Face identity verification security automated detection system and method

Abstract

The present application provides a kind of face identity check security automation detection system, it includes controller, image acquisition equipment and equipment mechanical arm;Image acquisition equipment is loaded on equipment mechanical arm, it is set to collect image, and response to face recognition trigger instruction, carries out face recognition;Equipment mechanical arm response to relative pose adjustment instruction, move to adjust the relative pose of image acquisition equipment and face recognition attack material to corresponding relative pose point;Controller is set to each relative pose point, generates the relative pose adjustment instruction adjusted to this relative pose point, generates face recognition trigger instruction to obtain face recognition result when adjusting to this relative pose point, to detect the attack resistance of image acquisition equipment.The present application also provides corresponding method.The face identity check security automation detection system of the present application moves the equipment to be measured by mechanical arm, relative pose is automatically variable in space, and realizes all-around automation.

Description

Technical Field

[0001] This invention belongs to the field of data security, specifically relating to an automated detection method and system for facial recognition security. Background Technology

[0002] The security of facial recognition is a perennial issue in the industry. Currently, the industry generally uses manual testing methods to simulate presentation attacks in order to test the security of facial recognition. Among them, presentation attacks refer to an attacker using technical means or other media (images / videos, etc.) to present the image to the device's camera in order to deceive the user. This is also a mainstream testing method to detect whether a device has the ability (size) to determine the authenticity of a face.

[0003] Manual bio-nuclei often face many challenges, such as long testing cycles, high labor costs, low testing accuracy, and difficulty in reproducing problems.

[0004] Specifically, in traditional facial recognition security testing, a person typically holds a facial recognition attack device and presents it to the device to see if it can successfully pass facial recognition verification. During manual testing, although the person holding the attack device may present it from different angles and distances, instability or shaking can cause what should have been a successful attack to fail. Furthermore, data recording during manual testing is asynchronous, introducing subjective human factors and increasing the risk of incorrect data recording and analytical bias.

[0005] Therefore, in order to achieve automated detection of multiple face IDs under multiple angles, lighting conditions, and multiple lighting conditions, and to save on manual testing cycles, an automated detection system for face verification and biometric security is needed. Summary of the Invention

[0006] The purpose of this application is to build an automated facial recognition security detection system that combines security, robotics, and vision technologies to achieve automation.

[0007] To achieve the above objectives, the present invention provides an automated face verification security detection system, which includes a controller, a robotic arm, and an image acquisition device to be tested.

[0008] The image acquisition device is mounted on the robotic arm of the device and is configured to: acquire images; and perform face recognition in response to a face recognition trigger command issued by the controller, and obtain a face recognition result;

[0009] The device mechanical arm is configured to, in response to a relative pose adjustment instruction issued by the controller, move to adjust the relative pose between the image acquisition device and the face recognition attack material to a corresponding relative pose point.

[0010] The controller is configured to, for each relative pose point between the image acquisition device and the face recognition attack material, generate a relative pose adjustment instruction for adjusting to the relative pose point and send the instruction to the device mechanical arm, generate a face recognition trigger instruction and send the instruction to the image acquisition device when the relative pose between the image acquisition device and the face recognition attack material is adjusted to the relative pose point, obtain a face recognition result at the relative pose point, and detect the attack resistance of the image acquisition device according to the face recognition results at the relative pose points.

[0011] Preferably, the number of at least one of the image acquisition device and the face recognition attack material is multiple; and

[0012] The controller is further configured to select the current image acquisition device and the current face recognition attack material according to a preset test sequence, so that the current image acquisition device is loaded on the device mechanical arm.

[0013] Preferably, the image acquisition device includes multiple devices of the same model, and for multiple image acquisition devices of the same model, each image acquisition device only records the face of one tested person;

[0014] The controller is further configured to, after the test is completed, for image acquisition devices of the same model, statistically record test data of each face recognition attack material, each image acquisition device and each relative pose, to obtain statistical results of the image acquisition device as a detection result of the attack resistance of the image acquisition device.

[0015] Preferably, the relative pose points are obtained by: pre-setting multiple relative pose points, selecting the current image acquisition device, taking the face of one person object as a reference, sequentially setting all face recognition attack materials corresponding to the current image acquisition device as the current face recognition attack material, and each time testing according to the pre-set multiple relative pose points; statistically recording test data of all relative pose points to obtain vulnerable points, and taking the vulnerable points as actual relative pose points.

[0016] Preferably, the number of image acquisition devices is multiple, each image acquisition device is placed in a device placement area when not loaded, and each image acquisition device corresponds to a unique device identification label and a unique device id.

[0017] The controller is further configured to, after selecting the current image acquisition device, if the current image acquisition device has not been loaded on the device mechanical arm, send a device loading instruction with information of the current image acquisition device to the device mechanical arm, the information of the image acquisition device including a device ID of the image acquisition device and a device placement area of the image acquisition device.

[0018] The device mechanical arm is provided with a carrier device and a camera, and is further configured to, in response to the device loading instruction, move the device mechanical arm so that the camera thereof is downwardly aligned with the device placement area of the required image acquisition device; acquire the device identification tag by using the camera to obtain the device ID by the controller; when the obtained device ID is the device ID of the required image acquisition device, move according to the position of the device identification tag so that the carrier device thereof is aligned with and contacts the image acquisition device; and drive the carrier device of the device mechanical arm to load the current image acquisition device on the device mechanical arm.

[0019] Preferably, the controller is further configured to, after the selection of the current image acquisition device and the current face recognition attack material is completed and the current image acquisition device is loaded on the device mechanical arm, generate a test preparation instruction before generating the relative pose adjustment instruction.

[0020] The device mechanical arm is further configured to, in response to the test preparation instruction, move so that the image acquisition device is located at an initial test pose facing the face recognition attack material.

[0021] Preferably, the device mechanical arm moves to adjust the relative pose between the image acquisition device and the face recognition attack material to the corresponding relative pose point, specifically including: based on the initial test pose, adjusting the relative angle between the image acquisition device and the face recognition attack material by making the image acquisition device perform spherical motion around the face of the face recognition attack material, and adjusting the relative distance between the image acquisition device and the face recognition attack material by performing linear motion on the line connecting the image acquisition device and the face recognition attack material.

[0022] Preferably, the spherical motion around the face of the face recognition attack material specifically includes: establishing a meridian circle and a parallel circle with the face recognition attack material as the center, moving the image acquisition device to the intersection of the meridian circle and the parallel circle of the face recognition attack material, and keeping the image acquisition device in a posture of directly facing the face recognition attack material.

[0023] Preferably, the controller is further configured to, after selecting the current face recognition attack material, determine whether the current face recognition attack material is a mechanical arm presentation material before generating the relative pose adjustment instruction.

[0024] If so, after selecting the current facial recognition attack material, before generating the relative pose adjustment instruction, the current robot arm is caused to present the material loading on the material robot arm, and after generating the relative pose adjustment instruction, the generated relative pose adjustment instruction is sent to the material robot arm; otherwise, after generating the relative pose adjustment instruction, the generated relative pose adjustment instruction is sent to the device robot arm.

[0025] The material robot arm is configured to move to adjust the relative pose between the image acquisition device and the facial recognition attack material to the relative pose point in response to the relative pose adjustment instruction.

