Monitoring, identification and analysis method of production area of ​​pumped storage power station based on machine vision

By analyzing the historical acquisition screen of the visual sensor in the production area of ​​the pumped storage power station, identifying and adjusting the affected visual sensors, the problems of visual sensor perception stability and identification accuracy are solved, and the accuracy and response speed of monitoring and identification are improved. Reduced maintenance costs.

CN118799820BActive Publication Date: 2024-11-15COLORFUL GUIZHOU IMPRESSION NETWORK MEDIA CO LTD
View PDF 8 Cites 0 Cited by

Patent Information

Application Number
CN202411276190.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-12
Publication Date
2024-11-15
Estimated Expiration
2044-09-12

AI Technical Summary

Technical Problem

In the monitoring and identification of pumped storage power station production areas, the perceived stability and identification accuracy of visual sensors are affected by the vibration of the turbine and the bad weather, resulting in misjudgment and response hysteresis, affecting the safety and stability of the power station.

Method used

By analyzing the acquisition range changes and image quality changes of the historical acquisition screen of the vision sensor, the visual sensors to be adjusted are selected, and calibrated and corrected based on the correlation analysis results to optimize the acquisition screen to improve the recognition accuracy.

Benefits of technology

Reduce the incidence of misjudgment, improve the accuracy and timeliness of abnormal identification, and reduce maintenance and time costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118799820B_ABST
    Figure CN118799820B_ABST
Patent Text Reader

Abstract

The present invention belongs to the technical field of monitoring, identifying and analyzing the hydropower station, and specifically discloses a method for monitoring, identifying and analyzing the production area of ​​a pumped-storage power station based on machine vision. The method performs collection range change analysis and image quality change analysis on the historical collection pictures corresponding to each visual sensor arranged in the production area of ​​the pumped-storage power station by retrieving the historical collection pictures, thereby screening out the visual sensors to be adjusted based on the analysis results, and then performing correlation analysis on the collection range change and image quality change of the visual sensors to be adjusted, thereby automatically performing feedback preventive processing based on the correlation analysis results, and optimizing the collection pictures of the visual sensors in advance, greatly reducing the occurrence rate of misjudgment, and improving the accuracy and timeliness of abnormal identification in the production area. Meanwhile, compared with processing only when problems occur in the collection pictures, unnecessary maintenance costs and time costs can be reduced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of monitoring, identifying and analyzing a hydropower station, and specifically discloses a method for monitoring, identifying and analyzing a production area of ​​a pumped-storage power station based on machine vision. Background Art

[0002] A pumped-storage power station is an important power system that uses changes in water levels to adjust the load balance of the power grid and is a key facility for achieving electric energy storage and dispatch. In the production area of ​​a pumped-storage power station, there are hydroelectric generating devices, dam structures, water pumps and other related equipment. The operating status of these equipment will have a great impact on the power supply effect of the power station. Therefore, monitoring and identification of the production area of ​​a pumped-storage power station is very necessary.

[0003] In order to provide real-time and intuitive monitoring data, the existing monitoring and identification of the production area of ​​the pumped-storage power station usually adopts an image monitoring method, that is, by deploying visual sensors in the production area to collect operating pictures of the deployment location to provide real-time visual data. There are also technical solutions involving the use of visual data to analyze the operating status of the hydropower station. For example, the application with publication number CN118233597A discloses an intelligent inspection method, system and storage medium for a hydropower station. Through intelligent video inspection of major equipment and major areas, the equipment in the major areas is captured and identified according to the inspection process, thereby screening out dangerous areas. When the inspection personnel enter the dangerous area, they should be given timely boundary alarms or inspection precautions reminders, which greatly improves the efficiency and intelligence of the inspection tasks. This invention does not take into account the perception stability and recognition accuracy of the visual sensor when using the formed video images to identify hazards in hydropower stations. The operation of turbines or water pumps in the production area of ​​a pumped-storage power station usually generates vibrations, which will more or less cause changes in the field of view of the visual sensor, making its perception unstable. At the same time, it will affect the image quality of the visual sensor in bad weather, reducing its recognition accuracy. When this influence is ignored and the images collected by the visual sensor are directly used for hazard identification, it is easy to cause misjudgment and fail to promptly discover actual problems. This situation may delay the response time to potential problems and affect the safety and stability of the power station.

[0004] In the prior art, the processing of abnormal perception and recognition of visual sensors is mostly performed when problems occur in the images collected by the visual sensors. This processing method requires time to troubleshoot the problem, which increases processing time and maintenance costs, resulting in delayed response to abnormal situations, which can easily lead to accidents or increased losses and is not conducive to preventive maintenance. Summary of the invention

[0005] To this end, one purpose of an embodiment of the present application is to provide a method for monitoring, identifying and analyzing the production area of ​​a pumped-storage power station based on machine vision. By retrieving historical acquisition images, correlation analysis is performed on problems that appear in the historical acquisition images of visual sensors in the production area of ​​the pumped-storage power station, and then preventive processing is performed based on the correlation analysis results, thereby effectively solving the problems mentioned in the background technology.

[0006] The purpose of the present invention can be achieved through the following technical scheme: A method for monitoring, identifying and analyzing the production area of ​​a pumped-storage power station based on machine vision, comprising the following steps: S1, locating the setting position of each visual sensor in the production area of ​​the pumped-storage power station, and numbering the visual sensors accordingly.

[0007] S2. Retrieving the historical acquisition images of each visual sensor from the historical acquisition records of the visual sensor to perform acquisition range change analysis and image quality change analysis respectively.

[0008] S3. Based on the analysis results of changes in the acquisition range of each visual sensor in the historical acquisition screen, a first visual sensor to be adjusted is identified. Meanwhile, based on the analysis results of changes in the acquisition quality of each visual sensor in the historical acquisition screen, a second visual sensor to be adjusted is identified.

