A face recognition method, device, equipment and storage medium for RGB-NIR different imaging for intelligent networked face lock

Abstract

The present invention relates to the field of image recognition technology, and in particular to a face recognition method, device, equipment and storage medium for RGB-NIR different imaging for intelligent networked face lock. The method comprises: obtaining RGB image data, performing face detection and face feature extraction in the cloud, obtaining high-dimensional features that can represent the face, and binding the high-dimensional features with flags according to binding information; unbinding the high-dimensional features according to the binding information, and storing them in the feature access stack of the face lock end; activating the face lock camera and obtaining camera data, and performing face detection and feature extraction on the NIR image data, obtaining high-dimensional features to be identified that represent the face, and calculating the similarity between the high-dimensional features to be identified and the high-dimensional features stored in the feature access stack of the face lock end to obtain the similarity. Finally, an update mechanism for the feature access stack of the face lock end is added. The present invention can improve the recognition accuracy of the face lock.

Description

Technical Field

[0001] The present invention relates to the field of image recognition technology, and in particular to a face recognition method, device, equipment and storage medium for RGB-NIR different imaging for intelligent networked face lock. Background Art

[0002] With the popularization of smart face locks and the widespread application of Internet and cloud computing technologies, smart face locks have developed to realize functions such as remote control, data interaction and analysis, further improving user experience and the intelligence level of products.

[0003] Traditional face locks only provide an application method for adding, deleting, modifying and checking faces before the face is locked. At the same time, traditional face locks only support traditional near-infrared imaging face recognition, which greatly reduces the development and convenience of face locks in intelligent applications.

[0004] Therefore, there is an urgent need for a face recognition method, device, equipment and storage medium for RGB-NIR different imaging for smart networked face lock to solve the above problems. Summary of the invention

[0005] The purpose of the present invention is to provide a face recognition method, device, equipment and storage medium for RGB-NIR different imaging for intelligent networked face lock: to solve the technical problems of inaccurate recognition, low intelligence and low convenience in the face lock recognition process of the existing solutions.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] On the one hand, a face recognition method using RGB-NIR different imaging for a smart networked face lock, the method comprising:

[0008] S1: Obtain image data, where the image data is RGB image data uploaded by different channels, and perform RGB-NIR conversion processing on the RGB image data to obtain a NIR image that is homologous to the face lock;

[0009] S2: The cloud front-end sending module performs face detection and facial feature extraction on the NIR image to obtain high-dimensional features that can represent the face, and binds the high-dimensional features to the flag bits according to the binding information;

[0010] S3: The cloud front-end sending module sends the high-dimensional features after the binding flag to the cloud back-end;

[0011] S4: The cloud backend processing module receives the high-dimensional features, unbinds the high-dimensional features according to the binding information, and saves them in the feature access stack of the face lock end;

[0012] S5: Activate the face lock camera and obtain camera data, wherein the camera data is NIR image data, and perform face detection and feature extraction on the NIR image data to obtain high-dimensional features to be identified that represent the face, perform similarity calculation on the high-dimensional features to be identified and the high-dimensional features stored in the feature access stack of the face lock end to obtain the similarity, and determine whether the requirements for generating an unlocking instruction are met based on the similarity. If so, generate an unlocking instruction.

[0013] Furthermore, the RGB image data uploaded by different channels include visible light facial images collected from mobile phone cameras or multiple facial images uploaded in batches in specific businesses.

[0014] Furthermore, the RGB image data is converted from RGB to NIR to obtain a NIR image that is homologous to the face lock, including the following process:

[0015] Collect multiple sets of paired RGB and NIR face data as data sets for training the generative adversarial network CycleGan, where the generative adversarial network CycleGan is an adversarial network included in the image generator;

[0016] Generate MBF-CycleGan adversarial network structure model based on adversarial network CycleGan: change Generator and change The original encoding model in the generator is changed to the MBF model structure for face feature extraction;

[0017] The collected multiple sets of paired RGB and NIR face data are input into the MBF-CycleGan adversarial network structure model for training to obtain a visible light RGB to NIR image converter model.

