CN114758369A - Living body detection model training method, system, equipment and storage medium - Google Patents

Abstract

The present invention provides a training method, system, device and storage medium for a living body detection model based on training data augmentation, comprising the following steps: reading depth data and corresponding RGB images; performing face detection on the RGB images to generate a benchmark A face frame, a random face frame is constructed according to the reference face frame; a negative sample training set is generated by taking a screenshot of the depth region corresponding to the depth data according to the random face frame. The 3D living body detection model is trained according to the negative sample training set to generate the target 3D living body detection model. In the present invention, a plurality of random face frames are generated by performing face detection on RGB images to generate a reference face frame, and then a screenshot of the depth region corresponding to the depth data is performed according to the random face frame to generate a negative sample training set. The sample training set trains the 3D live detection model, which can reduce the instability caused by the false face detection to the 3D live detection model.

Description

Liveness detection model training method, system, device and storage medium

technical field

The present invention relates to face detection, and in particular, to a training method, system, device and storage medium of a living body detection model based on training data augmentation.

Background technique

Face detection methods can be roughly divided into two categories: face detection based on 2D face images and face detection based on 3D face images. Among them, 2D face detection is made by 2D camera plane imaging, which cannot receive the third information in the physical world (geometric data such as size and distance). High, it can be easily cracked through photos, videos, makeup, human skin masks, etc., which cannot meet the requirements of the security level of smartphones.

3D face detection is through 3D camera stereo imaging, which can detect the three-dimensional coordinate information of each point in the field of view, so that the computer can obtain the 3D data of the space and restore the complete three-dimensional world, and realize various intelligent three-dimensional positioning. . To put it simply, the machine obtains more information, and the accuracy of analysis and judgment has been greatly improved. The face detection function can distinguish plane images, videos, makeup, leather masks, and twins. It is suitable for the financial field and smart phones. and other application scenarios with high security level requirements.

As the development of face detection, face alignment and other technologies on RGB images is becoming more and more mature, RGB is usually used for face detection and face alignment in the design process of 3D in vivo algorithms, and then aligned to the depth map to obtain the face. , and then perform 3D face living recognition.

Due to the complexity of the scene where the 3D in vivo algorithm is used, during the use of the 3D in vivo algorithm, there will be a large number of false detections in the face area, as shown in Figure 1. Due to the false detection of the face area, the area entering the 3D living body determination algorithm may not be other areas outside the target area, such as the background, head area, part of the face, etc. As shown in Figure 2, it will give the face The use of liveness determination algorithms has unpredictable consequences.

SUMMARY OF THE INVENTION

In view of the defects in the prior art, the purpose of the present invention is to provide a training method, system, device and storage medium of a living body detection model based on training data augmentation.

The method for training a living body detection model based on training data augmentation provided by the present invention includes the following steps:

Step S1: read the depth data and the corresponding RGB image;

Step S2: performing face detection on the RGB image to generate a reference face frame, and constructing a random face frame according to the reference face frame;

Step S3: taking a screenshot of the depth region corresponding to the depth data according to the random face frame to generate a negative sample training set;

Step S4: Train the 3D living body detection model according to the negative sample training set to generate a target 3D living body detection model.

Preferably, the target 3D living detection model is generated by training a neural network model;

The neural network model includes a feature extraction layer, a full connection layer, a random drop layer, a classification layer and a loss calculation layer that are set in sequence.

Preferably, the step S2 includes the following steps:

Step S201: Perform face detection on the RGB image to generate a reference face frame, where the reference face frame is expressed as g=(x1, y1, W, H), and (x1, y1) is the reference face frame The coordinates of a corner point, W is the width of the reference face frame, and H is the height of the reference face frame;

Step S202: take the smaller value of the width and height of the reference face frame as the maximum size S for generating a random face frame;

Step S203: randomly obtain a value nx in the range of [0, W-S), randomly obtain a value ny in the range of [0, H-S), a group of target face frames generated in the RGB image c=( nx,ny,S,S), wherein, (nx,ny) is the corner coordinate of the target face frame, and S is the width and height of the target face frame;

Step S204: Determine whether the intersection ratio between the target face frame and the reference face frame is less than a preset intersection ratio threshold, and when the intersection ratio between the target face frame and the reference face frame is The target face frame is determined to be a random face frame when the intersection ratio is less than a preset intersection ratio threshold.