[0009] S4. Filter out the changed images from the historical collection images of the first visual sensor to be adjusted, and retrieve the vibration data of the turbine according to the collection time of the changed images, thereby performing a correlation analysis between the change of the collection range of the first visual sensor to be adjusted and the vibration of the turbine.

[0010] S5. Based on the correlation analysis result between the acquisition range change and the turbine vibration, the calibration timing of the first visual sensor to be adjusted is controlled.

[0011] S6. Filter out low-quality images from the historical images collected by the second visual sensor to be adjusted, and retrieve weather data based on the collection time of the low-quality images, thereby performing a correlation analysis between the low quality of the images collected by the second visual sensor to be adjusted and the bad weather.

[0012] S7. Based on the correlation analysis result between the low quality of the collected images and the bad weather, the collected images of the second visual sensor to be adjusted are corrected and adjusted.

[0013] The acquisition range change analysis is as follows: edge cutting and target detection are performed on the retrieved historical acquisition pictures respectively to obtain the field of view boundary contour of the historical acquisition pictures and the boundary contour of the detection target.

[0014] The historical acquisition pictures retrieved by the same visual sensor are arranged in chronological order of acquisition time, and each historical acquisition picture is extracted in turn as the main picture according to the arrangement result, and then the field of view boundary contour of the main picture is overlapped and compared with the field of view boundary contours of other historical acquisition pictures. At the same time, the detection target of the main picture is mapped with the detection target of other historical acquisition pictures, and then the boundary contours of the detection target are overlapped and compared. The two overlap comparison results are combined to calculate the similarity of the acquisition range between the main picture and the other historical acquisition pictures, and compared with the set similarity threshold at the same time. Other historical acquisition pictures with acquisition range similarity greater than or equal to the similarity threshold and the main picture constitute a collection range picture group. Based on this, the retrieved historical acquisition pictures are grouped according to the similarity of the collection range to obtain a number of collection range picture groups.

[0015] The number of picture groups in the acquisition range formed by each visual sensor is counted and compared with the number of picture groups formed when the acquisition range does not change. If the number of picture groups in the acquisition range formed by a certain visual sensor is greater than the number of picture groups formed when the acquisition range does not change, it is analyzed that there is a change in the acquisition range of the visual sensor, and it is used as the first visual sensor to be adjusted.

[0016] The image quality change analysis comprises the following steps: extracting image quality parameters from the historical captured images of each visual sensor, specifically including contrast and signal-to-noise ratio.

[0017] The image quality parameters of the historical collection images are compared with the specification image quality parameters of the corresponding visual sensor to calculate the image quality coefficient of each historical collection image. The specific calculation formula is: , where Indicates the image quality coefficient, , Respectively represent the contrast and signal-to-noise ratio of the historical acquisition images. , They represent specification contrast and specification signal-to-noise ratio respectively. The specification image quality parameters of the visual sensor can be obtained from the instruction manual and technical manual of the visual sensor.

[0018] Compare the image quality coefficients of each historical acquisition screen corresponding to the same visual sensor, and extract the maximum image quality coefficient through the expression Get the image quality difference of the visual sensor , where , Represent the maximum image quality coefficient and the minimum image quality coefficient respectively.

[0019] The image quality difference corresponding to each visual sensor is compared with the set allowable difference. If the image quality difference corresponding to a visual sensor is greater than the allowable difference, it is analyzed that there is a change in the image quality of the captured image corresponding to the visual sensor, and it is used as the second visual sensor to be adjusted.

[0020] The method of selecting the changed pictures from the historical acquisition pictures of the first visual sensor to be adjusted is as follows: the proportion of the historical acquisition pictures existing in each acquisition range picture group formed by the first visual sensor to be adjusted is counted, and the acquisition range picture group corresponding to the maximum historical acquisition picture proportion value is taken as the dominant picture group, and the historical acquisition pictures existing in other picture groups are taken as the changed pictures.

[0021] The correlation analysis between the change of the acquisition range of the first visual sensor to be adjusted and the vibration of the turbine is implemented as follows: the field of view boundary contour of the screened changing picture is overlapped and compared with the field of view boundary contour of the dominant picture group to obtain the change degree of the changing picture.

[0022] A two-dimensional coordinate system is constructed with the acquisition time as the horizontal coordinate and the change degree as the vertical coordinate, thereby marking several points in the constructed two-dimensional coordinate system according to the acquisition time and change degree of the first visual sensor to be adjusted corresponding to the changed picture to form a picture change curve.

[0023] A two-dimensional coordinate system is constructed with the acquisition time as the horizontal coordinate and the vibration data as the vertical coordinate, thereby marking several points in the constructed two-dimensional coordinate system for the acquisition time of the corresponding changing picture of the first visual sensor to be adjusted and the turbine vibration data corresponding to the corresponding acquisition time to form a turbine vibration change curve.

[0024] The correlation coefficient between the picture change curve and the turbine vibration change curve is calculated to obtain the correlation between the change in the acquisition range of the first visual sensor to be adjusted and the turbine vibration.

[0025] The process of screening out low-quality images from the historical acquisition images of the second visual sensor to be adjusted is as follows: comparing the image quality coefficient of each historical acquisition image corresponding to the second visual sensor to be adjusted with the standard image quality coefficient, and screening out the historical acquisition images with a lower than standard image quality coefficient as low-quality images, wherein the standard image quality coefficient is the image quality coefficient that needs to be achieved based on the design specifications of the visual sensor.

[0026] The correlation analysis between the low quality of the image collected by the second visual sensor to be adjusted and the bad weather is implemented as follows: the weather data of the corresponding collection time is retrieved for the collection time of the low quality image, and the weather severity is calculated based on the weather data. The specific calculation method is to compare the light intensity, rainfall and visibility with the light intensity, rainfall and visibility that the visual sensor can receive when working normally. The calculation formula is: , represents the severity of the weather, where , , Respectively represent light intensity, rainfall, and visibility. , , They respectively represent the light intensity, rainfall, and visibility that the visual sensor can receive for normal operation.