[0026] Preferably, each robot arm presentation material is placed in a material placement area corresponding to the specific type of the robot arm presentation material when the robot arm presentation material is not loaded, and a material identification tag is attached to each robot arm presentation material, each robot arm presentation material corresponding to a unique material identification tag and a unique material id.

[0027] If the current facial recognition attack material is a robot arm presentation material and has not been loaded on the material robot arm, the current robot arm presentation material is caused to be loaded on the material robot arm, specifically including: sending a material loading instruction with information of the current facial recognition attack material to the material robot arm to cause the robot arm presentation material to be loaded on the material robot arm, the information of the facial recognition attack material including a material id and a material placement area.

[0028] The material robot arm is provided with a material loading device and a camera, and the material robot arm is configured to move to align the camera downward to the material placement area corresponding to the required robot arm presentation material in response to the material loading instruction; obtain the material identification tag with the camera to obtain the material id by the controller; when the obtained material id is the material id of the required robot arm presentation material, move according to the position of the material identification tag to align and contact the robot arm presentation material with the material loading device; drive the material loading device of the material robot arm to load the facial recognition attack material on the material robot arm.

[0029] Preferably, robot arm presentation materials of the same specific type are placed in the same material placement area in a stacked manner, and the material placement area has at least two stacking positions.

[0030] The camera aligns downward to the material placement area corresponding to the required robot arm presentation material, specifically including: the camera aligns downward to one of the stacking positions of the material placement area as the current inspection stacking position.

[0031] The material robot arm is further configured to, after obtaining the material id,

[0032] If the obtained material id is not the material id of the required robotic arm presenting material, move according to the position of the material identification tag to align and contact the material presenting device of the robotic arm with the material presenting material;

[0033] Drive the material presenting device of the robotic arm to load the face recognition attack material on the robotic arm;

[0034] Move to another stacking position other than the current inspection stacking position;

[0035] Drive the material presenting device of the robotic arm to unload the face recognition attack material;

[0036] Move to align the downward camera of the current inspection stacking position, repeat the steps of obtaining the material id and loading and unloading the face recognition attack material until the obtained material id is the material id of the required robotic arm presenting material.

[0037] Preferably, the material presenting device is a suction pump.

[0038] Preferably, the controller is further configured to, for each relative pose point, after adjusting to the relative pose point, adjust the lighting conditions to generate face recognition trigger instructions under different lighting conditions respectively, and obtain face recognition results under multiple lighting conditions of the relative pose point, wherein the lighting system includes light position, illumination and color temperature.

[0039] In another aspect, the present application provides a face recognition security automation detection method, comprising:

[0040] Providing an image acquisition device loaded on a device robotic arm; for each relative pose point between the image acquisition device and the face recognition attack material, generating a relative pose adjustment instruction adjusting to the relative pose point by a controller and sending it to the device robotic arm or the material robotic arm loaded with the face recognition attack material;

[0041] In response to the relative pose adjustment instruction, moving by the device robotic arm to adjust the relative pose between the image acquisition device and the face recognition attack material to the relative pose point, or moving by the material robotic arm to adjust the relative pose between the image acquisition device and the face recognition attack material to the relative pose point;

[0042] When adjusting to the relative pose point, generating a face recognition trigger instruction by the controller;

[0043] In response to the face recognition trigger instruction, performing face recognition by the image acquisition device to be tested;

[0044] The controller obtains the face recognition result of the relative pose point, and detects the attack resistance of the image acquisition device according to the face recognition result of each relative pose point.

[0045] Preferably, the number of at least one of the image acquisition device and the face recognition attack material is multiple;

[0046] And the image acquisition device loaded on the device mechanical arm is provided, and specifically comprises: selecting the current image acquisition device and the current face recognition attack material according to the preset test sequence, so that the current image acquisition device is loaded on the device mechanical arm.

[0047] Preferably, the number of the image acquisition device is multiple, each image acquisition device is placed in a device placement area when not loaded, and each image acquisition device corresponds to a unique device identification label and a unique device id; the device mechanical arm is provided with a loading device and a camera;

[0048] After selecting the current image acquisition device, if the current image acquisition device has not been loaded on the device mechanical arm, the current image acquisition device is loaded on the device mechanical arm, specifically including:

[0049] The controller sends a device loading instruction with the information of the current image acquisition device to the device mechanical arm, and the information of the image acquisition device includes the device id and the device placement area of the image acquisition device;

[0050] The device mechanical arm responds to the device loading instruction and moves so that its camera is downwardly aligned with the device placement area of the required image acquisition device; the device identification label is acquired by the camera to obtain the device id by the controller; when the obtained device id is the device id of the required image acquisition device, the device identification label is moved to align and contact the image acquisition device; the loading device of the device mechanical arm is driven to load the current image acquisition device on the device mechanical arm.

[0051] Preferably, after selecting the current face recognition attack material, before generating the relative pose adjustment instruction, the controller further judges whether the current face recognition attack material is a mechanical arm presentation material;

[0052] If yes, after selecting the current face recognition attack material, before generating the relative pose adjustment instruction, the current mechanical arm presentation material is loaded on the material mechanical arm, and after generating the relative pose adjustment instruction, the controller sends the generated relative pose adjustment instruction to the material mechanical arm, and the material mechanical arm responds to the relative pose adjustment instruction to adjust the relative pose between the image acquisition device and the face recognition attack material to the relative pose point.

[0053] Otherwise, after generating the relative pose adjustment instruction, the generated relative pose adjustment instruction is sent to the device mechanical arm by the controller, and the device mechanical arm moves in response to the relative pose adjustment instruction to adjust the relative pose between the image acquisition device and the facial recognition attack material to the relative pose point.

[0054] Preferably, each mechanical arm presentation material is placed in the material placement area corresponding to the specific type of the mechanical arm presentation material when the mechanical arm presentation material is not loaded, and a material identification tag is attached to each mechanical arm presentation material, each mechanical arm presentation material corresponding to a unique material identification tag and a unique material id; the material mechanical arm is provided with a loading device and a camera;

[0055] If the current facial recognition attack material has not been loaded on the material mechanical arm, the current mechanical arm presentation material is loaded on the material mechanical arm, specifically including:

[0056] The controller sends a material loading instruction with information of the current facial recognition attack material to the material mechanical arm, the information of the facial recognition attack material including the material id and the material placement area;

[0057] The material mechanical arm moves in response to the material loading instruction to align its camera downward to the material placement area corresponding to the required mechanical arm presentation material; acquires the material identification tag with its camera to obtain the material id by the controller; when the obtained material id is the material id of the required mechanical arm presentation material, moves according to the position of the material identification tag to align and contact the mechanical arm presentation material with its loading device; drives its loading device to load the facial recognition attack material on the material mechanical arm.

