Embedded face recognition device and method
By using an embedded face recognition device and method, combined with the assembly structure of the base and the device body and a convolutional neural network, the problems of poor face recognition accuracy and high false recognition rate in the prior art are solved, achieving higher recognition accuracy and faster recognition speed, while extending the service life of the illumination unit.
Patent Information
- Application Number
- CN202410993492.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-24
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2044-07-24
AI Technical Summary
Current facial recognition technology suffers from poor accuracy and long unlocking time, and there are serious misidentification problems, especially in practical applications.
An embedded face recognition device and method are adopted, which combines the assembly structure of the base and the device body, and sets up an image acquisition component, a recognition unit and an illumination unit. Convolutional neural network is used for feature extraction, and key facial points are highlighted by active adjustment of the illumination unit and light matching, thereby improving recognition accuracy and speed.
It effectively reduces the false recognition rate, improves the accuracy and speed of face recognition, extends the lifespan of the illumination unit, and provides multiple unlocking methods to enhance applicability.
Smart Images

Figure CN118522060B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of face recognition devices, and in particular to an embedded face recognition device and method. Background Art
[0002] Face recognition is the use of biometric information of human face to identify the identity of the corresponding person through certain technologies. For example, after obtaining the face image of the target person, the face image that matches the target person can be determined from a large number of face images to be determined. Face recognition is an important research work in the field of computer vision and pattern recognition, and it also has a wealth of real-life application scenarios.
[0003] Face recognition technology has been developed for decades, and many related machine learning algorithms were proposed in the early years, including methods based on geometric features and statistical methods. However, affected by issues such as light brightness, presentation posture, facial makeup, and environmental occlusion, the recognition accuracy of these early machine learning methods is not ideal. Once the environment or the makeup of the person changes significantly, it is difficult to accurately recognize the face. In recent years, methods based on convolutional neural networks have been widely used in face recognition and have achieved good recognition accuracy. The main process is to train the convolutional neural network with a training set, and use the trained convolutional neural network to extract features to obtain the vectors of the target image and the image to be determined. Then, through a vector-based similarity determination method, the image to be determined with the highest similarity can be determined as the target person.
[0004] However, although the existing residual convolutional neural network has achieved excellent accuracy on the face recognition dataset LFW, the neural network's ability to distinguish faces in actual application scenarios is still unsatisfactory. This is mainly manifested in the serious misidentification problem for untrained people, resulting in the fact that in actual applications, one person's face information is often mistakenly identified as another person. Summary of the invention
[0005] Technical issues solved:
[0006] In view of the above-mentioned shortcomings of the prior art, the present invention provides an embedded face recognition device and method, which can effectively solve the problems of poor accuracy and poor unlocking time in the prior art.
[0007] Technical Solution
[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions:
[0009] The present invention provides an embedded face recognition device, comprising a base; a magnetic card induction component is arranged on the base, a connecting piece is arranged on one side of the base, a pin is arranged on the connecting piece, a hinge seat is arranged on the outer side of the pin, a device body is arranged on one side of the hinge seat, an image acquisition component, a recognition unit, a comparison unit and an illumination unit are arranged on the top of the front of the device body, and a control component is arranged in the device body; a device body structure with the image acquisition component, the recognition unit, the comparison unit and the illumination unit is arranged, face recognition can be automatically started when there is a face recognition demand, monitoring can be performed when there is no face recognition demand, and the illumination unit can be prevented from being turned on untimely and too frequently, thereby extending the service life of the illumination unit, and cooperating with an internal recognition method to improve the accuracy and speed of face recognition.
[0010] A bottom mounting assembly is provided at the bottom end of the base, and the bottom mounting assembly includes a truss, a movable block arranged at the top end of the truss, a movable bolt arranged in the movable block and a positioning block arranged outside the movable bolt, a rotating shaft is fixed between the positioning blocks, two groups of sleeves are arranged on the outside of the rotating shaft, a mounting frame is fixed on the outside of each group of the sleeves, and a positioning bolt is penetrated on one side of the sleeve, and a positioning hole is arranged on one side of the mounting frame on both sides, and the base is fixedly connected to the positioning hole by a locking bolt.
[0011] Furthermore, a charging port is provided on one side of the base, and multiple groups of cable interfaces are provided on the other side; the device can be connected and fixed to an external device via a cable, and the charging port can facilitate electrical connection between an external power source and the device body.
[0012] Furthermore, a power amplifier, a switch button and a volume adjustment button are arranged on the back of the device body, and the hinge seat and the pin shaft are fixedly connected to the device body and the base respectively through connecting pieces.