Preferably, the step S3 includes the following steps:

Step S301: According to the random face frame, the depth region corresponding to the depth data is intercepted to generate 2D depth data;

Step S302: obtaining preset key point information;

Step S303: Corresponding the key point information to the 2D depth data, and then performing normalization processing to generate the negative sample training set.

Preferably, it also includes the following steps:

- Step S4 : training the 3D living body detection model according to the negative sample training set to generate the target 3D living body detection model.

Preferably, the step S4 includes the following steps:

Step S401: perform key point detection on the RGB image, and determine a plurality of face key points in the reference face frame;

Step S402: Map the face key points in the RGB image to the depth data normalized to a preset size to generate a positive sample training set;

Step S403: Train a 3D living body detection model according to the negative sample training set and the positive sample training set to generate a target 3D living body detection model.

Preferably, the step S204 includes the following steps:

Step S2041: obtaining a target face frame;

Step S2042: Obtain a preset intersection ratio threshold, determine whether the intersection ratio between the target face frame and the reference face frame is less than the preset intersection ratio threshold, when the target face frame and Step S2043 is triggered when the intersection ratio between the reference face frames is less than or equal to the preset intersection ratio threshold, and when the intersection ratio between the target face frame and the reference face frame is greater than the preset intersection ratio Step S2041 is triggered when the cross-union ratio is the threshold;

Step S2043: Determine that the target face frame is a random face frame.

Preferably, the step S301 includes the following steps:

Step S3011: Determine the depth region data corresponding to the depth data according to a random face frame;

Step S3012: judging whether there is zero-valued data in the depth direction in the depth area data, when the zero-valued data exists and the ratio of the number of the zero-valued data to the total data of the depth area data is greater than a preset proportional threshold , then discard the random face frame and return to step S3011, otherwise trigger step S3013;

Step S3013: Intercept the depth region corresponding to the depth data according to the random face frame to generate 2D depth data.

The training system for living detection model based on training data augmentation provided by the present invention includes the following modules:

Data reading module, used to read depth data and corresponding RGB images;

a random face frame generation module, for performing face detection on the RGB image to generate a reference face frame, and constructing a random face frame according to the reference face frame;

A training set generation module is used to generate a negative sample training set by taking a screenshot of the depth region corresponding to the depth data according to the random face frame;

The model training module is used for training the 3D living body detection model according to the negative sample training set to generate the target 3D living body detection model.

The live detection model training device based on training data augmentation provided according to the present invention includes:

processor;

a memory in which executable instructions for the processor are stored;

Wherein, the processor is configured to execute the steps of the method for training a living body detection model based on training data augmentation by executing the executable instructions.

The computer-readable storage medium provided according to the present invention is used for storing a program, and when the program is executed, the steps of the method for training a living body detection model based on training data augmentation are implemented.

Compared with the prior art, the present invention has the following beneficial effects:

In the present invention, a plurality of random face frames are generated by performing face detection on RGB images to generate a reference face frame, and then a negative sample training set is generated by taking a screenshot of the depth region corresponding to the depth data according to the random face frame. The sample training set trains the 3D live detection model, which can reduce the instability caused by the false face detection to the 3D live detection model.

