Face shielding recognition method and system, face recognition mathematical model construction method and system, electronic equipment and storage medium

Abstract

The invention discloses a human face shielding recognition method and system, a human face recognition mathematical model construction method and system, electronic equipment and a storage medium. The human face shielding recognition method comprises the steps that an artificial intelligence AI model and human face key point detection are used for manufacturing a simulated mouth and nose shielding data set; inputting the shielded data set and the original data set which is not shielded by the shielding object into a classification network for training, and learning the characteristics of the face when the face is shielded and not shielded, so as to construct a face recognition mathematical model; and acquiring a face image, inputting the acquired face image into the face recognition mathematical model, and judging whether the face is shielded or not. According to the face shielding recognition method and system, the face recognition mathematical model construction method and system, the electronic equipment and the storage medium, whether the face is shielded or not can be recognized.

Description

Technical Field

[0001] The present invention belongs to the technical field of face recognition, and relates to a face occlusion recognition method, in particular to a face occlusion recognition method and system, a face recognition mathematical model construction method and system, an electronic device and a storage medium. Background Art

[0002] Face recognition technology is becoming increasingly mature. This technology is being used to unlock mobile phones, detect crimes, identify specific groups of people, and can also be used to monitor our emotional states, etc. And this technology is widely used in the field of smart cities and has more application space in home care, such as the care of infants and young children. The baby monitor product can help parents see the baby even when they are not at home. If the baby is sleeping in another room, the baby monitor can detect whether the baby is crying and remind the parents through messages.

[0003] However, when a baby sleeps alone, there may still be a risk of mouth and nose occlusion. Therefore, a lot of mouth and nose occlusion data is needed for the training of the AI model. However, baby face data is very private and difficult to obtain, and there are many possibilities for occluders, including various clothes, dolls, pillows, etc. It is very difficult to obtain sufficient and diverse mouth and nose occlusion data.

[0004] In view of this, there is an urgent need to design a new face recognition method to overcome at least some of the above defects existing in the existing face recognition methods. Summary of the Invention

[0005] The present invention provides a face occlusion recognition method and system, a face recognition mathematical model construction method and system, an electronic device and a storage medium, which can identify whether a face is occluded.

[0006] To solve the above technical problems, according to one aspect of the present invention, the following technical solution is adopted:

[0007] A face occlusion recognition method, the face occlusion recognition method comprising:

[0008] Steps for making a mouth and nose occlusion data set; using an artificial intelligence (AI) model and face key point detection to make a simulated mouth and nose occlusion data set; synthesizing a face image through the AI model, and using the face key point detection to find the mouth and nose positions of the face image; placing the image of the target occluder at the corresponding face key point coordinates to generate a valid mouth and nose occlusion data set;

[0009] Steps for constructing a mathematical model; inputting the occlusion data set and the original data set not occluded by an occluder into a classification network for training, learning the features of the face when occluded and not occluded, so as to construct a face recognition mathematical model;

[0010] Steps for face occlusion recognition: Obtain a face image, input the obtained face image into a face recognition mathematical model, and determine whether the face is occluded.

[0011] As an implementation manner of the present invention, in the step of making the mouth and nose occlusion dataset, the face key point detection model outputs a face detection frame and the coordinates of five-point face key points; the size and position of the occluder are adjusted using the face detection frame and the coordinates of the five key points;

[0012] The occluder is scaled in an appropriate proportion; the range for occluding the mouth and nose is the area from the nose key point to the key points of both corners of the mouth, and the length of the upper half of the occluder should just cover the mouth and nose;

[0013] First, scale the width of the occluder to the same width as the face detection frame, then take the midpoint position of the upper 1 / 4 height from top to bottom as the reference point, and use the length of this 1 / 4 height (referred to as the reference length) as the basis to calculate the distance from the nose key point to the midpoint of the key points of both corners of the mouth (referred to as the mouth and nose distance). The reference length should be twice the mouth and nose distance. If it is less than twice or greater than twice, calculate the ratio of the reference length to the mouth and nose distance, and scale the occluder so that 1 / 4 of the occluder covers the upper half of the nose to the mouth, and the remaining 3 / 4 of the occluder covers the lower half of the face, thus generating an effective mouth and nose occlusion dataset.