[0013] Furthermore, a ball head is fixed on the inner side of the positioning block, and the ball head and the positioning block are an integrated structure, and spherical spaces matching the size of the ball head are opened on both sides of the rotating shaft; before use, the device body and the base can be assembled and fixed first, and then the device body is electrically connected to the external power supply and external device through the charging port and the cable interface, and then the base structure and the truss form a skeleton structure, and the movable block can fix the moving bolt, and the positioning block and the ball head structure fixed at the end of the moving bolt can be connected to the rotating shaft, so when the sleeve is fixed to the rotating shaft, it can be rotated on the outside of the rotating shaft, and after rotating to the required angle, the sleeve and the ball head can be fixed through the positioning bolts on the outside, thereby achieving the effect of fixing the angle.
[0014] Furthermore, the side walls of the mounting frames on both sides are provided with sliding grooves, one side of the sleeve is fixedly connected to the triangular plate, and the two sides of the triangular plate are arranged in the sliding grooves by positioning bolts.
[0015] Furthermore, a magnetic card sensing component is arranged on the outer side of the base; the user can also unlock the panel identification by means of a magnetic card, and the application scope of the device is improved through different and multiple unlocking methods.
[0016] Furthermore, a switch button is arranged on one side of the back side of the device body, and a volume adjustment button is arranged on the other side, and the switch button is arranged vertically, and the volume adjustment button is arranged horizontally.
[0017] In order to achieve the above objectives, the present invention provides a technical method to achieve:
[0018] An embedded face recognition method, characterized in that it includes an embedded face recognition device, and the specific steps include: S1: preprocessing multiple existing face image data to obtain training images; S2: building a convolutional neural network model; S3: inputting the preprocessed training images into the convolutional neural network model in S2; S4: calculating the angle error and its standard deviation; S5: modifying parameters according to the target error; S6: detecting key position points of the face; S7: processing the image to be processed according to the key position points in S6; S8: using the circular frame in the network model and the dark side of the image to highlight the key points of the face position point; S9: extracting the feature vector of the feature point through the obtained key position feature; S10: judging the accuracy of the image according to the feature vector in S9; therefore, the convolutional neural network used in the present invention is based on the structure of the residual-based convolutional neural network that has previously performed well in face recognition. Therefore, the method can further ensure the accuracy of face recognition in the process of reducing the misrecognition rate in the face recognition process, and in the final step of processing the image to be processed, the light coordination method is adopted, and the circular frame, light and shadow dark side method is used to highlight the key position points of the face, thereby speeding up the face recognition time.
[0019] Furthermore, the facial image data in step S1 are pre-processed training images with uniform size and corresponding to the existing facial images respectively.
[0020] Further, in steps S1 to S9, the illumination unit continuously exposes within the illumination threshold.
[0021] Beneficial Effects
[0022] Compared with the known public technology, the technical solution provided by the present invention has the following beneficial effects:
[0023] 1. The present invention provides a base and a device body structure. Before use, the device body and the base can be first assembled and fixed. After that, the device body is electrically connected to an external power source and an external device through a charging port and a cable interface. After that, the base structure and the truss form a skeleton structure, in which the movable block can fix the movable bolt, and the positioning block and the ball head structure fixed at the end of the movable bolt can be connected to the rotating shaft. Therefore, when the sleeve is fixed to the rotating shaft, it can be rotated on the outside of the rotating shaft. After rotating to the required angle, the sleeve and the ball head can be fixed by the positioning bolts on the outside thereof, thereby achieving the function of fixing the angle. On the mounting frames on both sides, the positioning holes and the positioning bolts can be used to fix the device body, so that the device body can be installed at the required position.
[0024] 2. This device is provided with a device body structure having an image acquisition component, a recognition unit, a comparison unit and an illumination unit. It can automatically start face recognition when there is a need for face recognition, and perform monitoring when there is no need for face recognition. It can also prevent the illumination unit from being turned on untimely or too frequently, thereby extending the service life of the illumination unit and cooperating with the internal recognition method to improve the accuracy and speed of face recognition.
[0025] 3. In the present method, by adopting the method and device of the convolutional neural network model based on deep learning, since in the stage of parameter updating, the update is performed based on the angle error optimized by the standard error, the convolutional neural network finally obtained can not only achieve good results in classification (i.e., different people are divided into different categories), but also in clustering (i.e., the probability of misidentification between different people is reduced). At the same time, since the convolutional neural network used in the present invention is based on the structure of the residual-based convolutional neural network that has previously performed well in face recognition, the present method can further ensure the accuracy of face recognition in the process of reducing the misidentification rate in the face recognition process, and in the final step of processing the image to be processed, the light coordination method is adopted, and the circular frame and the light and shadow dark side method are used to highlight the key position points of the face, thereby speeding up the face recognition time.
