CN-121998852-A - Pipeline deformation image noise reduction method

CN121998852ACN 121998852 ACN121998852 ACN 121998852ACN-121998852-A

Abstract

The invention provides a noise reduction method for a pipeline deformation image, and relates to the technical field of pipeline deformation measurement; the method comprises the following steps: making and pasting a mark target, calibrating image noise, constructing a reference image set and a noise image set, constructing and training a digital image noise model, and outputting a target pipeline noise reduction image through the trained digital image noise model; the invention recognizes corresponding pipeline deformation images under different noise environments through the mark target information, constructs the digital image noise model capable of automatically carrying out noise reduction treatment on the pipeline deformation images, efficiently and accurately realizes accurate deformation calculation of the pipeline deformation images under different environment states, and can be suitable for optical deformation measurement under complex illumination environments and deformation measurement under other photoelectric noise interference.

Inventors

Bi Fengdong
LIU LEI
FENG LIDE
REN GUOQI
YANG FENGPING
ZHANG LIANG
WANG MINHUI
HUANG RONG
KONG XIANGYU

Assignees

中国石油天然气股份有限公司
中石油昆仑燃气有限公司

Dates

Publication Date: 20260508
Application Date: 20241106

Claims (10)

1. The noise reduction method for the pipeline deformation image is characterized by comprising the following steps of: s1, manufacturing a plurality of mark targets for identifying different pipelines to be tested, respectively adhering the mark targets to the surfaces of the different pipelines to be tested and shooting to obtain a plurality of original images of the pipelines to be tested containing the mark targets; S2, measuring the distance between the pipelines to be measured through a distance measuring tool, measuring the strain data of the pipelines to be measured through strain gauges which are preset on the pipelines to be measured containing the mark targets, obtaining deformation distance data and deformation strain data corresponding to the pipelines to be measured, and shooting the deformed pipelines to be measured according to different shooting parameter combinations to obtain deformation images of the pipelines to be measured corresponding to the different shooting parameter combinations; S3, strain calculation is carried out based on all original images of the pipeline to be measured and deformation images of the pipeline to be measured corresponding to the original images, deformation interval data and deformation strain data corresponding to the deformation images of the pipeline to be measured obtained through calculation are compared with deformation interval data and deformation strain data corresponding to the pipeline to be measured, which are measured by the same pipeline to be measured, the deformation images of the pipeline to be measured, which are similar to the measurement data, are selected from comparison results to serve as reference images, a reference image set is built through a plurality of reference images, the deformation images of the rest pipeline to be measured are used as noise images, and a noise image set is built through a plurality of noise images; S4, constructing a digital image noise model based on a neural network, and training the digital image noise model through the reference image set and the noise image set to obtain a trained digital image noise model; S5, carrying out noise reduction processing on the imported target pipeline image through the trained digital image noise model, and outputting the target pipeline noise reduction image.
2. The method of noise reduction of pipeline deformation images according to claim 1, wherein the marking targets manufactured for identifying different pipelines to be tested comprise: And arranging the two strips with different colors in a preset sequence along the horizontal axial direction of the pipeline to be tested to obtain the mark target.
3. The method of pipeline deformation image noise reduction according to claim 1, wherein the different combinations of photographing parameters include combinations of different parameters of focal length, exposure and gain.
4. The method for reducing noise in a deformed image of a pipeline according to claim 1, wherein the performing strain calculation based on all original images of the pipeline to be measured and the deformed images of the pipeline to be measured corresponding to the original images of the pipeline to be measured includes: strain calculation is carried out on an original image of a pipeline to be measured and a deformation image of the pipeline to be measured corresponding to the original image through a strain mapping formula, deformation interval data and deformation strain data corresponding to the deformation image of the pipeline to be measured are obtained, and the strain mapping formula is as follows: Wherein x i 'is the x coordinate of the deformed image, y i ' is the y coordinate of the deformed image, d x is the x-direction gradient deformation of the pipeline to be tested, d y is the y-direction gradient deformation of the pipeline to be tested, u is the x-direction displacement deformation of the pipeline to be tested, v is the y-direction displacement deformation of the pipeline to be tested, For the x-direction rotation deformation of the pipeline to be tested relative to the x-axis, For the rotational deformation of the y direction of the pipeline to be tested relative to the x axis, For the rotational deformation of the x direction of the pipeline to be tested relative to the y axis, The y direction of the pipeline to be tested is deformed in a rotating way relative to the y axis.
