CN-121982574-A - Hydraulic engineering leakage detection method based on computer vision

CN121982574ACN 121982574 ACN121982574 ACN 121982574ACN-121982574-A

Abstract

A hydraulic engineering leakage detection method based on computer vision relates to the technical field of engineering detection, and has the advantages of extracting historical remote sensing parameters, generating parameter fluctuation ranges of different pixels in a target dam region, extracting current remote sensing parameters, comparing the current remote sensing parameters with the parameter fluctuation ranges to obtain parameter deviation degrees of the different pixels, extracting abnormal pixels in the target dam region, performing spatial clustering treatment on the abnormal pixels to obtain abnormal pixel regions, respectively obtaining morphological risk indexes and deviation risk indexes of the abnormal pixel regions, obtaining comprehensive risk indexes of the abnormal pixel regions, obtaining leakage risk grades according to the comprehensive risk indexes, and combining a preset dam region map to generate a leakage early warning map of the target dam region and feeding back.

Inventors

LIU JINPENG
PANG SONG
LIN JIAZHEN
ZHANG JINGPING
SONG HAONAN
JING ZHENZHEN
ZHAO FEI
LIN LI

Assignees

三门峡金鼎机械设备有限公司

Dates

Publication Date: 20260505
Application Date: 20260131

Claims (10)