[0026] 4. In this method, a light brightness adjustment mechanism is applied. First, the device body 3 senses the approach of a human body and actively turns on the lighting unit 7. The lighting unit 7 is pre-adjusted to a certain threshold by the user, and after being reflected at key positions on the face, feedback is received. After the face recognition mode is turned on, the light brightness can be actively adjusted through feedback, and the threshold of light can be actively adjusted to reduce the discomfort caused by excessive light on the face during the entire recognition process and the impact of overexposure on the recognition process. BRIEF DESCRIPTION OF THE DRAWINGS
[0027] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0028] Figure 1 It is one of the structural schematic diagrams of the present invention;
[0029] Figure 2 This is the second structural schematic diagram of the present invention;
[0030] Figure 3 This is one of the structural decomposition diagrams in the present invention;
[0031] Figure 4 This is the second structural split diagram of the present invention;
[0032] Figure 5 It is a structural schematic diagram of the bottom mounting assembly in the present invention;
[0033] Figure 6 This is one of the structural disassembly diagrams of the bottom mounting assembly in the present invention;
[0034] Figure 7 This is the second structural disassembly diagram of the bottom mounting assembly in the present invention;
[0035] Figure 8 It is a flowchart of the method in the present invention;
[0036] Fig. 9 It is a structural connection block diagram of the device body in the present invention;
[0037] Fig.10 This is a flow chart of the illumination unit adjustment method of the present invention.
[0038] The numbers in the figure represent: 1. base; 2. magnetic card sensing component; 3. device body; 4. image acquisition component; 5. recognition unit; 6. comparison unit; 7. illumination unit; 8. bottom mounting component; 81. truss; 82. movable block; 83. moving bolt; 84. positioning block; 841. ball head; 85. rotating shaft; 86. mounting bracket; 861. slide groove; 862. positioning hole; 87. sleeve; 88. positioning bolt; 9. switch button; 10. volume adjustment button; 11. power amplifier device; 12. connecting piece; 13. pin shaft; 14. hinge seat; 15. charging port. DETAILED DESCRIPTION
[0039] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are 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.
[0040] The present invention will be further described below in conjunction with the embodiments.
[0041] Embodiment 1: An embedded face recognition device, referring to the attached Figure 1 -Attached Figure 7 , including a base 1; a magnetic card sensing component 2 is arranged on the base 1, a connecting piece 12 is arranged on one side of the base 1, and a pin shaft 13 is arranged on the connecting piece 12, a hinge seat 14 is arranged on the outer side of the pin shaft 13, and a device body 3 is arranged on one side of the hinge seat 14, and an image acquisition component 4, a recognition unit 5, a comparison unit 6 and an illumination unit 7 are arranged on the top of the front of the device body 3. The device body structure with the image acquisition component 4, the recognition unit 5, the comparison unit 6 and the illumination unit 7 is provided, which can automatically start face recognition when there is a need for face recognition, and monitor when there is no need for face recognition, and can prevent the illumination unit from being turned on untimely and too frequently, thereby extending the service life of the illumination unit, and cooperating with the internal recognition method to improve the accuracy and speed of face recognition.
[0042] A control assembly is provided in the equipment body 3, and a bottom mounting assembly 8 is provided at the bottom end of the base 1. The bottom mounting assembly 8 includes a truss 81, a movable block 82 provided at the top of the truss 81, a moving bolt 83 provided in the movable block 82, and a positioning block 84 provided on the outside of the moving bolt 83. A rotating shaft 85 is fixed between the positioning blocks 84, and two groups of sleeves 87 are provided on the outside of the rotating shaft 85. A mounting frame 86 is fixed on the outside of each group of sleeves 87, and a positioning bolt 88 is provided through one side of the sleeve 87. A positioning hole 862 is provided on one side of the mounting frames 86 on both sides, and the base 1 is fixedly connected to the positioning hole 862 by a locking bolt.
[0043] A charging port 15 is provided on one side of the base 1, and a plurality of cable interfaces are provided on the other side.