[0058] Preferably, the same specific type of mechanical arm presentation material is placed in the same material placement area in a stacked manner, and the material placement area has at least two stacking positions;

[0059] The camera aligns downward to the material placement area corresponding to the required mechanical arm presentation material, specifically including: the camera aligns downward to one of the stacking positions of the material placement area, as the current inspection stacking position;

[0060] After obtaining the material id, before the obtained material id is the material id of the required mechanical arm presentation material, further including:

[0061] If the obtained material id is not the material id of the required mechanical arm presentation material, the material mechanical arm moves according to the position of the material identification tag to align and contact the mechanical arm presentation material with its loading device;

[0062] driving its carrier device by the material robot to load the face recognition attack material on the material robot;

[0063] moving by the material robot to another stacking position other than the current inspection stacking position;

[0064] driving its carrier device by the material robot to unload the face recognition attack material;

[0065] moving by the material robot to align its camera downward to the current inspection stacking position, repeating the steps of resolving the material id and loading and unloading the face recognition attack material until the resolved material id is the material id of the desired robot presenting material.

[0066] Preferably, for each relative pose point, after adjusting to the relative pose point, further comprising: adjusting by the controller the lighting conditions to generate face recognition trigger instructions respectively under different lighting conditions, obtaining the face recognition results of the relative pose point under multiple lighting conditions, the lighting system including light position, illuminance and color temperature.

[0067] The face recognition security automatic detection system of the present application moves the device under test by the robot, the relative pose is automatically variable in space, and then attacks the face recognition system of the image acquisition device; at the same time, the attack material can also be moved by the robot. In addition, the face recognition security automatic detection system of the present application can automatically load the device and material, and can automatically control the light source, the mobile phone face recognition function, and automatically obtain the detection results. The face recognition security automatic detection system of the present application comprehensively utilizes the technologies of robot control, light source control, mobile phone automatic control and biological safety detection, realizes all-round automation, solves the problem of automatic detection of multiple lighting scenes, multiple mobile phones and multiple Apps, and records the multi-dimensional factors of the scene data (angle, distance, light, material, etc.) of the detection process, automatically summarizes the attack pass rate, and automatically generates a detection report. BRIEF DESCRIPTION OF DRAWINGS

[0068] Figure 1 is a whole structure diagram of a face recognition security automatic detection system of the present application.

[0069] Figure 2 is a flow chart of a face recognition security automatic detection method of the present application. DETAILED DESCRIPTION

[0070] The present application will be further described below in conjunction with specific embodiments. It should be understood that the following embodiments are only used to illustrate the present application and are not used to limit the scope of the present application.

[0071] The automated facial recognition security testing system of this invention primarily tests various devices with facial authentication functions, such as mobile terminals, combinations of visual sensors and facial recognition algorithms, and various devices with facial recognition algorithm SDKs. This automated facial recognition security testing system mainly serves various testing protocols, such as the Google Biometric Security Protocol, the IIFAA Testing Protocol, and the IBASS Testing Protocol. These protocols specify the threshold range of biometric security detection results that devices must meet, and define the relationship between security levels and the threshold range of biometric security detection results. This invention is used to test the biometric security detection results of various devices, achieving the quantification of attack materials, testing processes, and test results. The test results obtained using this automated facial recognition security testing system can be sent to the sender in the form of biometric security data reports, evaluation reports, and evaluation scores, such as OEM manufacturers, chip manufacturers, algorithm manufacturers, IIFAA alliance members, and third parties authorized by the tester.

[0072] First embodiment: Automated facial recognition security detection system

[0073] like Figure 1 The diagram shows an automated face verification security detection system according to a first embodiment of the present invention, which includes a controller 10, at least one image acquisition device 20 to be tested, and a robotic arm 30, and is used to test the security of the image acquisition device 20 using various face recognition attack materials 60.

[0074] The image acquisition device 20 is configured to: acquire images; and, in response to the face recognition trigger command of the controller 10, perform face recognition and obtain face recognition results.

[0075] The robotic arm 30 is configured to move in response to a relative pose adjustment command to adjust the relative pose (i.e., spatial position and attitude) between the image acquisition device 20 and the face recognition attack material 60 to the corresponding relative pose point.

[0076] The controller 10 is configured to: generate a relative pose adjustment command to adjust to the relative pose point for each relative pose point between the image acquisition device 20 and the face recognition attack material 60 and send it to the device robotic arm or the material robotic arm; when adjusted to the relative pose point, generate a face recognition trigger command to obtain the face recognition result of the relative pose point; and detect the anti-attack capability of the image acquisition device based on the face recognition results of each relative pose point.

[0077] Therefore, by setting up the robotic arm, image shaking and offset caused by manually holding the image acquisition device 20 are avoided.

[0078] In the embodiment, the face verification security automatic detection system further comprises a material mechanical arm 40, which is configured to adjust the relative pose (i.e. spatial position and attitude) between the image acquisition device 20 and the face recognition attack material 60 to the relative pose point in response to the relative pose adjustment instruction. Thus, by the configuration of the mechanical arm, the material shaking and deviation caused by manually holding the face recognition attack material 60 are avoided.

[0079] The device mechanical arm 30 and the material mechanical arm 40 are both provided with a carrying device and a camera. The carrying device of the device mechanical arm 30 is configured to load and unload the image acquisition device 20 in response to a driving signal. The carrying device of the material mechanical arm 40 is configured to load and unload the mechanical arm presentation material in response to a driving signal.

[0080] In the embodiment, the carrying devices of the device mechanical arm 30 and the material mechanical arm 40 are both suction pumps arranged at the top ends of the device mechanical arm 30 and the material mechanical arm 40, so as to load the image acquisition device 20 and the face recognition attack material 60 by suction. In other embodiments, the suction pump can also be a clamping device or other device for loading.

[0081] In the embodiment, the number of at least one of the image acquisition device 20 and the face recognition attack material 60 is multiple. Accordingly, the controller 10 is further configured to: preset a test sequence according to the information of the image acquisition device 20 and the face recognition attack material 60; and according to the preset test sequence, when a device-material switching condition is met, select a current image acquisition device and a current face recognition attack material according to the preset test sequence, so that the current image acquisition device is loaded on the device mechanical arm, and the relative pose between the current image acquisition device and the current face recognition attack material is adjusted to the corresponding relative pose point. The device-material switching condition includes: the test is started, or the face recognition results of each relative pose point between the current image acquisition device and the current face recognition attack material are all obtained and the test is not ended.

[0082] The image acquisition device 20 includes a device with a face verification function, such as a mobile phone, a tablet, etc. In the embodiment, considering that the same device cannot record the faces of multiple test persons, the image acquisition device 20 includes multiple devices of the same model and / or multiple devices of different models, and for multiple image acquisition devices 20 of the same model, each image acquisition device 20 only records the face of one test person. It should be noted that different test protocols have different requirements for the number of test persons. Re-selecting the current image acquisition device 20 means that all face recognition attack materials 60 for testing the face of the next test person can be tested.

[0083] Each image capture device 20 is placed in a device placement area when not loaded.

[0084] The face recognition attack materials 60 include a plurality of face recognition attack materials 60 for biological identity verification. The face recognition attack materials 60 are classified into robot arm presentation materials and non-robot arm presentation materials according to whether they can be presented by the robot arm 40.

[0085] The types of the face recognition attack materials 60 include screen type materials, paper type materials, etc., and can be further subdivided into various specific types. The number of materials of each specific type is a plurality, and each of the materials of the same specific type corresponds to a testee.

[0086] Each robot arm presentation material is placed in a material placement area corresponding to the specific type of the robot arm presentation material when not loaded. In the embodiment, the robot arm presentation materials of the same specific type are placed in the same material placement area in a stacked manner, and the material placement area has at least two stacking positions.

