CN-122024041-A - Iron tower rust identification method, system, equipment, medium and program product

CN122024041ACN 122024041 ACN122024041 ACN 122024041ACN-122024041-A

Abstract

The application discloses a method, a system, equipment, a medium and a program product for identifying rust of an iron tower, which belong to the technical field of power equipment state monitoring, and the method comprises the steps of collecting iron tower image data, marking the iron tower image data and constructing a data set; training a pre-constructed recognition model by using a data set to obtain a trained recognition model, carrying out corrosion recognition on an image to be recognized by using the trained recognition model, outputting a corrosion recognition result and quantifying, carrying out evaluation analysis on all the corrosion recognition results and quantification results of the single iron tower, and generating an operation and maintenance scheme of the iron tower. The application realizes the comprehensive improvement of the detection efficiency, the identification precision and the operation and maintenance intelligent level of the iron tower.

Inventors

DAI GANG
GENG RUI
LI XIAODAN
LI JIUSHENG
CUI LIYING
LU HAIYU
TAO YE
Men Shengrui
XUE NAN
Ning Yedong
SHEN GANG
PANG XUEFENG
HE XUN
FENG ZHE
QI BAOJIN
YAN XIAOPENG
MA LIGANG
HAN JING
LOU ZIQIANG
YAN SIYUAN
BAI XU

Assignees

中国铁塔股份有限公司

Dates

Publication Date: 20260512
Application Date: 20260107

Claims (12)