[0014] As an implementation manner of the present invention, in the step of making the mouth and nose occlusion dataset, input the text description of the target into the AI model, the network generates the corresponding face image, and input the generated face image into the face key point detection network; obtain the nose coordinates and the coordinates of the two corner points of the mouth; scale the target occluder to an appropriate size, align the coordinates of the three-point face key points, and overlap it with the original image to form an occluded face dataset.

[0015] According to another aspect of the present invention, the following technical solution is adopted: A method for constructing a face recognition mathematical model, the method for constructing the face recognition mathematical model includes:

[0016] Steps for making the mouth and nose occlusion dataset; Use the artificial intelligence AI model and face key point detection to make a simulated mouth and nose occlusion dataset; Synthesize a face image through the AI model, and use the face key point detection on the face image to find the positions of the mouth and nose; Place the image of the target occluder on the corresponding face key point coordinates to generate an effective mouth and nose occlusion dataset;

[0017] Steps for constructing the mathematical model; Input the occlusion dataset and the original dataset without being occluded by the occluder into the classification network for training, learn the features of the face when occluded and not occluded, so as to construct a face recognition mathematical model.

[0018] According to another aspect of the present invention, the following technical solution is adopted: A face occlusion recognition system, the face occlusion recognition system includes:

[0019] A mouth and nose occlusion dataset production module, used to produce a simulated mouth and nose occlusion dataset using an artificial intelligence AI model and face key point detection; synthesize a face image through the AI model, and use the face image to make the face key point detection find the positions of the mouth and nose; place the image of the target occluder on the corresponding face key point coordinates to generate an effective mouth and nose occlusion dataset;

[0020] A mathematical model construction module, used to input the occlusion dataset and the original dataset without being occluded by the occluder into a classification network for training, learn the characteristics of the face when occluded and not occluded, so as to construct a face recognition mathematical model;

[0021] A face occlusion recognition module, used to obtain a face image, input the obtained face image into the face recognition mathematical model, and judge whether the face is occluded.

[0022] As an implementation manner of the present invention, the mouth and nose occlusion dataset production module is used to input the text description of the target into the AI model, and the network will generate the corresponding face image, and input the generated face image into the face key point detection network; obtain the nose coordinates and the coordinates of the two corner points of the mouth, and then scale the target occluder to an appropriate size, and overlap it with the original image after aligning the three-point face key point coordinates to form an occluded face dataset.

[0023] As an implementation manner of the present invention, the mouth and nose occlusion dataset production module will output a face detection frame and five-point face key point coordinates through the face key point detection model; use the face detection frame and the five-point key point coordinates to adjust the size and position of the occluder;

[0024] Perform appropriate scaling on the occluder; the occluded area of the mouth and nose is the area from the nose key point to the two corner key points of the mouth, and the length of the upper half of the occluder needs to just cover the mouth and nose;

[0025] First, scale the width of the occluder to the same width as the face detection frame, and then take the midpoint position of the height from top to bottom as the reference point, and use the length of this 1 / 4 height as the reference length (referred to as the reference length here), and calculate the distance from the nose key point to the midpoint of the two corner key points of the mouth (referred to as the mouth and nose distance here). The reference length needs to be twice the mouth and nose distance. If it is less than twice or greater than twice, calculate the ratio of the reference length to the mouth and nose distance, and scale the occluder so that 1 / 4 of the occluder covers the upper half of the nose to the mouth, and the remaining 3 / 4 of the occluder covers the lower half of the face, so as to generate an effective mouth and nose occlusion dataset.

