Face recognition interaction method and system based on privacy protection

Abstract

The invention relates to the technical field of face recognition interaction, and particularly discloses a face recognition interaction method and system based on privacy protection, and the method comprises the following steps: carrying out the grid division of a detection image, obtaining grid regions, and numbering the grid regions; determining a normal grid based on the similarity degree of the grid areas with the same number in the standard image and the comparison image; grouping the normal grids, and determining normal areas according to groups; and acquiring a face image of the user through the normal area with the largest area, and performing privacy protection processing on the face image. According to the invention, the accuracy of face recognition is improved while the privacy of the user is protected.

Description

Technical Field

[0001] The present invention relates to the technical field of face recognition interaction, and particularly relates to a face recognition interaction method and system based on privacy protection. Background Art

[0002] Face recognition interaction is a human-computer interaction method based on biometric recognition technology. It automatically captures and analyzes the facial features of users through cameras and algorithms to achieve identity verification or behavior authorization, and is widely used in fields such as mobile phone unlocking, payment authentication, security monitoring, and intelligent access control. With the continuous development of face recognition technology, its interaction experience has become more fluent and efficient, and at the same time, higher requirements have been put forward for privacy protection and data security.

[0003] A Chinese patent with the patent announcement number CN113313026B discloses a face recognition interaction method, device, and equipment based on privacy protection. The main steps in its solution include: collecting the face image of a user; determining the face key points in the face image; according to the face key points, determining a partial area in the face area of the face image as the face key area; performing privacy protection processing on the background area in the face image and the face area outside the face key area; and presenting the face key area and the area after privacy protection processing to the user to complete the face recognition process of the user.

[0004] However, with the continuous increase in the usage time, the device for collecting face images may experience varying degrees of wear under long-term operation and the influence of the external environment. For example, physical damages such as scratches on the surface of the lens may occur, resulting in unclear phenomena such as blurring, uneven brightness, or missing details in certain areas of the captured face image. Furthermore, the problem that the face key area in the obtained face image is unclear may occur, affecting the face recognition effect. Summary of the Invention

[0005] The purpose of the present invention is to provide a face recognition interaction method and system based on privacy protection to solve the following technical problems: With the continuous increase in the usage time, the device for collecting face images may experience varying degrees of wear under long-term operation and the influence of the external environment. For example, physical damages such as scratches on the surface of the lens or sensor may occur, resulting in unclear phenomena such as blurring, uneven brightness, or missing details in certain areas of the captured face image. Furthermore, the problem that the face key area in the obtained face image is unclear may occur, affecting the face recognition effect.

[0006] The purpose of the present invention can be achieved through the following technical solutions: A face recognition interaction method based on privacy protection includes the following steps: Obtain an image of the front of the lens of the face image acquisition device during a period when face recognition is not performed, record it as a detection image, perform grid division on the detection image to obtain a number of grid regions, the length and width of a single grid region are both preset values, and number the grid regions; Mark the detection image corresponding to the time point when the face image acquisition device starts to be put into use as a standard image, and mark the detection image corresponding to the current time as a comparison image; Denote the grid region numbered i in the standard image and the comparison image as grid Di and grid Di' respectively, obtain the similarity degree Xi between the grid Di and grid Di', and when the similarity degree is greater than a preset similarity degree threshold, mark the grid Di' as a normal grid; Group the normal grids, and there is at least one same side between the normal grids in the same group and at least one of the normal grids in this group. Obtain the area composed of all the normal grids in the same group, and record it as a normal area; When the user performs face recognition, collect the user's face image through the normal area with the largest area, and perform privacy protection processing on the face image.

[0007] As a further solution of the present invention: Collecting the user's face image through the normal area with the largest area includes: Mark the normal area with the largest area as the target area. When the face key area is not completely included in the face image obtained from the target area, perform the following steps: Mark the center point of the part of the face key area that is not included as point A1, mark the center point of the face key area as point A2, and mark the direction from point A2 to point A1 as the target direction; Control the face image acquisition device to move along the target direction until the face key area is completely included in the face image obtained from the target area.

