CN-122024154-A - Power station equipment safety monitoring method and device based on visual identification

CN122024154ACN 122024154 ACN122024154 ACN 122024154ACN-122024154-A

Abstract

The application discloses a power station equipment safety monitoring method and device based on visual identification, which relate to the technical field of visual monitoring and are used for collecting image identification data and non-visual data of a target power station area, extracting and quantitatively identifying image characteristics of the image identification data to obtain image characteristic information, optimizing the image characteristic information to improve night brightness and reduce reflection noise through a reflection component separation noise algorithm to obtain data to be detected, carrying out behavior identification and equipment state identification on the data to be detected and the non-visual data to obtain an identification result, judging an alarm grade on the basis of the identification result, and carrying out corresponding technical response on the basis of the alarm grade. In this way, the multi-source data can be combined for research and judgment, so that the monitoring of the power station is more accurate and comprehensive.

Inventors

HE QIANSHENG
ZHAO LEIHONG
GONG ZIQIANG
SONG FANGYU

Assignees

大唐玉曲河水电开发有限公司

Dates

Publication Date: 20260512
Application Date: 20251225

Claims (10)

1. The utility equipment safety monitoring method based on visual identification is characterized by comprising the following steps: collecting image identification data and non-visual data of a target power station area; extracting and quantitatively identifying image features of the image identification data to obtain image feature information; optimizing the image characteristic information to improve night brightness and reduce reflection noise through a reflection component separation noise algorithm to obtain data to be detected; performing behavior recognition and equipment state recognition based on the data to be detected and the non-visual data to obtain a recognition result; and judging an alarm level based on the identification result, and carrying out corresponding technical response based on the alarm level.
2. The method of claim 1, wherein the step of performing image feature extraction and quantization identification on the image recognition data to obtain image feature information comprises: Determining a real-time video stream, a high resolution still image and an instrument feature image based on the image identification data; constructing an image dataset based on the real-time video stream, the high resolution still image, and the meter close-up image; And extracting the characteristics of the image data set to obtain image characteristic information.
3. The method of claim 2, wherein the step of extracting features from the image dataset to obtain image feature information comprises: Splitting the image data set based on a data source type to obtain type data, wherein the data type comprises a circular instrument panel data type, a liquid level scale data type, a status lamp data type and a running and leaking data type; and respectively carrying out feature extraction and quantitative identification on the type data to obtain image feature information.
4. The method of claim 1, wherein the optimizing the image feature information to enhance night brightness and reduce reflection noise by a reflection component separation noise algorithm, the step of obtaining the data to be detected comprises: performing weighted gamma correction based on the image characteristic information, and performing contrast-limited self-adaptive histogram equalization through a weighted cumulative distribution function to obtain equalization characteristic information; filtering the equalization characteristic information through a reflection component separation noise algorithm to obtain filtering characteristic information; And carrying out fusion enhancement and edge enhancement based on the filtering characteristic information to obtain data to be detected.
5. The method of claim 4, wherein the step of performing fusion enhancement and edge enhancement based on the filtering characteristic information to obtain the data to be detected comprises: Constructing a target pyramid model aiming at the filtering characteristic information, and carrying out multi-scale fusion enhancement based on the target pyramid model to obtain a reconstructed image; And carrying out edge enhancement on the reconstructed image to obtain data to be detected.
6. The method of claim 1, wherein the step of performing behavior recognition and device state recognition based on the data to be detected and the non-visual data to obtain a recognition result comprises: performing equipment state identification and abnormality detection based on the data to be detected and the non-visual data to obtain an equipment abnormality detection result; Performing behavior recognition and anomaly detection based on the data to be detected, the non-visual data and the behavior rules of the target power station area to obtain a behavior anomaly detection result; and determining a recognition result according to the equipment abnormality detection result and the behavior abnormality detection result.
7. The method of claim 6, wherein the step of performing device state recognition and anomaly detection based on the data to be detected and the non-visual data to obtain a device anomaly detection result comprises: performing data synchronization and alignment based on the data to be detected and the non-visual data to obtain alignment data; And carrying out abnormality judgment of multi-source data fusion on the alignment data to obtain an equipment abnormality detection result.
8. The method of claim 6, wherein the step of performing behavior recognition and anomaly detection based on the behavior rules of the to-be-detected data and the non-visual data and the target power station area, and obtaining a behavior anomaly detection result comprises: determining detection behavior information based on the data to be detected and the non-visual data; and performing behavior recognition and abnormality detection according to the detected behavior information and the behavior rule of the target power station area to obtain a behavior abnormality detection result.
9. The method of claim 8, wherein the step of performing behavior recognition and anomaly detection according to the detected behavior information and the behavior rules of the target power station area, and obtaining a behavior anomaly detection result comprises: determining an abnormal behavior list according to the behavior rule of the target power station area; comparing the detected behavior information with the abnormal behavior list and detecting the abnormality, and determining a behavior abnormality detection result based on the detection result.
10. A power plant equipment safety monitoring device based on visual identification, the device comprising: The data acquisition module is used for acquiring image identification data and non-visual data of the target power station area; the data processing module is used for carrying out image feature extraction and quantitative recognition on the image recognition data to obtain image feature information; the data optimization module is used for optimizing the image characteristic information to improve night brightness and reduce reflection noise through a reflection component separation noise algorithm to obtain data to be detected; the identification alarm module is used for carrying out behavior identification and equipment state identification based on the data to be detected and the non-visual data to obtain an identification result; and the alarm response module is used for judging the alarm grade based on the identification result and carrying out corresponding technical response based on the alarm grade.