[0026] According to another aspect of the present invention, the following technical solution is adopted: A face recognition mathematical model construction system, the face recognition mathematical model construction system includes:

[0027] A mouth and nose occlusion dataset production module, used to produce a simulated mouth and nose occlusion dataset using an artificial intelligence AI model and face key point detection; synthesize a face image through the AI model, and use the face key point detection on the face image to find the positions of the mouth and nose; place the image of the target occluder on the corresponding face key point coordinates to generate a valid mouth and nose occlusion dataset;

[0028] A mathematical model construction module, used to input the occlusion dataset and the original dataset without being occluded by an occluder into a classification network for training, learn the characteristics of the face when occluded and not occluded, so as to construct a face recognition mathematical model.

[0029] According to another aspect of the present invention, the following technical solution is adopted: An electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the computer program, the steps of the above method are implemented.

[0030] According to another aspect of the present invention, the following technical solution is adopted: A storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the steps of the above method are implemented.

[0031] The beneficial effect of the present invention is that: The face occlusion recognition method and system, the face recognition mathematical model construction method and system, the electronic device and the storage medium proposed by the present invention can identify whether the face is occluded.

[0032] In a usage scenario of the present invention, the present invention uses a synthetic image to simulate the situation where a baby's mouth and nose are occluded and only the upper half or the left and right half of the face is exposed. First, use the AI model to generate a synthetic baby face, and through face key point detection, locate the positions of the two eyes, nose, and mouth of the face, and synthesize various possible occluders onto the mouth and nose positions of the face key points or any desired positions. The present invention can obtain a large amount of valid mouth and nose occlusion data, and finally through AI model training, we can know whether the current face state of the baby is occluded. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a flowchart of the face occlusion recognition method in an embodiment of the present invention.

[0034] Figure 2 It is a schematic composition diagram of the face occlusion recognition system in an embodiment of the present invention.

[0035] Figure 3Flowchart for creating a simulated infant mouth and nose occlusion dataset in an embodiment of the present invention.

[0036] Figure 4 Flowchart for generating an infant face using an AI model in an embodiment of the present invention.

[0037] Figure 5 Schematic diagram for training a mouth and nose occlusion model using a simulated infant mouth and nose occlusion dataset in an embodiment of the present invention.

[0038] Figure 6 Schematic diagram of the composition of an electronic device in an embodiment of the present invention. Figure 7 Flowchart of the steps for creating a nose occlusion dataset in an embodiment of the present invention. Detailed implementation manners

[0039] The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0040] To further understand the present invention, the preferred implementation manners of the present invention will be described below in conjunction with embodiments. However, it should be understood that these descriptions are only for further explaining the features and advantages of the present invention, rather than limiting the claims of the present invention.

[0041] The description of this part only focuses on several typical embodiments, and the present invention is not limited to the scope described in the embodiments. The mutual replacement of the same or similar prior art means and some technical features in the embodiments is also within the scope of the description and protection of the present invention.

[0042] The expression of the steps in each embodiment in the specification is only for convenience of description, and the implementation manner of the present application is not limited by the order of step implementation.

[0043] "Coupled" or "connected" in the specification includes both direct connection and indirect connection.

[0044] The present invention discloses a face occlusion recognition method. Figure 1 Flowchart of the face occlusion recognition method in an embodiment of the present invention; please refer to Figure 1 , the face occlusion recognition method includes:

[0045] [Step S1] Steps for creating a mouth and nose occlusion dataset; using an artificial intelligence (AI) model and face key point detection to create a simulated mouth and nose occlusion dataset; synthesizing face images through the AI model (such as non-repeating face images), and using face key point detection on the face images to find the positions of the mouth and nose; placing the images of the target occluder at the corresponding face key point coordinates to generate a valid mouth and nose occlusion dataset.

