A smart access control management device for smart computer rooms

Abstract

The present invention discloses a smart access control management device for a smart computer room, which relates to the technical field of computer room access control, and includes a body arranged on the wall of the smart computer room access control, a face acquisition unit arranged on the front of the body, and the face acquisition unit includes a target image acquisition module, a feature extraction module and a feature comparison module; the body is also provided with a face tracker and a rotation follow-up structure; the present invention accelerates the efficiency of face recognition by arranging the face tracker and the rotation follow-up structure on the access control device. At the same time, by interfering and generating the face image of the preset object by the generative adversarial network, an interference image after interference is obtained, and the preset object can be accurately recognized regardless of whether the preset object has a complete frontal face image exposed or is not exposed and is affected by interference factors, thereby optimizing the effect of face recognition.

Description

Technical Field

[0001] The present invention relates to the technical field of computer room access control, and in particular to intelligent access control management equipment for a smart computer room. Background Art

[0002] With the development of IT technology and information technology, the number of IT infrastructure in data center computer rooms is also increasing day by day, and at the same time, the supporting environmental equipment is also increasing. At present, the data center computer room has become an important hub for data operation. Once the computer room environmental equipment fails, it will affect the operation of the facilities in the computer room. If the fault is not eliminated as soon as possible, it will cause unplanned downtime of the facilities in the computer room, and even cause damage to the hardware equipment.

[0003] Computer room security is crucial to computer room management, so access control equipment needs to be installed in the computer room. It is used to perform facial recognition on staff entering the computer room. Only after successful recognition can they enter the computer room. Existing access control equipment is generally fixed on the wall at the entrance of the computer room. However, due to the different heights of different staff members, no matter where the facial recognition of the staff members is fixed, some employees will need to adjust the recognition angle during face recognition, which takes a certain amount of time.

[0004] In addition, the existing access control management equipment may not capture unobstructed frontal facial images of staff members. For example, the user may wear glasses, or the face is not in the center of the camera or only half of the face is exposed, or there is light and shadow blocking the face. At this time, the traditional face recognition rate in these scenarios will be very poor, and effective face recognition cannot be completed, which affects the staff's entry into the computer room. Summary of the invention

[0005] In order to solve the problems of the prior art, the present invention provides an intelligent access control management device for a smart computer room, comprising a body arranged on the wall at the access control of the smart computer room, a face acquisition unit is arranged on the front of the body, and the face acquisition unit comprises a target image acquisition module, a feature extraction module and a feature comparison module;

[0006] The target image acquisition module is used to acquire the target face image of the target object entering the smart computer room;

[0007] The feature extraction module is used to extract features from the target face image to obtain target features; the feature comparison module is used to compare the target features with features in the face recognition matching library, and determine that the target object is a preset object according to the comparison result;

[0008] The body is also provided with a facial tracker and a rotation following structure. The facial tracker is used to drive the rotation following structure. The rotation following structure consists of a carrier and a driving part. The driving part is arranged on the carrier. The carrier is arranged between the body and the wall. The driving part is used to drive the body to rotate through the carrier.

[0009] A further solution is that the face recognition matching library is constructed by a face recognition matching library construction unit, and the face recognition matching library construction unit includes an image acquisition module, an image generation module and a feature construction module;

[0010] The image acquisition module is used to acquire a facial image of a preset object; the image generation module is used to input the facial image into a generative adversarial network, and the generative adversarial network generates interference on the facial image to obtain at least one interference image; the feature construction module is used to construct a facial recognition matching library based on the image features of at least one interference image and facial image corresponding to the preset object, and the facial recognition matching library is used for facial recognition of the preset object.

[0011] A further solution is that the generative adversarial network includes a generator and a discriminator, the face image is input into the generative adversarial network, and the generative adversarial network interferes with the face image to generate at least one interference image, including:

[0012] Inputting interference noise into the generator to obtain at least one interference image to be determined output by the generator;

[0013] Inputting the interference image to be judged into the discriminator to obtain a first discrimination result output by the discriminator, wherein the first discrimination result is used to indicate the probability that the interference image to be judged is an image of a real human face;

[0014] Inputting the face image into the discriminator to obtain a second discrimination result output by the discriminator, wherein the second discrimination result is used to indicate the probability that the face image is a real face image;

[0015] According to the first discrimination result, calculating the generator network loss of the generator, and adjusting the network parameters of the generator according to the generator network loss;

[0016] Calculating a discriminator network loss of the discriminator according to the first discrimination result and the second discrimination result, and adjusting a network parameter of the discriminator according to the discriminator network loss;

[0017] When a preset training end condition is reached, at least one interference image output by the generator is obtained.

