CN-121999360-A - High-definition image-based extraction method and device for surface thickened oil steam injection pipe network, electronic equipment and storage medium

CN121999360ACN 121999360 ACN121999360 ACN 121999360ACN-121999360-A

Abstract

The invention relates to the technical field of pipeline identification, in particular to a method, a device, electronic equipment and a storage medium for extracting a surface thickened oil steam injection pipe network based on a high-definition image. According to the invention, the convolutional neural network is used as a main network and is fused with various network structures to establish a feature segmentation model, so that the characteristic representation capability is improved, an accurate feature segmentation image can be extracted from an oilfield remote sensing image to be identified, and then the feature segmentation image is analyzed based on edge detection and Hough transformation to obtain an accurate detection result of the surface thickened oil steam injection pipeline in the feature segmentation image, and reliable data support is provided for the surface thickened oil steam injection pipeline to manage.

Inventors

ZHAO JINLING
HUANG ZIXUAN
JIN TAO
YU LU
SUN YINGFANG
MAO XIN
HUANG JIAN
LIANG ZIJUN
WANG HAN
YANG YONGGANG
XIE MENGTAO
XIN LILI
GE XINGXING
Bahezati Nurtay

Assignees

中国石油天然气股份有限公司

Dates

Publication Date: 20260508
Application Date: 20241108

Claims (10)

1. The method for extracting the surface thickened oil steam injection pipe network based on the high-definition image is characterized by comprising the following steps of: acquiring an oilfield remote sensing image to be identified, inputting the oilfield remote sensing image to be identified into a feature segmentation model to obtain a feature segmentation image, wherein the feature segmentation model is obtained by deep learning a fusion neural network model by utilizing a plurality of samples, each sample in the plurality of samples comprises the oilfield remote sensing image and a corresponding feature segmentation image identifier, the fusion neural network model aims at improving the representation capability of the feature, and a convolutional neural network is used as a main network and is fused with a model of various network structures; And analyzing the characteristic segmentation image based on edge detection and Hough transformation to obtain a detection result of the surface thickened oil steam injection pipeline in the characteristic segmentation image.
2. The method for extracting the ground surface thickened oil steam injection pipe network based on the high-definition images according to claim 1, further comprising the steps of obtaining a plurality of remote sensing images to be identified of a set area, obtaining ground surface thickened oil steam injection pipe detection results of each remote sensing image to be identified, combining all the ground surface thickened oil steam injection pipe detection results according to coordinates, and extracting the ground surface thickened oil steam injection pipe network image of the set area.
3. The method for extracting the surface thickened oil steam injection pipe network based on the high-definition images according to claim 1 or 2, wherein the feature segmentation model construction process comprises the following steps: Acquiring a plurality of historical oilfield remote sensing images, and preprocessing each historical oilfield remote sensing image, wherein the preprocessing comprises binarization and image segmentation; carrying out data amplification on the preprocessed historical oilfield remote sensing images, and dividing all the historical oilfield remote sensing images subjected to data amplification into a training sample set and a testing sample set according to a proportion; Training an R2AU-Net model by using a training sample set, introducing a loss function during training, and ending training when the value of the loss function is stable to obtain a characteristic segmentation model, wherein the R2AU-Net model is formed by embedding a cyclic residual error module and an attention module in a U-Net model; And respectively testing the trained feature segmentation model by using the test set, optimizing model parameters of the feature segmentation model, and outputting the feature segmentation model meeting the test evaluation requirement.
4. The method for extracting the surface thickened oil steam injection pipe network based on the high-definition images according to any one of claims 1 to 3, wherein the analyzing the feature segmentation image based on edge detection and hough transformation to obtain the surface thickened oil steam injection pipe detection result in the feature segmentation image comprises the following steps: performing edge extraction on the feature segmentation image by using an edge detection operator, and outputting an edge contour image; And (5) carrying out straight line detection on the edge profile image by using Hough transformation, and outputting a detection result of the surface thickened oil steam injection pipeline.
5. The method for extracting the surface thickened oil steam injection pipe network based on the high-definition images according to claim 4, wherein the method for performing straight line detection on the edge profile image by using Hough transformation and outputting the surface thickened oil steam injection pipe network detection result comprises the following steps: Selecting any straight line in the edge contour image, converting the pixel point coordinates of any point where the straight line passes into a parameter space through parameter transformation, and obtaining an accumulator of a two-dimensional array; Traversing all points through which the straight line passes, and performing accumulated voting; And determining the accumulated value of all the straight lines in the edge profile image, and selecting the straight lines larger than a set threshold value as the detection result of the surface thickened oil steam injection pipeline.
6. The method for extracting the high-definition image-based surface thickened oil steam injection pipe network according to claim 4, wherein the edge detection operator is a Canny edge detection operator.
7. A high definition image-based surface thickened oil steam injection pipe network extraction device applying the method as claimed in any one of claims 1 to 6, which is characterized by comprising: The segmentation unit is used for acquiring an oilfield remote sensing image to be identified and inputting the oilfield remote sensing image to a feature segmentation model to obtain a feature segmentation image, wherein the feature segmentation model is obtained by deep learning a fusion neural network model by utilizing a plurality of samples, each sample in the plurality of samples comprises the oilfield remote sensing image and a corresponding feature segmentation image identifier, the fusion neural network model is a model which aims at improving the representation capability of the feature, takes a convolutional neural network as a main network and fuses various network structures; And the detection unit is used for analyzing the characteristic segmentation image based on edge detection and Hough transformation to obtain a detection result of the surface thickened oil steam injection pipeline in the characteristic segmentation image.
8. The high-definition image-based earth surface thickened oil steam injection pipe network extraction device according to claim 7, further comprising a pipe network image forming unit, wherein a plurality of to-be-identified remote sensing images of a set area are obtained, earth surface thickened oil steam injection pipe detection results of each to-be-identified remote sensing image are obtained, all earth surface thickened oil steam injection pipe detection results are combined according to coordinates, and earth surface thickened oil steam injection pipe network images of the set area are extracted.
9. An electronic device comprising a processor and a memory, the memory having stored therein a computer program that is loaded and executed by the processor to implement the steps in the method of any of claims 1 to 6.
10. A storage medium having stored thereon a computer program readable by a computer, the computer program being arranged to perform the steps of the method according to any of claims 1 to 6 when run.