[0046] In an embodiment of the present invention, in the steps for creating the mouth and nose occlusion dataset, the followingFigure 7 , the face key point detection model outputs a face detection frame and the coordinates of five-point face key points; the size and position of the occluder are adjusted using the face detection frame and the coordinates of the five-point key points;

[0047] Since the occluder may be of various sizes, in order to place it at the mouth and nose position in a suitable size, the occluder needs to be scaled by an appropriate ratio first; the range of covering the mouth and nose is the area from the nose key point to the key points of both mouth corners, and the length of the upper half of the occluder needs to just cover the mouth and nose;

[0048] First, scale the width of the occluder to the same width as the face detection frame, and then take the midpoint position of the height from top to bottom as the reference point. Based on the length of this 1 / 4 height (referred to as the reference length here), calculate the distance from the nose key point to the midpoint of the key points of both mouth corners (referred to as the mouth and nose distance here). The reference length needs to be twice the mouth and nose distance. If it is less than twice or greater than twice, calculate the ratio of the reference length to the mouth and nose distance, and scale the occluder so that 1 / 4 of the occluder covers the part from the upper half of the nose to the mouth, and the remaining 3 / 4 of the occluder covers the lower half of the face, thus generating an effective mouth and nose occlusion dataset.

[0049] In an embodiment of the present invention, the text description of the target can be input into the AI model, and the network generates a corresponding face image. The generated face image is input into the face key point detection network; the nose coordinates and the coordinates of the two mouth corner points on both sides are obtained; the target occluder is scaled to an appropriate size, and after aligning the coordinates of the three-point face key points, it is overlapped with the original image to form an occluded face dataset.

[0050] In an embodiment, an AI generation network with a GAN architecture is used to generate an image, and the input of this network is a text description. For example, "a baby girl with blond hair and blue eyes" is the target text description, and an image of a baby girl with blond hair and blue eyes is needed. Inputting this text description into the generation network will generate a target face that conforms to the description. Since the generated face is randomly generated, it will not be repeated.

[0051]

Step S2

[0052]

Step S3

[0053] The present invention also discloses a face occlusion recognition system, Figure 2 is a schematic diagram of the composition of the face occlusion recognition system in an embodiment of the present invention; please refer to Figure 2, the face occlusion recognition system includes: a mouth and nose occlusion dataset production module 1, a mathematical model construction module 2, and a face occlusion recognition module 3.

[0054] The mouth and nose occlusion dataset production module 1 is used to produce a simulated mouth and nose occlusion dataset using an artificial intelligence (AI) model and face key point detection; synthesize non-repeating face images through the AI model, and use the face key point detection to find the mouth and nose positions on the face images; place the image of the target occlusion object at the corresponding face key point coordinates to generate a valid mouth and nose occlusion dataset.

[0055] In an embodiment of the present invention, the mouth and nose occlusion dataset production module outputs a face detection frame and five-point face key point coordinates through a face key point detection model; adjusts the size and position of the occlusion object using the face detection frame and five-point key point coordinates;

[0056] Since the occlusion object may be of various sizes, in order to place it at the mouth and nose position in a suitable size, it is necessary to first scale the occlusion object by an appropriate ratio; the area for occluding the mouth and nose is the area from the nose key point to the key points of both corners of the mouth, and the length of the upper half of the occlusion object needs to just cover the mouth and nose;

[0057] First, scale the width of the occlusion object to the same width as the face detection frame, and then take the midpoint position of the height from top to bottom as the reference point. Using the length of this 1 / 4 height (referred to as the reference length here), calculate the distance from the nose key point to the midpoint of the key points of both corners of the mouth (referred to as the mouth and nose distance here). The reference length needs to be twice the mouth and nose distance. If it is less than twice or greater than twice, calculate the ratio of the reference length to the mouth and nose distance, and scale the occlusion object to use 1 / 4 of the occlusion object to cover the upper half of the nose to the mouth part, and the remaining 3 / 4 of the occlusion object to cover the lower half of the face, so as to generate a valid mouth and nose occlusion dataset.

[0058] In an embodiment of the present invention, the mouth and nose occlusion dataset production module 1 is used to input the target text description into the AI model, and the network will generate the corresponding face image. Input the generated face image into the face key point detection network; obtain the nose coordinates and the coordinates of the two corner points of the mouth. Next, scale the target occlusion object to an appropriate size, align the three-point face key point coordinates, and overlap it with the original image to form an occluded face dataset.