[0008] As a further solution of the present invention: When the ratio of the part of the face key area that is not included to the face key area exceeds a preset ratio threshold, send a warning message to prompt the user.

[0009] As a further solution of the present invention: Performing privacy protection processing on the face image includes: Cropping the part of the face image except the face key area.

[0010] As a further solution of the present invention: When the area of the target area is less than a preset area threshold, send a warning message to a preset manager.

[0011] As a further solution of the present invention: Obtain new normal grids after a preset time interval.

[0012] A face recognition interaction system based on privacy protection, comprising: Acquisition module: Obtain an image in front of the lens of the face image acquisition device during a period when face recognition is not performed, record it as a detection image, divide the detection image into a number of grid regions, the length and width of a single grid region are both preset values, and number the grid regions; Positioning module: Mark the detection image corresponding to the time point when the face image acquisition device starts to be put into use as a standard image, and mark the detection image corresponding to the current time as a comparison image; Denote the grid region numbered i in the standard image and the comparison image as grid Di and grid Di' respectively, obtain the similarity degree Xi between the grid Di and grid Di', and when the similarity degree is greater than a preset similarity degree threshold, mark the grid Di' as a normal grid; Group the normal grids, there is the same side between the normal grids in the same group and at least one of the normal grids in this group, obtain the area composed of all the normal grids in the same group, and denote it as a normal area; Privacy module: When the user performs face recognition, collect the user's face image through the normal area with the largest area, and perform privacy protection processing on the face image.

[0013] Advantages of the present invention: Compared with the prior art: 1) By obtaining the lens detection image and dividing the grid during the non-recognition period, and comparing the similarity of the corresponding grids in the standard image and the current image, the present invention can dynamically detect and identify the image degradation areas caused by wear, stains or aging in the lens, and accordingly delimit the available normal areas, avoiding the dependence on the degraded areas during the face image acquisition process, and fundamentally solving the problem of recognition failure caused by blurred images, uneven brightness or missing details due to long-term use of the device; 2) In the prior art, more emphasis is placed on desensitizing the image content, for example, only retaining the key areas and masking the background areas, but ignoring the possible impact of image quality changes on the face key areas themselves, there is a risk of formal privacy protection and recognition failure; On the basis of ensuring privacy protection, the present invention proposes to collect faces by dynamically detecting normal areas and accurately identifying the largest effective area, and at the same time combine the integrity check of the face key areas with the device automatic adjustment compensation mechanism to ensure that the finally collected image contains complete and clear key areas. By cropping non-key parts and adjusting the acquisition perspective in real time, not only the user's privacy information is not leaked, but also the image quality is effectively guaranteed to be in an available state, enabling the system to achieve a technical balance between privacy protection and recognition accuracy and improving the actual usability. Description of the Drawings

[0014] The present invention will be further described below with reference to the accompanying drawings.

[0015] Figure 1 It is a schematic flow chart of a face recognition interaction method based on privacy protection according to the present invention. Specific embodiments