Description of drawings

In order to explain the embodiments of the present invention or the technical solutions in the prior art more clearly, the following briefly introduces the accompanying drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only It is an embodiment of the present invention. For those of ordinary skill in the art, other drawings can also be obtained according to the provided drawings without creative work. Other features, objects and advantages of the present invention will become more apparent by reading the detailed description of non-limiting embodiments with reference to the following drawings:

Fig. 1 is the schematic diagram of the false detection of human face area during human face detection in the prior art;

Fig. 2 is a schematic diagram of the influence of false detection of face region on the training of 3D living body detection model in the prior art;

3 is a flowchart of steps of a training method for a living body detection model based on training data augmentation in an embodiment of the present invention;

4 is a flow chart of steps for constructing a random face frame according to the reference face frame in an embodiment of the present invention

5 is a flowchart of steps for generating a negative sample training set in an embodiment of the present invention;

6(a) is a schematic structural diagram of a training stage of a living body detection model according to an embodiment of the present invention;

FIG. 6(b) is a schematic structural diagram of the inference stage of the in vivo detection model in the embodiment of the present invention;

7 is a flowchart of steps for generating a target 3D living detection model in an embodiment of the present invention;

8 is a flowchart of steps for determining a random face frame in an embodiment of the present invention;

9 is a flowchart of steps for generating 2D depth data in an embodiment of the present invention;

10 is a schematic diagram of a random face frame in a depth image according to an embodiment of the present invention;

11 is a flowchart of steps of an offline augmentation method in an embodiment of the present invention;

12 is a flowchart of steps of an online augmentation method in an embodiment of the present invention;

13 is a schematic diagram of a module of a training system based on training data augmentation for a living body detection model training system according to an embodiment of the present invention;

14 is a schematic structural diagram of an apparatus for training a living body detection model based on training data augmentation according to an embodiment of the present invention; and

FIG. 15 is a schematic structural diagram of a computer-readable storage medium in an embodiment of the present invention.

Detailed ways

The present invention will be described in detail below with reference to specific embodiments. The following examples will help those skilled in the art to further understand the present invention, but do not limit the present invention in any form. It should be noted that, for those skilled in the art, several modifications and improvements can be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

The terms "first", "second", "third", "fourth", etc. (if present) in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It is to be understood that the data so used may be interchanged under appropriate circumstances so that the embodiments of the invention described herein can, for example, be practiced in sequences other than those illustrated or described herein. Furthermore, the terms "comprising" and "having" and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those expressly listed Rather, those steps or units may include other steps or units not expressly listed or inherent to these processes, methods, products or devices.

The technical solutions of the present invention will be described in detail below with specific examples. The following specific embodiments may be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.

The method for training a living body detection model based on training data augmentation provided by the present invention aims to solve the problems existing in the prior art.

The technical solutions of the present invention and how the technical solutions of the present application solve the above-mentioned technical problems will be described in detail below with specific examples. The following specific embodiments may be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. Embodiments of the present invention will be described below with reference to the accompanying drawings.

3 is a flowchart of steps of a training method for a living body detection model based on training data augmentation in an embodiment of the present invention. As shown in FIG. 3 , the training method for a living body detection model based on training data augmentation provided by the present invention includes the following steps:

Step S1: read the depth data and the corresponding RGB image;

In an embodiment of the present invention, the depth data is aligned at the pixel level of the RGB image.

Step S2: performing face detection on the RGB image to generate a reference face frame, and constructing a random face frame according to the reference face frame;

4 is a flowchart of steps for constructing a random face frame according to the reference face frame in an embodiment of the present invention. As shown in FIG. 4 , the step S2 includes the following steps:

Step S201: Perform face detection on the RGB image to generate a reference face frame, where the reference face frame is expressed as g=(x1, y1, W, H), and (x1, y1) is the reference face frame The coordinates of a corner point, W is the width of the reference face frame, and H is the height of the reference face frame;

In the embodiment of the present invention, the (x1, y1) is the coordinates of the upper corner of the reference face frame.

Step S202: take the smaller value of the width and height of the reference face frame as the maximum size S for generating a random face frame;

Step S203: randomly obtain a value nx in the range of [0, W-S), randomly obtain a value ny in the range of [0, H-S), a group of target face frames generated in the RGB image c=( nx,ny,S,S), wherein, (nx,ny) is the corner coordinate of the target face frame, and S is the width and height of the target face frame;

Step S204: Determine whether the intersection ratio between the target face frame and the reference face frame is less than a preset intersection ratio threshold, and when the intersection ratio between the target face frame and the reference face frame is The target face frame is determined to be a random face frame when the intersection ratio is less than a preset intersection ratio threshold.

