An intelligent face recognition method and system based on internet of things and big data analysis

Abstract

The application relates to an intelligent face recognition method and system based on the Internet of Things and big data analysis, wherein a value range of an original setting attribute label is evenly divided according to a dynamic interval M, a derivative attribute label vector is determined by two points based on the dynamic interval M, and the dynamic interval M is adjusted through output recognition precision feedback of a face recognition neural network model, so that the derivative attribute label which contributes more obviously to the recognition precision can be found, and the output recognition precision of the face recognition neural network model based on the same can be obviously improved.

Description

Technical Field

[0001] This invention relates to an intelligent face recognition method and system based on the Internet of Things and big data analysis. Background Technology

[0002] The use of multiple attributes to improve the accuracy of face recognition is already very common. For example, the relevant patent CN111507263A discloses a face recognition technology based on multiple attributes. The core of this technology is to generate derived label values ​​for attributes and calculate derived label vectors through two-point representation. Then, the loss is calculated for the derived label vectors. However, since the derived label calculation is based on the average division by fixed interval values, the derived labels are also relatively fixed, and the face recognition effect based on this is not ideal. Summary of the Invention

[0003] To overcome the shortcomings of existing technologies, this invention provides an intelligent face recognition method and system based on the Internet of Things and big data analysis.

[0004] The technical solution adopted by this invention to solve its technical problem is:

[0005] A smart face recognition method based on the Internet of Things and big data analysis includes the following steps:

[0006] The system acquires target face images for face recognition via IoT cameras; these images are then processed into a mature dataset; a face recognition neural network model is constructed, including a dedicated neural network model for attribute features; the attributes corresponding to these features include regression attributes, in which original and derived attribute labels are obtained; the derived attribute labels are derived from the original attribute labels; specifically, the value range of the original attribute labels is divided equally according to a dynamic interval M, and the derived attribute label vector is determined using a two-point method based on the dynamic interval M; the dynamic interval M is adjusted based on the recognition accuracy feedback from the output of the face recognition neural network model.

[0007] Furthermore, based on the dynamic interval M, the derived attribute label vector is determined by a two-point formula. Specifically, if the truth value of the original attribute label is set to T, then the following equation is constructed:

[0008]

[0009]

[0010]

[0011] The derived attribute label vector has weights P1 and P2 at positions x1 and x2, respectively. x1 and x2 are the marker points after average division. The derived attribute label vector is determined by x1, x2, P1, and P2.

[0012] Furthermore, the dynamic interval M is adjusted by the feedback of the recognition accuracy of the face recognition neural network model. Specifically, a feedback adjustment time period from t1 to t2 is initially set.

[0013] Initially set a dynamic interval M and obtain the corresponding recognition accuracy Z of the face recognition neural network model; if the recognition accuracy Z is not the maximum value, then perform the following calculations in a loop:

[0014] Modify the dynamic interval M value, and change the increment of the dynamic interval M value from time period t1 to t2 as follows:

[0015]

[0016] Where t is a general variable from t1 to t2, and Q, Q1, Q2, and Q3 are all constants;

[0017] Then calculate the recognition accuracy Z of the face recognition neural network model corresponding to the modified dynamic interval M;

[0018] As the dynamic interval M changes from time t1 to t2, calculate whether the change in Z is positively correlated with the cumulative effect parameter M1 of the dynamic interval M.

[0019]

[0020] Furthermore, determine whether the recognition accuracy Z is the maximum value:

[0021] The values ​​of all recognition accuracies Z are obtained from time period t1 to t2. Then, they are arranged in time sequence and fitted as a function of time. The value of the maximum recognition accuracy Z can be determined by taking the second derivative of the fitted function.

[0022] An intelligent face recognition system based on the Internet of Things (IoT) and big data analytics includes a connected IoT-enabled camera, a dataset construction unit, a recognition model construction unit, and a derived attribute label generation unit.

[0023] The dataset construction unit is used to acquire target face images for face recognition through IoT cameras; then the target face images are processed into a mature dataset.

[0024] The recognition model building unit is used to build a face recognition neural network model, which includes a face recognition neural network model specifically for attribute features; the attributes corresponding to the attribute features include regression attributes, and the original attribute labels and derived attribute labels are obtained in the regression attribute building process;

[0025] The derived attribute label generation unit is used to generate derived attribute labels, which are derived from the original attribute labels. Specifically, the value range of the original attribute labels is divided equally according to the dynamic interval M, and the derived attribute label vector is determined by a two-point method based on the dynamic interval M. It is also used to adjust the dynamic interval M through the feedback of the recognition accuracy of the face recognition neural network model.