Description

Power station equipment safety monitoring method and device based on visual identification Technical Field The application relates to the technical field of visual monitoring, in particular to a power station equipment safety monitoring method and device based on visual identification. Background The current energy storage power station inspection system covers a booster station and key equipment monitoring of an energy storage area, but still has a scene of partially relying on manual auxiliary judgment, such as easy deviation of equipment state (dial value and interface leakage) identification under a strong light/low illumination environment, difficulty in capturing potential problems (such as main light and micro vibration and hidden discharge of GIS equipment) in equipment by single vision monitoring, dependence on single source data on abnormal early warning, and easy hidden danger disposal lag caused by misjudgment and missed judgment. Meanwhile, although the existing inspection data record is digitalized, the existing inspection data record lacks multi-source data association analysis, so that the equipment health trend research and judgment are difficult to support, and the requirement of 'lean management' cannot be completely supported. The foregoing is provided merely for the purpose of facilitating understanding of the technical solutions of the present application and is not intended to represent an admission that the foregoing is prior art. Disclosure of Invention The application mainly aims to provide a power station equipment safety monitoring method and device based on visual identification, and aims to solve the technical problem that the monitoring of power station equipment lacks the associated monitoring and studying of multi-source data. In order to achieve the above purpose, the application provides a power station equipment safety monitoring method based on visual identification, which comprises the following steps: collecting image identification data and non-visual data of a target power station area; extracting and quantitatively identifying image features of the image identification data to obtain image feature information; optimizing the image characteristic information to improve night brightness and reduce reflection noise through a reflection component separation noise algorithm to obtain data to be detected; performing behavior recognition and equipment state recognition based on the data to be detected and the non-visual data to obtain a recognition result; and judging an alarm level based on the identification result, and carrying out corresponding technical response based on the alarm level. Optionally, the step of extracting and quantitatively identifying the image features of the image identification data to obtain image feature information includes: Determining a real-time video stream, a high resolution still image and an instrument feature image based on the image identification data; constructing an image dataset based on the real-time video stream, the high resolution still image, and the meter close-up image; And extracting the characteristics of the image data set to obtain image characteristic information. Optionally, the step of extracting features from the image dataset to obtain image feature information includes: Splitting the image data set based on a data source type to obtain type data, wherein the data type comprises a circular instrument panel data type, a liquid level scale data type, a status lamp data type and a running and leaking data type; and respectively carrying out feature extraction and quantitative identification on the type data to obtain image feature information. Optionally, the step of optimizing the image feature information by using a reflected component separation noise algorithm to improve night brightness and reduce reflection noise, and obtaining the data to be detected includes: performing weighted gamma correction based on the image characteristic information, and performing contrast-limited self-adaptive histogram equalization through a weighted cumulative distribution function to obtain equalization characteristic information; filtering the equalization characteristic information through a reflection component separation noise algorithm to obtain filtering characteristic information; And carrying out fusion enhancement and edge enhancement based on the filtering characteristic information to obtain data to be detected. Optionally, the step of obtaining the data to be detected by performing fusion enhancement and edge enhancement based on the filtering characteristic information includes: Constructing a target pyramid model aiming at the filtering characteristic information, and carrying out multi-scale fusion enhancement based on the target pyramid model to obtain a reconstructed image; And carrying out edge enhancement on the reconstructed image to obtain data to be detected. Optionally, the step of performing behavior recogni