CN-121982212-A - Cable three-dimensional imaging method based on tensor coupling and physical constraint

CN121982212ACN 121982212 ACN121982212 ACN 121982212ACN-121982212-A

Abstract

The cable three-dimensional imaging method based on tensor coupling and physical constraint comprises the steps of performing tensor decomposition on terahertz time-domain spectrum data and X-ray projection data of a cable to obtain a core tensor and a modal base matrix, registering terahertz three-dimensional imaging data and X-ray three-dimensional imaging data of the cable based on the core tensor and the modal base matrix, generating three-dimensional potential distribution of the cable by utilizing a pre-built joint optimization objective function for physical constraint according to the registered terahertz three-dimensional imaging data and the registered X-ray three-dimensional imaging data, generating dynamic weights through a double-branch convolution network based on the three-dimensional potential distribution, and repairing by adopting a weighted difference value and a context sensing network according to the dynamic weights to obtain a three-dimensional image of the cable. Therefore, deep cooperative information among different modal data is fully mined, and the structural integrity and detail definition of the image are obviously enhanced.

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. A method of cable three-dimensional imaging based on tensor coupling and physical constraints, the method comprising: Performing tensor decomposition on terahertz time-domain spectrum data and X-ray projection data of the cable to obtain a core tensor and a modal base matrix; registering terahertz three-dimensional imaging data and X-ray three-dimensional imaging data of the cable based on the core tensor and the modal base matrix; generating three-dimensional potential distribution of the cable by utilizing a pre-constructed joint optimization objective function for physical constraint according to the registered terahertz three-dimensional imaging data and the registered X-ray three-dimensional imaging data; Based on the three-dimensional potential distribution, generating dynamic weights through a double-branch convolution network, and repairing by adopting a weighted difference value and a context awareness network according to the dynamic weights to obtain a three-dimensional image of the cable.
2. The method of cable three-dimensional imaging based on tensor coupling and physical constraints of claim 1, wherein the step of registering terahertz three-dimensional imaging data and X-ray three-dimensional imaging data of the cable based on the core tensor and the modal basis matrix comprises: Performing wavelet decomposition on the terahertz three-dimensional imaging data and the X-ray three-dimensional imaging data to obtain a terahertz local coefficient and an X-ray local coefficient; based on the core tensor, calculating wavelet domain self-adaptive weights by adopting the terahertz local coefficients and the X-ray local coefficients; utilizing the modal base matrix to align the frequency domain phase of the terahertz three-dimensional imaging data with the frequency domain phase of the X-ray three-dimensional imaging data, and then calculating a frequency domain phase cross-correlation function by combining the wavelet domain self-adaptive weight; And performing frequency domain displacement in the frequency domain phase cross-correlation function to obtain registered terahertz three-dimensional imaging data and registered X-ray three-dimensional imaging data.
3. The method for three-dimensional imaging of a cable based on tensor coupling and physical constraints according to claim 2, characterized in that the expression of the frequency domain phase cross correlation function is: Wherein, the Representing the frequency domain phase cross-correlation function, The frequency coordinates are represented as such, Representing a three-dimensional fourier transform, Representing the inverse three-dimensional fourier transform, Representing the terahertz three-dimensional imaging data, Representing the X-ray three-dimensional imaging data, Representing the spectral complex conjugate of the X-ray three-dimensional imaging data, The product of the Hadamard is represented, Representing wavelet domain adaptive weights, and , Representing the terahertz local coefficient, Representing the local coefficients of said X-rays, The scale parameters representing the wavelet decomposition are, Representing the local contrast standard deviation.
4. The method of claim 1, wherein the step of generating the three-dimensional potential distribution of the cable from the registered terahertz three-dimensional imaging data and the registered X-ray three-dimensional imaging data using a pre-constructed joint optimization objective function for physical constraints comprises: extracting terahertz frequency domain dielectric constants from the registered terahertz three-dimensional imaging data; And carrying out iterative solution by using a pre-constructed joint optimization objective function through an alternate direction multiplier method according to the registered X-ray three-dimensional imaging data and the terahertz frequency domain dielectric constant until convergence, so as to obtain the three-dimensional potential distribution of the cable.
5. The method of cable three-dimensional imaging based on tensor coupling and physical constraints according to claim 1, wherein the step of generating dynamic weights by a two-branch convolution network based on the three-dimensional potential distribution comprises: And respectively extracting terahertz characteristics and X-ray characteristics from the three-dimensional potential distribution by utilizing the double-branch convolution network, and performing channel splicing on the terahertz characteristics and the X-ray characteristics, and then obtaining the dynamic weight through an activation function.
6. The method for three-dimensional imaging of a cable based on tensor coupling and physical constraints according to claim 1, wherein said step of obtaining a three-dimensional image of said cable using weighted difference and context-aware network repair according to said dynamic weights comprises: carrying out weighted summation on the registered terahertz three-dimensional imaging data and the registered X-ray three-dimensional imaging data by utilizing the dynamic weights, and carrying out convolution smoothing processing on the weighted summation result by combining a kernel function to obtain a low-resolution fusion image; And converting the low-resolution fusion image into a high-resolution three-dimensional image of the cable based on a generator objective function of the context-aware network.
7. The method of cable three-dimensional imaging based on tensor coupling and physical constraints according to claim 6, wherein the expression of the kernel function is: Wherein, the The function of the kernel is represented by a function of the kernel, Representing pixel offset, for controlling interpolation weight distribution, Representing the offset distance of the current interpolation point along the x-axis relative to the nearest integer pixel, Representing the offset distance of the current interpolation point along the y-axis relative to the nearest integer pixel.
8. A cable three-dimensional imaging apparatus based on tensor coupling and physical constraints, the apparatus comprising: the core tensor determining module is used for performing tensor decomposition on terahertz time-domain spectrum data and X-ray projection data of the cable to obtain a core tensor and a modal base matrix; An imaging data registration module for registering terahertz three-dimensional imaging data and X-ray three-dimensional imaging data of the cable based on the core tensor and the modal base matrix; the potential distribution generation module is used for generating three-dimensional potential distribution of the cable by utilizing a pre-constructed joint optimization objective function for physical constraint according to the registered terahertz three-dimensional imaging data and the registered X-ray three-dimensional imaging data; And the three-dimensional image determining module is used for generating dynamic weights through a double-branch convolution network based on the three-dimensional potential distribution, and adopting weighted difference values and context awareness network restoration according to the dynamic weights to obtain the three-dimensional image of the cable.
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 tensor coupling and physical constraints of any one of claims 1 to 7.
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 tensor coupling and physical constraints as recited in any one of claims 1 to 7.

