Search

CN-121838283-B - Image recognition method for cross-domain access authentication

CN121838283BCN 121838283 BCN121838283 BCN 121838283BCN-121838283-B

Abstract

The invention relates to the technical field of image recognition and discloses an image recognition method for cross-domain access authentication, which comprises the steps of obtaining an image to be authenticated uploaded by an image acquisition node, transforming the image to be authenticated to a frequency domain, extracting complex response signals under multiple preset directions and scales, calculating a phase consistency feature map by utilizing real parts and imaginary part response components, generating a phase consistency feature map, counting local phase distribution of the phase consistency feature map, determining phase distribution entropy representing physical properties of an access medium, reducing weight of the phase consistency feature map according to the phase distribution entropy, inhibiting weight of the phase consistency feature map, extracting geometric topological relation of structural anchor points in the inhibited feature map, constructing an identity characterization vector, comparing the identity characterization vector, and identifying and inhibiting phase interference introduced by an unnatural display medium.

Inventors

  • CHEN YING
  • FAN YE
  • LEI JUNLI

Assignees

  • 陕西路恒电子科技有限公司

Dates

Publication Date
20260512
Application Date
20260316

Claims (10)

  1. 1. An image recognition method for cross-domain access authentication, comprising the steps of: step S1, an image to be authenticated uploaded by an image information acquisition node is obtained; s2, transforming an image to be authenticated from a spatial domain to a frequency domain, and performing filtering operation on the frequency domain image by utilizing a multi-scale logarithmic Gaussian filter bank to extract complex frequency response signals of the image to be authenticated in a plurality of preset directions and at a plurality of preset scales, wherein the complex frequency response signals consist of a real part response component and an imaginary part response component; Step S3, calculating the sum of local energy of the real part response component and the imaginary part response component as the local total energy, calculating a modulus value of the sum of response vectors of the real part response component and the imaginary part response component at each pixel position, determining the ratio of the modulus value to the local total energy as the phase consistency measure of the image to be authenticated at each pixel position, and generating a phase consistency feature map according to the phase consistency measure of each pixel position; Step S4, counting phase angle distribution of the phase consistency feature map at each pixel position along with the change of a preset scale, calculating phase distribution entropy representing physical properties of an access medium, performing numerical subtraction operation on the phase distribution entropy and a preset medium discrimination threshold value, and mapping the difference value obtained by the operation into a feature suppression factor for each pixel position; And S5, extracting a local extremum point set from the suppressed feature map to serve as a structure anchor point, constructing an identity characterization vector according to the geometric distribution topological relation of the structure anchor point, executing matching operation on the identity characterization vector and a preset registration template, and outputting an authentication instruction aiming at the image information acquisition node according to a matching operation result.
  2. 2. The method for identifying images for cross-domain access authentication according to claim 1, wherein the step S3 of calculating the phase consistency measure further comprises the steps of obtaining the sum of the component amplitudes of the real part response component and the imaginary part response component under each preset scale, and accumulating the sum of the component amplitudes under each preset scale to obtain the local total energy, wherein the modulus of the sum of the local response vectors is obtained by calculating the sum root of the sum of the square of the sum of the real part response component and the square of the sum of the imaginary part response component under each preset scale, so as to characterize the edge structure of the image to be authenticated by calculating the phase resonance intensity between each preset scale.
  3. 3. The method according to claim 1, wherein in step S2, each filter included in the multi-scale logarithmic gaussian filter set is geometrically multiple distributed in the logarithmic frequency space, and the center frequencies of each filter are mutually covered to lock the invariant feature of the image to be authenticated under different optical resolutions.
  4. 4. The method for cross-domain access authentication as claimed in claim 1, further comprising the step of S6, obtaining a focal length parameter of an optical imaging module in the image information acquisition node, and determining a radial distance from the pixel point to the center of the image Calculating each preset scale Compensation phase term of lower filtering operator The formula is expressed as: , wherein, And (3) with And performing phase pre-hedging processing on the complex frequency response signals by using a compensation phase term to correct phase offset caused by radial distortion of the optical imaging module.
  5. 5. The method according to claim 1, wherein in step S4, the phase distribution entropy is obtained by calculating the phase angle variance of each pixel position at each preset scale, so as to quantify the phase arrangement consistency of the image to be authenticated in the frequency domain space.
  6. 6. The method according to claim 1, wherein in step S4, the step of performing the weight-down operation using the feature suppression factor includes determining that the value of the feature suppression factor is smaller than 1 when the phase distribution entropy is larger than the medium discrimination threshold, and multiplying the feature amplitude of the corresponding position by the feature suppression factor to attenuate the periodic streak feature introduced by the unnatural display medium.
  7. 7. The method for identifying images for cross-domain access authentication according to claim 1, wherein in step S5, constructing an identity token vector based on the geometric distribution topological relation of the structure anchor points comprises extracting local phase information and coordinate information of each structure anchor point, and encoding the local phase information as characteristic elements into the identity token vector according to the arrangement sequence of the coordinate information.
  8. 8. The method for identifying an image for cross-domain access authentication according to claim 1, wherein before step S1, the method further comprises identifying a device type tag of the image information collection node, and calling a corresponding preset channel noise model according to the device type tag, and performing gray-scale average value equalization processing on an original image stream to be collected.
  9. 9. The method for identifying the image used for cross-domain access authentication according to claim 1, wherein the step S2 specifically comprises the steps of mapping an image to be authenticated from a space domain to a frequency domain by utilizing fast Fourier transform, performing point-by-point product operation on the mapped image signal and a logarithmic Gaussian filter operator in the frequency domain, and performing inverse Fourier transform on a result of the product operation to obtain a real response component and an imaginary response component of corresponding directions and scales.
  10. 10. The method for cross-domain access authentication according to claim 1, further comprising step S7 of generating an authorization message containing the encrypted access token and issuing the authorization message to the image information collection node when the authentication command is authentication pass.