[0018] Furthermore, binding the high-dimensional features to the flag bits according to the binding information includes the following process:

[0019] In extracting high-dimensional features After that, the binding information includes an identity ID and a representative flag, based on high-dimensional features. and binding information to form a new vector .

[0020] Furthermore, the high-dimensional feature vector to be identified is stored in the feature access stack of the face lock end and sent from the cloud. The high-dimensional feature vector of the face extracted by the face lock end through the camera NIR data The similarity calculation includes the following process:

[0021] ;

[0022] in, is the similarity, Represents the i-th eigenvector value of the high-dimensional feature vector extracted from the camera data, It represents the i-th feature vector value of the high-dimensional feature vector stored in the feature access stack of the face lock end, and n is the dimension size of the feature vector.

[0023] Furthermore, judging whether the requirements for generating unlocking instructions are met based on the similarity includes the following process:

[0024] It is determined whether the similarity exceeds a preset threshold. If so, the requirement for generating an unlock instruction is met; if not, the requirement for generating an unlock instruction is not met.

[0025] Furthermore, the method further comprises:

[0026] Save and update the high-dimensional features extracted by the face camera and the high-dimensional features sent by the cloud backend. The execution steps are as follows:

[0027] After the requirements for generating unlocking instructions are met, the ID of the current inspector is obtained;

[0028] Find the corresponding face feature access stack S according to the ID of the current detected person, and check whether the size of the stack meets the set size m. If it does, the high-dimensional features extracted by the face camera are stored in the stack. Record as temporary storage, and the m facial features previously saved in stack S Calculate and compare the similarities one by one, and find the one corresponding to the minimum similarity value in stack S , is the face feature with the smallest similarity value in the corresponding stack, and uses the temporarily saved Replace it and save it to the face feature access stack.

[0029] On the other hand, a face recognition device with different RGB-NIR imaging for smart networked face lock, the device comprising:

[0030] An image conversion module is used to obtain image data, wherein the image data is RGB image data uploaded by different channels, and performs RGB-NIR conversion processing on the RGB image data to obtain a NIR image that is homologous to the face lock;

[0031] The high-dimensional feature binding module is used to perform face detection and facial feature extraction on NIR images, obtain high-dimensional features that can represent faces, and bind the high-dimensional features to the flags according to the binding information;

[0032] The cloud front-end sending module sends the high-dimensional features after the binding flag to the cloud back-end;

[0033] The cloud backend processing module is used to receive high-dimensional features, unbind the high-dimensional features according to the binding information, and save them in the feature access stack of the face lock end;

[0034] The similarity calculation module is used to activate the face lock camera and obtain camera data, wherein the camera data is NIR image data, and perform face detection and feature extraction on the NIR image data to obtain high-dimensional features to be identified that represent the face, perform similarity calculation on the high-dimensional features to be identified and the high-dimensional features stored in the feature access stack of the face lock end to obtain the similarity, and judge whether the requirements for generating an unlocking instruction are met based on the similarity. If so, generate an unlocking instruction.

[0035] On the other hand, a device includes a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute any one of the above-mentioned RGB-NIR different imaging face recognition methods for smart networked face locks.

[0036] Finally, a storage medium stores computer instructions, wherein the computer instructions enable a computer to execute any of the above-mentioned face recognition methods using RGB-NIR different imaging for a smart networked face lock.

[0037] Compared with the existing solutions, the present invention achieves the following beneficial effects:

[0038] The present invention enables the addition, deletion, modification and query of personnel information of the face lock to be improved by sending it through the network, which greatly increases the wide range of application scenarios of the face lock.

[0039] The encoding module of the generator in the CycleGan generative adversarial network adopts the same module in the face feature extraction module, which ensures the consistency of the generated image and the face feature extraction model and improves the recognition efficiency of face recognition.