[0087] It should be noted that different image capture devices 20 can correspond to different protocols, and thus the test requirements and the required face recognition attack materials 60 can be different.

[0088] Each image capture device 20 is pasted with a device identification label, and each image capture device 20 corresponds to a unique device identification label and a unique device id. Each face recognition attack material 60 is pasted with a material identification label, and each face recognition attack material 60 corresponds to a unique material identification label and a unique material id. In the embodiment, the device identification label and the material identification label are two-dimensional codes.

[0089] Therefore, the information of the image capture device 20 includes the device id and the device placement area of the image capture device 20. The information of the face recognition attack material 60 includes the material id and the material placement area.

[0090] The current image capture device 20 and the face recognition attack material 60 are selected, specifically including: the controller 10 determines, according to the test sequence, whether all face recognition attack materials 60 have been tested for the current image capture device 20, if not, the current face recognition attack material 60 is reselected, if yes, the current image capture device 20 is reselected when all image capture devices 20 have not been tested, and the test is ended when all image capture devices 20 are tested.

[0091] The controller 10 is further configured to, after ending the test, record test data of each face recognition attack material 60, each image acquisition device 20 and each relative pose of the same model of image acquisition device 20, to obtain statistical results of the image acquisition device 20, as a detection result of the attack resistance of the image acquisition device 20.

[0092] In the embodiment, the statistical results of the image acquisition device 20 correspond to the same model of image acquisition device 20, which include the average SAR of each face recognition attack material 60 obtained by transverse statistics, and the total attack pass rate SAR of all face recognition attack materials 60 obtained by longitudinal statistics.

[0093] The relative pose points are obtained by the following method: a plurality of relative pose points are set in advance, a current image acquisition device 20 is selected, the face of a person object is taken as a reference, all face recognition attack materials 60 corresponding to the current image acquisition device 20 are sequentially set as the current face recognition attack material, and each time the test is performed according to the plurality of relative pose points set in advance; the test data of all relative pose points are counted to obtain vulnerable points, and the vulnerable points are taken as actual relative pose points. Thus, the detected vulnerable points of one person object can be used for the test of multiple person objects.

[0094] For details, refer to the existing artificial face test process. After the material is prepared, the vulnerable point screening is first performed. In this stage, one person is taken as a reference, and a mechanical arm is used to test and screen 54 calibration points of each material of the person. Each calibration point is attacked once, and finally a point set that can be attacked and passed is output. Then, the attack test is performed. In this stage, the points screened in the first stage are respectively rechecked on 10 persons. A mechanical arm is used to restore the attack position according to the material and the relative pose point, and 48 attacks are performed. Finally, the attack pass rate SAR of each material, each relative pose point and each person is output. Finally, the test report is obtained: the average SAR of each material is transversely counted, and the total attack pass rate SAR of 10 persons is longitudinally counted.

[0095] The process that the device mechanical arm 30 loads the image acquisition device 20 is described in detail below.

[0096] The controller 10 is further configured to select a current image acquisition device 20, and if the current image acquisition device has not been loaded on the device mechanical arm, send a device loading instruction with information of the current image acquisition device 20 to the device mechanical arm 30. The current image acquisition device 20 is reselected, which further includes sending a device unloading instruction to the device mechanical arm 30. The information of the image acquisition device 20 includes the device id and the device placement area of the image acquisition device 20.

[0097] Accordingly, the device mechanical arm 30 is configured to: in response to the device loading instruction, move the device mechanical arm 30 so that the camera thereof is downwardly aligned with the device placement area of the required image acquisition device 20; acquire the device identification tag by using the camera to be parsed by the controller 10 to obtain the device id, and when the obtained device id is the device id of the required image acquisition device 20, move according to the position of the device identification tag to align and contact the image acquisition device 20 by the carrying device thereof; and drive the carrying device of the device mechanical arm 30 to load the current image acquisition device 20 on the device mechanical arm 30.

[0098] In the present embodiment, the face recognition attack material 60 is divided into mechanical arm presented material and non-mechanical arm presented material, wherein the mechanical arm presented material needs to be loaded on the material mechanical arm 40, and the non-mechanical arm presented material cannot be loaded on the material mechanical arm 40, but must be fixed in place. Therefore, it is necessary to judge whether the face recognition attack material 60 is the mechanical arm presented material.

[0099] Specifically, the controller is further configured to: after selecting the current face recognition attack material 60, before generating the relative pose adjustment instruction, judge whether the current face recognition attack material 60 is the mechanical arm presented material, if yes, after selecting the current face recognition attack material, before generating the relative pose adjustment instruction, load the mechanical arm presented material on the material mechanical arm 40, and after generating the relative pose adjustment instruction, send the generated relative pose adjustment instruction to the material mechanical arm 40; otherwise, after generating the relative pose adjustment instruction, send the generated relative pose adjustment instruction to the device mechanical arm 30.

[0100] In other embodiments, whether the current face recognition attack material 60 is the mechanical arm presented material can not be distinguished, and all face recognition attack materials 60 are executed according to the test method of the mechanical arm presented material.

[0101] The process of loading the mechanical arm presented material by the material mechanical arm 40 will be described in detail below.

[0102] The loading of the material robot 40 is achieved by the material id described above. Specifically, if the current facial recognition attack material 60 is a robot presented material and has not been loaded on the material robot, the current robot presented material is loaded on the material robot, specifically including: sending a material loading instruction with the information of the current facial recognition attack material to the material robot, so that the robot presented material is loaded on the material robot. When other robot presented materials are loaded on the material robot, it also includes: when the test completed facial recognition attack material 60 is a robot presented material, sending a material unloading instruction with the information of the current facial recognition attack material 60 to the material robot 40. The information of the facial recognition attack material 60 includes the material id and the material placement area.

[0103] The material robot 40 is configured to: in response to the material loading instruction, move so that its camera is downwardly aligned with the material placement area corresponding to the required robot presented material; use the camera to obtain the material identification tag to obtain the material id by the controller 10, and when the obtained material id is the material id of the required robot presented material, move according to the position of the material identification tag to align and contact the robot presented material with the material loading device; drive the material loading device of the material robot 40 to load the facial recognition attack material 60 on the material robot 40.

[0104] In the embodiment, since the same specific kind of robot presented material is placed in a stacked manner in the same material placement area, and the material placement area has at least two stacking positions, the camera downwardly aligning with the material placement area corresponding to the required robot presented material specifically includes: the camera downwardly aligning with one of the stacking positions of the material placement area as the current inspection stacking position; the material robot 40 is also configured to: after obtaining the material id, if the obtained material id is not the material id of the required robot presented material before the obtained material id is the material id of the required robot presented material, move according to the position of the material identification tag to align and contact the robot presented material with the material loading device; drive the material loading device of the material robot 40 to load the facial recognition attack material 60 on the material robot 40; move to another stacking position other than the current inspection stacking position; drive the material loading device of the material robot 40 to unload the facial recognition attack material 60; move so that its camera downwardly aligns with the current inspection stacking position, and repeat the steps of obtaining the material id and loading and unloading the facial recognition attack material 60 until the obtained material id is the material id of the required robot presented material.

[0105] Thus, the test materials stacked together can be inspected in a similar way to searching.