5. The method for reducing noise of a deformed image of a pipeline according to claim 1, wherein comparing the deformation interval data and the deformation strain data corresponding to the deformed image of the pipeline to be measured obtained by calculation with the deformation interval data and the deformation strain data corresponding to the pipeline to be measured obtained by measuring the same pipeline to be measured respectively, and selecting the deformed image of the pipeline to be measured close to the measured data from the comparison result as the reference image, includes: And respectively carrying out difference operation on deformation interval data and deformation strain data corresponding to the calculated deformation images of the pipeline to be detected and deformation interval data and deformation strain data corresponding to the pipeline to be detected, which are measured by the same pipeline to be detected, taking the absolute value of the difference value to obtain the absolute value of the difference value corresponding to each deformation image of the pipeline to be detected, respectively selecting the absolute value of the difference value corresponding to each deformation image of the pipeline to be detected and a preset judgment threshold value, and selecting the deformation image of the pipeline to be detected corresponding to the absolute value of the difference value which does not exceed the judgment threshold value as a reference image.
6. The method for noise reduction of a pipeline deformation image according to claim 1, wherein the constructing a digital image noise model based on a convolutional neural network, training the digital image noise model through the reference image set and the noise image set, and obtaining a trained digital image noise model comprises: constructing a digital image noise model based on a neural network, wherein the digital image noise model comprises three layers of neural networks, and the first layer of neural network and the second layer of neural network are respectively connected with a third layer of neural network; Taking the reference image set as the input of a first layer neural network, extracting a mark target image in each reference image of the reference image set through a convolution kernel function and a linear transformation layer function of the first layer neural network, and inputting the mark target image in each reference image and a corresponding reference image into a third layer neural network; Taking the noise image set as the input of a second layer neural network, extracting a mark target image in each noise image of the noise image set through a convolution kernel function and a linear conversion layer function of the second layer neural network, and inputting the mark target image in each noise image and the corresponding noise image into a third layer neural network; And the third layer neural network performs noise filtering training on the mark target images in each reference image and the corresponding reference images thereof and the mark target images in each noise image and the corresponding noise images thereof through a noise filtering function to obtain a trained digital image noise model.
7. The method of noise reduction of a pipeline deformation image according to claim 6, wherein the convolution kernel is: Where i=0,..h-s, j=0,..w-s, t=0,..n-1, s are window function sizes of convolution kernel functions of the convolution layers in the first layer neural network or the second layer neural network, n is a convolution level of the first layer neural network or the second layer neural network, m is a layer number of a previous layer, x is an output of the previous layer, K is an output of the current layer, I is an image layer convolution result, h and w are resolutions of an input reference image or noise image, b is convolution deviation, and r, K and l are a level of the convolution neural network and window function parameters, respectively.
8. The method of pipeline deformation image noise reduction according to claim 6, wherein the linear transformation layer function is: Where α is a translation process of the image when x is less than 0, and x is a retention of the original data when x is greater than or equal to 0, which is randomly generated by gaussian to increase the convergence speed.
9. The method of noise reduction of a pipeline deformation image according to claim 6, wherein the noise filtering function is: Wherein M is the total layer number, M is the current layer number, ω is the weight matrix of the reference image or the noise image, and x is the hierarchical output matrix of the reference image or the noise image.
10. The method for noise reduction of pipeline deformation image according to claim 1, wherein the measuring the distance between the pipelines to be measured by the distance measuring tool, and measuring the strain data of the pipelines to be measured by the strain gauge preset on the pipelines to be measured containing the mark target, to obtain the deformation distance data and the deformation strain data corresponding to the pipelines to be measured, further comprises: The method comprises the steps of measuring the distance between all pipelines to be measured through a distance measuring tool according to set times, measuring the strain data of all the pipelines to be measured through strain gauges which are preset on all the pipelines to be measured containing mark targets according to set times, obtaining a plurality of deformation distance initial data and a plurality of deformation strain data corresponding to all the pipelines to be measured, averaging the deformation distance initial data to obtain deformation distance data corresponding to all the pipelines to be measured, and averaging the deformation strain data to obtain the deformation strain data corresponding to all the pipelines to be measured.

