Data security management method and system based on image recognition, electronic equipment, computer readable storage medium and computer program product

Abstract

The invention discloses a data security management method and system based on image recognition, electronic equipment, a computer readable storage medium and a computer program product, and the method comprises the steps: collecting the face image information of a login person, carrying out the image processing of the collected face image information of the login person, and obtaining the processed face image information of the login person; comparing the processed registrant face image information with authorized face image information in a face image library, acquiring registrant reading device image information only when the processed registrant face image information is consistent with the authorized face image information in the face image library, and performing image processing on the acquired registrant reading device image information to obtain processed registrant reading device image information; judging whether the reading device is an optical disk reading device or not based on the processed logger reading device image information, and if yes, allowing the logger to read data; otherwise, an alarm signal is sent; wherein the same image acquisition equipment is adopted to acquire the face image information of the logger and the image information of the logger reading device.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of data security management, and particularly relates to a data security management method and system based on image recognition, an electronic device, a computer readable storage medium and a computer program product. BACKGROUND

[0002] With the rapid development of information technology, especially big data, cloud computing and mobile internet technology, data storage devices have become one of the core media for information bearing and transmission. In enterprises, governments, medical institutions, financial institutions and individual users, a large amount of sensitive information, business secrets and personal privacy are stored, processed and transmitted in the form of data, so the security management of data has become a crucial issue in the field of information security.

[0003] The data security management method in the prior art mainly guarantees data security through encryption of data and authentication of a loginer. The encryption of data is performed by encrypting the data to be saved through a special algorithm and then decrypting the data through an algorithm. With the rapid development of artificial intelligence technology, the security of the encryption algorithm cannot be guaranteed. In the authentication of the loginer, the precision of the face image recognition of the loginer also limits the development of the technology. SUMMARY

[0004] The application aims to solve the defects of the data security management method in the prior art, and provides a data security management method and system based on image recognition, an electronic device, a computer readable storage medium and a computer program product.

[0005] Technical scheme: In a first aspect, the application provides a data security management method based on image recognition, which comprises the following steps: Collecting face image information of a loginer, performing image processing on the collected face image information of the loginer, and obtaining processed face image information of the loginer; Comparing the processed face image information of the loginer with authorized face image information in a face image library, and only when the two are consistent, collecting loginer reading device image information, performing image processing on the collected loginer reading device image information, and obtaining processed loginer reading device image information; Based on the processed loginer reading device image information, determining whether the reading device is a CD reading device, and if so, allowing the loginer to read data; otherwise, issuing an alarm signal; In the application, the same image collection device is used to collect the face image information of the loginer and the reading device image information of the loginer.

[0006] Further, the image processing on the collected face image information of the loginer comprises: Step 1: the feature image f(x, y) extracted from the collected login person face image information L is represented as: ; Wherein, L i represents the i-th blurred image, represents a filter using regional covariance; By removing the feature image from the collected login person face image information L, the texture image is obtained, which is represented as: ; Step 2: by smoothing the collected login person face image information L, a smoothed image g is obtained, and L0 gradient is minimized by the smoothed image g, and then the image is obtained, which is represented as: ; Wherein, is the L2 norm, is the L0 norm, represents the gradient of L, represents a parameter for controlling the sparsity of the signal L; The image is decomposed into a detail layer and a base layer, and the deblurring image V i is calculated by obtaining the blurred image s, the blur kernel u and the image prior weight : ; Wherein, F represents the discrete Fourier transform, F -1 represents the inverse discrete Fourier transform, represents the conjugate of the discrete Fourier transform; G i represents the detail layer; Wherein, the blurred image s and the blur kernel u are determined according to the image background noise; the image prior weight is a preset weight parameter determined according to the requirements of the target image; Step 3: using the split Bregman filter to deblur the deblurring image V i , the image B i is obtained, which is represented as: ; The image B i is enhanced by using the deconvolution based on Laplace prior, and the image is obtained, which is represented as: ; Wherein, represents the weight of the deconvolution based on Laplace prior, represents an application of B i a bilateral filter for performing smoothing processing; adding a texture image to an image to obtain a processed login person face image information A i represents: ; wherein, represents a weighting factor.