In the embodiment of the present invention, the Intersection over Union (IoU) is: (W1∩W2)/(W1+W2-W1∩W2), W1 may be the target face frame, and W2 may be the reference person face frame. In this embodiment of the present invention, the preset intersection ratio threshold is any value between 0.2 and 0.4, preferably 0.3.

In the embodiment of the present invention, about 10 random face frames are generated based on the standard face frame.

Step S3: taking a screenshot of the depth region corresponding to the depth data according to the random face frame to generate a negative sample training set.

FIG. 5 is a flowchart of steps for generating a negative sample training set in an embodiment of the present invention. As shown in FIG. 5 , the step S3 includes the following steps:

Step S301: According to the random face frame, the depth region corresponding to the depth data is intercepted to generate 2D depth data;

Step S302: obtaining preset key point information;

Step S303: Corresponding the key point information to the 2D depth data, and then performing normalization processing to generate the negative sample training set.

In the embodiment of the present invention, the normalized uniform size is 180 pixels wide and 220 pixels high. The preset key point information can be set according to the characteristics of the face, such as taking four positions (70.5, 116.5), (109.5, 116.5), (90.5, 137.5), (90, 159) as the left eye, right eye, The position of the nose, mouth, and pre-set key points.

Step S4: Train the 3D living body detection model according to the negative sample training set to generate a target 3D living body detection model.

Fig. 6(a) is a schematic structural diagram of the training stage of the living body detection model in the embodiment of the present invention, and Fig. 6(b) is a schematic structural diagram of the inference stage of the living body detection model according to the embodiment of the present invention, as shown in Figs. 6(a) and 6( b), in the embodiment of the present invention, the target 3D living body detection model is generated by training a neural network model, and the neural network model includes a feature extraction layer, a fully connected layer, a random discarding layer, and a classification layer that are set in sequence. and the loss computation layer.

FIG. 7 is a flowchart of steps for generating a target 3D live detection model in an embodiment of the present invention. As shown in FIG. 7 , the step S4 includes the following steps:

Step S401: perform key point detection on the RGB image, and determine a plurality of face key points in the reference face frame;

Step S402: Map the face key points in the RGB image to the depth data normalized to a preset size to generate a positive sample training set;

Step S403: Train a 3D living body detection model according to the negative sample training set and the positive sample training set to generate a target 3D living body detection model.

FIG. 8 is a flowchart of steps for determining a random face frame in an embodiment of the present invention. As shown in FIG. 8 , the step S204 includes the following steps:

Step S2041: obtaining a target face frame;

Step S2042: Obtain a preset intersection ratio threshold, determine whether the intersection ratio between the target face frame and the reference face frame is less than the preset intersection ratio threshold, when the target face frame and Step S2043 is triggered when the intersection ratio between the reference face frames is less than or equal to the preset intersection ratio threshold, and when the intersection ratio between the target face frame and the reference face frame is greater than the preset intersection ratio Step S2041 is triggered when the cross-union ratio is the threshold;

Step S2043: Determine that the target face frame is a random face frame.

FIG. 10 is a schematic diagram of a random face frame in a depth image according to an embodiment of the present invention. As shown in FIG. 10 , a random face frame generated according to a reference face frame can be seen.

FIG. 9 is a flowchart of steps for generating 2D depth data in an embodiment of the present invention. As shown in FIG. 9 , the step S301 includes the following steps:

Step S3011: Determine the depth region data corresponding to the depth data according to a random face frame;

Step S3012: judging whether there is zero-valued data in the depth direction in the depth area data, when the zero-valued data exists and the ratio of the number of the zero-valued data to the total data of the depth area data is greater than a preset proportional threshold , then discard the random face frame and return to step S3011, otherwise trigger step S3013;

Step S3013: Intercept the depth region corresponding to the depth data according to the random face frame to generate 2D depth data.

In this embodiment of the present invention, the preset ratio threshold is 20% to 40%, preferably 30%.

