Search

CN-121982211-A - Cable three-dimensional imaging method based on multi-modal characteristics and physical constraints

CN121982211ACN 121982211 ACN121982211 ACN 121982211ACN-121982211-A

Abstract

The cable three-dimensional imaging method based on the multi-modal characteristics and the physical constraints comprises the steps of generating a three-dimensional frequency domain-space domain coupling field according to a terahertz image and an X-ray image for detecting a cable, performing tensor decomposition on the three-dimensional frequency domain-space domain coupling field, taking a core tensor as a cross-modal characteristic, calculating fusion weights through the cross-modal characteristic, weighting and fusing the terahertz image and an X-ray data image by using the fusion weights to generate a fusion image, determining an objective function for the physical constraints, performing iterative optimization on the fusion image until convergence by using the fusion weights based on the objective function to obtain the objective image, and performing parallel acceleration of objective image reconstruction by using quantum derivatization to obtain the three-dimensional image of the cable. Therefore, the depth integration of the structural information and the material attribute in the images of different modes is realized, and the imaging quality of the three-dimensional image is improved.

Inventors

  • TANG QI
  • CHEN ZHIPING

Assignees

  • 广东电网有限责任公司佛山供电局

Dates

Publication Date
20260505
Application Date
20260127

Claims (10)

  1. 1. A cable three-dimensional imaging method based on multi-modal characteristics and physical constraints, the method comprising: Generating a three-dimensional frequency domain-space domain coupling field according to a terahertz image and an X-ray image for detecting the cable, performing tensor decomposition on the three-dimensional frequency domain-space domain coupling field, and taking a core tensor as a cross-modal characteristic; After the cross-modal feature is used for calculating the fusion weight, the terahertz image and the X-ray data image are subjected to weighted fusion by using the fusion weight to generate a fusion image; determining an objective function for physical constraint, and performing iterative optimization on the fusion image by adopting the fusion weight based on the objective function until convergence to obtain an objective image; and accelerating the reconstruction of the target image by utilizing quantum derivatization in parallel to obtain a three-dimensional image of the cable.
  2. 2. The method of claim 1, wherein the step of generating a three-dimensional frequency-spatial coupled field from the terahertz image and the X-ray image for detecting the cable comprises: Acquiring original terahertz data and original X-ray data for detecting the cable; Constructing a terahertz attenuation model based on the original terahertz data, and solving a terahertz image for representing the cable frequency domain attenuation; Constructing an X-ray absorption model based on the original X-ray data, and solving an X-ray image for representing the spatial absorption of the cable; And coupling the terahertz image and the X-ray image to obtain the three-dimensional frequency domain-space domain coupling field.
  3. 3. The method of cable three-dimensional imaging based on multi-modal features and physical constraints of claim 1, wherein the step of computing fusion weights from the cross-modal features comprises: calculating an adjustment factor according to the cross-modal characteristics; And based on the adjusting factors, carrying out self-adaptive matching on the terahertz image and the X-ray image to obtain the fusion weight.
  4. 4. A method of cable three-dimensional imaging based on multi-modal features and physical constraints according to claim 3, wherein the step of calculating an adjustment factor from the cross-modal features comprises: the adjustment factor is calculated using the following expression: Wherein, the The said regulation factor is indicated as such, Representing the hyperbolic tangent function, Characteristic weights representing the terahertz images, Characteristic weights representing the X-ray images, And From the cross-modal characteristics of the device, Representing the local gradient of the X-ray image, Representing the local gradient standard deviation.
  5. 5. The method for cable three-dimensional imaging based on multi-modal features and physical constraints according to claim 3, wherein the step of adaptively matching the terahertz image and the X-ray image based on the adjustment factors to obtain the fusion weights comprises: the fusion weights were calculated using the following expression: Wherein, the The fusion weights are represented as such and, The said regulation factor is indicated as such, Representing the local gradient of the X-ray image, Representing the coefficient of smoothing and the coefficient of smoothing, Representing the terahertz image.
  6. 6. The method of three-dimensional imaging of a cable based on multimodal features and physical constraints according to claim 1, characterized in that the expression of the objective function is: Wherein, the The image of the fusion is represented by a representation, The representation of the observation model is given, Representing the actual observed data of the image, The weight parameters representing the total variation regularization term, Represents the total variation regularization term, an , The fusion weights are represented as such and, Representation of A gradient in the x-direction, Representation of A gradient in the y-direction of the sample, Representation of A gradient in the z-direction is provided, The weight coefficients representing the a priori constraint terms, Representing the cylindrical symmetry constraint operator of the cable, Representing a priori model of cable insulation thickness, etc.
  7. 7. The method for three-dimensional imaging of a cable based on multimodal features and physical constraints according to claim 1, characterized in that said step of accelerating the reconstruction of said target image with quantum derivatization in parallel, obtaining a three-dimensional image of said cable, comprises: Mapping the target image to a quantum state space, and carrying out parallel optimization on the quantum amplitude of each quantum state in the quantum state space based on a quantum search algorithm; and carrying out acceleration measurement on the optimized quantum state, and reconstructing a three-dimensional image of the cable according to a measurement result.
  8. 8. A cable three-dimensional imaging device based on multi-modal characteristics and physical constraints, the device comprising: The cross-modal feature determining module is used for generating a three-dimensional frequency domain-space domain coupling field according to the terahertz image and the X-ray image for detecting the cable, performing tensor decomposition on the three-dimensional frequency domain-space domain coupling field, and taking a core tensor as a cross-modal feature; The fusion image generation module is used for weighting and fusing the terahertz image and the X-ray data image by utilizing the fusion weight after calculating the fusion weight through the cross-modal characteristic to generate a fusion image; the target image determining module is used for determining a target function for physical constraint, and carrying out iterative optimization on the fusion image by adopting the fusion weight based on the target function until convergence to obtain a target image; and the three-dimensional image determining module is used for accelerating the reconstruction of the target image in parallel by utilizing quantum derivatization to obtain a three-dimensional image of the cable.
  9. 9. A storage medium having stored therein computer readable instructions which, when executed by one or more processors, cause the one or more processors to perform the steps of the cable three-dimensional imaging method based on multimodal features and physical constraints as in any one of claims 1 to 7.
  10. 10. A computer device includes one or more processors and a memory; Stored in the memory are computer readable instructions which, when executed by the one or more processors, perform the steps of the cable three-dimensional imaging method based on multimodal features and physical constraints as claimed in any one of claims 1 to 7.