[0007] Further, the image processing of the collected login person reading device image information to obtain the processed login person reading device image information specifically includes: Suppose that the collected original reading device image is f(x), introduce a degradation image g(x) and a point spread function h(x), and let f'(x), g'(x) and h'(x) represent the values normalized to the sum, represented as: ; ; ; When the probability density of the light spot existing at point x1 in the original reading device image is: ; The probability density of the light spot image formed at point x2 is: ; According to the image background noise, the corresponding blurred image is determined; If there is a light spot at point x1 in the original reading device image, the probability density of the blurred image imaging at point x2 is , represented as: ; The probability density of the original image of the light spot image formed at point x2 of the blurred image is represented by Bayes' theorem as: ; In the formula, is the probability density of the original image of the light spot image formed at point x2; Only when is greater than a preset value, the reading device used by the login person is a CD reading device.

[0008] In a second aspect, the present application provides a data security management system based on image recognition, comprising: An image acquisition module is used to acquire login person face image information, or to acquire login person reading device image information; The first image processing module is configured to perform image processing on the collected loginer face image information to obtain processed loginer face image information. The face comparison module is configured to compare the processed loginer face image information with authorized face image information in a face image library, and only when the comparison result is consistent, the image acquisition module is called to collect loginer reading device image information. The second image processing module is configured to perform image processing on the collected loginer reading device image information to obtain processed loginer reading device image information. The reading device judgment module is configured to judge whether the reading device is a CD reading device based on the processed loginer reading device image information, and if so, the loginer is allowed to read data; otherwise, an alarm signal is sent.

[0009] Further, in the first image processing module, the following steps are performed: Step 1: Extract the feature image f(x, y) from the collected loginer face image information L, which is represented as: ; Wherein, L i represents the i-th blurred image, represents a filter using regional covariance; By removing the feature image from the collected loginer face image information L, a texture image is obtained, which is represented as: ; Step 2: By performing smoothing processing on the collected loginer face image information L, a smoothed image g is obtained, and L0 is minimized by the smoothed image g, and then an image is obtained, which is represented as: ; Wherein, is the L2 norm, is the L0 norm, represents the gradient of L, represents a parameter controlling the sparsity of the signal L; The image is decomposed into a detail layer and a base layer, and the deblurred image V i is calculated by obtaining the blurred image s, the blur kernel u and the image prior weight : ; Wherein, F represents the discrete Fourier transform, F -1 represents the inverse discrete Fourier transform, represents the conjugate of the discrete Fourier transform; G irepresenting a detail layer; Wherein, the blurred image s and the blur kernel u are determined according to image background noise; the image prior weight is a preset weight parameter determined according to target image requirements; Step 3: using a split Bregman filter To the deblurred image V i , the image B i is obtained by deblurring processing. ; The image B i is enhanced by using deconvolution based on Laplace prior, and the image B is obtained, which is represented as: ; Wherein, represents the weight of deconvolution based on Laplace prior, represents a bilateral filter applied to B i for smoothing processing. The texture image is added to the image B , and the processed login person face image information A i is obtained, which is represented as: ; Wherein, represents a weighting factor.

[0010] Further, in the second image processing module, the following steps are performed: The collected login reader device image information is processed to obtain the processed login reader device image information, specifically including: Assuming that the collected original reader device image is f(x), introducing a degraded image g(x) and a point spread function h(x), let f'(x), g'(x) and h'(x) represent the values normalized to the total, represented as: ; ; ; When the probability density of the light spot located at point x1 in the original reader device image is: ; The probability density of the light spot image formed at point x2 is: ; According to the image background noise, the corresponding blurred image is determined; If there is a light spot at point x1 of the original reading device image, the probability density of the blurred image imaging at point x2 is , which is expressed as: ; The probability density of the original image forming a light spot image at point x2 of the blurred image is expressed by Bayes' theorem as: ; In the formula, is the probability density of the original image forming a light spot image at point x2; Only when is greater than a preset value, the reading device used by the login user is an optical disc reading device.