[0040] After the face lock terminal performs face comparison, the face features stored in the access stack are updated, the face features with the smallest similarity value are removed and replaced, and the best face features are always saved. This ensures that each comparison can quickly and effectively identify the face, and improves the recognition speed of face recognition in the face lock. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0042] Figure 1 It is a flow chart of a face recognition method of RGB-NIR different imaging for a smart networked face lock according to an embodiment of the present invention;

[0043] Figure 2 It is a flow chart of another face recognition method of RGB-NIR different imaging for smart networked face lock according to an embodiment of the present invention;

[0044] Figure 3 It is a structural block diagram of a face recognition device with different RGB-NIR imaging for a smart networked face lock according to an embodiment of the present invention. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] In addition, the described features, structures or characteristics may be combined in one or more example embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the example embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure may be practiced while omitting one or more of the specific details, or other methods, components, steps, etc. may be adopted. In other cases, well-known structures, methods, implementations or operations are not shown or described in detail to avoid obscuring various aspects of the present disclosure.

[0047] This embodiment provides a face recognition method using different RGB-NIR imaging for a smart networked face lock. Figure 1 : is a flow chart of a face recognition method of RGB-NIR different imaging for smart networked face lock according to an embodiment of the present invention, such as Figure 1 As shown, the method comprises the following steps:

[0048] S1: Obtain image data, where the image data is RGB image data uploaded by different channels, and perform RGB-NIR conversion processing on the RGB image data to obtain a NIR image that is homologous to the face lock;

[0049] S2: The cloud front-end sending module performs face detection and facial feature extraction on the NIR image to obtain high-dimensional features that can represent the face, and binds the high-dimensional features to the flag bits according to the binding information;

[0050] S3: The cloud front-end sending module sends the high-dimensional features after the binding flag to the cloud back-end;

[0051] S4: The cloud backend processing module receives the high-dimensional features, unbinds the high-dimensional features according to the binding information, and saves them in the feature access stack of the face lock end;

[0052] S5: Activate the face lock camera and obtain camera data, wherein the camera data is NIR image data, and perform face detection and feature extraction on the NIR image data to obtain high-dimensional features to be identified that represent the face, perform similarity calculation on the high-dimensional features to be identified and the high-dimensional features stored in the feature access stack of the face lock end to obtain the similarity, and determine whether the requirements for generating an unlocking instruction are met based on the similarity. If so, generate an unlocking instruction.

[0053] In summary, the present invention enables the addition, deletion, modification and query of personnel information of the face lock to be improved by means of network distribution, which greatly increases the extensiveness of the application scenarios of the face lock; the encoding module of the generator in the CycleGan generative adversarial network adopts the same module in the face feature extraction module, which ensures the consistency of the image after image generation and the face feature extraction model, and improves the recognition efficiency of face recognition; after the face comparison is performed at the face lock terminal, the face features stored in the access stack are updated, the face features with the smallest similarity value are eliminated and replaced, and the best face features are always preserved; it ensures that each comparison can quickly and effectively identify the face, and improves the recognition speed of face recognition in the face lock.

[0054] In some embodiments, the RGB image data uploaded by different channels include visible light face images collected from a mobile phone camera or multiple face images uploaded in batches in a specific business, wherein the specific business may be multiple face images uploaded in advance.

[0055] In some embodiments, Figure 2 FIG. 1 is a flow chart of another face recognition method using RGB-NIR different imaging for a smart networked face lock according to an embodiment of the present invention. Figure 2 As shown, performing RGB-NIR conversion processing on the RGB image data to obtain a NIR image homologous to the face lock includes the following steps:

[0056] Step S201: collecting multiple sets of paired RGB and NIR face data as data sets for training a generative adversarial network CycleGan, wherein the generative adversarial network CycleGan is a generative adversarial network included in an image generator;

[0057] Step S202: Generate MBF-CycleGan adversarial network structure model based on adversarial network CycleGan: change Generator and change The original encoding module in the generator is modified to the MBF model structure for face feature extraction;

[0058] It is worth noting that the data set in this embodiment contains 14K groups of paired face data. The faces in each group are RGB and NIR imaging face images belonging to the same person, and the size of each face image is uniformly scaled to 512X512.