Description

Image recognition method for cross-domain access authentication Technical Field The invention relates to an image recognition method for cross-domain access authentication, belonging to the technical field of image recognition. Background In the prior art, a deep neural network is generally adopted to perform feature mapping in an image airspace, in order to attempt to convert pixel distribution of different sensors into a unified feature space through nonlinear transformation, as the access terminal shows hardware heterogeneous characteristics, the sensor sensitivity and the lens resolution have performance differences, so that nonlinear amplitude distortion is generated in the original image, the amplitude response deviation caused by hardware enables the airspace feature extraction process to depend on massive cross-domain alignment data, when the terminal to be authenticated is in a low-illumination environment or an optical resolution limited working condition, the airspace feature mapping function is easy to generate deviation, the recognition accuracy is reduced, the airspace feature distribution is fitted by increasing the model depth or introducing domain countermeasure network, the computing load of edge side equipment is increased, and the interference of hardware optical gain on the image structure semantics cannot be eliminated from the physical mechanism aspect. Besides the inherent performance difference of front-end hardware, the existing access control method also shows logic limitation when processing the consistency of cross-domain signals, for example, china patent with the authority bulletin number of CN114037457B discloses an industrial complex product terminal cross-domain access authentication method based on identity identification, and the registration flow in the terminal moving process is optimized on the protocol layer by utilizing the identification password technology and session key negotiation mechanism, so that the resource loss of inter-domain authentication is effectively reduced, however, the control logic is always preset with the standardization and purity of front-end acquired data, the physical gain interference in the imaging process is ignored, when the signal to be authenticated is influenced by the thermal noise of a sensor, the directional motion blur or the geometric distortion of a wide-angle lens, the handshake logic simply relying on an upper protocol is difficult to compensate the precision shrinkage of an original image on the physical characterization, meanwhile, the prior scheme generally lacks the perception capability of the physical attribute of an access medium, the security is not enough when the system is faced with the physical layer attack of a distributed heterogeneous environment, and the problem that the system is incapable of identifying and blocking the injection of non-natural phase signals such as the inversion of a screen in an authentication frame is solved, the reliability is caused, and the reliability of the distributed heterogeneous environment is improved, and the reliability is not stable, and the reliability of the spatial gain is improved, and the reliability of the spatial domain is improved. Therefore, how to construct an image recognition mechanism which can decouple the optical characteristics of terminal hardware and has environmental adaptability, and realize identity authentication in heterogeneous terminal environments becomes the technical problem to be solved by the invention. Disclosure of Invention In order to solve the problems in the background technology, the technical scheme of the invention is as follows, an image recognition method for cross-domain access authentication comprises the following steps: step S1, an image to be authenticated uploaded by an image information acquisition node is obtained; s2, transforming an image to be authenticated from a spatial domain to a frequency domain, and performing filtering operation on the frequency domain image by utilizing a multi-scale logarithmic Gaussian filter bank to extract complex frequency response signals of the image to be authenticated in a plurality of preset directions and at a plurality of preset scales, wherein the complex frequency response signals consist of a real part response component and an imaginary part response component; Step S3, calculating the sum of local energy of the real part response component and the imaginary part response component as the local total energy, calculating a modulus value of the sum of response vectors of the real part response component and the imaginary part response component at each pixel position, determining the ratio of the modulus value to the local total energy as the phase consistency measure of the image to be authenticated at each pixel position, and generating a phase consistency feature map according to the phase consistency measure of each pixel position; Step S4, counting phase angle distribution of the p