[0011] In a third aspect, the present application provides an electronic device, which comprises: at least one processor; and a memory in communication connection with the at least one processor; wherein the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to execute the above-mentioned image recognition-based data security management method.

[0012] In a fourth aspect, the present application provides a computer readable storage medium, which stores computer instructions for enabling a processor to implement the above-mentioned image recognition-based data security management method when executed.

[0013] In a fifth aspect, the present application provides a computer program product comprising a computer program, which implements the above-mentioned image recognition-based data security management method when executed by a processor.

[0014] Advantages: Compared with the prior art, the present application has the following advantages: (1) The method of the present application can realize the recognition of the login user's face information and the recognition of the reading device used by the login user through the same set of image acquisition unit, which not only ensures the reliability of data reading authorization, but also ensures the security of data transmission after reading; (2) The method of the present application uses the linear combination of texture extraction and improved Fourier transform to better defocus and deblur the image, linearly combines the texture to obtain a deblurred image, and eliminates artifacts while preserving edges in the deblurring process, thereby improving the reliability of face recognition; (3) The method of the present application improves the efficiency of the reading device identification by analyzing the light spot to determine whether to use the optical disc reading device, and effectively overcomes the image distortion problem caused by the image acquisition unit. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 A principle block diagram of a data security management system based on image recognition is provided for the second embodiment; Figure 2 A principle block diagram of a first level security identification module is provided for the second embodiment; Figure 3 A principle block diagram of a second level security identification module is provided for the second embodiment. DETAILED DESCRIPTION

[0016] The technical solutions of the present application will be further described in combination with the drawings and embodiments.

[0017] The data security management method based on image recognition provided by the present application will be described in detail in combination with the drawings and embodiments.

[0018] The flowcharts and block diagrams in the drawings of the present application show the possible implementation architecture, function and operation of the method and system according to various embodiments of the present disclosure. It should be noted that each block in the block diagram can represent a module, a program segment, or a part of code, which can include one or more executable instructions for implementing the logical functions defined in various embodiments. It should also be noted that in some alternative implementations, the functions indicated in the blocks can also occur in an order different from that indicated in the drawings. For example, two blocks indicated in succession can actually be executed substantially in parallel, or they can sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the flowchart and / or block diagram, and the combination of blocks in the flowchart and / or block diagram, can be implemented using a dedicated hardware-based system that performs the specified functions or operations, or can be implemented using a combination of dedicated hardware and computer instructions.

[0019] The terms "include", "contain" and similar terms should be understood as open terms, i.e. "including / containing but not limited to", indicating that other contents can also be included. The same reference signs in the embodiments and the drawings refer to the same or similar parts or structural features, and secondly, when one part is described as being arranged on another part, it means that one part is directly arranged on another part, or one part is indirectly arranged on another part, i.e. one or more parts are arranged between the two parts, and when one part is described as being directly arranged on another part, it means that there is no other part between the two parts.

[0020] In any embodiment, data is collected under the premise of legality and compliance.

[0021] Example 1: This invention proposes a data security management method based on image recognition, which mainly includes the following steps: The system collects the user's facial image information and compares it with facial images in the facial image database. If they match, it collects the user's reading device image information and determines whether the user is using a CD reading device based on the collected reading device image information. If the user is using a CD reading device, the system allows the user to read data; otherwise, it issues an alarm signal.

[0022] In this embodiment of the invention, regardless of whether the collected facial image information of the registrant matches the facial image information in the facial image database, and regardless of whether the registrant uses an optical disc reading device, the registrant's facial image information and the reading device's image information will be saved.