[0059] CycleGan Generative Adversarial Network contains change R generator, change Generators, each generator contains an encoding module, a conversion module, and a decoding module. In order to improve the overall image generation quality and efficiency, the face feature extraction in the final generator can be consistent with the face feature extraction module. change R generator, change The encoding model in the generator adopts a part of the improved structure of the MBF model for face feature extraction, mainly the input layer input image size is 512x512x3; the first layer uses a standard convolution layer output channel number of 64, the convolution kernel size is 3x3, the step size is 2x2, and the padding is 1x1; the second layer uses a variant of the depthwise separable convolution, setting 64 groups, the convolution kernel size is 3x3, the step size is 2x2, and the padding is 1x1; the third layer is a residual block, which contains 4 internal convolution blocks, each block contains a convolution layer, a batch normalization layer, and a ReLU activation function; the fourth layer is another depthwise separable convolution, which increases the number of channels to 1 28, the size remains unchanged, the fifth layer is a residual block, which contains 6 internal blocks. The convolution kernel size in each block is 3x3, the step size is 1x1, the padding is 1x1, the number of output channels is 128, and the size remains unchanged; the fifth layer is a depth-wise separable convolution, the number of channels is increased to 256, the sixth layer is a residual block, which contains two internal blocks. The convolution kernel size in each block is 3x3, the step size is 1x1, the padding is 1x1, the final output channel number is 256, and the size remains 64x64. The final output layer is a 64x64x256 feature vector that can represent the face, and finally forms a new generative adversarial network structure model of MBF-CycleGan.

[0060] Step S203: input the collected multiple groups of paired RGB and NIR face data into the MBF-CycleGan adversarial network structure model for training to obtain a visible light RGB to NIR image converter model.

[0061] It is worth noting that in the fine-tuning training, some model parameters of CycleGan were used to initialize the parameters of the MBF-CycleGan structure model, and only the initialization parameters other than the MBF encoding module in the generator were retained, focusing on training the encoder. During the process, the number of training iterations was set to 2000, the number of pictures input to the network each time was 64 groups, and the training test learning rate was 0.0005. To ensure better model parameters, the learning rate will be reduced by 0.1 times at 500 and 1500 training times. After the set 2000 training iterations, the final visible light RGB to NIR image converter model is obtained. The converter converts the uploaded face image from RGB to NIR to obtain a NIR image that is homologous to the face lock.

[0062] In some embodiments, binding the high-dimensional features to the flag bits according to the binding information includes the following process:

[0063] In extracting high-dimensional features After that, the binding information includes an identity ID and a representative flag, based on high-dimensional features. and binding information to form a new vector .

[0064] In some embodiments, the high-dimensional feature vector to be identified and stored in the feature access stack of the face lock end sent from the cloud is The high-dimensional feature vector of the face extracted by the face lock end through the camera NIR data The similarity calculation includes the following process:

[0065] ;

[0066] in, is the similarity, Represents the i-th eigenvector value of the high-dimensional feature vector extracted from the camera data, It represents the i-th feature vector value of the high-dimensional feature vector stored in the feature access stack of the face lock end, and n is the dimension size of the feature vector.

[0067] Judging whether the requirements for generating unlocking instructions are met based on similarity includes the following process:

[0068] It is determined whether the similarity exceeds a preset threshold. If so, the requirement for generating an unlock instruction is met; if not, the requirement for generating an unlock instruction is not met.