1. An iron tower rust identification method, which is characterized by comprising the following steps: acquiring iron tower image data, marking the iron tower image data and constructing a data set; Training the pre-constructed recognition model by utilizing the data set to obtain a trained recognition model; Rust identification is carried out on the image to be identified by utilizing the trained identification model, and a rust identification result is output and quantized; and evaluating and analyzing all corrosion identification results and quantitative results of the single iron tower, and generating an operation and maintenance scheme of the iron tower.
2. The method for identifying rust of iron towers according to claim 1, wherein the steps of collecting the image data of the iron towers, labeling the image data of the iron towers and constructing a data set comprise: Dynamically scheduling an unmanned aerial vehicle-mounted camera and a ground camera to perform multi-band collaborative acquisition operation on a target iron tower based on a preset environmental parameter threshold value, and obtaining a multi-band image sequence of the target iron tower; performing quality evaluation on a multiband image sequence of a target iron tower, screening effective multiband images in the multiband image sequence, and triggering an automatic re-acquisition mechanism on ineffective multiband images to obtain an effective image set; preprocessing an effective image set, segmenting an iron tower region, and rust labeling the processed effective image set to construct a data set; Wherein the onboard cameras include high definition industrial cameras and multispectral cameras.
3. The method for identifying rust of iron tower according to claim 2, wherein the dynamically scheduling the unmanned aerial vehicle on-board camera and the ground camera to perform multi-band collaborative acquisition operation on the target iron tower based on the preset environmental parameter threshold value, to obtain a multi-band image sequence of the target iron tower comprises: Monitoring environmental parameters of an operation site in real time, wherein the environmental parameters comprise illumination intensity, air visibility and wind speed; When the illumination intensity is lower than the illumination intensity threshold value, the onboard camera is controlled to start a light supplementing function, when the air visibility is lower than the visibility threshold value, a polarization filtering mode of the multispectral camera is started to collect images of a target iron tower, and when the wind speed is higher than the wind speed threshold value, the collection task of the onboard camera is paused and is switched to a ground camera to conduct leading collection; and after the acquisition is completed, obtaining the multiband image sequence of the target iron tower.
4. A method for identifying rust of iron towers according to claim 2 or 3, wherein the steps of preprocessing the effective image set and dividing the iron tower area, and performing rust marking on the processed effective image set to construct a data set comprise the steps of: sequentially performing denoising processing and contrast enhancement processing on the effective multiband images in the effective image set to obtain an enhanced effective multiband image; Semantic segmentation is carried out on the basis of the effective multiband image after enhancement processing so as to extract a main body area of the iron tower and remove a background area, and an iron tower area image without the background is obtained; Performing multidimensional corrosion labeling on the iron tower area image, constructing a data set, and dividing the data set into a training set, a verification set and a test set; Wherein, the multidimensional corrosion mark content comprises the outline boundary, the corrosion grade and the corrosion type of each corrosion area.
5. The method for identifying rust of iron tower according to claim 1, wherein training the pre-constructed identification model by using the data set to obtain a trained identification model comprises: constructing an identification model based on a convolutional neural network and a attention mechanism; expanding the training set by adopting a multi-mode data enhancement technology, and training the recognition model by utilizing the expanded training set; in the training process, optimizing the performance of the recognition model by adopting a cross-region learning mechanism and a knowledge distillation technology to obtain a trained recognition model; Wherein, adopt multimode data enhancement technique to expand training set, include: Performing basic enhancement processing of random geometric transformation on training samples in a training set to generate a basic enhancement sample set; based on the basic enhanced sample set, generating simulated rust samples with different forms by adopting a generation countermeasure network to obtain a generated sample set; Based on the basic enhanced sample set and the generated sample set, adopting CutMix technology to select image samples with different rust grades from the basic enhanced sample set and carrying out region fusion to generate a new image sample and a corresponding mixed label thereof, and obtaining a new image sample set; And combining the basic enhanced sample set, the generated sample set and the new image sample set to obtain a final training set, and expanding the training set.
6. The iron tower rust identification method according to claim 1, wherein the performing rust identification on the image to be identified by using the trained identification model, outputting the rust identification result and quantifying the rust identification result comprises: Inputting the image to be identified into a trained identification model, and extracting features of the image to be identified by the identification model to obtain a feature map; the recognition model focuses on a suspected rusted area in the feature map through an attention mechanism, and a pixel-level coordinate range of the suspected rusted area is primarily judged; Based on the suspected rust area and the pixel level coordinate range thereof, the recognition model further extracts multidimensional features in the feature map for classification recognition so as to further determine the rust area and the boundary pixel coordinates, the rust grade and the rust type of the rust area in the image to be recognized; And counting the total number of pixels of the rusted area according to the boundary pixel coordinates of the rusted area to obtain the pixel area of the rusted area, and calculating the actual physical area of the rusted area by combining an image calibration technology and the geometric characteristics of the component.
7. The method for identifying rust in iron towers according to claim 6, wherein the step of calculating the actual physical area of the rust area by combining the image calibration technology with the geometric features of the components comprises the steps of: Selecting a reference member with known actual size from the image to be identified, measuring the pixel length of the reference constructed in the feature map, and calculating the pixel size coefficient k of the reference member; Determining a curvature correction factor c according to the curved surface characteristics of the iron tower component where the rust area is located; Based on the pixel size coefficient k and the curvature correction factor c, converting the pixel area of the rusted area into a corresponding actual physical area through double integral operation is as follows: , In the formula, Indicating the actual physical area of the rust area, Representing the curvature correction factor as such, The pixel size coefficient representing the reference member, And Respectively representing the maximum pixel coordinate and the minimum pixel coordinate of the rusted region in the X axis of the image to be identified, And Respectively representing the maximum pixel coordinate and the minimum pixel coordinate of the rusted region in the Y axis of the image to be identified, Indicating the function.
8. The method for identifying rust of iron tower according to claim 1, wherein the evaluating and analyzing all rust identification results and quantization results thereof of the single iron tower to generate the operation and maintenance scheme of the iron tower comprises: The method comprises the steps of merging multiple rust identification results of a single iron tower in a detection period, and determining the overall rust grade of the single iron tower by adopting a voting mechanism; Based on the multiple rust identification results and the quantitative results of the single iron tower in the detection period, carrying out dynamic risk assessment on all key components of the single iron tower, obtaining total rust risk values of all key components and determining corresponding risk levels; based on the whole corrosion grade of the single iron tower and the total corrosion risk value of all key components, constructing a corrosion development rate prediction model by combining historical corrosion data and environmental parameters, and predicting the corrosion development rate of the single iron tower by the corrosion development rate prediction model so as to adjust the detection period of the iron tower; Based on the overall corrosion grade of the iron tower and the total corrosion risk value of all key components thereof, the historical corrosion data of the iron tower and the corrosion development rate of the iron tower, the space-time visualization of the corrosion information of the iron tower is realized on an operation and maintenance management platform, and an operation and maintenance decision scheme is generated.
9. An iron tower rust identification system, the system comprising: the acquisition and construction unit is used for acquiring the image data of the iron tower, labeling the image data of the iron tower and constructing a data set; the training unit is used for training the pre-constructed recognition model by utilizing the data set to obtain a trained recognition model; the recognition unit is used for carrying out rust recognition on the image to be recognized by utilizing the trained recognition model, outputting a rust recognition result and quantifying; and the analysis unit is used for evaluating and analyzing all the rust identification results and the quantitative results thereof of the single iron tower and generating an operation and maintenance scheme of the iron tower.
10. An electronic device comprising a memory, a processor and a computer program stored on the memory, the processor executing the computer program or instructions to implement a method of iron tower rust identification as claimed in any one of claims 1 to 8.
11. A computer readable storage medium, wherein a computer program or instructions is stored in the computer readable storage medium, which when executed by a processor, implements a method for identifying rust of an iron tower according to any one of claims 1 to 8.
12. A computer program product comprising computer programs or instructions which when executed by a processor implement a method of iron tower rust identification as claimed in any one of claims 1 to 8.