[0023] Because the distance between the user and the image acquisition unit varies each time, some points in the acquired image may be in focus while others may be out of focus, resulting in blurriness. This is caused by the non-constant focal length. Therefore, in this embodiment of the invention, the user's facial image information is sequentially processed by texture extraction, modified Fourier transform, and merged extraction. The modified Fourier transform is used to deblur the multispectral image. Finally, the deblurred image is enhanced by removing low-amplitude structures and artifacts and merging the extracted textures. The specific steps are as follows: Step 1: Assume the collected user face image information L is a set of multispectral defocused blurred images of size X×Y, where x=1,2,..., X represents rows, y=1,2,..., Y represents columns, and the images contain structural covariance extracted from pixels. The covariance matrix is ​​used to represent texture information through second-order statistical relationships between features. The feature image f(x,y) extracted from the collected user face image information L is represented as: ; Among them, L i This represents the i-th blurred image, i.e., the i-th multispectral image. th band, This indicates a filter that uses regional covariance.

[0024] The texture image is obtained by removing feature images from the collected face image information L of the logged-in user. , represented as: ; Here, This refers to the texture extracted from a blurred image by removing features from the image.

[0025] Step 2: modifying the Fourier transform, modifying the Fourier transform by combining edge preservation and detail enhancement, in order to preserve the edge, by smoothing the collected login face image information L to obtain the smooth image g, minimizing L0 gradient through the smooth image g, sharpening the main edge and eliminating the low amplitude structure, and then obtaining the image information image , expressed as: ; Wherein, is the L2 norm, is the L0 norm, indicates the gradient of L, indicates the parameter of controlling the sparsity of L.

[0026] In order to preserve the clear edge, the image is decomposed into a detail layer and a base layer, and since the image enhancement effect is still not enough and some details are missing, the detail layer will be further enlarged to generate an enhanced detail layer G i , the deblurring image V i is calculated by obtaining the blurred image s, the blur kernel u and the image prior weight : ; Wherein, F represents the discrete Fourier transform, F -1 represents the inverse discrete Fourier transform, represents the conjugate of the discrete Fourier transform; G i represents the detail layer; in this case, the Fourier transform is better than the wavelet transform, because the wavelet transform will produce artifacts and over-saturated results.

[0027] Wherein, the blurred image s and the blur kernel u are determined according to the image background noise; the image prior weight is a preset weight parameter determined according to the requirements of the target image; Step 3: merging and extracting, the modified Fourier transform image V i contains low amplitude structure, which is removed by smoothing, and the deblurring image V i is deblurred using a split Bregman filter to obtain the image B i , expressed as: ; In order to enhance the image B i and remove artifacts, deconvolution based on Laplace prior is used to enhance the image B i to obtain the image , expressed as: ; wherein, represents the weight of deconvolution based on Laplace prior, can be =0.08, represents the bilateral filter applied to B i The bilateral filter is used for smoothing processing.

[0028] The texture image is added to the image to obtain the processed login person face image information A i is represented as: ; wherein, represents a weighting factor, =0.5.

[0029] The embodiment of the application uses the linear combination of texture extraction and improved Fourier transform to better defocus deblurring technology of the image, extracts the texture, uses L0 to modify the guide image, gradient projection is used as the input of Fourier transform, after denoising, the texture is linearly combined to obtain a deblurred image, and the edge is retained while the artifact is eliminated in the deblurring process.

[0030] Since the optical disc reading device has strong reflection in the image, the embodiment of the application determines whether it is an optical disc reading device by calculating the probability density of the light spot. Specifically, the collected login reading device image information is processed to obtain the processed login reading device image information, specifically including: Suppose that the collected original reading device image is f(x), the degraded image g(x) and the point spread function h(x) are introduced, and f'(x), g'(x) and h'(x) represent the values normalized to the total, and are represented as: ; ; ; The degraded image and the point spread function are tool functions introduced in the image processing process. Specifically, during the collection, transmission and recording of the image, the quality will be reduced due to various factors (such as lens defocus, camera shaking, atmospheric turbulence, noise, etc.). This process is called degradation. The introduction of these two concepts can establish a mathematical model to describe this process. Once the point spread function that causes degradation is known or estimated, the degradation process can be reversed through algorithms such as inverse filtering, Wiener filtering, Richardson-Lucy iteration, etc. to try to reconstruct the clearest original image from the observed degraded image.