[0069] In some embodiments, the high-dimensional features extracted by the face camera and the high-dimensional features sent by the cloud backend are saved and updated, and the execution steps are as follows:

[0070] After the requirements for generating unlocking instructions are met, the ID of the current inspector is obtained;

[0071] Find the corresponding face feature access stack S according to the ID of the current detected person, and check whether the size of the stack meets the set size m. If it does, the high-dimensional features extracted by the face camera are stored in the stack. Record as temporary storage, and the m facial features previously saved in stack S Calculate and compare the similarities one by one, and find the one corresponding to the minimum similarity value in stack S , is the face feature with the smallest similarity value in the corresponding stack, and uses the temporarily saved Replace it and save it to the face feature access stack.

[0072] on the other hand, Figure 3 : is a structural block diagram of a face recognition device with different RGB-NIR imaging for a smart networked face lock according to an embodiment of the present invention. Figure 3 As shown, the device comprises:

[0073] An image conversion module is used to obtain image data, wherein the image data is RGB image data uploaded by different channels, and performs RGB-NIR conversion processing on the RGB image data to obtain a NIR image that is homologous to the face lock;

[0074] The high-dimensional feature binding module is used to perform face detection and facial feature extraction on NIR images, obtain high-dimensional features that can represent faces, and bind the high-dimensional features to the flags according to the binding information;

[0075] The cloud front-end sending module is used to send the high-dimensional features after the binding flag to the cloud back-end;

[0076] The cloud backend processing module is used to receive high-dimensional features, unbind the high-dimensional features according to the binding information, and save them in the feature access stack of the face lock end;

[0077] The similarity calculation module is used to activate the face lock camera and obtain camera data, wherein the camera data is NIR image data, and perform face detection and feature extraction on the NIR image data to obtain high-dimensional features to be identified that represent the face, perform similarity calculation on the high-dimensional features to be identified and the high-dimensional features stored in the feature access stack of the face lock end to obtain the similarity, and judge whether the requirements for generating an unlocking instruction are met based on the similarity. If so, generate an unlocking instruction.

[0078] On the other hand, the present invention also provides a device, comprising a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute any RGB-NIR different imaging face recognition method for a smart networked face lock.

[0079] Finally, the present invention also provides a storage medium, which stores computer instructions, and the computer instructions enable the computer to execute any RGB-NIR different imaging face recognition method for an intelligent networked face lock.

[0080] It is worth noting that the facial images and identity IDs used in this application are agreed upon by the parties and uploaded to the cloud, in compliance with relevant laws and regulations.

[0081] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0082] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0083] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0084] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only some logical function divisions. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0085] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0086] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A face recognition method using RGB-NIR different imaging for intelligent networked face lock, characterized in that: Methods include: S1: Obtain image data, where the image data is RGB image data uploaded by different channels, perform RGB-NIR conversion on the RGB image data, and obtain a NIR image that is the same as the face lock: Collect multiple sets of paired RGB and NIR face data as data sets for training the generative adversarial network CycleGan, where the adversarial network CycleGan is a generative adversarial network that includes an image generator; Generate MBF-CycleGan adversarial network structure model based on generative adversarial network CycleGan: change Generator and change The original encoding module structure in the generator is changed to the MBF model structure for face feature extraction; The collected multiple sets of paired RGB and NIR face data are input into the MBF-CycleGan adversarial network structure model for fine-tuning training to obtain a visible light RGB to NIR image converter model; S2: The cloud front-end sends a module to perform face detection and facial feature extraction on the NIR image to obtain high-dimensional features that can represent the face, and binds the high-dimensional features to the flag bits according to the binding information: In extracting high-dimensional feature vectors After that, the binding information includes an identity ID and a representative flag, based on the high-dimensional feature vector and binding information to form a new vector ; S3: The cloud front-end sending module sends the high-dimensional features after the binding flag to the cloud back-end; S4: The cloud backend processing module receives the high-dimensional features, unbinds the high-dimensional features according to the binding information, and saves them in the feature access stack of the face lock end; S5: Activate the face lock camera and obtain camera data, wherein the camera data is NIR image data, and perform face detection and feature extraction on the NIR image data to obtain high-dimensional features to be identified that represent the face, perform similarity calculation on the high-dimensional features to be identified and the high-dimensional features stored in the feature access stack of the face lock end to obtain similarity, and determine whether the requirements for generating an unlock instruction are met based on the similarity, and if so, generate an unlock instruction; Among them, the high-dimensional feature vector to be identified is stored in the feature access stack of the face lock end and sent from the cloud. The high-dimensional feature vector of the face extracted by the face lock end through the camera NIR data The similarity calculation includes the following process: ； in, is the similarity, Represents the i-th eigenvector value of the high-dimensional feature vector extracted from the camera data, It represents the i-th feature vector value of the high-dimensional feature vector stored in the feature access stack of the face lock end, and n is the dimension size of the feature vector.