[0018] A further solution is that constructing a face recognition matching library according to the image features of at least one interference image and a face image corresponding to the preset object includes:

[0019] respectively extracting image features of the at least one interference image to obtain at least one first feature;

[0020] Extracting image features of the face image to obtain a second feature;

[0021] A face recognition matching library is constructed according to the matching relationship between the at least one first feature and the second feature and the preset object.

[0022] A further solution is that constructing a face recognition matching library according to the image features of at least one interference image and a face image corresponding to the preset object includes:

[0023] Using a convolutional neural network, extracting features from at least one interference image and a face image to obtain fusion features;

[0024] According to the matching relationship between the fusion feature and the preset object, a face recognition matching library is constructed.

[0025] A further solution is that the interference image includes at least one of the following interference factors: the preset object wears glasses; the preset object's side face; the preset object's face has light and shadow.

[0026] A further solution is that the face acquisition unit includes a camera, the body is also provided with a display screen, and an LED fill light is embedded on the body above the camera.

[0027] A further solution is that the carrier includes a transmission body and a U-shaped seat, the transmission body is integrally connected to the back of the body, the U-shaped seat is rotatably connected to one end of the transmission body, and a mounting plate is integrally connected to the U-shaped seat.

[0028] A further solution is that the driving part includes a motor and a gear set, the motor is installed in the U-shaped seat, the gear set is installed outside the U-shaped seat and is respectively connected to the transmission body and the motor, and the motor is electrically connected to the facial tracker.

[0029] A further solution is that the gear set is wrapped with a baffle, the motor is covered with a baffle, and the baffle and the baffle are both connected to the U-shaped seat.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] The present invention arranges a face tracker and a rotational following structure on the access control device so that when the staff recognizes the face of a person, the face tracker drives the rotational following structure to drive the camera of the body to actively adjust and aim at the face of the staff, thereby avoiding passive adjustment of the staff when the person is recognized and captured by the camera, and accelerating the efficiency of face recognition.

[0032] The present invention generates interference on the face image of a preset object by a generative adversarial network to obtain an interference image after interference, thereby constructing a face recognition matching library according to the image features of the interference image and the face image, so that when the face recognition matching library is used for face recognition, it can accurately recognize the face of the preset object regardless of whether the preset object reveals a complete frontal face image or does not reveal a complete frontal face but is affected by interference factors, thereby optimizing the face recognition effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 A schematic diagram of the structure of an intelligent access control management device for a smart computer room provided by an embodiment of the present invention;

[0034] Figure 2 A structural schematic diagram of a face acquisition unit provided in an embodiment of the present invention;

[0035] Figure 3 A schematic diagram of the structure of a face recognition matching library construction unit provided in an embodiment of the present invention;

[0036] Figure 4 A schematic diagram of the body and the rotation follower structure provided by an embodiment of the present invention;

[0037] Figure 5 A schematic structural diagram of a rotation follower structure provided by an embodiment of the present invention;

[0038] The attached drawings are marked as follows: 1-body; 10-camera; 11-display screen; 12-LED fill light; 2-face tracker; 30-transmission body; 31-U-shaped seat; 32-mounting plate; 40-motor; 50-shield; 51-baffle. DETAILED DESCRIPTION

[0039] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0040] like Figure 1-5 As shown, one embodiment of the present invention discloses an intelligent access control management device for a smart computer room, including a body 1 arranged on the wall of the access control area of ​​the smart computer room, and a face collection unit is arranged on the front of the body 1.

[0041] The body 1 is also provided with a face tracker 2 and a rotation follower structure. The face tracker 2 is used to drive the rotation follower structure. The rotation follower structure is composed of a carrier and a driving part. The driving part is arranged on the carrier. The carrier is arranged between the body 1 and the wall. The driving part is used to drive the body 1 to rotate through the carrier. The face acquisition unit includes a camera 10. The body 1 is also provided with a display screen 11. The face tracker is a sensor component that can sense the face of a human body and output a control signal to drive the rotation follower structure to drive the body to swing up and down until the camera on the body can record the face of the staff and display it on the display screen.