[0031] When the probability density of the light spot existing in the original reading device image at point x1 is: ; The probability density of the light spot image formed at point x2 of the blurred image is: ; The corresponding blurred image is determined according to the image background noise; If there is a light spot at point x1 of the original reading device image, the probability density of the blurred image imaged at point x2 is , which is expressed as: ; The probability density of the original image of the light spot image formed at point x2 of the blurred image is expressed by Bayes' theorem as: ; In the formula, is the probability density of the original image of the light spot image formed at point x2; Only when is greater than a preset value, the reading device used by the login user is a CD reading device.

[0032] Embodiment two: The embodiment of the present application proposes a data security management system based on image recognition, which is composed of a first-level security identification module, a second-level security identification module and a remote monitoring control module, wherein the first-level security identification module is connected with the second-level security identification module, and the first-level security identification module and the second-level security identification module are connected with the remote monitoring control module, and the second-level security identification module is connected with a data storage device (such as a computer, etc.).

[0033] The first-level security identification module is used to identify the face image information of the login user, and the first-level security identification module transmits the collected image information to the remote monitoring control module. If the image information collected by the first-level security identification module is consistent with the face image library information stored therein, the first-level security identification module sends an instruction to the second-level security identification module. The second-level security identification module is used to identify whether the login user uses a CD reading device, and the second-level security identification module transmits the collected image information to the remote monitoring control module. If the second-level security identification module judges that the login user uses a CD reading device, the second-level security identification module sends an instruction to the data storage device, and the data storage device allows the login user to read data.

[0034] Specifically, regardless of whether the first-level security identification module issues an instruction to the second-level security identification module, the first-level security identification module transmits the collected image information to the remote monitoring and control module. Similarly, regardless of whether the second-level security identification module issues an instruction to the data storage device, the second-level security identification module transmits the collected image information to the remote monitoring and control module. Thus, the remote monitoring and control module can record the information of the person attempting to log in and the information of the reading device used.

[0035] Specifically, the remote monitoring and control module saves the information received each time, and if the user logs in without authorization to read the data, an alarm signal is issued.

[0036] The first-level security identification module includes an image acquisition unit, an image processing unit, a face image library, and an instruction sending unit. These units are connected sequentially. The image acquisition unit acquires the original face image of the user and transmits it to the image processing unit. The image processing unit processes the received original image and then transmits it to the face image library, which stores the face image information of authorized users. The face image library matches the received image information. If a match is successful, the instruction sending unit sends an instruction to the second-level security identification module.

[0037] Specifically, since the distance between the user and the image acquisition unit varies each time, some points in the acquired image may be in focus while others may be out of focus, resulting in blurring. This is caused by the non-constant focal length. Therefore, in this invention, the image processing unit sequentially performs texture extraction, modifies the Fourier transform, and merges the extracted textures from the received image. The Fourier transform is modified to deblur the multispectral image. Finally, the deblurred image is enhanced by removing low-amplitude structures and artifacts and merging the extracted textures. The specific steps are as follows: Step 1: Assume the collected user face image information L is a set of multispectral defocused blurred images of size X×Y, where x=1,2,..., X represents rows, y=1,2,..., Y represents columns, and the images contain structural covariance extracted from pixels. The covariance matrix is ​​used to represent texture information through second-order statistical relationships between features. The feature image f(x,y) extracted from the collected user face image information L is represented as: ; Among them, L i This represents the i-th blurred image, i.e., the i-th multispectral image. th band, This indicates a filter that uses regional covariance.

[0038] The texture image is obtained by removing the feature image from the collected login face image information L , is expressed as: ; Here, indicates the texture extracted from the blurred image by removing the features from the image.