[0026] Beneficial effects

[0027] This application divides the value range of the original attribute labels into equal parts by a dynamic interval M. Based on the dynamic interval M, a derived attribute label vector is determined by a two-point method. The dynamic interval M is then adjusted by the feedback of the recognition accuracy of the face recognition neural network model. This allows us to find derived attribute labels that contribute more significantly to the recognition accuracy. Based solely on this, the output recognition accuracy of the face recognition neural network model can be significantly improved. Detailed Implementation

[0028] This application discloses an intelligent face recognition method based on the Internet of Things and big data analysis, including the following steps:

[0029] The system acquires target face images for face recognition via IoT cameras; these images are then processed into a mature dataset; a face recognition neural network model is constructed, including a dedicated neural network model for attribute features; the attributes corresponding to these features include regression attributes, in which original and derived attribute labels are obtained; the derived attribute labels are derived from the original attribute labels; specifically, the value range of the original attribute labels is divided equally according to a dynamic interval M, and the derived attribute label vector is determined using a two-point method based on the dynamic interval M; the dynamic interval M is adjusted based on the recognition accuracy feedback from the output of the face recognition neural network model.

[0030] Accordingly, this application also discloses an intelligent face recognition system based on the Internet of Things and big data analysis, which includes a connected IoT-enabled camera, a dataset construction unit, a recognition model construction unit, and a derived attribute label generation unit.

[0031] The dataset construction unit is used to acquire target face images for face recognition through IoT cameras; then the target face images are processed into a mature dataset.

[0032] The recognition model building unit is used to build a face recognition neural network model, which includes a face recognition neural network model specifically for attribute features; the attributes corresponding to the attribute features include regression attributes, and the original attribute labels and derived attribute labels are obtained in the regression attribute building process;

[0033] The derived attribute label generation unit is used to generate derived attribute labels, which are derived from the original attribute labels. Specifically, the value range of the original attribute labels is divided equally according to the dynamic interval M, and the derived attribute label vector is determined by a two-point method based on the dynamic interval M. It is also used to adjust the dynamic interval M through the feedback of the recognition accuracy of the face recognition neural network model.

[0034] The IoT-connected camera refers to a camera connected to the Internet of Things (IoT). The target face image of this camera can be output to a big data analysis server via the IoT. The intelligent face recognition system is built on the big data analysis server. This application divides the value range of the originally set attribute labels into an average of dynamic intervals M. Based on the dynamic interval M, a derived attribute label vector is determined by a two-point method. The dynamic interval M is adjusted by the feedback of the recognition accuracy of the face recognition neural network model. In this way, a derived attribute label that contributes more significantly to the recognition accuracy can be found. Only on this basis can the output recognition accuracy of the face recognition neural network model be significantly improved.

[0035] In specific implementation, the derived attribute label vector is determined by a two-point formula based on the dynamic interval M. Specifically, if the truth value of the original attribute label is set to T, then the following equation is constructed:

[0036]

[0037]

[0038] The derived attribute label vector has weights P1 and P2 at positions x1 and x2, respectively. x1 and x2 are the marker points after average division. The derived attribute label vector is determined by x1, x2, P1, and P2.

[0039] Preferably, in the implementation, the dynamic interval M is adjusted by the feedback of the recognition accuracy of the face recognition neural network model. Specifically, a feedback adjustment time period t1 to t2 is initially set.

[0040] Initially set a dynamic interval M and obtain the corresponding recognition accuracy Z of the face recognition neural network model; if the recognition accuracy Z is not the maximum value, then perform the following calculations in a loop:

[0041] Modify the dynamic interval M value, and change the increment of the dynamic interval M value from time period t1 to t2 as follows:

[0042]

[0043] Where t is a general variable from t1 to t2, and Q, Q1, Q2, and Q3 are all constants;

[0044] Then calculate the recognition accuracy Z of the face recognition neural network model corresponding to the modified dynamic interval M;

[0045] As the dynamic interval M changes from time t1 to t2, calculate whether the change in Z is positively correlated with the cumulative effect parameter M1 of the dynamic interval M.

[0046]

[0047] Specifically, to determine if the recognition accuracy Z is the maximum value:

[0048] The values ​​of all recognition accuracies Z are obtained from time period t1 to t2. Then, they are arranged in time sequence and fitted as a function of time. The value of the maximum recognition accuracy Z can be determined by taking the second derivative of the fitted function.