FIG. 11 is a flow chart of steps of an offline augmentation method in an embodiment of the present invention. As shown in FIG. 11 , as shown in FIG. 11 , in the process of offline augmentation, the process is performed based on the RGB image corresponding to each depth data in the training set. A separate augmentation generates 2D depth data, and the augmented negative samples are added to the training list and the training data set, and a new negative sample training set can be obtained. Based on this new larger training set, training is performed. A more robust algorithm against false detection can be obtained.

FIG. 12 is a flowchart of steps of an online augmentation method in an embodiment of the present invention. As shown in FIG. 12 , as shown in FIG. 12 , the online augmentation mode is to perform data augmentation in real time during the training process, and no additional The memory is spent to store the augmented samples. In the online augmentation mode, a random control switch is required, and the random control switch is turned on and off randomly according to a certain probability. When the random switch is turned on, the first channel and the second channel are turned on at the same time, and based on the pre-stored depth data, the positive samples and augmented prosthesis samples based on the real life can be generated simultaneously; when the random switch is off, only the second channel is Open, data augmentation of negative samples is not performed during the training process, and only the depth data of real people is used to participate in training as positive samples.

13 is a schematic diagram of a module of a training system for a living body detection model based on training data augmentation in an embodiment of the present invention. As shown in FIG. 13 , the living body detection model training system based on training data augmentation provided by the present invention includes the following modules:

Data reading module, used to read depth data and corresponding RGB images;

a random face frame generation module, for performing face detection on the RGB image to generate a reference face frame, and constructing a random face frame according to the reference face frame;

A training set generation module is configured to take a screenshot of the depth region corresponding to the depth data according to the random face frame to generate a negative sample training set.

The model training module is used for training the 3D living body detection model according to the negative sample training set to generate the target 3D living body detection model.

An embodiment of the present invention further provides a training device for training a living body detection model based on training data augmentation, including a processor. A memory in which executable instructions for the processor are stored. Wherein, the processor is configured to execute the steps of the training method for training a living body detection model based on training data augmentation by executing the executable instructions.

As above, in this embodiment, a plurality of random face frames can be generated by performing face detection on an RGB image to generate a reference face frame, and then a negative sample training set can be generated by taking a screenshot of the depth region corresponding to the depth data according to the random face frame, Using the negative sample training set to train the 3D live detection model can reduce the instability caused by the false face detection to the 3D live detection model.

As will be appreciated by one skilled in the art, various aspects of the present invention may be implemented as a system, method or program product. Therefore, various aspects of the present invention can be embodied in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software aspects, which may be collectively referred to herein as implementations "Circuit", "Module" or "Platform".

FIG. 14 is a schematic structural diagram of an apparatus for training a living body detection model based on training data augmentation in an embodiment of the present invention. The electronic device 600 according to this embodiment of the present invention is described below with reference to FIG. 14 . The electronic device 600 shown in FIG. 14 is only an example, and should not impose any limitation on the function and scope of use of the embodiments of the present invention.

As shown in FIG. 14, electronic device 600 takes the form of a general-purpose computing device. Components of the electronic device 600 may include, but are not limited to, at least one processing unit 610, at least one storage unit 620, a bus 630 connecting different platform components (including the storage unit 620 and the processing unit 610), a display unit 640, and the like.

Wherein, the storage unit stores program codes, and the program codes can be executed by the processing unit 610, so that the processing unit 610 executes the various exemplary embodiments of the present invention described in the above-mentioned section of the training method for training a living body detection model based on training data augmentation. A step of. For example, the processing unit 610 may perform the steps shown in FIG. 1 .

The storage unit 620 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 6201 and/or a cache storage unit 6202 , and may further include a read only storage unit (ROM) 6203 .

The storage unit 620 may also include a program/utility 6204 having a set (at least one) of program modules 6205 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, An implementation of a network environment may be included in each or some combination of these examples.

The bus 630 may be representative of one or more of several types of bus structures, including a memory cell bus or memory cell controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local area using any of a variety of bus structures. bus.