[0106] In summary, the device loading instructions and the device unloading instructions are executed by the device robot arm 30, and the material loading and material unloading instructions are executed by the material robot arm 40. In other embodiments, the device loading instructions and the device unloading instructions and the material loading and material unloading instructions can be displayed by a display screen connected to the controller 10, and the image acquisition device 20 or the facial recognition attack material 60 is loaded and unloaded manually.

[0107] Before the image acquisition device 20 is loaded, the display screen further displays an interface instruction that the test mobile phone is being sucked.

[0108] After the image acquisition device 20 is loaded, the display screen further displays an instruction that the image acquisition device 20 is being carried and the test angle is being adjusted.

[0109] After the facial recognition attack material 60 is loaded, the display screen further displays an instruction that the facial recognition attack material 60 is being carried and the test angle is being adjusted.

[0110] If the face is not located at the center of the image captured by the image acquisition device 20, the face is easily recognized by the facial recognition software, and the face verification security detection is invalid.

[0111] Therefore, in the present embodiment, the controller 10 is further configured to generate a test preparation instruction before generating the relative pose adjustment instruction after the current image acquisition device and the current facial recognition attack material are selected and the current image acquisition device is loaded on the device robot arm. Accordingly, the device robot arm 30 is further configured to move to an initial test pose of the image acquisition device 20 facing the facial recognition attack material 60 in response to the test preparation instruction. Thus, it is ensured that the image acquisition device 20 and the facial recognition attack material 60 are aligned with each other at the initial time after being loaded.

[0112] In the present embodiment, the initial test pose is not only the pose of the image acquisition device 20 facing the facial recognition attack material 60. At the initial test pose, the camera of the image acquisition device 20 and the facial recognition attack material 60 are on the same horizontal plane, and the face of the facial recognition attack material 60 is located at the center of the image captured by the image acquisition device 20.

[0113] Correspondingly, the controller 10 is configured to: when the image acquisition device 20 faces the face recognition attack material 60, receive an image acquired by the image acquisition device 20, determine the position of the face of the face recognition attack material 60 according to the acquired image, and determine the movement parameter of the device mechanical arm 30 as a test preparation instruction according to the position of the face, and the device mechanical arm 30 is configured to move to the initial test pose according to the movement parameter of the device mechanical arm 30.

[0114] The following describes the movement mode of the mechanical arm when switching between the relative pose points.

[0115] For the non-mechanical arm presentation material, the material mechanical arm 40 is stationary, and the device mechanical arm 30 moves. For the mechanical arm presentation material, the material mechanical arm 40 moves, and the device mechanical arm 30 is stationary.

[0116] Specifically, the controller is configured to: if the current face recognition attack material 60 is a mechanical arm presentation material, send the generated relative pose adjustment instruction to the material mechanical arm 40; otherwise, send the generated relative pose adjustment instruction to the device mechanical arm 30; and the material mechanical arm 40 is configured to: in response to the relative pose adjustment instruction, adjust the relative pose between the image acquisition device 20 and the face recognition attack material 60 to the relative pose point.

[0117] The device mechanical arm 30 moves to adjust the relative pose between the image acquisition device 20 and the face recognition attack material 60 to the relative pose point, specifically including: based on the initial test pose, adjusting the relative angle between the image acquisition device 20 and the face recognition attack material 60 by making the image acquisition device 20 perform spherical motion around the face of the face recognition attack material 60, and adjusting the relative distance between the image acquisition device 20 and the face recognition attack material 60 by making the image acquisition device 20 perform linear motion on the line connecting the image acquisition device 20 and the face recognition attack material 60.

[0118] The spherical motion around the face of the face recognition attack material 60 specifically includes: establishing a meridian circle and a parallel circle with the face recognition attack material 60 as the center, and moving the image acquisition device 20 to the intersection of the meridian circle and the parallel circle of the face recognition attack material 60 and keeping the image acquisition device 20 in a posture of directly facing the face recognition attack material 60. This posture, i.e., the plane of the image acquisition device 20 should be located on the tangent plane formed by the meridian circle and the parallel circle. The distance between the meridian circle and the parallel circle is artificially set, and the specific coordinates of the intersection of the meridian circle and the parallel circle can be obtained by calculation.

[0119] The material mechanical arm 40 moves to adjust the relative pose between the image acquisition device 20 and the face recognition attack material 60 to the relative pose point, specifically including: on the basis of the initial test pose, adjusting the relative angle between the image acquisition device 20 and the face recognition attack material 60 by making the face recognition attack material 60 perform spherical motion around the camera of the image acquisition device 20, and adjusting the relative distance between the image acquisition device 20 and the face recognition attack material 60 by performing linear motion on the line connecting the image acquisition device 20 and the face recognition attack material 60.

[0120] The spherical motion around the camera of the image acquisition device 20 is also performed in a similar way to establish meridian and parallel circles.

[0121] In the embodiment, each test corresponds to a combination of lighting conditions and the relative pose of the face recognition attack material 60 and the image acquisition device 20. In the embodiment, for any face recognition attack material 60, the corresponding relative pose information is consistent, because the number of relative poses is sufficient to cover all possible cases. The number of relative poses is 54, that is, all combinations of 3 lighting conditions, 2 distances and 9 angles. In other embodiments, the number of lighting conditions, distances and angles can be changed (usually increased) as needed. The lighting angle includes the position of the light, the illuminance and the color temperature. The illuminance is generally 0lx-10000lx, and the color temperature is generally 2700k to 5400k.

[0122] Correspondingly, the controller is further configured to, for each relative pose point, after adjusting to the relative pose point, adjust the lighting conditions, and generate face recognition trigger instructions respectively under different lighting conditions to obtain face recognition results under multiple lighting conditions of the relative pose point, wherein the lighting system includes light position, illuminance and color temperature.

[0123] Since the device mechanical arm 30 and the material mechanical arm 40 are fixedly installed above the automatic guide rail, the light is installed on the automatic guide rail, and the light and the automatic guide rail are in communication connection with the face check security automation detection system. Therefore, the face check security automation detection system can change the position of the light by driving the automatic guide rail, and can drive the light to change the illuminance and the color temperature.

[0124] In other embodiments, if it is not necessary to test multiple lighting conditions and distances, the information of each relative pose can only include the angle of the face recognition attack material 60 relative to the image acquisition device 20.

[0125] The face recognition result of the relative pose point is obtained, specifically including: acquiring a screenshot or video data of the current image acquisition device 20 in real time, and obtaining a face recognition result according to the screenshot or video data and OCR information extraction.

[0126] Specifically, because the user interface (UI) is different when the face recognition succeeds or fails, especially the text on the user interface is different, the face recognition result can be obtained according to the screenshot or video data and OCR information extraction.

[0127] Second embodiment face check-in security automatic detection method

[0128] The application provides a face check-in security automatic detection method, which is applied to a face check-in security automatic detection system in communication connection with an image acquisition device 20, a device mechanical arm 30, a material mechanical arm 40 and an illumination system, and is used for testing the security of the image acquisition device 20 through various face recognition attack materials 60.

[0129] As shown in Figure 2 The face check-in security automatic detection method includes:

[0130] 100: selecting a current image acquisition device 20 and a current face recognition attack material 60 according to a preset test sequence, so that the current image acquisition device 20 is loaded on the device mechanical arm 30; thereby, the image acquisition device 20 loaded on the device mechanical arm 30 is provided.

