic device implements the methods in the embodiments of the present application will not be described in detail here. As long as the devices used by those skilled in the art to implement the methods in the embodiments of the present application fall within the scope of the intended protection of the present application.

Based on the same inventive concept, the present application provides a storage medium corresponding to the first embodiment. For details, refer to the fourth embodiment.

Embodiment 4

This embodiment provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, any one of the implementation manners in the first embodiment can be implemented.

The main idea of the present invention is based on the continuity of the fake face attack on the time axis. In actual scenarios, fake face attacks usually continue to attack continuously for a period of time. If there are continuous attacks for a period of time, the probability of continuing attacks is very high. At this time, the attack can be temporarily blocked, greatly increasing the cost of fake face attack.

The key point of the present invention is to design a list of "attack suspects", and screen recent attackers based on the similarity of face recognition feature values. If a person's facial features appear fake face attack for many times, the facial feature value will be recorded, referred to as "attack suspect". The list holds face feature values and confidences that may be fake faces. The results of the live detection need to be compared with the list of "attack suspects" to determine that they are not "attack suspects" before proceeding to the next step of face recognition.

The probability of "attacking the suspect" changes with the environment: when a single-frame face is judged to be false, the confidence of the "attacking suspect" increases; List comparison. If there is a list of "attack suspects" and the confidence level is above the threshold, the live body does not pass. At the same time, the confidence level of the "attack suspect" will also drop. When it drops to a certain level, the "attack suspect" is deleted. The "attack suspect" confidence level decays over time. Delete the "attack suspect" when a time is not triggered.

The above descriptions are only examples of preferred implementations of the present invention, and are not intended to limit the protection scope of the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims (8)

1. A living body detection auxiliary method for a face recognition system is characterized in that: the method comprises the following steps:

step S1: establishing a 'suspect of attack' list, each suspect of attack containing a confidence level P_attackInitializing two fields of human face characteristics, and initializing the fields to be empty initially;

step S2: extracting a face region by using a face detection algorithm;

step S3: calculating the living body probability of a single frame of human face by using a neural network model, and setting the probability as P_live；

Step S4: calculating a face characteristic value;

step S5: according to the face characteristic value, matching 'attack suspects';

step S6: updating confidence P of' current attack suspect_attackAccording to the probability P of the suspect living body_liveAnd a set threshold value P_{live_threshold}Are compared with each other to adjust the confidence coefficient P_attackTo determine whether the current suspect is true or false;

step S7: confidence P of attack suspect_attackDecay with time: traversing all 'attack suspects' at intervals of a period of time T, and obtaining a confidence level P_attack-attenuation step Δ d2, up to 0.0 cut-off; when the confidence degree P_attackTo 0, delete the "suspect of attack"; go to step S2.

2. The live-detection assisting method for a face recognition system as claimed in claim 1, wherein: the step S5 specifically includes:

step S51: traversing the 'attack suspect' list, comparing the similarity of the face features, and keeping the similarity higher than a threshold value P_{live_threshold}Locked as "suspect of current attack";

step S52: if not, adding 'attack suspect' with confidence P_attack0, set to "suspect currently attacked".

3. The live body detection assistance method for a face recognition system according to claim 1, characterized in that: the step S6 specifically includes:

step S61: if the probability of suspect living body P_liveLess than a threshold value P_{live_threshold}Confidence level P of suspect_attack+ update the step Δ d1 until the confidence level P_attackCut off by 1.0; if the current suspect confidence level P_attackGreater than a threshold value P_{attack_threshold}Returning the result as false;

step S62: if the probability of suspect living body P_liveIs greater than or equal to threshold value P_{live_threshold}Confidence level P of suspect_attackUpdate the step size Δ d1 until the confidence P_attackCut off to 0.0; if the current suspect confidence level P_attackLess than a threshold value P_{attack_threshold}The return result is true, otherwise it is false.

4. A live body detection assisting apparatus for a face recognition system, characterized in that: the method comprises the following steps:

the 'attack suspects' list initialization module is used for establishing an 'attack suspects' list, and each attack suspects comprises a confidence degree P_attackInitializing two fields of human face characteristics, and initializing to be null;

the face region extraction module is used for extracting a face region by adopting a face detection algorithm;

the living body probability calculation module of the single-frame face is used for calculating the living body probability of the single-frame face by adopting a neural network model and is set as P_live；

The face characteristic value calculating module is used for calculating a face characteristic value;

the 'attack suspect' matching module is used for matching 'attack suspect' according to the face characteristic value,

confidence P of' current attack suspect_attackAn updating module for realizing the probability P of the suspect living body_liveAnd a set threshold value P_{live_threshold}Are compared with each other to adjust the confidence coefficient P_attackSo as to judge whether the current suspect is true or false;

confidence P of attack suspect_attackA time-dependent attenuation module for traversing all 'attack suspects' at intervals of a period of time T and obtaining a confidence level P_attack-attenuation step Δ d2, up to 0.0 cut-off; when the confidence degree P_attackThe confidence is 0, and the "suspect of attack" is deleted.

5. A live body detection assisting apparatus for a face recognition system as set forth in claim 4, wherein: the matching module for the 'attack suspect' specifically comprises:

traversing the 'attack suspect' list, comparing the similarity of the face features, and keeping the similarity higher than a threshold value P_{live_threshold}Locked as "suspect of current attack";

if not, adding 'attack suspect' with confidence P_attack0, set to "suspect currently attacked".

6. The live body detection assisting apparatus for use in a face recognition system as set forth in claim 4, wherein: confidence P of' current attack suspect_attackThe update module specifically includes:

if the probability of suspect living body P_liveLess than a threshold value P_{live_threshold}Confidence level P of suspect_attack+ update the step Δ d1 until the confidence level P_attackCut off by 1.0; if the current suspect confidence level P_attackGreater than a threshold value P_{attack_threshold}Returning the result as false;

if the probability of suspect living body P_liveIs greater than or equal to threshold value P_{live_threshold}Confidence level P of suspect_attackUpdate the step size Δ d1 until the confidence P_attackCut off to 0.0; if the current suspect confidence level P_attackLess than a threshold value P_{attack_threshold}The return result is true, otherwise it is false.

7. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the method according to any of claims 1 to 3 when executing the program.

8. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the method according to any one of claims 1 to 3.

Images