[0042] The embodiment of the present invention can realize, through the above-mentioned setting, that when it is necessary to enter the smart machine room, the staff's face is facing the machine body, the camera takes a picture of the staff's face and displays it on the display screen, and performs facial recognition on the staff's face.

[0043] In this embodiment, an LED fill light 12 is embedded on the body 1 above the camera 10 .

[0044] The embodiment of the present invention can realize the use of LED fill light to clearly illuminate the facial state of the staff through the above-mentioned setting when the lighting conditions are insufficient, so that the camera can more clearly record the details of the staff's face, which helps to speed up facial recognition.

[0045] In this embodiment, the carrier includes a transmission body 30 and a U-shaped seat 31 . The transmission body 30 is integrally connected to the back of the body 1 . The U-shaped seat 31 is rotatably connected to one end of the transmission body 30 . A mounting plate 32 is integrally connected to the U-shaped seat 31 .

[0046] The embodiment of the present invention can achieve the door access control device of this embodiment to be fixed on a wall at a specific height through the above arrangement, that is, connected to the wall by means of the mounting plate on the U-shaped seat.

[0047] Because the carrier is composed of a transmission body and a U-shaped seat that are rotatably connected to each other, the orientation angle of the body on the wall can be adjusted, and the access control device automatically completes the adjustment of the up and down angles by receiving the instructions of the facial tracker through the driving part.

[0048] Of course, the driving part structure can also be removed to make a damping rotation connection between the transmission body and the U-shaped seat, so that the orientation angle of the body can be manually adjusted according to usage requirements.

[0049] In this embodiment, the driving part includes a motor 40 and a gear set. The motor 40 is installed in the U-shaped seat 31, and the gear set is installed outside the U-shaped seat 31 and connected to the transmission body 30 and the motor 40 respectively. The motor 40 is electrically connected to the face tracker 2.

[0050] The embodiment of the present invention can realize the operation of the motor started by the face tracker through the above arrangement, so that the motor drives the gear set to rotate, and then drives the transmission body to rotate up and down in the U-shaped seat through the gear set. Since the transmission body is integrally connected with the machine body, the camera on the machine body can actively find the face of the staff member, and then perform facial recognition and comparison work, thereby avoiding passive adjustment when the face of the staff member is captured, and can further accelerate the facial recognition efficiency of the face.

[0051] In this embodiment, a shield 50 is wrapped around the gear set, and a baffle 51 is shielded on the motor 40 . Both the shield 50 and the baffle 51 are connected to the U-shaped seat 31 .

[0052] The embodiment of the present invention can be used to shield and protect the driving part by providing a shield and a baffle.

[0053] The face acquisition unit includes a target image acquisition module, a feature extraction module and a feature comparison module;

[0054] The target image acquisition module is used to acquire the target face image of the target object entering the smart computer room;

[0055] The feature extraction module is used to extract features from the target face image to obtain target features; the feature comparison module is used to compare the target features with the features in the face recognition matching library, and determine that the target object is a preset object based on the comparison results;

[0056] In this embodiment, the face recognition matching library is constructed by a face recognition matching library construction unit, and the face recognition matching library construction unit includes an image acquisition module, an image generation module, and a feature construction module;

[0057] The embodiment of the present invention can realize interference generation of the face image of the preset object by the generative adversarial network through the above-mentioned setting, and obtain the interference image after interference, so as to construct a face recognition matching library according to the image features of the interference image and the face image, so that when the face recognition matching library is used for face recognition, it can accurately recognize the face of the preset object regardless of whether the preset object reveals a complete frontal face image or does not reveal a complete frontal face but is affected by interference factors, thereby optimizing the face recognition effect.

[0058] The image acquisition module is used to acquire the face image of a preset object; the image generation module is used to input the face image into a generative adversarial network, and the generative adversarial network interferes with the face image to generate at least one interference image; the feature construction module is used to construct a face recognition matching library based on the image features of at least one interference image and the face image corresponding to the preset object, and the face recognition matching library is used for face recognition of the preset object.

[0059] Among them, Generative Adversarial Networks (GAN) is a deep learning model framework that generates output through the mutual game learning of at least two modules in the framework: the generative model and the discriminative model.