[0059] In an embodiment, an AI generation network with a GAN architecture is used to generate images, and the input of this network is a text description. For example, "a baby girl with blond hair and blue eyes" is the target text description, and an image of a baby girl with blond hair and blue eyes is needed. Input this text description into the generation network, and it will generate a target face that meets the description. Since the generated face is randomly generated, it will not repeat.

[0060] The mathematical model construction module 2 is used to input the occluded data set and the original data set without occlusion by an occluder into a classification network for training, learn the features of a human face when occluded and not occluded, and thus construct a mathematical model for human face recognition;

[0061] The human face occlusion recognition module 3 is used to obtain a human face image, input the obtained human face image into the mathematical model for human face recognition, and determine whether the human face is occluded.

[0062] The present invention discloses a method for constructing a mathematical model for human face recognition. The method for constructing a mathematical model for human face recognition includes:

[0063]

Step S1

[0064]

Step S2

[0065] The present invention also discloses a system for constructing a mathematical model for human face recognition. The system for constructing a mathematical model for human face recognition includes: a mouth and nose occlusion data set making module and a mathematical model construction module.

[0066] The mouth and nose occlusion data set making module is used to make a simulated mouth and nose occlusion data set by using an artificial intelligence AI model and human face key point detection; Synthesize non-repeating human face images through the AI model, and use the human face key point detection to find the mouth and nose positions of the human face images; Place the image of the target occluder at the corresponding human face key point coordinates to generate a valid mouth and nose occlusion data set.

[0067] The mathematical model construction module is used to input the occluded data set and the original data set without occlusion by an occluder into a classification network for training, learn the features of a human face when occluded and not occluded, and thus construct a mathematical model for human face recognition.

[0068] Figure 3 It is a process for making a simulated baby mouth and nose occlusion data set. Synthesize non-repeating baby human face images through the AI model, obtain the nose coordinates and the coordinates of the two corners of the mouth on both sides through the human face key point detection network. Next, scale the target occluder to an appropriate size, align the three-point human face key point coordinates and overlap them with the original image to form an occluded human face data set. The target occluders include pillows, eye masks, scarves, clothes, blankets, hands, etc.

[0069] Figure 4 The process of generating baby faces using an AI model. First, input a string to describe the features of the desired face. Next, extract the features of the string through natural language processing and input them into a face generation network to generate a face that meets the requirements. Using this flexible method, baby faces with various angles, expressions, and even different skin colors can be generated, making the dataset rich and diverse.

[0070] Figure 5 Describe the training of a nose and mouth occlusion model using a simulated baby nose and mouth occlusion dataset. After creating the simulated baby nose and mouth occlusion dataset, input both the occlusion dataset and the synthetic baby face dataset without occlusion into a simple classification network for training. This can effectively learn the features of a baby's face when occluded and when not occluded. In actual application, capture a face image and input it into the classification network to determine whether there is occlusion.

[0071] The present invention also discloses an electronic device. Figure 6 It is a schematic diagram of the composition of the electronic device in an embodiment of the present invention; please refer to Figure 6 , at the hardware level, the electronic device includes a memory, a processor, and at least one network interface; the processor can be a microprocessor, and the memory can include internal memory, such as random access memory (RAM), and can also include non-volatile memory, etc. Of course, the electronic device can also be provided with other hardware as needed.

[0072] The processor, network interface, and memory can be interconnected through an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect Standard) bus, or an EISA (Extended Industry Standard Architecture) bus, etc.; the bus can include an address bus, a data bus, a control bus, etc. The memory is used to store programs (which can include an operating system program and application programs); the programs can include program code, and the program code can include computer operation instructions. The memory can include internal memory and non-volatile memory and provide instructions and data to the processor.