[0016] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0017] Please refer to Figure 1 As shown, the present invention is a face recognition interaction method based on privacy protection, including the following steps: During the time period when face recognition is not performed, an image in front of the lens of the face image acquisition device is acquired and recorded as a detection image. The detection image is divided into a number of grid regions. The length and width of a single grid region are both preset values, and the grid regions are numbered. It should be noted that during the non-face recognition period, the image captured by the lens of the face image acquisition device itself is acquired, that is, the detection image. Its purpose is not to identify the user, but to determine whether there are abnormal conditions in the lens, such as physical problems that cause the imaging quality to decline due to wear, scratches, stains, or aging. The system controls the camera to perform an imaging of the current environment during the time period without recognition tasks (such as early morning or idle time), and divides the image into a number of grid regions of equal size. For example, an image of 1920×1080 is divided into grids of 60×60 pixels, a total of 576. Each grid represents a part of the lens imaging area. The system assigns numbers to each grid for subsequent image quality comparison and anomaly detection. If there are slight scratches or local stains on the lens, usually only some image areas will be affected. By identifying whether there are physical anomalies in the lens, the accuracy and clarity of the subsequent captured face images can be ensured; The detection image corresponding to the time point when the face image acquisition device starts to be put into use is marked as a standard image, and the detection image corresponding to the current time is marked as a comparison image; the grid regions numbered i in the standard image and the comparison image are respectively denoted as grid Di and grid Di'. The similarity degree Xi between the grid Di and the grid Di' is obtained. When the similarity degree is greater than a preset similarity degree threshold, the grid Di' is marked as a normal grid; It can be understood that when the lens leaves the factory or the device is first put into use, an image of the lens state at that time is collected as a standard image, representing the initial state of the lens without scratches, pollution, and with good imaging quality. Subsequently, during the operation of the device, the system regularly collects the lens imaging images as comparison images for quality comparison with the standard image. Each image has been numbered according to a fixed grid rule. For example, for the grid area numbered i, it is extracted from the standard image and the comparison image respectively, denoted as grid Di and Di'. The similarity of these two corresponding grid areas is calculated. The method can adopt structural similarity index (SSIM), histogram comparison, edge information matching, etc. to obtain the similarity degree Xi of this area. If Xi exceeds the set threshold (for example, 0.9), it indicates that the imaging of this area in the current image is not much different from the initial state and can be considered not affected by lens wear. Therefore, this grid Di' is marked as a normal grid. Through this grid-by-grid comparison method, the system can carefully identify which positions in the lens imaging area are still clearly available, thereby dynamically constructing the collectable area and improving the monitoring accuracy of image quality and the reliability of subsequent face collection; It should be noted that the normal grid is re-acquired after a preset time interval.

[0018] Group the normal grids. There is at least one same side between the normal grids in the same group. Obtain the area composed of all the normal grids in the same group, denoted as the normal area; When the user performs face recognition, collect the user's face image through the normal area with the largest area, and perform privacy protection processing on the face image; In a preferred embodiment of the present invention, collecting the user's face image through the normal area with the largest area includes: Mark the normal area with the largest area as the target area. When the face key area is not completely included in the face image obtained from the target area, perform the following steps: Mark the center point of the part of the face key area that is not included as point A1, mark the center point of the face key area as point A2, and mark the direction from point A2 to point A1 as the target direction; Control the face image acquisition device to move along the target direction until the face key area is completely included in the face image obtained from the target area; It can be understood that when the area of the target area is less than the preset area threshold, send a warning message to the preset management personnel; It should be noted that after the similarity analysis of all grid regions is completed, the connectivity analysis will be performed on the regions determined to be normal grids, and the adjacent judgment method is used to group the connected normal grids into the same group. Specifically, if a normal grid shares an edge (adjacent in the up, down, left, or right direction) with another normal grid, they are considered to belong to the same connected region. The system traverses all normal grids to form multiple continuous block-shaped grid sets, called normal regions. Subsequently, the system compares the areas of each normal region and selects the region with the largest area as the target acquisition region for image acquisition in subsequent face recognition. If the face image captured in the target region fails to completely cover the key face regions (such as eyes, nose, mouth, etc.), the system will determine the position of the missing part through the edge of the key region. Taking the center point of the key face region as A2 and the center of the missing part as A1, the system calculates the vector direction from A2 to A1 as the target direction and controls the camera or the acquisition module to move along this direction until the target region in the newly captured image completely covers the key face region. During this process, the image acquisition structure is not changed, and compensation is achieved only by adjusting the acquisition angle or the composition position, so as to ensure that the face image not only avoids the damaged area of the lens but also contains complete key face information, providing a reliable basis for subsequent recognition and privacy processing; It should be noted that when the ratio of the part of the key face region that is not included to the key face region exceeds the preset ratio threshold, a warning message is sent to prompt the user.

[0019] In another preferred embodiment of the present invention, the privacy protection process for the face image includes: cropping the part of the face image other than the key face region.