[0027] A two-dimensional coordinate system is constructed with the acquisition time as the horizontal coordinate and the image quality coefficient as the vertical coordinate, thereby marking several points in the constructed two-dimensional coordinate system according to the acquisition time and image quality coefficient of the second visual sensor to be adjusted corresponding to the low-quality picture to form an image quality change curve.

[0028] A two-dimensional coordinate system is constructed with the acquisition time as the horizontal coordinate and the weather severity as the vertical coordinate. According to the acquisition time of the low-quality image corresponding to the second visual sensor to be adjusted and the weather severity corresponding to the corresponding acquisition time, several points are marked in the constructed two-dimensional coordinate system to form a weather change curve.

[0029] Similarly, referring to the correlation analysis method between the change in the acquisition range and the vibration of the turbine, the correlation between the low quality of the image collected by the second visual sensor to be adjusted and the bad weather is obtained.

[0030] The specific operation of controlling the calibration timing of the first visual sensor to be adjusted is as follows: comparing the correlation between the change in the acquisition range of the first visual sensor to be adjusted and the vibration of the turbine with the preset effective positive correlation, and selecting the first visual sensor to be adjusted whose correlation is greater than the effective positive correlation as the calibration object.

[0031] For the change degree of the change screen corresponding to the calibration object and the turbine vibration data, a correlation curve between the acquisition range change and the turbine vibration is formed in a two-dimensional coordinate system with the turbine vibration data as the horizontal coordinate and the change degree as the vertical coordinate, and the slope of each point on the change curve is obtained to determine the sensitive vibration data.

[0032] During the operation of the water turbine, the vibration data is feedback-connected to the calibration object, and when the vibration data of the water turbine reaches the sensitive vibration data, the calibration object is started for calibration.

[0033] The correction and adjustment of the captured image of the second visual sensor to be adjusted refers to the following process: the correlation between the low quality of the captured image of the second visual sensor to be adjusted and the bad weather is compared with the preset effective negative correlation, and the second visual sensor to be adjusted whose correlation is less than the effective negative correlation is selected as the correction object.

[0034] For the image quality coefficient and weather severity of the low-quality image corresponding to the correction object, a correlation curve between the low quality of the acquired image and the bad weather is formed in a two-dimensional coordinate system with the bad weather severity as the horizontal coordinate and the image quality coefficient as the vertical coordinate, and the slope of each point on the change curve is obtained. Similarly, the sensitive weather severity is determined by referring to the method for determining the sensitive vibration data.

[0035] An environmental perception sensor is installed at the location where the correction object is located, and the weather data released by the weather station is extracted in real time to form a feedback connection with the correction object. The weather severity is then calculated based on the released weather data and compared with the sensitive weather severity. If the sensitive weather severity is reached, the environmental perception sensor is activated to perceive the environment, so as to correct and adjust the image collected by the correction object.

[0036] The sensitive vibration data is determined as follows: the horizontal coordinate corresponding to the maximum slope point among the slopes corresponding to the points of the correlation curve between the acquisition range change and the turbine vibration is taken as the sensitive vibration data.

[0037] In combination with all the above-mentioned technical solutions, the positive effect possessed by the present invention is as follows: the present invention performs collection range change analysis and image quality change analysis on the historical collection images corresponding to each visual sensor deployed in the production area of ​​the pumped-storage power station, thereby screening out the visual sensors to be adjusted based on the analysis results, and then performs correlation analysis on the collection range change and image quality change of the visual sensors to be adjusted, thereby automatically providing feedback for preventive processing based on the correlation analysis results, and optimizing the collection images of the visual sensors in advance, greatly reducing the incidence of misjudgment, and improving the accuracy and timeliness of abnormality identification in the production area. At the same time, compared with processing only when problems occur in the collection images, it can reduce unnecessary maintenance costs and time costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The present invention is further described using the accompanying drawings, but the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative work.

[0039] Figure 1 The present invention is a flowchart of the steps for implementing the method. DETAILED DESCRIPTION

[0040] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] See also Figure 1 As shown, the present invention proposes a method for monitoring, identifying and analyzing the production area of ​​a pumped-storage power station based on machine vision, comprising the following steps: S1, locating the setting position of each visual sensor in the production area of ​​the pumped-storage power station, wherein the visual sensor can be a camera, and numbering the visual sensors accordingly.

[0042] In the specific application of the above scheme, a reference position may be selected for numbering the visual sensors, and the numbering may be performed in order from near to far based on the distance between the setting position of the visual sensor and the reference position.

[0043] S2. Retrieving the historical acquisition images of each visual sensor from the historical acquisition records of the visual sensor to perform acquisition range change analysis and image quality change analysis respectively.

[0044] It should be understood that before adjusting the visual sensors, it is necessary to first determine which sensors need to be adjusted. Not every sensor deployed in the production area of ​​a pumped storage power station needs to be adjusted. This is because there may be different regional characteristics and environmental conditions in the production area. The deployment locations of some visual sensors may be less affected by bad weather, such as being deployed in sheltered areas. The deployment locations of some visual sensors may be far away from the operating coverage of turbines or pumps. Therefore, for visual sensors in different deployment locations, the optimization requirements for acquisition range and image quality will also be different. Since acquisition range adjustment and image quality correction are operations that require resources and costs, in this case, specific sensors only need to be adjusted when there is an obvious need, so that resources can be used effectively and unnecessary adjustments can be avoided.