Description

Pipeline deformation image noise reduction method Technical Field The invention mainly relates to the technical field of pipeline deformation measurement, in particular to a pipeline deformation image noise reduction method. Background The pipeline system is an important component of an industrial system and is widely used in the fields of energy, chemical industry and the like. The existing detection method is mainly used for evaluating and detecting the health state of the pipeline by using traditional modes such as ultrasonic and contact sensors, and is low in detection efficiency and cannot meet the current rapid and efficient detection requirements for a complex pipeline system consisting of multiple pipelines by using an operating vehicle, a robot or a person. The existing optical non-contact measurement means can obtain deformation information of a plurality of pipelines in a measured area, but in the actual engineering process, due to the interference of ambient light and the interference of other photoelectric noise, image noise exists in the acquired images, so that the measurement accuracy is low, the actual engineering requirements cannot be met, and the development requirements of the domestic energy and chemical industry on pipeline safety are severely restricted. Disclosure of Invention The invention aims to solve the technical problem of providing a noise reduction method for a deformed image of a pipeline aiming at the defects in the prior art. The technical scheme for solving the technical problems is as follows, the pipeline deformation image noise reduction method comprises the following steps: s1, manufacturing a plurality of mark targets for identifying different pipelines to be tested, respectively adhering the mark targets to the surfaces of the different pipelines to be tested and shooting to obtain a plurality of original images of the pipelines to be tested containing the mark targets; S2, measuring the distance between the pipelines to be measured through a distance measuring tool, measuring the strain data of the pipelines to be measured through strain gauges which are preset on the pipelines to be measured containing the mark targets, obtaining deformation distance data and deformation strain data corresponding to the pipelines to be measured, and shooting the deformed pipelines to be measured according to different shooting parameter combinations to obtain deformation images of the pipelines to be measured corresponding to the different shooting parameter combinations; S3, strain calculation is carried out based on all original images of the pipeline to be measured and deformation images of the pipeline to be measured corresponding to the original images, deformation interval data and deformation strain data corresponding to the deformation images of the pipeline to be measured obtained through calculation are compared with deformation interval data and deformation strain data corresponding to the pipeline to be measured, which are measured by the same pipeline to be measured, the deformation images of the pipeline to be measured, which are similar to the measurement data, are selected from comparison results to serve as reference images, a reference image set is built through a plurality of reference images, the deformation images of the rest pipeline to be measured are used as noise images, and a noise image set is built through a plurality of noise images; S4, constructing a digital image noise model based on a neural network, and training the digital image noise model through the reference image set and the noise image set to obtain a trained digital image noise model; S5, carrying out noise reduction processing on the imported target pipeline image through the trained digital image noise model, and outputting the target pipeline noise reduction image. The method has the beneficial effects that the corresponding pipeline deformation images under different noise environments are identified through the mark target information, the digital image noise model capable of automatically carrying out noise reduction treatment on the pipeline deformation images is constructed, the accurate deformation calculation of the pipeline deformation images under different environment states is efficiently and accurately realized, and the method is suitable for optical deformation measurement under complex illumination environments and deformation measurement under other photoelectric noise interference. Drawings FIG. 1 is a flowchart of a method for reducing noise of a deformed image of a pipeline according to an embodiment of the present invention; Fig. 2 is a schematic diagram of a marker target provided in an embodiment of the present invention. Detailed Description The principles and features of the present invention are described below with reference to the drawings, the examples are illustrated for the purpose of illustrating the invention and are not to be construed as limiting the sc