[0060] The interference image is an image with interference factors compared to the original face image. Exemplarily, the interference image includes at least one of the following interference factors: the preset object wears glasses; the preset object's side face; the preset object's face has light and shadow. Each interference image in at least one interference image may correspond to a different interference factor, or may correspond to multiple interference factors. For example, when the interference factor is the preset object wearing glasses, the interference image is an image of the face of the preset object wearing glasses. In addition, the interference image may also be an image of the face of the preset object after makeup, an image of the face of the preset object after tanning, an image of the face of the preset object making different expressions, etc.

[0061] Therefore, the face image is input into the discriminative model of the generative adversarial network, and interference input control is applied to the generative model. Then, the generative adversarial network is iteratively trained. After the training is completed, at least one interference image output by the generative model is obtained.

[0062] For example, during the iterative training process, the generative model adds specified interference input controls such as wearing glasses, side faces, and light and shadows on the face. The generative model tries to disguise and generate interference faces close to the real ones, while the discriminative model is used to learn the corresponding facial features in the face image. The discriminative model will discriminate the interference faces generated by the generative model, and the discrimination results will be passed back to the generator, and then the network parameters of the generative model and the discriminative model will be updated. After multiple rounds of mutual game, the image generated by the generative model will be very realistic, so realistic that even after adjusting its own network parameters, the discriminative model has no way to predict whether the image of the generative model is true or false. At this time, it is believed that the generative adversarial network has reached a certain equilibrium point of the binary minimax game. At this time, the image generated by the generative model can be considered to have generated interference images in a series of scenes such as real users wearing glasses, side faces, and light and shadows on the face, which are used for the next step of feature extraction.

[0063] In this embodiment, the generative adversarial network includes a generator and a discriminator. A face image is input into the generative adversarial network, and the generative adversarial network interferes with the face image to generate at least one interference image, including:

[0064] Inputting interference noise into a generator to obtain at least one interference image to be determined output by the generator;

[0065] Inputting the interference image to be judged into the discriminator, obtaining a first discrimination result output by the discriminator, wherein the first discrimination result is used to indicate the probability that the interference image to be judged is an image of a real face;

[0066] Inputting the face image into the discriminator to obtain a second discrimination result output by the discriminator, wherein the second discrimination result is used to indicate the probability that the face image is a real face image;

[0067] According to the first discrimination result, calculating the generator network loss of the generator, and adjusting the network parameters of the generator according to the generator network loss;

[0068] Calculating a discriminator network loss of the discriminator according to the first discrimination result and the second discrimination result, and adjusting a network parameter of the discriminator according to the discriminator network loss;

[0069] When a preset training end condition is reached, at least one interference image output by the generator is obtained.

[0070] The interference noise is noise with input control, for example, adding specified input controls such as wearing glasses, side faces, and faces with light and shadows, so that the generator can disguise and generate interference images with the above interference factors as close to the real ones as possible, and the interference images to be judged include faces with interference factors. In other examples, the generated data of the generator can be calibrated or human intervention can be added to allow the generator's generation strategy to slowly shift to the data set with interference factors, and the data set includes face images.

[0071] In this embodiment, a face recognition matching library is constructed based on the image features of at least one interference image and a face image corresponding to a preset object, including:

[0072] respectively extracting image features of at least one interference image to obtain at least one first feature;

[0073] Extracting image features of the face image to obtain a second feature;

[0074] A face recognition matching library is constructed according to a matching relationship between at least one first feature and a second feature and a preset object.

[0075] In this embodiment, a face recognition matching library is constructed based on the image features of at least one interference image and a face image corresponding to a preset object, including:

[0076] Using a convolutional neural network, extracting features from at least one interference image and a face image to obtain fusion features;

[0077] Based on the matching relationship between the fusion features and the preset objects, a face recognition matching library is constructed.

[0078] In this embodiment, the interference image includes at least one of the following interference factors: the preset object wears glasses; the preset object has a side face; and there are light shadows on the face of the preset object.

[0079] Finally, it should be noted that the above only describes in detail the specific embodiments of the present invention. However, the present invention is not limited to the specific embodiments described above. Equivalent modifications and substitutions made by those skilled in the art to the present invention are also within the scope of the present invention. Therefore, equal changes and modifications made without departing from the spirit and scope of the present invention are all within the scope of the present invention.