Description

Cable three-dimensional imaging method based on multi-modal characteristics and physical constraints Technical Field The application relates to the technical field of underground cables, in particular to a cable three-dimensional imaging method based on multi-mode characteristics and physical constraints. Background In recent years, urban underground space resources are largely developed and utilized, the laying scale of underground cables is continuously enlarged, and the operation and maintenance difficulties are also remarkably increased. The traditional detection means have the problems of low efficiency, incomplete information acquisition and the like in a complex environment, so that more researches are promoted to focus on intelligent perception technology based on images. In this context, it has become a mainstream trend that a plurality of sensors cooperatively collect different types of image data and enhance the recognition capability of a cable target through fusion processing. The existing method tries to fuse the modal data such as thermal infrared, visible light, geological radar and the like, extracts multi-source characteristics through a depth network, and realizes comprehensive judgment of cable states and surrounding environments. Despite the advances made in the prior art, there is still room for optimization in the fusion strategy. Some methods have insufficient modeling of heterogeneity among the modal data, and the information advantages of the heterogeneous modeling can not be effectively integrated, so that the expression capacity of the fusion image is limited. In addition, in actual deployment, the problems of insufficient image detail recovery and slow model response speed still exist, and the application value of the method in high-efficiency and real-time cable detection is restricted. Disclosure of Invention The present application aims to solve at least one of the above technical drawbacks, and in particular, to fully utilize complementarity of multi-modal data in the prior art. In a first aspect, the present application provides a cable three-dimensional imaging method based on multi-modal characteristics and physical constraints, the method comprising: Generating a three-dimensional frequency domain-space domain coupling field according to the terahertz image and the X-ray image for detecting the cable, performing tensor decomposition on the three-dimensional frequency domain-space domain coupling field, and taking a core tensor as a cross-modal characteristic; after the cross-modal characteristics are used for calculating the fusion weight, the terahertz image and the X-ray data image are subjected to weighted fusion by using the fusion weight, so that a fusion image is generated; Determining an objective function for physical constraint, and performing iterative optimization on the fusion image by adopting fusion weights based on the objective function until convergence to obtain an objective image; And (3) accelerating target image reconstruction by utilizing quantum derivative parallelism to obtain a three-dimensional image of the cable. In one embodiment, the step of generating a three-dimensional frequency-space domain coupled field from the terahertz image and the X-ray image for detecting the cable includes: acquiring original terahertz data and original X-ray data for detecting the cable; Constructing a terahertz attenuation model based on original terahertz data, and solving a terahertz image for representing cable frequency domain attenuation; Constructing an X-ray absorption model based on original X-ray data, and solving an X-ray image for representing the spatial absorption of a cable; and coupling the terahertz image with the X-ray image to obtain a three-dimensional frequency domain-space domain coupling field. In one embodiment, the step of computing the fusion weights by cross-modal features includes: Calculating an adjustment factor according to the cross-modal characteristics; and carrying out self-adaptive matching on the terahertz image and the X-ray image based on the adjustment factors to obtain fusion weights. In one embodiment, the step of calculating the adjustment factor based on the cross-modal characteristics comprises: The adjustment factor is calculated using the following expression: Wherein, the The expression "adjustment factor" is used to indicate,Representing the hyperbolic tangent function,The feature weights of the terahertz images are represented,Representing the characteristic weights of the X-ray image,AndDerived from the cross-modal characteristics,Representing the local gradient of the X-ray image,Representing the local gradient standard deviation. In one embodiment, the step of adaptively matching the terahertz image and the X-ray image based on the adjustment factor to obtain the fusion weight includes: the fusion weights were calculated using the following expression: Wherein, the The fusion weights are represented as such,T