[0045] In the preferred implementation of the above scheme, the acquisition range change analysis is carried out as follows: the retrieved historical acquisition pictures are subjected to edge cutting and target detection respectively to obtain the field of view boundary contour of the historical acquisition picture and the boundary contour of the detection target, wherein the field of view boundary contour reflects the acquisition range of the visual camera, and the detection target is a specific object or building in the acquisition picture, which can help determine the acquisition range of the visual camera based on the position and size of these objects.

[0046] Furthermore, in the above, the detection targets existing in the production area of ​​the pumped-storage power station include but are not limited to water pumps, power generation equipment, buildings, etc. These objects are usually fixedly installed in specific positions. By detecting these objects from the collected images, the collection range of the visual camera can be effectively determined.

[0047] The present invention determines the acquisition range of the visual camera comprehensively based on the field of view boundaries of historical acquisition images and the boundaries of each detection target from the overall and local aspects. It can cope with slight changes in the acquisition range, is more detailed, and helps to improve the accuracy of determining the acquisition range.

[0048] The historical acquisition images retrieved by the same visual sensor are arranged in the order of acquisition time, and each historical acquisition image is extracted as the main image in turn according to the arrangement results, and then the field of view boundary contour of the main image is overlapped and compared with the field of view boundary contour of other historical acquisition images to obtain the overlapping field of view boundary contour area, and divided by the field of view boundary contour area of ​​the main image to obtain the field of view boundary overlap degree. At the same time, the detection target of the main image and the detection target of other historical acquisition images are mapped, and the boundary contour of the detection target is overlapped and compared to obtain the contour overlap area of ​​each detection target, and divided by the contour area of ​​the detection target in the main image to obtain the contour overlap degree of each detection target. Therefore, the acquisition degree of the main image and other historical acquisition images is calculated by combining the two overlap comparison results. The specific calculation formula is to calculate the mean of the contour overlap of each detection target and the weighted average of the overlap of the field of view boundary and then take the percentage to obtain the collection range similarity between the main picture and other historical collection pictures. By way of example, the weight factors of the detection target contour overlap and the field of view boundary overlap are 0.6 and 0.4 respectively, and then compared with the set similarity threshold, wherein the similarity threshold can be set to 85%, and the similarity threshold is initially set. The purpose of setting is to assist in the division of the collection range picture group, so that other historical collection pictures with collection range similarity greater than or equal to the similarity threshold and the main picture constitute a collection range picture group, thereby grouping the retrieved historical collection pictures by collection range similarity to obtain a number of collection range picture groups.

[0049] The number of picture groups in the acquisition range formed by each visual sensor is counted and compared with the number of picture groups formed when the acquisition range does not change. If the number of picture groups in the acquisition range formed by a certain visual sensor is greater than the number of picture groups formed when the acquisition range does not change, it is analyzed that there is a change in the acquisition range of the visual sensor, and it is used as the first visual sensor to be adjusted.

[0050] In the above, when the acquisition range of the visual sensor has not changed, it indicates that the acquisition range formed by each historical acquisition screen corresponding to the visual sensor is consistent. In this case, there should be only one acquisition range screen group formed. By comparing the number of acquisition range screen groups formed by each visual sensor with the number of screen groups formed when the acquisition range does not change, it is possible to effectively identify whether the acquisition range of the visual sensor has changed.

[0051] In another preferred implementation of the above scheme, the image quality change analysis is performed as follows: image quality parameters are extracted from the historical acquisition images of each visual sensor, specifically including contrast and signal-to-noise ratio.

[0052] It should be added that the reason for choosing contrast and signal-to-noise ratio as image quality parameters is that bad weather conditions will affect the contrast and signal-to-noise ratio of the captured image, which is specifically reflected in the following aspects: 1. The signal-to-noise ratio is a parameter that describes the ratio of useful information to noise in an image or signal, and is one of the important indicators for evaluating image quality. Under bad weather conditions, such as low light conditions, dense fog, and heavy rain, the visual sensor may need to increase the ISO sensitivity or use a longer exposure time to capture enough light, which may lead to an increase in the noise level in the image. At the same time, under bad light conditions, the response of the visual sensor may vary, which may introduce additional noise or reduce the clarity of the image, thereby affecting the signal-to-noise ratio. 2. Contrast refers to the measure of the brightness difference between different areas in an image, which affects the visual clarity and detail performance of the image. Under bad weather conditions, such as in dense fog or heavy rain, light will reflect and scatter on water droplets or haze particles, causing the propagation of light to become uneven, which may narrow the brightness range of the overall scene, thereby reducing the contrast of the image. For example, cloudy, gray weather conditions reduce the intensity and directionality of light, causing the overall lighting conditions to become uniform and soft. In this case, the brightness difference between the highlight and shadow areas in the image will be reduced, thereby reducing contrast.

[0053] The image quality parameters of the historical collection images are compared with the specification image quality parameters of the corresponding visual sensor to calculate the image quality coefficient of each historical collection image. The specific calculation formula is: , where Indicates the image quality coefficient, , Respectively represent the contrast and signal-to-noise ratio of the historical acquisition images. , They represent specification contrast and specification signal-to-noise ratio respectively.

[0054] The above-mentioned specifications and image quality parameters of the visual sensor are obtained from the instruction manual and technical manual of the visual sensor, and can be understood as the ideal image quality parameters that the visual sensor can achieve.

[0055] Compare the image quality coefficients of each historical acquisition screen corresponding to the same visual sensor, and extract the maximum image quality coefficient through the expression Get the image quality difference of the visual sensor , where , They represent the maximum image quality coefficient and the minimum image quality coefficient respectively. The greater the difference between the maximum image quality coefficient and the minimum image quality coefficient, the greater the image quality difference.

[0056] The image quality difference corresponding to each visual sensor is compared with the set allowable difference. For example, the allowable difference is 0.2, where the allowable difference is initially set. The purpose of setting is to assist in judging whether the image quality has changed. If the image quality difference corresponding to a visual sensor is greater than the allowable difference, it is analyzed that the image quality of the captured image corresponding to the visual sensor has changed, and it is used as the second visual sensor to be adjusted.