Description

Iron tower rust identification method, system, equipment, medium and program product Technical Field The application belongs to the technical field of power equipment state monitoring, and particularly relates to an iron tower rust identification method, an iron tower rust identification system, an iron tower rust identification equipment, an iron tower rust identification medium and a program product. Background The iron tower is used as a key supporting structure of a power transmission and communication network, and the rust state of the iron tower is directly related to safe and stable operation of public infrastructure. Traditionally, iron tower corrosion detection mainly relies on manual inspection, and detection personnel need to climb an iron tower or observe closely with the help of equipment of ascending a height, and this method not only intensity of labour is big, inefficiency, and exists high altitude construction risk, receives the subjective experience influence of personnel simultaneously, and detection data's uniformity and accuracy are difficult to guarantee. In order to improve the detection efficiency, the prior art introduces an automatic method based on image recognition. However, the methods depend on simple threshold segmentation or basic machine learning models, and have obvious limitations that firstly, the identification accuracy is insufficient, the failure rate of light rust (the area ratio is less than 5%) and hidden rust under the coverage of component gaps and dust is high, secondly, the model generalization capability is weak, training data are single, the difference of rust forms under different regional environments (such as coastal high salt mist and inland dry areas) is difficult to adapt, thirdly, the output information is brief, only rough rust grades can be provided generally, and the refined analysis on the actual rust area, type division and development trend is lacking, so that effective support can not be provided for differentiated operation and maintenance decisions. With the continuous expansion of the scale of the power grid and the communication network, the number of iron towers is more than ten millions, the distribution environment is increasingly complex, and the conventional method and the prior art cannot meet the requirements of large-scale, high-precision and low-cost corrosion monitoring. Therefore, there is a need for a rust identification method that can reduce human dependency, achieve refined identification, and enable intelligent prediction and decision-making in combination with environmental factors. Disclosure of Invention In order to solve the problems, the application provides a method, a system, equipment, a medium and a program product for identifying rust of an iron tower, which realize comprehensive improvement of the detection efficiency, the identification precision and the operation and maintenance intelligent level of the iron tower. In order to achieve the above purpose, the present application provides the following technical solutions: in a first aspect, an embodiment of the present application provides a method for identifying rust of an iron tower, where the method includes: acquiring iron tower image data, marking the iron tower image data and constructing a data set; Training the pre-constructed recognition model by utilizing the data set to obtain a trained recognition model; Rust identification is carried out on the image to be identified by utilizing the trained identification model, and a rust identification result is output and quantized; and evaluating and analyzing all corrosion identification results and quantitative results of the single iron tower, and generating an operation and maintenance scheme of the iron tower. Further, the collecting the image data of the iron tower, labeling the image data of the iron tower and constructing a data set includes: Dynamically scheduling an unmanned aerial vehicle-mounted camera and a ground camera to perform multi-band collaborative acquisition operation on a target iron tower based on a preset environmental parameter threshold value, and obtaining a multi-band image sequence of the target iron tower; performing quality evaluation on a multiband image sequence of a target iron tower, screening effective multiband images in the multiband image sequence, and triggering an automatic re-acquisition mechanism on ineffective multiband images to obtain an effective image set; preprocessing an effective image set, segmenting an iron tower region, and rust labeling the processed effective image set to construct a data set; Wherein the onboard cameras include high definition industrial cameras and multispectral cameras. Further, based on a preset environmental parameter threshold, dynamically scheduling the unmanned aerial vehicle onboard camera and the ground camera to perform multiband collaborative acquisition operation on the target iron tower to obtain a multiband image sequence of