[0131] According to the information of the image acquisition device and the face recognition attack material, the test sequence is preset; when the device material switching condition is met, the current image acquisition device and the current face recognition attack material are selected according to the preset test sequence, and the device material switching condition includes: the test is started, or the face recognition result of each relative pose point between the current image acquisition device and the current face recognition attack material is obtained.

[0132] The number of the image acquisition devices 20 is multiple, each image acquisition device 20 is placed in a device placement area when not being loaded, each image acquisition device 20 corresponds to a unique device identification label and a unique device id; the device mechanical arm 30 is provided with a loading device and a camera;

[0133] After selecting the current image acquisition device 20, if the current image acquisition device has not been loaded on the device mechanical arm, the current image acquisition device is loaded on the device mechanical arm, specifically including:

[0134] The controller 10 sends a device loading instruction with the information of the current image acquisition device 20 to the device robot arm 30, the information of the image acquisition device 20 includes the device id and the device placement area of the image acquisition device 20;

[0135] The device robot arm 30 moves to align the camera downward to the device placement area of the required image acquisition device 20 in response to the device loading instruction, acquires the device identification label by the camera to obtain the device id by the controller 10, moves to align and contact the image acquisition device 20 according to the position of the device identification label when the obtained device id is the device id of the required image acquisition device 20, and drives the carrying device of the device robot arm 30 to load the current image acquisition device 20 on the device robot arm 30.

[0136] Each robot arm presentation material is placed in the material placement area corresponding to the specific type of the robot arm presentation material when it is not loaded, and the material identification label is pasted on all robot arm presentation materials, each robot arm presentation material corresponds to a unique material identification label and a unique material id, robot arm presentation materials of the same specific type are placed in the same material placement area in a stacked manner, and the material placement area has at least two stacking positions, and the material robot arm 40 is provided with a carrying device and a camera.

[0137] After selecting the current face recognition attack material 60, generating the relative pose adjustment instruction further includes: determining by the controller 10 whether the current face recognition attack material 60 is a robot arm presentation material, if so, after selecting the current face recognition attack material 60, generating the relative pose adjustment instruction, making the current robot arm presentation material 60 loaded on the material robot arm 40, and after the controller 10 generates the relative pose adjustment instruction, sending the generated relative pose adjustment instruction to the material robot arm 40 by the controller 10, and moving by the material robot arm 40 in response to the relative pose adjustment instruction to adjust the relative pose between the image acquisition device and the face recognition attack material to the relative pose point;

[0138] Otherwise, after generating the relative pose adjustment instruction, the generated relative pose adjustment instruction is sent to the device robot arm 30, and the device robot arm 30 moves in response to the relative pose adjustment instruction to adjust the relative pose between the image acquisition device 20 and the face recognition attack material 60 to the relative pose point.

[0139] After selecting the current image acquisition device 20, if the current image acquisition device 20 has not been loaded on the device mechanical arm, the current mechanical arm is caused to present the material 20 to be loaded on the material mechanical arm, specifically including: the controller 10 sends a material loading instruction with the information of the current face recognition attack material 60 to the material mechanical arm 40, the information of the face recognition attack material 60 includes the material id and the material placement area;

[0140] The material mechanical arm 40 moves to align the camera downward to the material placement area corresponding to the required mechanical arm presentation material in response to the material loading instruction; the material identification tag is acquired by the camera to obtain the material id by the controller 10; when the obtained material id is the material id of the required mechanical arm presentation material, the material mechanical arm 40 moves to align and contact the mechanical arm presentation material according to the position of the material identification tag; the material mechanical arm 40 drives the material loading device to load the face recognition attack material 60 on the material mechanical arm 40.

[0141] The camera aligns downward to the material placement area corresponding to the required mechanical arm presentation material, specifically including: the camera aligns downward to one of the stacking positions of the material placement area as the current inspection stacking position;

[0142] After obtaining the material id, before the obtained material id is the material id of the required mechanical arm presentation material, further including:

[0143] If the obtained material id is not the material id of the required mechanical arm presentation material, the material mechanical arm 40 moves according to the position of the material identification tag to align and contact the mechanical arm presentation material;

[0144] The material mechanical arm 40 drives the material loading device to load the face recognition attack material 60 on the material mechanical arm 40;

[0145] The material mechanical arm 40 moves to another stacking position other than the current inspection stacking position;

[0146] The material mechanical arm 40 drives the material loading device to unload the face recognition attack material 60;

[0147] The material mechanical arm 40 moves to align the camera downward to the current inspection stacking position, and repeats the steps of obtaining the material id and loading and unloading the face recognition attack material 60 until the obtained material id is the material id of the required mechanical arm presentation material.

[0148] 110: generate test preparation instructions;

[0149] After selecting the current face recognition attack material 60, before generating the test preparation instruction, further comprising: judging whether the current face recognition attack material 60 is a mechanical arm presentation material, if so, loading the mechanical arm presentation material on the material mechanical arm 40;

[0150] 120: moving the image acquisition device 20 to an initial test pose facing the face recognition attack material 60 in response to the test preparation instruction by the device mechanical arm 30;

[0151] At the initial test pose, the camera of the image acquisition device 20 and the face recognition attack material 60 are on the same horizontal plane, and the face of the face recognition attack material 60 is located at the center of the image captured by the image acquisition device 20.

[0152] 130: For each relative pose point between the image acquisition device 20 and the face recognition attack material 60, the controller 10 generates a relative pose adjustment instruction adjusted to the relative pose point and sends it to the device mechanical arm 30 or the material mechanical arm 40 loaded with the face recognition attack material 60.

[0153] 140: In response to the relative pose adjustment instruction, the relative pose between the image acquisition device 20 and the face recognition attack material 60 is adjusted to the relative pose point by the device mechanical arm 30, or the relative pose between the image acquisition device 20 and the face recognition attack material 60 is adjusted to the relative pose point by the material mechanical arm 40;

[0154] Specifically, if the current face recognition attack material 60 is a mechanical arm presentation material, the generated relative pose adjustment instruction is sent to the material mechanical arm 40, and the relative pose between the image acquisition device 20 and the face recognition attack material 60 is adjusted to the relative pose point by the material mechanical arm 40; otherwise, the generated relative pose adjustment instruction is sent to the device mechanical arm 30, and the relative pose between the image acquisition device 20 and the face recognition attack material 60 is adjusted to the relative pose point by the device mechanical arm 30.

[0155] 150: Generating a face recognition trigger instruction by the controller 10 when adjusting to the relative pose point;

[0156] Wherein, for each relative pose point, after adjusting to the relative pose point, further comprising: adjusting the lighting conditions to generate face recognition trigger instructions under different lighting conditions respectively, obtaining face recognition results under multiple lighting conditions of the relative pose point, and the lighting system includes light position, illuminance and color temperature.

[0157] 160: performing face recognition by the image acquisition device 20 to be tested in response to the face recognition trigger instruction, to obtain a face recognition result;

[0158] 170: obtaining the face recognition result of the relative pose point by the controller 10, and detecting the attack resistance of the image acquisition device 20 according to the face recognition result of each relative pose point;

[0159] Specifically, for the same model of image acquisition device 20, the test data of each face recognition attack material 60, each image acquisition device 20 and each relative pose are recorded to obtain the statistical result of the image acquisition device 20 as the detection result of the attack resistance of the image acquisition device 20.