2. The face recognition method of RGB-NIR different imaging for intelligent networked face lock according to claim 1 is characterized in that: The RGB image data uploaded by different channels include visible light facial images collected from mobile phone cameras or multiple facial images uploaded in batches in specific businesses.

3. The face recognition method of RGB-NIR different imaging for intelligent networked face lock according to claim 1 is characterized in that: Judging whether the requirements for generating unlocking instructions are met based on similarity includes the following process: It is determined whether the similarity exceeds a preset threshold. If so, the requirement for generating an unlock instruction is met; if not, the requirement for generating an unlock instruction is not met.

4. The face recognition method of RGB-NIR different imaging for intelligent networked face lock according to claim 1 is characterized in that: The method also includes: Save and update the high-dimensional features extracted by the face camera and the high-dimensional features sent by the cloud backend. The execution steps are as follows: After the requirements for generating unlocking instructions are met, the ID of the current inspector is obtained; Find the corresponding face feature access stack S according to the ID of the current detected person, and check whether the size of the stack meets the set size m. If it does, the high-dimensional features extracted by the face camera are stored in the stack. Record as temporary storage, and the m facial features previously saved in stack S Calculate and compare the similarities one by one, and find the one corresponding to the minimum similarity value in stack S , is the face feature with the smallest similarity value in the corresponding stack, and uses the temporarily saved Replace it and save it to the face feature access stack.

5. A face recognition device with different RGB-NIR imaging for smart networked face lock, applicable to a face recognition method with different RGB-NIR imaging for smart networked face lock as claimed in any one of claims 1 to 4, characterized in that: The device includes: An image conversion module is used to obtain image data, wherein the image data is RGB image data uploaded by different channels, and performs RGB-NIR conversion processing on the RGB image data to obtain a NIR image that is homologous to the face lock; The high-dimensional feature binding module is used to perform face detection and facial feature extraction on NIR images, obtain high-dimensional features that can represent faces, and bind the high-dimensional features to the flags according to the binding information; The cloud front-end sending module is used to send the high-dimensional features after the binding flag to the cloud back-end; The cloud backend processing module is used to receive high-dimensional features, unbind the high-dimensional features according to the binding information, and save them in the feature access stack of the face lock end; The similarity calculation module is used to activate the face lock camera and obtain camera data, wherein the camera data is NIR image data, and perform face detection and feature extraction on the NIR image data to obtain high-dimensional features to be identified that represent the face, perform similarity calculation on the high-dimensional features to be identified and the high-dimensional features stored in the feature access stack of the face lock end to obtain the similarity, and judge whether the requirements for generating an unlocking instruction are met based on the similarity. If so, generate an unlocking instruction.

6. A device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to run the computer program to execute the face recognition method of RGB-NIR different imaging for a smart networked face lock according to any one of claims 1 to 4.

7. A storage medium, characterized in that: The storage medium stores computer instructions, which enable the computer to execute the face recognition method of RGB-NIR different imaging for smart networked face lock as described in any one of claims 1 to 4.

Images