[0057] In the above, when the image quality of the picture collected by the visual sensor has not changed, it means that the image quality coefficients of the visual sensor corresponding to each historical collection picture are consistent. In this case, the image quality difference formed should be close to 0. By comparing the image quality difference corresponding to each visual sensor with the allowed difference, it is possible to effectively identify that there is a change in the image quality of the picture collected by the visual sensor.

[0058] S3. Based on the analysis results of changes in the acquisition range of each visual sensor in the historical acquisition screen, a first visual sensor to be adjusted is identified. Meanwhile, based on the analysis results of changes in the acquisition quality of each visual sensor in the historical acquisition screen, a second visual sensor to be adjusted is identified.

[0059] It should be understood that after the first visual sensor to be adjusted is selected from the analysis results of the visual sensor's acquisition range change, and the second visual sensor to be adjusted is selected from the acquisition image quality change analysis results, it is shown that the acquisition range change of the selected first visual sensor to be adjusted may be affected by the operating vibration of the turbine or water pump, causing it to move or shake slightly, thereby changing the acquisition range of the visual sensor, and the acquisition image quality change of the selected second visual sensor to be adjusted may be affected by bad weather. Next, it is necessary to conduct correlation verification on this impact, and the visual sensor cannot be directly adjusted according to this possible impact.

[0060] S4. Filter out the changed images from the historical collection images of the first visual sensor to be adjusted, and retrieve the vibration data of the turbine or water pump according to the collection time of the changed images, wherein the vibration data may be the vibration frequency or the vibration amplitude, thereby performing a correlation analysis between the change in the collection range of the first visual sensor to be adjusted and the vibration of the turbine or water pump.

[0061] Applied to the above embodiment, the changed pictures are screened out from the historical acquisition pictures of the first visual sensor to be adjusted as follows: the proportion of historical acquisition pictures existing in each acquisition range picture group formed by the first visual sensor to be adjusted is counted, and the acquisition range picture group corresponding to the maximum historical acquisition picture proportion is taken as the dominant picture group, and the historical acquisition pictures existing in other picture groups are taken as the changed pictures.

[0062] It needs to be explained that when filtering the changed pictures, it is necessary to select the pictures with unchanged collection range from the historical collection pictures as the dominant picture. The present invention selects the collection range picture group corresponding to the largest historical collection picture ratio as the dominant picture group because the turbines and pumps of pumped-storage power stations are usually operated on demand rather than in real time. This makes the collection range of the visual sensor unaffected by the vibration of the turbine or pump in most cases. Therefore, the collection range picture group corresponding to the largest historical collection picture ratio is very likely to reflect the collection range of the visual sensor under normal circumstances, and it can be used as the dominant picture group.

[0063] Further applied to the above embodiment, the correlation analysis between the change of the acquisition range of the first visual sensor to be adjusted and the vibration of the turbine is implemented as follows: the field of view boundary contour of the screened changing picture is overlapped and compared with the field of view boundary contour of the dominant picture group. Specifically, the field of view boundary contour of the changing picture can be overlapped and compared with the field of view boundary contour of the main picture in the dominant picture group to obtain the field of view boundary overlap degree of the changing picture, and then the field of view boundary overlap degree of the changing picture is subtracted from the value 1 to obtain the change degree of the changing picture.

[0064] A two-dimensional coordinate system is constructed with the acquisition time as the horizontal coordinate and the change degree as the vertical coordinate, thereby marking several points in the constructed two-dimensional coordinate system according to the acquisition time and change degree of the first visual sensor to be adjusted corresponding to the changed picture to form a picture change curve.

[0065] A two-dimensional coordinate system is constructed with the acquisition time as the horizontal coordinate and the vibration data as the vertical coordinate, thereby marking several points in the constructed two-dimensional coordinate system for the acquisition time of the corresponding changing picture of the first visual sensor to be adjusted and the vibration data of the turbine or water pump corresponding to the corresponding acquisition time to form a vibration change curve of the turbine or water pump.

[0066] The correlation coefficient between the picture change curve and the turbine or water pump vibration change curve is calculated to obtain the correlation between the change in the acquisition range of the first visual sensor to be adjusted and the turbine or water pump vibration.

[0067] In the specific implementation of the above scheme, the correlation coefficient usually refers to the Pearson correlation coefficient, which can measure the strength and direction of the linear relationship between two variables. If the correlation coefficient is close to 1, it means that there is a strong positive correlation between the two curves, that is, they change in the same direction. If the correlation coefficient is close to -1, it means that there is a strong negative correlation between the two curves, that is, they change in the opposite direction. If the correlation coefficient is close to 0, it means that there is no correlation between the two curves.

[0068] When calculating the Pearson correlation coefficient between the picture change curve and the turbine or pump vibration change curve, the curve can be regarded as time series data of two variables, which have corresponding values ​​at the same time point. Therefore, the two curves are divided into two arrays according to the corresponding values ​​at the same time point, and then the covariance and standard deviation of the two arrays are calculated respectively, and finally the covariance and standard deviation are used to calculate the Pearson correlation coefficient.

[0069] When analyzing the correlation between the change in the visual sensor acquisition range and the vibration of a turbine or a water pump, the present invention uses a correlation coefficient to analyze the correlation by using curve comparison, which can intuitively display the correlation between the visual sensor acquisition range and the vibration of a turbine or a water pump, and provides an effective tool for quantifying the correlation between the change in the visual sensor acquisition range and the vibration of a turbine or a water pump.