[0073] In one embodiment, the processor can read the corresponding program from the non-volatile memory into the internal memory and then run it; the processor can execute the programs stored in the memory and is specifically used to perform the following operations (as shown in Figure 1 ):

[0074]

Step S1

[0075]

Step S2

[0076]

Step S3

[0077] The present invention further discloses a storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the following steps of the method of the present invention are implemented (as Figure 1 shown):

[0078]

Step S1

[0079]

Step S2

[0080]

Step S3

[0081] In summary, the face occlusion recognition method and system, the face recognition mathematical model construction method and system, the electronic device and the storage medium proposed by the present invention can identify whether a face is occluded.

[0082] In a usage scenario of the present invention, the present invention uses a synthetic image to simulate the situation where the baby's mouth and nose are blocked, showing only the upper half of the face or the left or right half of the face. First, an AI model is used to generate a synthetic baby face, and the positions of the two eyes, nose, and mouth of the face are located through facial key point detection. Then, various possible occluders are synthesized at the mouth and nose positions of the facial key points or any desired positions. The present invention can obtain a large amount of effective mouth and nose occlusion data. Finally, through the training of the AI model, we can know whether the current state of the baby's face is occluded.

[0083] The technical features of the above-described embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as within the scope described in this specification.

[0084] The description and application of the present invention here are illustrative and are not intended to limit the scope of the present invention to the above embodiments. The effects or advantages involved in the embodiments may not be reflected in the embodiments due to various factors. The description of the effects or advantages is not used to limit the embodiments. The deformations and changes of the embodiments disclosed here are possible, and the substitutions and equivalent components of the embodiments are well known to those of ordinary skill in the art. Those skilled in the art should clearly understand that the present invention can be implemented in other forms, structures, arrangements, proportions, and with other components, materials, and parts without departing from the spirit or essential characteristics of the present invention. Other deformations and changes can be made to the embodiments disclosed here without departing from the scope and spirit of the present invention.

Claims

1. A method for face occlusion recognition, characterized in that, The described face occlusion recognition method includes: Steps for making a nose and mouth occlusion dataset: Use an artificial intelligence (AI) model and face key point detection to make a simulated nose and mouth occlusion dataset. Synthesize a face image through the AI model, and use the face key point detection on the face image to find the positions of the nose and mouth. Place the image of the target occlusion object at the corresponding face key point coordinates to generate a valid nose and mouth occlusion dataset. Steps for constructing a mathematical model: Input the occlusion dataset and the original dataset without being occluded by the occlusion object into a classification network for training, learn the features of the face when occluded and not occluded, and thus construct a face recognition mathematical model. Steps for face occlusion recognition: Obtain a face image, input the obtained face image into the face recognition mathematical model, and determine whether the face is occluded.

2. The face occlusion recognition method according to claim 1, wherein: In the steps for making the nose and mouth occlusion dataset, the face key point detection model outputs a face detection frame and the coordinates of five-point face key points. Use the face detection frame and the coordinates of the five key points to adjust the size and position of the occlusion object. Scale the occlusion object by an appropriate ratio. The area to occlude the nose and mouth is the area from the nose key point to the key points of both corners of the mouth. The length of the upper half of the occlusion object should just cover the nose and mouth. Scale the width of the occlusion object to be equal to the width of the face detection frame, and then take the midpoint position of the height from top to bottom as the reference point, and use the length of this 1 / 4 height as the reference length. Calculate the distance from the nose key point to the midpoint of the key points of both corners of the mouth as the nose and mouth distance. The reference length should be twice the nose and mouth distance. If it is less than twice or greater than twice, calculate the ratio of the reference length to the nose and mouth distance, and scale the occlusion object so that 1 / 4 of the occlusion object covers the upper half of the nose to the mouth, and the remaining 3 / 4 of the occlusion object covers the lower half of the face, thus generating a valid nose and mouth occlusion dataset.

3. The face occlusion recognition method according to claim 1, wherein: In the steps for making the nose and mouth occlusion dataset, input the text description of the target into the AI model, the network generates the corresponding face image, and input the generated face image into the face key point detection network. Obtain the nose coordinates and the coordinates of the two corner points of the mouth. Scale the target occlusion object to an appropriate size, align the three-point face key point coordinates, and overlap it with the original image to form an occluded face dataset.