[0020] A face recognition interaction system based on privacy protection includes: Acquisition module: Obtain the image in front of the lens of the face image acquisition device during the time period when face recognition is not performed, denoted as the detection image. Divide the detection image into several grid regions, the length and width of a single grid region are both preset values, and number the grid regions; Positioning module: Mark the detection image corresponding to the time point when the face image acquisition device starts to be put into use as the standard image, and mark the detection image corresponding to the current time as the comparison image; Denote the grid regions numbered i in the standard image and the comparison image as grid Di and grid Di' respectively, obtain the similarity degree Xi between grid Di and grid Di', and when the similarity degree is greater than the preset similarity degree threshold, mark grid Di' as a normal grid; Group the normal grids. The normal grids in the same group share at least one side with at least one of the normal grids in the group. Obtain the area formed by all the normal grids in the same group, which is denoted as the normal area. Privacy module: When the user performs face recognition, collect the user's face image through the normal area with the largest area, and perform privacy protection processing on the face image.

[0021] The above has described an embodiment of the present invention in detail, but the content is only a preferred embodiment of the present invention and cannot be considered as limiting the scope of implementation of the present invention. All equivalent changes and improvements made according to the scope of the present invention application should still fall within the scope covered by the present invention.

Claims

1. A face recognition interaction method based on privacy protection, characterized in that, It includes the following steps: Obtain an image of the front of the lens of the face image acquisition device during a period when face recognition is not being performed, denote it as the detection image, perform grid division on the detection image to obtain a number of grid regions, the length and width of a single grid region are both preset values, and number the grid regions; Mark the detection image corresponding to the time point when the face image acquisition device starts to be put into use as the standard image, and mark the detection image corresponding to the current time as the comparison image; Denote the grid region numbered i in the standard image and the comparison image as grid Di and grid Di' respectively, obtain the similarity degree Xi between grid Di and grid Di', and when the similarity degree is greater than the preset similarity degree threshold, mark grid Di' as a normal grid; Group the normal grids, and there is at least one same side between the normal grids in the same group and at least one of the normal grids in the group, obtain the area composed of all the normal grids in the same group, and denote it as the normal area; When the user performs face recognition, collect the user's face image through the normal area with the largest area, and perform privacy protection processing on the face image.

2. The face recognition interaction method based on privacy protection according to claim 1, wherein Collecting the user's face image through the normal area with the largest area includes: Mark the normal area with the largest area as the target area, and when the face key area is not completely included in the face image obtained from the target area, perform the following steps: Mark the center point of the part of the face key area that is not included as point A1, mark the center point of the face key area as point A2, and mark the direction from point A2 to point A1 as the target direction; Control the face image acquisition device to move along the target direction until the face key area is completely included in the face image obtained from the target area.

3. The face recognition interaction method based on privacy protection according to claim 2, wherein, When the ratio of the part of the face key area that is not included to the face key area exceeds the preset ratio threshold, send a warning message to prompt the user.

4. A face recognition interaction method based on privacy protection according to claim 1, characterized in that Performing privacy protection processing on the face image includes: cropping the part of the face image except the face key area.

5. The face recognition interaction method based on privacy protection according to claim 2, wherein When the area of the target area is less than the preset area threshold, send a warning message to the preset management personnel.

6. The face recognition interaction method based on privacy protection according to claim 1, wherein Re-obtain the normal grids after a preset time interval.

7. A face recognition interaction system based on privacy protection, characterized in that, It includes: Collection module: Obtain an image of the front of the lens of the face image acquisition device during a period when face recognition is not being performed, denote it as the detection image, perform grid division on the detection image to obtain a number of grid regions, the length and width of a single grid region are both preset values, and number the grid regions; Positioning module: Mark the detection image corresponding to the time point when the face image acquisition device starts to be put into use as the standard image, and mark the detection image corresponding to the current time as the comparison image; Denote the grid region numbered i in the standard image and the comparison image as grid Di and grid Di' respectively, obtain the similarity degree Xi between grid Di and grid Di', and when the similarity degree is greater than the preset similarity degree threshold, mark grid Di' as a normal grid; Group the normal grids. The normal grids in the same group share at least one side with at least one of the normal grids in the group. Obtain the area composed of all the normal grids in the same group, which is denoted as the normal area. Privacy module: When the user performs face recognition, collect the user's face image through the normal area with the largest area, and perform privacy protection processing on the face image.

Images