[0039] Step 2: Modify the Fourier transform, modify the Fourier transform by combining edge preservation and detail enhancement, in order to preserve the edge, by smoothing the collected login face image information L, to obtain the smoothed image g, minimize L0 by the smoothed image g, sharpen the main edge and eliminate the low-amplitude structure, and further obtain the image information image , is expressed as: ; where, is the L2 norm, is the L0 norm, indicates the gradient of L, indicates the parameter controlling the sparsity of the signal L.

[0040] In order to preserve the clear edge, the image is decomposed into a detail layer and a base layer, since the image enhancement effect is still not enough, and some details are missing, the detail layer will be further enlarged to generate an enhanced detail layer G i , the deblurring image V i is calculated by obtaining the blurred image s, the blur kernel u and the image prior weight : ; where F represents the discrete Fourier transform, F -1 represents the inverse discrete Fourier transform, represents the conjugate of the discrete Fourier transform; G i represents the detail layer; in this case, the Fourier transform is better than the wavelet transform, because the wavelet transform will produce artifacts and over-saturated results.

[0041] where the blurred image s and the blur kernel u are determined according to the image background noise; the image prior weight is a preset weight parameter determined according to the requirements of the target image; Step 3: Merge extraction, the modified Fourier transform image V i contains low-amplitude structures, which are removed using smoothing processing, and the split Bregman filter is used to deblur the deblurred image V i to obtain the image B i , is expressed as: In order to enhance the image B i , and remove artifacts, the image B i is enhanced by deconvolution based on Laplacian prior to obtain the image , which is expressed as: wherein, represents the weight of deconvolution based on Laplacian prior, which can be =0.08, represents a bilateral filter applied to B i for smoothing processing.

[0042] The texture image is added to the image to obtain the processed login person face image information A i , which is expressed as: wherein, represents a weighting factor, =0.5.

[0043] The image processing unit in the first level security recognition module in the application proposes a technology of using texture extraction and improved linear combination of Fourier transform to better defocus and deblur the image, extracts the texture, uses L0 to modify the guide image, gradient projection is used as the input of Fourier transform, after denoising, the texture is linearly combined to obtain a deblurred image, and the edge is retained while the artifacts are eliminated in the deblurring process.

[0044] The second level security recognition module includes an image acquisition unit, a light point probability analysis unit and an instruction sending unit, the first level security recognition module and the second level security recognition module use the same image acquisition unit, the image acquisition unit, the light point probability analysis unit and the instruction sending unit are connected in sequence, the image acquisition unit acquires an image including not only the login person face information but also the image information of the reading device used by the login person, since the optical disc reading device has strong reflectivity in the image, therefore, the light point probability analysis unit judges whether it is the optical disc reading device by calculating the light point existence probability density, if yes, the instruction sending unit sends an instruction to the data storage device.

[0045] Specifically, the collected login person reading device image information is subjected to image processing to obtain the processed login person reading device image information, which specifically includes: Suppose that the collected original reading device image is f(x), the degradation image g(x) and the point spread function h(x) are introduced, f'(x), g'(x) and h'(x) respectively represent the values normalized to the total sum, which is expressed as: ​​​ ; ; ; wherein the degraded image and the point spread function are introduced tools in the image processing process. Specifically, the image will be degraded in quality due to various factors (such as lens defocus, camera shake, atmospheric turbulence, noise, etc.) in the acquisition, transmission, recording process, this process is called degradation, the introduction of these two concepts can establish a mathematical model to describe this process, once the point spread function that causes degradation is known or estimated, the degradation process can be reversed by deconvolution algorithm (such as inverse filtering, Wiener filtering, Richardson-Lucy iteration, etc.) to try to restore the original image as clear as possible from the observed degraded image.

[0046] When the probability density of the light spot existing at point x1 in the original reading device image is: ; The probability density of the light spot image formed at point x2 is: ; According to the image background noise, the corresponding blurred image is determined; If there is a light spot at point x1 in the original reading device image, the probability density of the blurred image imaging at point x2 is , expressed as: ; The probability density of the original image of the light spot image formed at point x2 in the blurred image is expressed by Bayes theorem as: ; In the formula, is the probability density of the original image of the light spot image formed at point x2; Only when is greater than a preset value, the reading device used by the login user is a CD reading device.