Description

Cable three-dimensional imaging method based on tensor coupling and physical constraint Technical Field The application relates to the technical field of cable detection, in particular to a cable three-dimensional imaging method based on tensor coupling and physical constraint. Background With the continuous improvement of urban underground cable laying density, efficient and accurate detection of the urban underground cable laying density becomes an important means for guaranteeing stable operation of power and communication systems. Therefore, the multi-mode data fusion technology is introduced into the field of underground cable detection and is used for integrating sensor data of different types such as infrared, electromagnetic, ultrasonic and the like so as to improve the comprehensiveness and accuracy of fault identification. The existing multi-mode fusion method improves the quality of the detected image to a certain extent and enhances the perception capability of abnormal characteristics of the cable. However, the method often fails to deeply mine the cooperative information among various modal data, so that the fusion result has the problem of low information utilization rate. In addition, the partial fusion algorithm has a disadvantage in the definition of the generated image. Disclosure of Invention The present application aims to solve at least one of the above technical drawbacks, and in particular, to solve the technical drawbacks of the prior art that the fusion result has low information utilization rate and insufficient image definition. In a first aspect, the present application provides a cable three-dimensional imaging method based on tensor coupling and physical constraints, the method comprising: Performing tensor decomposition on terahertz time-domain spectrum data and X-ray projection data of the cable to obtain a core tensor and a modal base matrix; registering terahertz three-dimensional imaging data and X-ray three-dimensional imaging data of the cable based on the core tensor and the modal basis matrix; generating three-dimensional potential distribution of the cable by utilizing a pre-constructed joint optimization objective function for physical constraint according to the registered terahertz three-dimensional imaging data and the registered X-ray three-dimensional imaging data; based on three-dimensional potential distribution, generating dynamic weights through a double-branch convolution network, and repairing by adopting a weighted difference value and a context awareness network according to the dynamic weights to obtain a three-dimensional image of the cable. In one embodiment, the step of registering terahertz three-dimensional imaging data and X-ray three-dimensional imaging data of the cable based on the core tensor and the modal basis matrix comprises: performing wavelet decomposition on the terahertz three-dimensional imaging data and the X-ray three-dimensional imaging data to obtain a terahertz local coefficient and an X-ray local coefficient; Based on the core tensor, calculating wavelet domain self-adaptive weights by adopting terahertz local coefficients and X-ray local coefficients; The frequency domain phase of the terahertz three-dimensional imaging data and the frequency domain phase of the X-ray three-dimensional imaging data are aligned by using a modal base matrix, and then a frequency domain phase cross-correlation function is calculated by combining wavelet domain self-adaptive weights; and performing frequency domain displacement in the frequency domain phase cross-correlation function to obtain registered terahertz three-dimensional imaging data and registered X-ray three-dimensional imaging data. In one embodiment, the expression of the frequency domain phase cross correlation function is: Wherein, the Representing the frequency domain phase cross-correlation function,The frequency coordinates are represented as such,Representing a three-dimensional fourier transform,Representing the inverse three-dimensional fourier transform,Representing terahertz three-dimensional imaging data,Representing the three-dimensional imaging data of the X-rays,Representing the spectral complex conjugate of the X-ray three-dimensional imaging data,The product of the Hadamard is represented,Representing wavelet domain adaptive weights, and,Representing the terahertz local coefficient,The local coefficients of the X-rays are represented,The scale parameters representing the wavelet decomposition are,Representing the local contrast standard deviation. In one embodiment, the step of generating a three-dimensional potential distribution of the cable from the registered terahertz three-dimensional imaging data and the registered X-ray three-dimensional imaging data using a pre-constructed joint optimization objective function for physical constraints includes: extracting terahertz frequency domain dielectric constants from the registered terahertz three-dimensional imaging