4. A method for constructing a face recognition mathematical model, characterized in that, The method for constructing the face recognition mathematical model includes: Steps for making a nose and mouth occlusion dataset: Use an artificial intelligence (AI) model and face key point detection to make a simulated nose and mouth occlusion dataset. Synthesize a face image through the AI model, and use the face key point detection on the face image to find the positions of the nose and mouth. Place the image of the target occlusion object at the corresponding face key point coordinates to generate a valid nose and mouth occlusion dataset. Steps for constructing a mathematical model: Input the occlusion dataset and the original dataset without being occluded by the occlusion object into a classification network for training, learn the features of the face when occluded and not occluded, and thus construct a face recognition mathematical model.

5. A face occlusion recognition system, characterized in that, The described face occlusion recognition system includes: The mouth and nose occlusion dataset production module is used to produce a simulated mouth and nose occlusion dataset by using an artificial intelligence (AI) model and face key point detection. The face images are synthesized through the AI model, and the mouth and nose positions are found by using the face key point detection on the face images. The images of the target occluder are placed at the corresponding face key point coordinates to generate a valid mouth and nose occlusion dataset. The mathematical model construction module is used to input the occlusion dataset and the original dataset without being occluded by the occluder into a classification network for training, learn the features of the face when occluded and not occluded, so as to construct a face recognition mathematical model. The face occlusion recognition module is used to obtain a face image, input the obtained face image into the face recognition mathematical model, and determine whether the face is occluded.

6. The face occlusion recognition system according to claim 5, characterized in that: The mouth and nose occlusion dataset production module is used to input the text description of the target into the AI model, and the network will generate the corresponding face image, and input the generated face image into the face key point detection network. Obtain the nose coordinates and the coordinates of the two corners of the mouth, and then scale the target occluder to an appropriate size, and overlap it with the original image after aligning the three-point face key point coordinates to form an occluded face dataset.

7. The face occlusion recognition system according to claim 5, characterized in that: The mouth and nose occlusion dataset production module will output a face detection frame and five-point face key point coordinates through the face key point detection model; use the face detection frame and the five-point key point coordinates to adjust the size and position of the occluder; Scale the occluder by an appropriate ratio; the occluded area of the mouth and nose is the area from the nose key point to the two corners of the mouth key point, and the length of the upper part of the occluder should just cover the mouth and nose; Scale the width of the occluder to the same width as the face detection frame, and then take the midpoint position of the upper 1 / 4 height from top to bottom as the reference point, and use the length of this 1 / 4 height as the reference length; Calculate the distance from the nose key point to the midpoint of the two corners of the mouth key point as the mouth and nose distance; the reference length should be twice the mouth and nose distance. If it is less than twice or greater than twice, calculate the ratio of the reference length to the mouth and nose distance, and scale the occluder so that 1 / 4 of the occluder covers the upper half of the nose to the mouth, and the remaining 3 / 4 of the occluder covers the lower half of the face, so as to generate a valid mouth and nose occlusion dataset.

8. A face recognition mathematical model construction system, characterized in that, The face recognition mathematical model construction system includes: The mouth and nose occlusion dataset production module is used to produce a simulated mouth and nose occlusion dataset by using an artificial intelligence (AI) model and face key point detection. The face images are synthesized through the AI model, and the mouth and nose positions are found by using the face key point detection on the face images. The images of the target occluder are placed at the corresponding face key point coordinates to generate a valid mouth and nose occlusion dataset. The mathematical model construction module is used to input the occlusion dataset and the original dataset without being occluded by the occluder into a classification network for training, learn the features of the face when occluded and not occluded, so as to construct a face recognition mathematical model.

9. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the computer program, the steps of the method according to any one of claims 1 to 4 are implemented.

10. A storage medium having computer program instructions stored thereon, characterized in that, When executed by a processor, the computer program instructions implement the steps of the method according to any one of claims 1 to 4.

Images