[0044] The back of the device body 3 is provided with an amplifier device 11, a switch button 9 and a volume adjustment button 10, and the hinge seat 14 and the pin shaft 13 are fixedly connected to the device body 3 and the base 1 respectively through the connecting piece 12; a ball head 841 is fixed to the inner side of the positioning block 84, and the ball head 841 and the positioning block 84 are an integrated structure, and spherical spaces matching the size of the ball head 841 are provided on both sides of the rotating shaft 85; the side walls of the mounting frames 86 on both sides are provided with slide grooves 861, one side of the sleeve 87 is fixedly connected to the triangular plate, and the two sides of the triangular plate are arranged in the slide grooves 861 by positioning bolts 88; the present invention provides a base 1 and a device body 3 structure, and before use, the device body 3 and the base 1 can be first assembled and fixed, and then the device body 3 is electrically connected to an external power source and an external device through the charging port 15 and the cable interface, and then the base 1 structure and the truss 81 form a skeleton structure, wherein the movable block 82 can fix the movable bolt 83, and the positioning block 84 and the ball head 841 structure fixed at the end of the movable bolt 83 can be connected to the rotating shaft 85, so when the sleeve 87 is fixed to the rotating shaft 85, it can be rotated on the outside of the rotating shaft 85, and after rotating to the required angle, the sleeve 87 and the ball head 841 can be fixed by the positioning bolt 88 on the outside thereof, so as to achieve the function of fixing the angle, and on the mounting frames 86 on both sides, the positioning holes 862 and the positioning bolts 88 can be fixedly connected to the device body 3, so as to facilitate the installation of the device body 3 at the required position.
[0045] A magnetic card sensing component 2 is arranged on the outer side of the base 1, and the user can also unlock the panel identification by means of a magnetic card, thereby improving the application scope of the device through different and multiple unlocking methods.
[0046] A switch button 9 is provided on one side of the back of the device body 3, and a volume adjustment button 10 is provided on the other side. The switch button 9 is vertically arranged, and the volume adjustment button 10 is horizontally arranged. The identification unit 5, the comparison unit 6 and the illumination unit 7 are symmetrically arranged in two groups at the top of the front side of the device body 3. The power amplifier device 11 can broadcast voice in real time, and the broadcast volume can be adjusted by the volume adjustment button 10 on one side.
[0047] Embodiment 2: This embodiment proposes a face recognition method based on Embodiment 1:
[0048] like Figure 8-10As shown, an embedded face recognition method includes an embedded face recognition device in embodiment 1, including, S1: preprocessing multiple existing face image data and obtaining training images; S2: building a convolutional neural network model; S3: inputting the preprocessed training images into the convolutional neural network model in S2; S4: calculating the angle error and its standard deviation; S5: modifying parameters according to the target error; S6: detecting key position points of the face; S7: processing the image to be processed according to the key position points in S6; S8: using the circular frame in the network model and the dark side of the image to highlight the key position points of the face; S9: extracting the feature vector of the feature point through the obtained key position features; S10: judging the accuracy of the image according to the feature vector in S9.
[0049] The facial image data in step S1 is a pre-processed training image with uniform size and corresponding to the existing facial image respectively; in steps S1-S9, the illumination unit 7 is continuously exposed within the illumination threshold; by adopting the method and device of the convolutional neural network model based on deep learning, since in the parameter updating stage, the update is performed based on the angle error optimized by the standard error, the convolutional neural network finally obtained can not only achieve good results in classification (i.e., different people are divided into different categories), but also in clustering (i.e., the probability of misidentification between different people is reduced). At the same time, since the convolutional neural network used in the present invention is based on the structure of the residual-based convolutional neural network that has previously performed relatively well in face recognition, the method can further ensure the accuracy of face recognition in the process of reducing the misidentification rate in the face recognition process, and in the final step of processing the image to be processed, the light coordination method is adopted, and the circular frame, light and shadow dark side method is used to highlight the key position points of the face, thereby speeding up the face recognition time.
[0050] In the present method, a light brightness adjustment mechanism is applied. First, the device body 3 senses the approach of a human body and actively turns on the illumination unit 7. The illumination unit 7 is pre-adjusted to a certain threshold by the user, reflected at key positions on the face, and then receives feedback. After the face recognition mode is turned on, the light brightness can be actively adjusted through feedback, and the threshold of light can be actively adjusted to reduce the discomfort and overexposure caused by excessive light on the face during the entire recognition process. Compared with a face recognition device in which the illumination unit 7 is always turned on or the illumination unit 7 is manually turned on, in the face recognition device of the present invention, the illumination unit 7 can be turned on in time when there is a need for face recognition, and will not be turned on too frequently, thereby greatly reducing the total turn-on time of the illumination unit and extending the service life of the illumination unit 7.