[0160] In the embodiment, the statistical result of the image acquisition device 20 corresponds to the same model of image acquisition device 20, which includes the average SAR of each face recognition attack material 60 obtained by transverse statistics, and the total attack pass rate SAR of all face recognition attack materials 60 obtained by longitudinal statistics.

[0161] SAR (Spoof Accept Rate, spoof behavior acceptance rate) refers to the ratio of passing face recognition authentication through spoof attack forms. The calculation formula is SAR (%) = number of spoof attack pass acceptance / total number of spoof attacks x 100%.

[0162] The above is only a preferred embodiment of the present application, not to limit the scope of the present application, the above embodiment of the present application can also be made various changes. Any simple, equivalent changes and modifications made according to the content of the claims and the specification of the present application, all fall within the scope of the present application. The present application is not described in detail, which is the conventional technical content.

Claims

1. A face recognition security automation detection system, comprising a controller, a device mechanical arm and an image acquisition device to be tested; the image acquisition device is loaded on the device mechanical arm, and is configured to acquire images and perform face recognition in response to a face recognition trigger instruction sent by the controller to obtain a face recognition result; the device mechanical arm is configured to move to adjust a relative pose between the image acquisition device and face recognition attack materials to a corresponding relative pose point in response to a relative pose adjustment instruction sent by the controller; the controller is configured to generate a relative pose adjustment instruction for adjusting to each relative pose point between the image acquisition device and the face recognition attack materials and send the relative pose adjustment instruction to the device mechanical arm, generate a face recognition trigger instruction and send the face recognition trigger instruction to the image acquisition device when the relative pose between the image acquisition device and the face recognition attack materials is adjusted to the relative pose point to obtain a face recognition result of the relative pose point, and detect the attack resistance of the image acquisition device according to the face recognition results of the relative pose points; wherein the number of at least one of the image acquisition device and the face recognition attack materials is plural; the image acquisition device comprises a plurality of devices of the same model, and for the plurality of image acquisition devices of the same model, each image acquisition device only records the face of one person to be tested; the controller is further configured to, after the test is completed, for the image acquisition devices of the same model, statistically record the test data of each face recognition attack material, each image acquisition device and each relative pose to obtain a statistical result of the image acquisition device as a detection result of the attack resistance of the image acquisition device. 2.The face recognition security automation detection system of claim 1, wherein the controller is further configured to select a current image acquisition device and a current face recognition attack material according to a preset test sequence so that the current image acquisition device is loaded on the device mechanical arm. 3.The face recognition security automation detection system of claim 1, wherein the relative pose points are obtained by the following method: a plurality of relative pose points are preset, a current image acquisition device is selected, all face recognition attack materials corresponding to the current image acquisition device are set as the current face recognition attack materials in turn with the face of one person as a reference and each time the test is performed according to the preset plurality of relative pose points; test data of all relative pose points are statistically recorded to obtain vulnerable points, and the vulnerable points are taken as actual relative pose points. 4.The face recognition security automation detection system of claim 2, wherein the number of the image acquisition devices is plural, each image acquisition device is placed in a device placement area when not loaded, and each image acquisition device corresponds to a unique device identification label and a unique device id. The controller is further configured to, after selecting the current image acquisition device, if the current image acquisition device has not been loaded on the device mechanical arm, send a device loading instruction with information of the current image acquisition device to the device mechanical arm, the information of the image acquisition device including a device ID of the image acquisition device and a device placement area; The device mechanical arm is provided with a carrier device and a camera, and is further configured to, in response to the device loading instruction, move the device mechanical arm so that the camera thereof is downwardly aligned with the device placement area of the required image acquisition device; acquire the device identification tag by using the camera to obtain the device ID by the controller; when the obtained device ID is the device ID of the required image acquisition device, move according to the position of the device identification tag so that the carrier device thereof is aligned with and contacts the image acquisition device; and drive the carrier device of the device mechanical arm to load the current image acquisition device on the device mechanical arm.

5. The face recognition security automated detection system of claim 2, wherein the controller is further configured to, after the current image acquisition device and the current face recognition attack material are selected and the current image acquisition device is loaded on the device mechanical arm, generate a test preparation instruction before generating the relative pose adjustment instruction. The device mechanical arm is further configured to, in response to the test preparation instruction, move so that the image acquisition device is located at an initial test pose facing the face recognition attack material. 6.The face boarding security automated detection system of claim 5, wherein the device mechanical arm moves to adjust the relative pose between the image capturing device and the face recognition attack material to the corresponding relative pose point, in particular comprising: Based on the initial test pose, the image acquisition device is caused to perform spherical motion around a face of the face recognition attack material to adjust a relative angle between the image acquisition device and the face recognition attack material, and perform linear motion on a line connecting the image acquisition device and the face recognition attack material to adjust a relative distance between the image acquisition device and the face recognition attack material.

7. The face biometric security automated detection system of claim 6, wherein the face surrounding the face recognition attack material performs spherical motion, and specifically comprising: Meridians and parallels are established with the face recognition attack material as a spherical center, the image acquisition device is caused to move to an intersection of a meridian and a parallel of the face recognition attack material and maintain a posture of directly facing the face recognition attack material.

8. The face recognition security automated detection system of claim 2, wherein the controller is further configured to, after the current face recognition attack material is selected, judge whether the current face recognition attack material is a mechanical arm presented material before generating the relative pose adjustment instruction; if yes, after the current face recognition attack material is selected, cause the current mechanical arm presented material to be loaded on the material mechanical arm before generating the relative pose adjustment instruction, and send the generated relative pose adjustment instruction to the material mechanical arm after generating the relative pose adjustment instruction; otherwise, send the generated relative pose adjustment instruction to the device mechanical arm after generating the relative pose adjustment instruction; The material mechanical arm is configured to, in response to the relative pose adjustment instruction, move to adjust a relative pose between the image acquisition device and the face recognition attack material to the relative pose point.

9. The system of claim 8, wherein each of the robotic arm presentation materials is placed in a material placement area corresponding to a specific type of the robotic arm presentation material when the robotic arm presentation material is not loaded, and each of the robotic arm presentation materials has a material identification tag attached thereto, and each of the robotic arm presentation materials corresponds to a unique material identification tag and a unique material id. If the current face recognition attack material is a mechanical arm presentation material and has not been loaded on the material mechanical arm, the current mechanical arm presentation material is loaded on the material mechanical arm, specifically including: issuing a material loading instruction with information of a current facial recognition attack material to the material robotic arm to cause the robotic arm presentation material to be loaded on the material robotic arm, wherein the information of the facial recognition attack material includes the material id and the material placement area; the material robotic arm is configured to move to align the camera thereof downward to the material placement area corresponding to the desired robotic arm presentation material in response to the material loading instruction; acquiring the material identification tag with the camera to obtain the material id by the controller; when the obtained material id is the material id of the desired robotic arm presentation material, moving to align and contact the robotic arm presentation material with the material robotic arm; driving the material robotic arm to load the facial recognition attack material on the material robotic arm.