Claims

1. A smart access control management device for a smart computer room, comprising a body (1) arranged on the wall of the access control area of ​​the smart computer room, characterized in that: A face acquisition unit is arranged on the front of the body (1), and the face acquisition unit comprises a target image acquisition module, a feature extraction module and a feature comparison module; The target image acquisition module is used to acquire the target face image of the target object entering the smart computer room; the feature extraction module is used to extract features from the target face image to obtain target features; the feature comparison module is used to compare the target features with the features in the face recognition matching library, and determine that the target object is a preset object according to the comparison result; The machine body (1) is also provided with a face tracker (2) and a rotation following structure. The face tracker (2) is used to drive the rotation following structure. The rotation following structure is composed of a carrier and a driving part. The driving part is arranged on the carrier. The carrier is arranged between the machine body (1) and the wall. The driving part is used to drive the machine body (1) to rotate through the carrier.

2. According to claim 1, the intelligent access control management device for a smart computer room is characterized by: The face recognition matching library is constructed by a face recognition matching library construction unit, and the face recognition matching library construction unit includes an image acquisition module, an image generation module and a feature construction module; The image acquisition module is used to acquire a face image of a preset object; the image generation module is used to input the face image into a generative adversarial network, and the generative adversarial network performs interference generation on the face image to obtain at least one interference image; The feature construction module is used to construct a face recognition matching library according to the image features of at least one interference image and a face image corresponding to the preset object, and the face recognition matching library is used for face recognition of the preset object.

3. According to claim 1, the intelligent access control management device for a smart computer room is characterized by: The generative adversarial network includes a generator and a discriminator. The face image is input into the generative adversarial network, and the generative adversarial network interferes with the face image to generate at least one interference image, including: Inputting interference noise into the generator to obtain at least one interference image to be determined output by the generator; Inputting the interference image to be judged into the discriminator to obtain a first discrimination result output by the discriminator, wherein the first discrimination result is used to indicate the probability that the interference image to be judged is an image of a real human face; Inputting the face image into the discriminator to obtain a second discrimination result output by the discriminator, wherein the second discrimination result is used to indicate the probability that the face image is a real face image; According to the first discrimination result, calculating the generator network loss of the generator, and adjusting the network parameters of the generator according to the generator network loss; Calculating a discriminator network loss of the discriminator according to the first discrimination result and the second discrimination result, and adjusting a network parameter of the discriminator according to the discriminator network loss; When a preset training end condition is reached, at least one interference image output by the generator is obtained.

4. The intelligent access control management device for a smart computer room according to claim 1 is characterized in that: The step of constructing a face recognition matching library according to the image features of at least one interference image and a face image corresponding to the preset object comprises: respectively extracting image features of the at least one interference image to obtain at least one first feature; Extracting image features of the face image to obtain a second feature; A face recognition matching library is constructed according to the matching relationship between the at least one first feature and the second feature and the preset object.

5. According to claim 1, the intelligent access control management device for a smart computer room is characterized by: The step of constructing a face recognition matching library according to the image features of at least one interference image and a face image corresponding to the preset object comprises: Using a convolutional neural network, extracting features from at least one interference image and a face image to obtain fusion features; According to the matching relationship between the fusion feature and the preset object, a face recognition matching library is constructed.

6. The intelligent access control management device for a smart computer room according to claim 1 is characterized by: The interference image includes at least one of the following interference factors: the preset object wears glasses; the preset object has a side face; and the preset object has light and shadow on its face.

7. The intelligent access control management device for a smart computer room according to claim 1 is characterized by: The face acquisition unit comprises a camera (10); the body (1) is further provided with a display screen (11); and an LED fill light (12) is embedded on the body (1) above the camera (10).

8. The intelligent access control management device for a smart computer room according to claim 1 is characterized by: The carrier comprises a transmission body (30) and a U-shaped seat (31); the transmission body (30) is integrally connected to the back of the machine body (1); the U-shaped seat (31) is rotatably connected to one end of the transmission body (30); and a mounting plate (32) is integrally connected to the U-shaped seat (31).

9. The intelligent access control management device for a smart computer room according to claim 1 is characterized by: The driving part comprises a motor (40) and a gear set, wherein the motor (40) is installed in the U-shaped seat (31), the gear set is installed outside the U-shaped seat (31) and is respectively connected to the transmission body (30) and the motor (40), and the motor (40) is electrically connected to the facial tracker (2).

10. The intelligent access control management device for a smart computer room according to claim 1 is characterized by: The gear set is wrapped with a baffle (50), the motor (40) is shielded with a baffle (51), and both the baffle (50) and the baffle (51) are connected to the U-shaped seat (31).

Images