[0070] S5. Based on the correlation analysis results between the acquisition range change and the turbine or water pump vibration, the calibration timing of the first visual sensor to be adjusted is controlled. The specific operation is as follows: the correlation between the acquisition range change of the first visual sensor to be adjusted and the turbine or water pump vibration is compared with the preset effective positive correlation. For example, the effective positive correlation is 0.8, and the first visual sensor to be adjusted whose correlation is greater than the effective positive correlation is selected as the calibration object.

[0071] It should be understood that, considering that as the vibration of the turbine or pump increases, the acquisition range of the visual sensor will change accordingly, there is a positive correlation between the vibration of the turbine or pump and the acquisition range.

[0072] For the variation degree of the change screen corresponding to the calibration object and the vibration data of the turbine or water pump, a correlation curve between the collection range variation and the turbine or water pump vibration is formed in a two-dimensional coordinate system constructed with the turbine or water pump vibration data as the horizontal coordinate and the variation degree as the vertical coordinate, and the slope of each point on the change curve is obtained, and then the horizontal coordinate corresponding to the maximum slope point in the slopes corresponding to each point of the correlation curve between the collection range variation and the turbine or water pump vibration is taken as the sensitive vibration data.

[0073] During the operation of the turbine or water pump, the vibration data is feedback-connected to the calibration object, and when the vibration data of the turbine or water pump reaches the sensitive vibration data, the calibration object is started for calibration, wherein the calibration is to adjust the position or angle of the visual sensor to maintain the stability of the acquisition range.

[0074] S6. Filter out low-quality images from the historical images collected by the second visual sensor to be adjusted, and retrieve weather data based on the collection time of the low-quality images, thereby performing a correlation analysis between the low quality of the images collected by the second visual sensor to be adjusted and the bad weather.

[0075] In the specific implementation of the above scheme, the screening process of low-quality images is: comparing the image quality coefficient of each historical acquisition image corresponding to the second visual sensor to be adjusted with the standard image quality coefficient, wherein the standard image quality coefficient is the image quality coefficient that needs to be achieved based on the design specifications of the visual sensor. For example, the standard image quality coefficient can be 0.7, and the historical acquisition images with a quality coefficient lower than the standard image quality coefficient are screened out as low-quality images.

[0076] In further implementation, the correlation analysis between the low quality of the image collected by the second visual sensor to be adjusted and the bad weather is realized as follows: the weather data of the corresponding collection time is retrieved for the collection time of the low-quality image, and the influence of sunlight, rain and fog on the image quality is taken into account. The weather data includes light intensity, rainfall and visibility, and the weather severity is calculated based on the weather data. The specific calculation method is to compare the light intensity, rainfall and visibility with the light intensity, rainfall and visibility that the visual sensor can receive for normal operation. The calculation formula is: , represents the severity of the weather, where , , Respectively represent light intensity, rainfall, and visibility. , , They respectively represent the light intensity, rainfall, and visibility that the visual sensor can receive for normal operation.

[0077] A two-dimensional coordinate system is constructed with the acquisition time as the horizontal coordinate and the image quality coefficient as the vertical coordinate, thereby marking several points in the constructed two-dimensional coordinate system according to the acquisition time and image quality coefficient of the second visual sensor to be adjusted corresponding to the low-quality picture to form an image quality change curve.

[0078] A two-dimensional coordinate system is constructed with the acquisition time as the horizontal coordinate and the weather severity as the vertical coordinate. According to the acquisition time of the low-quality image corresponding to the second visual sensor to be adjusted and the weather severity corresponding to the corresponding acquisition time, several points are marked in the constructed two-dimensional coordinate system to form a weather change curve.

[0079] Similarly, referring to the correlation analysis method between the change in the acquisition range and the vibration of the turbine or pump, the correlation between the low quality of the image collected by the second visual sensor to be adjusted and the bad weather is obtained.

[0080] S7. Based on the correlation analysis results between the low quality of the collected images and the bad weather, the collected images of the second visual sensor to be adjusted are corrected and adjusted. The specific process is as follows: the correlation between the low quality of the collected images and the bad weather of the second visual sensor to be adjusted is compared with the preset effective negative correlation. For example, the effective negative correlation is -0.8. The second visual sensor to be adjusted whose correlation is less than the effective negative correlation is selected as the correction object. Here, the influence of negative values ​​on the contrast is taken into account during the comparison.

[0081] It should be understood that as the weather gets worse, the image quality of the images captured by the visual sensor will be affected by the bad weather, resulting in a smaller image quality coefficient. Therefore, there is a negative correlation between the severity of the weather and the quality of the captured images.

[0082] For the image quality coefficient of the low-quality image corresponding to the correction object and the weather severity, a correlation curve between the low quality of the acquired image and the bad weather is formed in a two-dimensional coordinate system with the weather severity as the horizontal coordinate and the image quality coefficient as the vertical coordinate, and the slope of each point on the change curve is obtained, and the absolute value of the slope of each point in the correlation curve between the low quality of the acquired image and the bad weather is compared, and the horizontal coordinate corresponding to the point with the maximum slope absolute value is extracted as the sensitive weather severity.

[0083] Environmental perception sensors, such as infrared cameras and radars, are installed at the location where the correction object is located, and weather data released by the meteorological station is extracted in real time to form a feedback connection with the correction object. The weather severity is then calculated based on the released weather data and compared with the sensitive weather severity. If the sensitive weather severity is reached, the environmental perception sensor is activated for environmental perception, so as to correct and adjust the image captured by the correction object.

[0084] It should be added that infrared thermal imaging can measure the surface temperature of the target object and generate corresponding thermal imaging images without relying on visible light. Radar emits electromagnetic waves and receives their reflected signals, so it is not affected by optical obstacles (such as haze and smoke). As a result, infrared cameras and radars are less affected by bad weather when working. When the image quality of the visual sensor's collected images is poor, the distance measurement capability provided by the radar can be used to accurately locate the target, and then the thermal imaging function of the infrared camera can be used for target recognition and feature extraction to correct the image collected by the visual sensor.