The electronic device 600 may also communicate with one or more external devices 700 (eg, keyboards, pointing devices, Bluetooth devices, etc.), with one or more devices that enable a user to interact with the electronic device 600, and/or with Any device (eg, router, modem, etc.) that enables the electronic device 600 to communicate with one or more other computing devices. Such communication may occur through input/output (I/O) interface 650 . Also, the electronic device 600 may communicate with one or more networks (eg, a local area network (LAN), a wide area network (WAN), and/or a public network such as the Internet) through a network adapter 660 . Network adapter 660 may communicate with other modules of electronic device 600 through bus 630 . It should be appreciated that, although not shown in FIG. 14, other hardware and/or software modules may be used in conjunction with electronic device 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tapes drives and data backup storage platforms, etc.

Embodiments of the present invention further provide a computer-readable storage medium for storing a program, and the steps of a training method for training a living body detection model based on training data augmentation are implemented when the program is executed. In some possible implementations, various aspects of the present invention can also be implemented in the form of a program product, which includes program code, when the program product runs on a terminal device, the program code is used to cause the terminal device to execute the above-mentioned description in this specification. The steps according to the various exemplary embodiments of the present invention are described in the section of training method for in vivo detection model augmentation based on training data.

As shown above, when the program of the computer-readable storage medium of this embodiment is executed, in the present invention, a plurality of random face frames are generated according to the face detection on the RGB image and the reference face frame is generated, and then according to the random face frame, a plurality of random face frames are generated. Taking a screenshot of the depth region corresponding to the depth data to generate a negative sample training set, and using the negative sample training set to train a 3D living body detection model can reduce the instability brought by false face detection to the 3D living body detection model.

FIG. 15 is a schematic structural diagram of a computer-readable storage medium in an embodiment of the present invention. Referring to FIG. 15, a program product 800 for implementing the above method according to an embodiment of the present invention is described, which can adopt a portable compact disk read only memory (CD-ROM) and include program codes, and can be stored in a terminal device, For example running on a personal computer. However, the program product of the present invention is not limited thereto, and in this document, a readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

The program product may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or a combination of any of the above. More specific examples (non-exhaustive list) of readable storage media include: electrical connections with one or more wires, portable disks, hard disks, random access memory (RAM), read only memory (ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, portable compact disk read only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

A computer-readable storage medium may include a data signal propagated in baseband or as part of a carrier wave with readable program code embodied thereon. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. A readable storage medium can also be any readable medium other than a readable storage medium that can transmit, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. Program code embodied on a readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

Program code for carrying out operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages—such as Java, C++, etc., as well as conventional procedural programming Language - such as the "C" language or similar programming language. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device and partly on a remote computing device, or entirely on the remote computing device or server execute on. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any kind of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (eg, using an Internet service provider business via an Internet connection).

In the embodiment of the present invention, a plurality of random face frames are generated according to face detection on an RGB image to generate a reference face frame, and then a negative sample training set is generated by taking a screenshot of the depth region corresponding to the depth data according to the random face frame, Using the negative sample training set to train the 3D living body detection model can reduce the instability brought by the false face detection to the 3D living body detection model.

The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same and similar parts between the various embodiments can be referred to each other. The above description of the disclosed embodiments enables any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

The specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the above-mentioned specific embodiments, and those skilled in the art can make various variations or modifications within the scope of the claims, which do not affect the essential content of the present invention.

Claims (10)

1. A living body detection model training method based on training data augmentation is characterized by comprising the following steps:

step S1: reading depth data and a corresponding RGB image;

step S2: performing face detection on the RGB image to generate a reference face frame, and constructing a random face frame according to the reference face frame;

step S3: screenshot is carried out on a depth area corresponding to the depth data according to the random face frame to generate a negative sample training set;

step S4: and training the 3D in-vivo detection model according to the negative sample training set to generate a target 3D in-vivo detection model.

2. The training data augmentation-based in vivo detection model training method according to claim 1, wherein the target 3D in vivo detection model is generated by adopting neural network model training;

the neural network model comprises a feature extraction layer, a full connection layer, a random discarding layer, a classification layer and a loss calculation layer which are sequentially arranged.