[0047] In the present application, the same set of image acquisition unit can realize the recognition of the login person's face information and the recognition of the reading device used, not only guaranteeing the reliability of data reading authorization, but also guaranteeing the security of data transmission after reading. In the first level security identification module, the image processing unit proposes a technology of using texture extraction and improved linear combination of Fourier transform to better defocus and deblur the image, linearly combining the texture to obtain a deblurred image, eliminating artifacts while preserving edges in the deblurring process, and improving the reliability of face recognition; in the second level security identification module, whether to use a disc reading device is judged through the analysis of light points, and the efficiency of reading device identification is improved. The two level security identification modules effectively overcome the image distortion problem caused by the image acquisition unit.

[0048] Finally, it is again stated that the constituent elements / components described in the foregoing embodiments of the present application are only for illustration and are not intended to limit the scope of the present application, and other equivalent components or changes should be within the scope of the present application.

Claims

1. A data security management method based on image recognition, characterized in that: Includes the following steps: Collect facial image information of the user, process the collected facial image information of the user to obtain the processed facial image information of the user; The processed user's facial image information is compared with the authorized facial image information in the facial image database. Only when they match, the user's reading device image information is collected. The collected user's reading device image information is then processed to obtain the processed user's reading device image information. Based on the processed image information of the user's device, determine whether the device is a CD / DVD reader. If so, allow the user to read data; otherwise, issue an alarm signal. The same image acquisition device is used to collect facial image information of the user and image information from the user's reading device.

2. The data security management method based on image recognition according to claim 1, characterized in that: The aforementioned image processing of the collected login user facial image information to obtain processed login user facial image information specifically includes: Step 1: The feature image f(x,y) extracted from the collected face image information L of the logged-in user is represented as: ； Among them, L i This represents the i-th blurred image. This indicates a filter that uses regional covariance; The texture image is obtained by removing feature images from the collected face image information L of the logged-in user. , represented as: ； Step 2: Smooth the collected face image information L of the logged-in user to obtain a smoothed image g. Minimize the gradient of L0 by smoothing the image g, and then obtain the image. , represented as: ； in, It is the L2 norm. It is the L0 norm. Denotes the gradient of L. Parameters representing the sparsity of the control signal L; Image The model is decomposed into detail layers and base layers, by obtaining the blurred image s, the blur kernel u, and the image prior weights. Calculate the deblurred image V i : ； Where F represents the Discrete Fourier Transform, F -1 This represents the inverse discrete Fourier transform. G represents the conjugate of the discrete Fourier transform; i Represents the detail layer; Among them, the blurred image s and the blurred kernel u are determined based on the background noise of the image; the prior image weights are preset weight parameters determined according to the requirements of the target image. Step 3: Use a split Bregman filter For the deblurred image V i Deblurring is performed to obtain image B. i , represented as: ； Image B is deconvolutioned using Laplacian prior. i Enhancement is performed to obtain the image. , represented as: ； in, This represents the weights of the deconvolution based on the Laplace prior. Indicates application to B i A bilateral filter that performs smoothing; Add texture image to image In the process, the processed facial image information of the logged-in user, A, is obtained. i Represented as: ； in, This represents the weighting factor.

3. The data security management method based on image recognition according to claim 1, characterized in that: The aforementioned image processing of the collected login user image information to obtain processed login user image information specifically includes: Assuming the acquired raw image from the reading device is f(x), and introducing the degraded image g(x) and the point spread function h(x), let f'(x), g'(x), and h'(x) represent the values ​​normalized to the sum, expressed as: ； ； ； The probability density of a light spot at point x1 in the original image read by the device is: ； The probability density of forming a spot image at point x2 is: ； The corresponding blurred image is determined based on the background noise of the image. If a light spot exists at point x1 in the original image from the reading device, the probability density of the blurred image forming at point x2 is: , represented as: ； The probability density of the original image that forms a spot image at point x2 in the blurred image is expressed by Bayes' theorem as: ； In the formula, The probability density of the original image from which the light spot image is formed at point x2; Only when When the value exceeds the preset value, the user uses a CD reading device.