[0051] Embodiment 3: This embodiment introduces another facial recognition method based on embodiment 1. The specific steps are as follows:
[0052] S1: Preprocess the existing multiple face image data and obtain training images; S2: Construct a convolutional neural network model; S3: Input the preprocessed training images into the convolutional neural network model in S2; S4: Calculate the angle error and its standard deviation; S5: Modify parameters according to the target error; S6: Detect the key position points of the face; S7: Process the image to be processed according to the key position points in S6; S8: Use the key position points of the face in the network model; S9: Extract the feature vector of the feature point through the obtained key position features; S10: Judge the accuracy of the image according to the feature vector in S9.
[0053] Compared with the second embodiment, the difference is that in step S8, the traditional facial recognition position point method is used for unlocking. Compared with the method of using the circular frame in the first embodiment to determine and highlighting the key position points through the illumination unit 7, when this method is used, compared with the existing method, the unlocking time of this device is slightly slower than the method of the first embodiment and slightly faster than the existing technical solution. Therefore, it is known that the facial recognition method of the first embodiment carried in the device body 3 can have a faster unlocking experience compared with the existing method and the solution of the third embodiment.
[0054] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the protection scope of the technical solutions of the embodiments of the present invention.
Claims
1. An embedded face recognition device, characterized in that: The invention comprises a base (1); a magnetic card sensing component (2) is arranged on the base (1); a connecting piece (12) is arranged on one side of the base (1); a pin shaft (13) is arranged on the connecting piece (12); a hinge seat (14) is arranged on the outer side of the pin shaft (13); a device body (3) is arranged on one side of the hinge seat (14); an image acquisition component (4), an identification unit (5), a comparison unit (6) and an illumination unit (7) are arranged on the top of the front side of the device body (3); a control component is arranged in the device body (3); a bottom mounting component (8) is arranged at the bottom end of the base (1); and the bottom mounting component (8) includes The truss (81) comprises a movable block (82) arranged at the top of the truss (81), a movable bolt (83) arranged in the movable block (82), and a positioning block (84) arranged outside the movable bolt (83), a rotating shaft (85) is fixed between the positioning blocks (84), two groups of sleeves (87) are arranged outside the rotating shaft (85), a mounting frame (86) is fixed to the outside of each group of sleeves (87), and a positioning bolt (88) is penetrated through one side of the sleeve (87), and a positioning hole (862) is arranged on one side of the mounting frame (86) on both sides, and the base (1) is fixedly connected to the positioning hole (862) by a locking bolt; A ball head (841) is fixed on the inner side of the positioning block (84), and the ball head (841) and the positioning block (84) are an integrated structure, and a spherical space matching the size of the ball head (841) is provided on both sides of the rotating shaft (85); The side walls of the mounting frames (86) on both sides are provided with sliding grooves (861), one side of the sleeve (87) is fixedly connected to the triangular plate, and the two sides of the triangular plate are arranged in the sliding grooves (861) via positioning bolts (88).
2. The embedded face recognition device according to claim 1, characterized in that: One side of the base (1) is provided with a charging port (15), and the other side is provided with multiple groups of cable interfaces.
3. The embedded face recognition device according to claim 1, characterized in that: The back of the device body (3) is provided with a power amplifier device (11), a switch button (9) and a volume adjustment button (10), and the hinge seat (14) and the pin shaft (13) are respectively fixedly connected to the device body (3) and the base (1) via a connecting piece (12).
4. The embedded face recognition device according to claim 1, characterized in that: A magnetic card sensing component (2) is arranged on the outer side of the base (1).
5. The embedded face recognition device according to claim 1, characterized in that: A switch button (9) is arranged on one side of the back of the device body (3), and a volume adjustment button (10) is arranged on the other side, wherein the switch button (9) is arranged vertically and the volume adjustment button (10) is arranged horizontally.
6. The embedded face recognition device according to claim 1, characterized in that: The identification unit (5), the comparison unit (6) and the illumination unit (7) are symmetrically arranged in two groups at the top of the front side of the device body (3).
7. An embedded face recognition method, characterized in that: An embedded face recognition device comprising any one of claims 1 to 6, Including, S1: preprocessing the existing multiple face image data and obtaining training images; S2: building a convolutional neural network model; S3: Input the preprocessed training image into the convolutional neural network model in S2; S4: Calculate the angle error and its standard deviation; S5: modify parameters according to target error; S6: detect key position points of the face; S7: Process the image to be processed according to the key position points in S6; S8: Use the circular frame in the network model and the dark side of the image to highlight the key positions of the face; S9: Extract the feature vector of the feature point through the obtained key position features; S10: Determine the accuracy of the image based on the feature vector in S9.
8. The embedded face recognition method according to claim 7, characterized in that: The facial image data in step S1 are pre-processed training images with uniform sizes and corresponding to the existing facial images respectively.