[0085] It is further necessary to add that when determining sensitive vibration data and sensitive weather severity, the point with the largest slope (or absolute value of the slope) is selected from the correlation curve. This is because the tangent slope in the above correlation curve reflects the sensitivity of vibration changes to changes in the acquisition range and weather severity to changes in image quality. The larger the tangent slope, the greater the sensitivity. When we find the point with the largest tangent slope on the curve, it means that under this condition, the sensitivity of turbine or pump vibration to changes in the acquisition range and weather to changes in image quality has reached the highest point. In other words, a small change under this condition will lead to the greatest change in the acquisition range and image quality. Therefore, the vibration data and weather severity corresponding to the point with the largest tangent slope are considered sensitive vibration data and sensitive weather severity.

[0086] The above contents are merely examples and explanations of the structure of the present invention. The technicians in this technical field may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the structure of the invention or exceed the scope defined by the present invention, they should all fall within the protection scope of the present invention.

Claims

1. A method for monitoring, identifying and analyzing the production area of ​​a pumped storage power station based on machine vision, characterized in that: The following steps are involved: S1. Locate the locations of the visual sensors in the production area of ​​the pumped storage power station and number the visual sensors accordingly; S2. Retrieving the historical acquisition images of each visual sensor from the historical acquisition records of the visual sensor to respectively perform acquisition range change analysis and image quality change analysis; S3, identifying the first visual sensor to be adjusted based on the analysis results of the acquisition range change of each visual sensor in the historical acquisition screen, and identifying the second visual sensor to be adjusted based on the analysis results of the acquisition image quality change of each visual sensor in the historical acquisition screen; S4, selecting a changed picture from the historical collection pictures of the first visual sensor to be adjusted, and retrieving the vibration data of the turbine according to the collection time of the changed picture, thereby performing a correlation analysis between the change of the collection range of the first visual sensor to be adjusted and the vibration of the turbine; S5. Controlling the calibration timing of the first visual sensor to be adjusted based on the correlation analysis result between the acquisition range change and the turbine vibration; S6, selecting low-quality images from the historical images collected by the second visual sensor to be adjusted, and retrieving weather data for the collection time of the low-quality images, thereby performing a correlation analysis between the low quality of the images collected by the second visual sensor to be adjusted and the bad weather; S7, based on the correlation analysis result between the low quality of the collected images and the bad weather, the collected images of the second visual sensor to be adjusted are corrected and adjusted; The image quality change analysis process is as follows: Extract image quality parameters from the historical images collected by each visual sensor, including contrast and signal-to-noise ratio; The image quality parameters of the historical collection images are compared with the specification image quality parameters of the corresponding visual sensor to calculate the image quality coefficient of each historical collection image. The specific calculation formula is: , where Indicates the image quality coefficient, , Respectively represent the contrast and signal-to-noise ratio of the historical acquisition images. , They represent the specification contrast and the specification signal-to-noise ratio respectively. The specification image quality parameters of the visual sensor can be obtained from the instruction manual and technical manual of the visual sensor. Compare the image quality coefficients of each historical acquisition screen corresponding to the same visual sensor, and extract the maximum image quality coefficient through the expression Get the image quality difference of the visual sensor , where , Respectively represent the maximum image quality coefficient and the minimum image quality coefficient; Compare the image quality difference corresponding to each visual sensor with the set allowable difference. If the image quality difference corresponding to a certain visual sensor is greater than the allowable difference, analyze whether the image quality of the captured image corresponding to the visual sensor has changed, and use it as the second visual sensor to be adjusted; The specific operation of controlling the timing of calibrating the first visual sensor to be adjusted is as follows: Compare the correlation between the change in the acquisition range of the first visual sensor to be adjusted and the vibration of the turbine with a preset effective positive correlation, and select the first visual sensor to be adjusted whose correlation is greater than the effective positive correlation as a calibration object; For the variation degree of the variation picture corresponding to the calibration object and the turbine vibration data, a correlation curve between the acquisition range variation and the turbine vibration is formed in a two-dimensional coordinate system constructed with the turbine vibration data as the abscissa and the variation degree as the ordinate, and the slope of each point on the correlation curve is obtained, thereby determining the sensitive vibration data; During the operation of the turbine, the vibration data is connected to the calibration object through feedback, and when the vibration data of the turbine reaches the sensitive vibration data, the calibration object is started for calibration; The correction and adjustment of the captured image of the second visual sensor to be adjusted refers to the following process: Compare the correlation between the low quality of the image collected by the second visual sensor to be adjusted and the bad weather with the preset effective negative correlation, and select the second visual sensor to be adjusted whose correlation is less than the effective negative correlation as the correction object; For the image quality coefficient and weather severity of the low-quality image corresponding to the correction object, a correlation curve between low-quality captured images and bad weather is formed in a two-dimensional coordinate system with weather severity as the horizontal coordinate and image quality coefficient as the vertical coordinate, and the slope of each point on the correlation curve is obtained to further determine the sensitive weather severity; An environmental perception sensor is installed at the location where the correction object is located, and the weather data released by the weather station is extracted in real time to form a feedback connection with the correction object. The weather severity is then calculated based on the released weather data and compared with the sensitive weather severity. If the sensitive weather severity is reached, the environmental perception sensor is activated to perceive the environment, so as to correct and adjust the image collected by the correction object.