4. A data security management system based on image recognition, characterized in that: include: The image acquisition module is used to acquire facial image information of the user or image information from the user's reading device; The first image processing module is used to process the collected facial image information of the registrant to obtain the processed facial image information of the registrant. The face comparison module is used to compare the processed face image information of the registrant with the authorized face image information in the face image database. Only when they match, the image acquisition module is called to acquire the image information of the registrant's reading device. The second image processing module is used to process the image information of the login device that has been collected, and to obtain the processed image information of the login device. The device reading judgment module is used to determine whether the reading device is a CD reading device based on the processed image information of the user reading device. If it is, the user is allowed to read data; otherwise, an alarm signal is issued.

5. A data security management system based on image recognition according to claim 4, characterized in that: In the first image processing module, the following steps are performed: Step 1: The feature image f(x,y) extracted from the collected face image information L of the logged-in user is represented as: ； Among them, L i This represents the i-th blurred image. This indicates a filter that uses regional covariance; The texture image is obtained by removing feature images from the collected face image information L of the logged-in user. , represented as: ； Step 2: Smooth the collected face image information L of the logged-in user to obtain a smoothed image g. Minimize the gradient of L0 by smoothing the image g, and then obtain the image. , represented as: ； in, It is the L2 norm. It is the L0 norm. Denotes the gradient of L. Parameters representing the sparsity of the control signal L; Image The model is decomposed into detail layers and base layers, by obtaining the blurred image s, the blur kernel u, and the image prior weights. Calculate the deblurred image V i : ； Where F represents the Discrete Fourier Transform, F -1 This represents the inverse discrete Fourier transform. G represents the conjugate of the discrete Fourier transform; i Represents the detail layer; Among them, the blurred image s and the blurred kernel u are determined based on the background noise of the image; the prior image weights are preset weight parameters determined according to the requirements of the target image. Step 3: Use a split Bregman filter For the deblurred image V i Deblurring is performed to obtain image B. i , represented as: ； Image B is deconvolutioned using Laplacian prior. i Enhancement is performed to obtain the image. , represented as: ； in, This represents the weights of the deconvolution based on the Laplace prior. Indicates application to B i A bilateral filter that performs smoothing; Add texture image to image In the process, the processed facial image information of the logged-in user, A, is obtained. i Represented as: ； in, This represents the weighting factor.

6. A data security management system based on image recognition according to claim 4, characterized in that: In the second image processing module, the following steps are performed: The aforementioned image processing of the collected login user image information to obtain processed login user image information specifically includes: Assuming the acquired raw image from the reading device is f(x), and introducing the degraded image g(x) and the point spread function h(x), let f'(x), g'(x), and h'(x) represent the values ​​normalized to the sum, expressed as: ； ； ； The probability density of a light spot at point x1 in the original image read by the device is: ； The probability density of forming a spot image at point x2 is: ； The corresponding blurred image is determined based on the background noise of the image. If a light spot exists at point x1 in the original image from the reading device, the probability density of the blurred image forming at point x2 is: , represented as: ； The probability density of the original image that forms a spot image at point x2 in the blurred image is expressed by Bayes' theorem as: ； In the formula, The probability density of the original image from which the light spot image is formed at point x2; Only when When the value exceeds the preset value, the user uses a CD reading device.

7. An electronic device, characterized in that, The electronic device includes: At least one processor; and a memory communicatively connected to the at least one processor; The memory stores a computer program that can be executed by the at least one processor, which is then executed by the at least one processor to enable the at least one processor to perform the data security management method based on image recognition as described in any one of claims 1-3.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the data security management method based on image recognition as described in any one of claims 1-3.

9. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the data security management method based on image recognition as described in any one of claims 1-3.

Images