Description

High-definition image-based extraction method and device for surface thickened oil steam injection pipe network, electronic equipment and storage medium Technical Field The invention relates to the technical field of pipeline identification, in particular to a method and a device for extracting a surface thickened oil steam injection pipe network based on high-definition images, electronic equipment and a storage medium. Background At present, in the field of ground engineering of oil fields, when partial old areas and newly-built energy-capacity surface thickened oil steam injection pipelines cannot be updated and put in storage in time, detection and identification are required to be carried out on the surface thickened oil steam injection pipelines, and the detection and identification method of the existing surface thickened oil steam injection pipelines comprises the following steps: (1) The traditional physical detection method is only suitable for identifying buried pipelines, needs manual execution and has limited efficiency; (2) The traditional mapping method of the satellite positioning system comprises the steps of firstly using a GPS positioning technology, placing handheld GPS equipment near a pipeline or on the pipeline for data acquisition, secondly using a total station for measurement, using the total station for measurement of specific points on the pipeline to obtain three-dimensional coordinates of the points, and then obtaining longitude and latitude coordinates through calculation and conversion. And further, the method is combined with RTK, laser range finder or online map service. If the method uses GPS positioning equipment, total station, RTK, laser range finder and other modes to identify the thick oil steam injection pipeline, the method is usually completed by means of a professional external team, and the method is high in identification accuracy, but large in personnel workload and long in acquisition period. If the thick oil steam injection pipeline is identified by using the map on-line service mode, the method has the following problems that firstly, the acquisition precision depends on the resolution ratio of image data, in other words, the higher the resolution ratio is, the higher the acquisition precision is, secondly, the measurement error exists by a manual click mode, in other words, pipeline click acquisition is carried out on a scale map with the resolution ratio of 1:500, if the screen error deviates from the pipeline by a few millimeters, the screen error possibly deviates from an actual site by 1-2 meters, so that the identification precision is easily influenced by the image resolution ratio and personnel click precision although the working efficiency is high; (3) The existing ground oil-gas pipeline detection method based on the algorithm model can realize detection and identification of the ground thickened oil steam injection pipeline of the remote sensing image, but because the pipeline has local characteristics with smaller size, such as junction points, and also has characteristics with larger size, such as long-distance oil gathering and transportation trunk lines, compared with the ground features of buildings, vehicles, vegetation and the like in the oilfield remote sensing image, the pipeline has multi-scale characteristics, so the oilfield remote sensing image comprises the characteristics of larger difference of pipeline target sizes, unbalanced dense distribution, smaller pixel proportion occupied by the pipeline targets, various types of ground feature information and the like, and the existing single neural network model (such as a common CNN convolutional neural network) cannot improve the representation capability of the features, so that under the condition that interference cannot be eliminated, effective features are extracted in the oilfield remote sensing image, thereby influencing the accuracy of pipeline detection and identification. Disclosure of Invention The invention provides a high-definition image-based method, a device, electronic equipment and a storage medium for extracting a ground surface thickened oil steam injection pipe network, which overcome the defects of the prior art, and can effectively solve the problems that the characteristic expression capability cannot be improved, the effective characteristic cannot be extracted from an oilfield remote sensing image, and the pipeline detection and identification accuracy is affected in the existing ground surface thickened oil steam injection pipe detection and identification method based on a single neural network model. The technical scheme of the invention is realized by the following measures that the method for extracting the ground surface thickened oil steam injection pipe network based on the high-definition image comprises the following steps: acquiring an oilfield remote sensing image to be identified, inputting the oilfield remote sensing image to be identified into