[0049] The program code used to implement the functions of the dataset construction unit, recognition model construction unit, and derived attribute label generation unit of the system in this application can be written in any combination of one or more programming languages. This program code can be provided to the processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The program code can be executed entirely on a machine, partially on a machine, partially on a machine and partially on a remote machine as a standalone software package, or entirely on a remote machine or server. In the context of this application, the program code for the functions of the system in this application is stored in a machine-readable medium, which can be a tangible medium that may contain or store programs for use by or in conjunction with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer 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.

[0050] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0051] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), the Internet, and blockchain networks.

[0052] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. A server can be a cloud server, also known as a cloud computing server or cloud host, a hosting product within the cloud computing service ecosystem, addressing the shortcomings of traditional physical hosts and VPS (Virtual Private Server, or simply "VPS") services, such as high management difficulty and weak business scalability. Servers can also be servers for distributed systems or servers incorporating blockchain technology.

[0053] According to embodiments of this application, this application also provides a computer program product, including a computer program, wherein the computer program, when executed by a processor, implements the functions of the system according to the above embodiments of this application. As is known from common technical knowledge, the present invention can be implemented through other embodiments that do not depart from its spirit or essential characteristics. The disclosed embodiments above are merely illustrative in all respects and are not exhaustive. All modifications within the scope of this invention or equivalent to the scope of this invention are included in this invention.

Claims

1. A smart face recognition method based on the Internet of Things and big data analysis, characterized in that, The steps include: The system acquires target face images for face recognition via IoT cameras; these images are then processed into a mature dataset; a face recognition neural network model is constructed, comprising a neural network model for attribute features; the attributes corresponding to these features include regression attributes, in which original and derived attribute labels are obtained; the derived attribute labels are derived from the original attribute labels; specifically, the value range of the original attribute labels is divided equally by a dynamic interval M, and the derived attribute label vector is determined using a two-point method based on the dynamic interval M; the dynamic interval M is adjusted by feedback from the recognition accuracy of the face recognition neural network model's output. The dynamic interval M is adjusted by the feedback of the recognition accuracy of the face recognition neural network model. Specifically, a feedback adjustment time period t1 to t2 is initially set. Initially set a dynamic interval M and obtain the corresponding recognition accuracy Z of the face recognition neural network model; if the recognition accuracy Z is not the maximum value, then perform the following calculations in a loop: Modify the dynamic interval M value, and change the increment of the dynamic interval M value from time period t1 to t2 as follows: ； Where t is a general variable from t1 to t2, and Q, Q1, Q2, and Q3 are all constants; Then calculate the recognition accuracy Z of the face recognition neural network model corresponding to the modified dynamic interval M; As the dynamic interval M changes from time t1 to t2, calculate whether the change in Z is positively correlated with the cumulative effect parameter M1 of the dynamic interval M. ； Specifically, to determine if the recognition accuracy Z is the maximum value: The values ​​of all recognition accuracies Z are obtained from time period t1 to t2. Then, they are arranged in time sequence and fitted as a function of time. The value of the maximum recognition accuracy Z can be determined by taking the second derivative of the fitted function.

2. The intelligent face recognition method based on the Internet of Things and big data analysis according to claim 1, characterized in that, Based on the dynamic interval M, the derived attribute label vector is determined by a two-point formula. Specifically, if the truth value of the original attribute label is set to T, then the following equation is constructed: ； ； ； The derived attribute label vector has weights P1 and P2 at positions x1 and x2, respectively. x1 and x2 are the marker points after average division. The derived attribute label vector is determined by x1, x2, P1, and P2.

3. A system applying the intelligent face recognition method based on the Internet of Things and big data analysis as described in claim 1, characterized in that, This includes connected IoT cameras, dataset construction units, recognition model construction units, and derived attribute label generation units. The dataset construction unit is used to acquire target face images for face recognition through IoT cameras; then the target face images are processed into a mature dataset. The recognition model building unit is used to build a face recognition neural network model, which includes a face recognition neural network model for attribute features; the attributes corresponding to the attribute features include regression attributes, and the original attribute labels and derived attribute labels are obtained in the regression attribute building process; The derived attribute tag generation unit is used to generate derived attribute tags, which are derived from the original attribute tags. Specifically, the value range of the original attribute labels is divided equally according to the dynamic interval M, and the derived attribute label vector is determined by two points based on the dynamic interval M; it is also used to adjust the dynamic interval M through the feedback of the recognition accuracy of the face recognition neural network model.

Images