2. The method for monitoring, identifying and analyzing the production area of ​​a pumped storage power station based on machine vision according to claim 1, characterized in that: The acquisition range change analysis process is as follows: The retrieved historical acquisition images are subjected to edge cutting and target detection respectively, so as to obtain the field of view boundary contour of the historical acquisition images and the boundary contour of the detection target; The historical acquisition pictures retrieved by the same visual sensor are arranged in the order of acquisition time, and each historical acquisition picture is extracted in turn as the main picture according to the arrangement result, and then the field of view boundary contour of the main picture is overlapped and compared with the field of view boundary contour of other historical acquisition pictures, and the detection target of the main picture is mapped with the detection target of other historical acquisition pictures, and then the boundary contour of the detection target is overlapped and compared, thereby combining the two overlap comparison results to calculate the acquisition range similarity between the main picture and the other historical acquisition pictures, and at the same time compare with the set similarity threshold, and other historical acquisition pictures with acquisition range similarity greater than or equal to the similarity threshold and the main picture constitute a collection range picture group, and accordingly the retrieved historical acquisition pictures are grouped by collection range similarity to obtain a number of collection range picture groups; The number of picture groups in the acquisition range formed by each visual sensor is counted and compared with the number of picture groups formed when the acquisition range does not change. If the number of picture groups in the acquisition range formed by a certain visual sensor is greater than the number of picture groups formed when the acquisition range does not change, it is analyzed that there is a change in the acquisition range of the visual sensor, and it is used as the first visual sensor to be adjusted.

3. The method for monitoring, identifying and analyzing the production area of ​​a pumped storage power station based on machine vision according to claim 2, characterized in that: The method of selecting the changed pictures from the historical collection pictures of the first visual sensor to be adjusted is as follows: The percentage of historical acquisition pictures in each acquisition range picture group formed by the first visual sensor to be adjusted is counted, and the acquisition range picture group corresponding to the maximum historical acquisition picture percentage is taken as the dominant picture group, and the historical acquisition pictures in other picture groups are taken as the changing pictures.

4. The method for monitoring, identifying and analyzing the production area of ​​a pumped storage power station based on machine vision as claimed in claim 3, characterized in that: The correlation analysis between the change in the acquisition range of the first visual sensor to be adjusted and the vibration of the turbine is implemented as follows: The field of view boundary contour of the selected changing picture is overlapped and compared with the field of view boundary contour of the dominant picture group to obtain the change degree of the changing picture; A two-dimensional coordinate system is constructed with the acquisition time as the horizontal coordinate and the change degree as the vertical coordinate, thereby marking a number of points in the constructed two-dimensional coordinate system according to the acquisition time and the change degree of the first visual sensor to be adjusted corresponding to the changed picture to form a picture change curve; A two-dimensional coordinate system is constructed with the acquisition time as the horizontal coordinate and the vibration data as the vertical coordinate, thereby marking a number of points in the constructed two-dimensional coordinate system for the acquisition time of the corresponding change screen of the first visual sensor to be adjusted and the turbine vibration data corresponding to the corresponding acquisition time to form a turbine vibration change curve; The correlation coefficient between the picture change curve and the turbine vibration change curve is calculated to obtain the correlation between the change in the acquisition range of the first visual sensor to be adjusted and the turbine vibration.

5. The method for monitoring, identifying and analyzing the production area of ​​a pumped storage power station based on machine vision according to claim 1, characterized in that: The process of selecting low-quality images from the historically collected images of the second visual sensor to be adjusted is as follows: The image quality coefficient of each historical acquisition picture corresponding to the second visual sensor to be adjusted is compared with the standard image quality coefficient, and the historical acquisition pictures with lower than the standard image quality coefficient are screened out as low-quality pictures. The standard image quality coefficient is the image quality coefficient that needs to be achieved based on the design specifications of the visual sensor.

6. The method for monitoring, identifying and analyzing the production area of ​​a pumped storage power station based on machine vision according to claim 5, characterized in that: The correlation analysis between the low quality of the image collected by the second visual sensor to be adjusted and the bad weather is implemented as follows: The weather data of the corresponding collection time is retrieved for the low-quality images, and the weather severity is calculated based on the weather data. The specific calculation method is to compare the light intensity, rainfall and visibility with the light intensity, rainfall and visibility that the visual sensor can receive when working normally. The calculation formula is: , represents the severity of the weather, where , , Respectively represent light intensity, rainfall, and visibility. , , They represent the light intensity, rainfall, and visibility that the visual sensor can receive for normal operation. A two-dimensional coordinate system is constructed with the acquisition time as the horizontal coordinate and the image quality coefficient as the vertical coordinate, thereby marking a number of points in the constructed two-dimensional coordinate system according to the acquisition time and the image quality coefficient of the second visual sensor to be adjusted corresponding to the low-quality picture to form an image quality change curve; A two-dimensional coordinate system is constructed with the acquisition time as the horizontal coordinate and the weather severity as the vertical coordinate, thereby marking a number of points in the constructed two-dimensional coordinate system according to the acquisition time of the low-quality image corresponding to the second visual sensor to be adjusted and the weather severity corresponding to the corresponding acquisition time to form a weather change curve; The correlation coefficient between the image change curve and the turbine vibration change curve is calculated to obtain the correlation between the low quality of the image collected by the second visual sensor to be adjusted and the bad weather.

7. The method for monitoring, identifying and analyzing the production area of ​​a pumped storage power station based on machine vision according to claim 1, characterized in that: The sensitive vibration data is determined as follows: The horizontal coordinate corresponding to the maximum slope point among the slopes corresponding to each point of the correlation curve between the acquisition range change and the turbine vibration is taken as the sensitive vibration data.

Citation Information

Patent Citations

  • Intelligent inspection method and system for hydropower station and storage medium

    CN118233597A

  • Indoor environment quality display method and terminal

    CN110146941A

  • Camera testing device and method, electronic equipment and computer readable storage medium

    CN114040093A

  • HiL test platform of photovoltaic power station and performance test method of PPC

    CN114721354A

  • Counting analysis system for photovoltaic panels in intelligent inspection scene of photovoltaic power station unmanned aerial vehicle

    CN118570680A