3. The training method for the living body test model augmented based on the training data as claimed in claim 1, wherein the step S2 comprises the steps of:

step S201: performing face detection on the RGB image to generate a reference face frame, wherein g is (x1, y1, W, H), (x1, y1) is an angular point coordinate of the reference face frame, W is the width of the reference face frame, and H is the height of the reference face frame;

step S202: taking the smaller value of the width and the height of the reference face frame as the maximum size S of the generated random face frame;

step S203: randomly acquiring a value nx in a range of [0, W-S), randomly acquiring a value ny in a range of [0, H-S), and generating a set of target face frames c ═ n (nx, ny, S) in the RGB image, wherein (nx, ny) is an angular point coordinate of the target face frame, and S is a width and a height of the target face frame;

step S204: and judging whether the intersection ratio between the target face frame and the reference face frame is smaller than a preset intersection ratio threshold value or not, and determining that the target face frame is a random face frame when the intersection ratio between the target face frame and the reference face frame is smaller than the preset intersection ratio threshold value.

4. The training method for the living body test model augmented based on the training data as claimed in claim 1, wherein the step S3 comprises the steps of:

step S301: intercepting a depth area corresponding to the depth data according to the random face frame to generate 2D depth data;

step S302: acquiring preset key point information;

step S303: and corresponding the key point information to the 2D depth data, and further performing normalization processing to generate the negative sample training set.

5. The training method for living body detection model based on training data augmentation of claim 1, wherein the step S4 comprises the steps of:

step S401: performing key point detection on the RGB image, and determining a plurality of face key points in the reference face frame;

step S402: mapping the key points of the human face in the RGB image to depth data normalized to a preset size to generate a positive sample training set;

step S403: and training the 3D in-vivo detection model according to the negative sample training set and the positive sample training set to generate a target 3D in-vivo detection model.

6. The method for training the living body detection model based on the training data augmentation of claim 2, wherein the step S204 comprises the steps of:

step S2041: acquiring a target face frame;

step S2042: acquiring a preset intersection ratio threshold, judging whether the intersection ratio between the target face frame and the reference face frame is smaller than the preset intersection ratio threshold, triggering a step S2043 when the intersection ratio between the target face frame and the reference face frame is smaller than or equal to the preset intersection ratio threshold, and triggering a step S2041 when the intersection ratio between the target face frame and the reference face frame is larger than the preset intersection ratio threshold;

step S2043: and determining the target face frame as a random face frame.

7. The method for training a living body test model based on training data augmentation of claim 3, wherein the step S301 comprises the steps of:

step S3011: determining depth area data corresponding to the depth data according to the random face frame;

step S3012: judging whether zero-value data exist in the depth region data in the depth direction, when the zero-value data exist and the ratio of the number of the zero-value data to the total data of the depth region data is larger than a preset proportional threshold, discarding the random face frame and returning to the step S3011, otherwise, triggering the step S3013;

step S3013: and intercepting a depth area corresponding to the depth data according to the random face frame to generate 2D depth data.

8. A living body detection model training system based on training data augmentation is characterized by comprising the following modules:

the data reading module is used for reading the depth data and the corresponding RGB image;

the random face frame generating module is used for carrying out face detection on the RGB image to generate a reference face frame and constructing a random face frame according to the reference face frame;

the training set generation module is used for carrying out screenshot on a depth area corresponding to the depth data according to the random face box to generate a negative sample training set;

and the model training module is used for training the 3D in-vivo detection model according to the negative sample training set to generate a target 3D in-vivo detection model.

9. A live body test model training apparatus based on training data augmentation, comprising:

a processor;

a memory having stored therein executable instructions of the processor;

wherein the processor is configured to perform the steps of the training method for a living body detection model augmented based on training data of any one of claims 1 to 7 via execution of the executable instructions.

10. A computer-readable storage medium storing a program which, when executed, implements the steps of the training method for a living body test model augmented based on training data of any one of claims 1 to 7.

Images