10. The system of claim 9, wherein the robotic arm presentation materials of the same specific type are placed in the same material placement area in a stacked manner, and the material placement area has at least two stacking positions; aligning the camera downward to the material placement area corresponding to the desired robotic arm presentation material specifically includes aligning the camera downward to one of the stacking positions of the material placement area as a current checking stacking position; the material robotic arm is further configured to, after obtaining the material id, before the obtained material id is the material id of the desired robotic arm presentation material, when the obtained material id is not the material id of the desired robotic arm presentation material, moving to align and contact the robotic arm presentation material with the material robotic arm; driving the material robotic arm to load the facial recognition attack material on the material robotic arm; moving to another stacking position other than the current checking stacking position; driving the material robotic arm to unload the facial recognition attack material; moving to align the camera downward to the current checking stacking position, repeating the steps of obtaining the material id and loading and unloading the facial recognition attack material until the obtained material id is the material id of the desired robotic arm presentation material.

11. The system of claim 4 or 9, wherein the material robotic arm is a suction pump.

12. The face biometric security automated detection system of claim 1, wherein the controller is further configured to, for each relative pose point, adjust lighting conditions after adjusting to the relative pose point to generate face recognition trigger instructions under different lighting conditions respectively, and obtain face recognition results of the relative pose point under multiple lighting conditions, the lighting conditions including light position, illumination, and color temperature.

13. A face biometric security automated detection method, comprising: providing an image acquisition device loaded on a device mechanical arm; for each relative pose point between the image acquisition device and a face recognition attack material, generating, by a controller, a relative pose adjustment instruction adjusting to the relative pose point and sending the relative pose adjustment instruction to the device mechanical arm or a material mechanical arm loaded with the face recognition attack material; in response to the relative pose adjustment instruction, adjusting, by the device mechanical arm, the relative pose between the image acquisition device and the face recognition attack material to the relative pose point, or adjusting, by the material mechanical arm, the relative pose between the image acquisition device and the face recognition attack material to the relative pose point; generating, by the controller, a face recognition trigger instruction when adjusting to the relative pose point; performing, by the image acquisition device to be tested, face recognition in response to the face recognition trigger instruction; obtaining, by the controller, face recognition results of the relative pose point, and detecting attack resistance of the image acquisition device according to face recognition results of each relative pose point; wherein the number of at least one of the image acquisition device and the face recognition attack material is multiple; the image acquisition device includes multiple devices of the same model, and for multiple image acquisition devices of the same model, each image acquisition device only records the face of one person to be tested; after the test is completed, for image acquisition devices of the same model, test data of each face recognition attack material, each image acquisition device, and each relative pose are recorded to obtain statistical results of the image acquisition device as the detection result of the attack resistance of the image acquisition device.

14. The face security automated detection method of claim 13, providing an image collection device loaded on a device mechanical arm, specifically comprising: selecting the current image acquisition device and the current face recognition attack material according to a preset test order, so that the current image acquisition device is loaded on the device mechanical arm.

15. The face biometric security automated detection method of claim 14, wherein the number of the image acquisition device is multiple, each image acquisition device is placed in a device placement area when not loaded, each image acquisition device corresponds to a unique device identification label and a unique device id, and the device mechanical arm is provided with a loading device and a camera; after the current image acquisition device is selected, if the current image acquisition device is not loaded on the device mechanical arm, the current image acquisition device is loaded on the device mechanical arm, specifically including: issuing, by the controller, a device loading instruction with information of the current image acquisition device to the device mechanical arm, the information of the image acquisition device including the device id and the device placement area of the image acquisition device; The device mechanical arm moves to align its camera downward to the device placement area of the required image acquisition device in response to a device loading instruction; the device identification tag is acquired by the camera to be parsed by the controller to obtain the device id; when the obtained device id is the device id of the required image acquisition device, the device mechanical arm moves to align and contact the image acquisition device according to the position of the device identification tag; the carrier of the device mechanical arm is driven to load the current image acquisition device on the device mechanical arm.

16. The method of claim 14, wherein after selecting the current face recognition attack material, before generating the relative pose adjustment instruction, the method further comprises: The controller determines whether the current face recognition attack material is a mechanical arm presentation material; If yes, after the current face recognition attack material is selected, the current mechanical arm presentation material is loaded on the material mechanical arm before a relative pose adjustment instruction is generated, and after the relative pose adjustment instruction is generated, the generated relative pose adjustment instruction is sent to the material mechanical arm by the controller, and the material mechanical arm moves to adjust the relative pose between the image acquisition device and the face recognition attack material to the relative pose point in response to the relative pose adjustment instruction; Otherwise, after the relative pose adjustment instruction is generated, the generated relative pose adjustment instruction is sent to the device mechanical arm by the controller, and the device mechanical arm moves to adjust the relative pose between the image acquisition device and the face recognition attack material to the relative pose point in response to the relative pose adjustment instruction.

17. The face recognition security automation detection method of claim 16, each mechanical arm presentation material is placed in a material placement area corresponding to the specific type of the mechanical arm presentation material when not loaded, and a material identification tag is attached to each mechanical arm presentation material, each mechanical arm presentation material corresponds to a unique material identification tag and a unique material id; the material mechanical arm is provided with a carrier and a camera; If the current face recognition attack material has not been loaded on the material mechanical arm, the current mechanical arm presentation material is loaded on the material mechanical arm, specifically including: The controller sends a material loading instruction with information of the current face recognition attack material to the material mechanical arm, the information of the face recognition attack material includes the material id and the material placement area; The material mechanical arm moves to align its camera downward to the material placement area corresponding to the required mechanical arm presentation material in response to the material loading instruction; the material identification tag is acquired by the camera to be parsed by the controller to obtain the material id; when the obtained material id is the material id of the required mechanical arm presentation material, the material mechanical arm moves to align and contact the mechanical arm presentation material according to the position of the material identification tag; the carrier of the material mechanical arm is driven to load the face recognition attack material on the material mechanical arm.

18. The face recognition security automation detection method of claim 17, the mechanical arm presentation materials of the same specific type are placed in the same material placement area in a stacked manner, and the material placement area has at least two stacking positions; The camera downwardly aims at a material placement area corresponding to the material presented by the required robot arm, specifically including: the camera downwardly aims at one of the stacking positions of the material placement area as the current checking stacking position; After the material id is parsed, before the obtained material id is the material id of the material presented by the required robot arm, further comprising: If the obtained material id is not the material id of the material presented by the required robot arm, the material robot arm moves according to the position of the material identification tag to align and contact the material presented by the robot arm with the carrying device of the material robot arm; The carrying device of the material robot arm is driven to load the face recognition attack material on the material robot arm; The material robot arm moves to another stacking position other than the current checking stacking position; The carrying device of the material robot arm is driven to unload the face recognition attack material; The material robot arm moves to align the camera downwardly aiming at the current checking stacking position, and the steps of parsing the material id and loading and unloading the face recognition attack material are repeated until the parsed material id is the material id of the material presented by the required robot arm.

19. The method of claim 13, further comprising, for each relative pose point, after adjusting to the relative pose point: The controller adjusts the lighting conditions to generate face recognition trigger instructions under different lighting conditions, respectively, and obtains face recognition results under multiple lighting conditions of the relative pose point, and the lighting conditions include light position, illuminance and color temperature.

Images