1. The hydraulic engineering leakage detection method based on computer vision is characterized by comprising the following steps of: Acquiring a historical remote sensing image, regional meteorological data and a current remote sensing image of a target dam region; extracting different types of historical remote sensing parameters according to the historical remote sensing images, and generating parameter fluctuation ranges at different pixels in a target dam area by combining regional meteorological data; Extracting different types of current remote sensing parameters according to the current remote sensing image, and comparing the current remote sensing parameters with the parameter fluctuation range to obtain parameter deviation degrees at different pixels; extracting abnormal pixels in the target dam area according to the parameter deviation degree, and performing spatial clustering on the abnormal pixels to obtain an abnormal pixel area; Acquiring the accumulated time length of the abnormal pixel in the abnormal pixel area within a preset time window, generating a deviation adjusting factor, and adjusting the parameter deviation degree according to the deviation adjusting factor to obtain a corrected deviation degree; Respectively obtaining a morphological risk index and a deviation risk index according to the geometric characteristics and the correction deviation degree of the abnormal pixel region, and obtaining a comprehensive risk index; And obtaining leakage risk grades corresponding to the abnormal pixel areas according to the comprehensive risk indexes, and generating a leakage early warning map of the target dam area by combining a preset dam area map and feeding back the leakage early warning map.
2. The computer vision-based hydraulic engineering leakage detection method according to claim 1, wherein the historical remote sensing parameters comprise historical vegetation index, historical surface temperature and historical soil humidity; Respectively extracting vegetation indexes, surface temperatures and soil humidity of each geographic pixel in different months according to the spectral characteristics of the historical remote sensing images to serve as the historical vegetation indexes, the historical surface temperatures and the historical soil humidity of the corresponding geographic pixels in the corresponding months; the geographic pixels refer to pixels corresponding to the same geographic position in the target dam area in the historical remote sensing image, and the average value of the same historical remote sensing parameters in the same month is obtained for each geographic pixel And standard deviation And generating a parameter fluctuation range of the corresponding historical remote sensing parameters of the geographic pixels in the corresponding months ; The said For the dynamic influence factor of the month corresponding to the parameter fluctuation range, The said For the weather influencing factors of the corresponding month, ; For the preset weight value to be set, For the precipitation of the corresponding month in the present year and its average precipitation over the years, both are obtained based on the regional weather data, the And the trend influence factor of the parameter fluctuation range corresponding to the historical remote sensing parameter is obtained by normalizing the average value of the change slope of the historical remote sensing parameter in the past year.
3. The computer vision-based hydraulic engineering leakage detection method according to claim 2, wherein the current remote sensing parameters comprise a current vegetation index, a current surface temperature and a current soil humidity; Respectively extracting the vegetation index, the surface temperature and the soil humidity of each geographic pixel in the current month according to the spectral characteristics of the current remote sensing image to serve as the current vegetation index, the current surface temperature and the current soil humidity of the corresponding geographic pixel in the current month; each current remote sensing parameter of a single geographic pixel is compared with the corresponding parameter fluctuation range of the current remote sensing parameter in the current month according to each geographic pixel, and if the current remote sensing parameter is in the parameter fluctuation range, the parameter deviation degree of the current remote sensing parameter is recorded as 0; And if the current remote sensing parameter is not in the parameter fluctuation range, acquiring the parameter deviation degree of the current remote sensing parameter, wherein the parameter deviation degree is equal to a value obtained by subtracting the intermediate value of the parameter fluctuation range from the current remote sensing parameter and dividing the difference value by the intermediate value, and acquiring the parameter deviation degree of the current remote sensing parameter of different types at each geographic pixel.
4. The computer vision-based hydraulic engineering leakage detection method according to claim 3, wherein the parameter deviation degree comprises vegetation parameter deviation degree, temperature parameter deviation degree and humidity parameter deviation degree; When the vegetation parameter deviation degree of a single geographic pixel is smaller than a preset vegetation threshold value, the temperature parameter deviation degree of the single geographic pixel is smaller than a preset low-temperature threshold value, and the humidity parameter deviation degree of the single geographic pixel is larger than a preset high-humidity threshold value, marking the single geographic pixel as an abnormal pixel, wherein the preset vegetation threshold value and the preset high-humidity threshold value are positive numbers, and the preset low-temperature threshold value is negative numbers; And carrying out spatial clustering on all abnormal pixels in the current month based on density by using a K-means clustering algorithm, classifying each abnormal pixel with a spatial distance smaller than a preset distance threshold and with the number of pixels in the cluster larger than a preset number threshold into a cluster, and respectively outputting each cluster into an abnormal pixel area.
5. The computer vision-based hydraulic engineering leakage detection method according to claim 4, wherein the accumulated time length T of each abnormal pixel determined to be in the abnormal pixel region in the preset time window is acquired for each abnormal pixel; the deviation adjustment factor K has a value of 0.5-2.0, And multiplying each parameter deviation value of the corresponding abnormal pixel by the corresponding deviation adjustment factor to obtain a corresponding correction deviation value for the total duration of the preset time window, wherein the correction deviation value comprises vegetation correction deviation degree, temperature correction deviation degree and humidity correction deviation degree.
6. The computer vision-based hydraulic engineering leakage detection method according to claim 5, wherein the morphological risk index of each abnormal pixel region is obtained respectively , For the preset weight value to be set, The area, perimeter and length-width ratio of the minimum circumscribing rectangle corresponding to the abnormal pixel area respectively, Respectively the functions of the three; obtaining deviation risk indexes of each abnormal pixel in each abnormal pixel area according to the corrected deviation degree of each abnormal pixel Obtaining the comprehensive risk index according to the morphological risk index and the deviation risk index of the same abnormal pixel region ; In order to adjust the factor(s), For the preset weight value to be set, Respectively the average value of vegetation correction deviation degree, temperature correction deviation degree and humidity correction deviation degree of each abnormal pixel in the corresponding abnormal pixel region, Respectively a preset vegetation threshold value, a preset low temperature threshold value and a preset high humidity threshold value.
7. The computer vision-based hydraulic engineering leakage detection method according to claim 6, wherein the function , For a preset area threshold, the function The function is ; The regulatory factor , For the preset basic adjustment factor, the control unit, The normalized area of the corresponding abnormal pixel area and the aspect ratio of the minimum circumscribed rectangle are respectively obtained.
8. The computer vision-based hydraulic engineering leakage detection method according to claim 1, wherein the comprehensive risk index of a single abnormal pixel area is calculated Respectively and preset risk ranges Comparing, wherein the leakage risk level comprises low risk, medium risk and high risk; If it is The abnormal pixel area is marked as low risk if The abnormal pixel area is marked as high risk if Marking the abnormal pixel area as a risk of stroke; and taking the spatial profile of each abnormal pixel area and the leakage risk level thereof as a layer, superposing the spatial profile and the leakage risk level thereof on a preset dam area map representing the geographic characteristics of the target dam area, differentiating different leakage risk levels by adopting different colors to generate a leakage early warning map of the target dam area, and feeding the leakage early warning map back to related personnel.
9. Hydraulic engineering seepage detection system based on computer vision, its characterized in that includes following module: The data acquisition module is used for acquiring historical remote sensing images, regional weather data and current remote sensing images of the target dam region; the parameter extraction module is used for extracting different types of historical remote sensing parameters according to the historical remote sensing images and generating parameter fluctuation ranges at different pixels in a target dam area by combining the regional meteorological data; The deviation analysis module is used for extracting different types of current remote sensing parameters according to the current remote sensing image, and comparing the current remote sensing parameters with the parameter fluctuation range to obtain parameter deviation degrees at different pixels; The region extraction module is used for extracting abnormal pixels in the target dam region according to the parameter deviation degree, and performing spatial clustering processing on the abnormal pixels to obtain an abnormal pixel region of the abnormal pixels; the deviation correction module is used for obtaining the accumulated duration of the abnormal pixel in the abnormal pixel area in a preset time window, generating a deviation adjustment factor, and adjusting the parameter deviation degree according to the deviation adjustment factor to obtain a correction deviation degree; The risk assessment module is used for respectively obtaining a morphological risk index and a deviation risk index according to the geometric characteristics and the correction deviation degree of the abnormal pixel region and obtaining a comprehensive risk index of the abnormal pixel region; And the map feedback module is used for obtaining the leakage risk grade of the corresponding abnormal pixel area according to the comprehensive risk index, and generating a leakage early warning map of the target dam area by combining a map of the preset dam area and feeding back the leakage early warning map.
10. A computer storage medium storing computer executable instructions which when executed implement the computer vision-based hydraulic engineering leak detection method of any one of claims 1-8.

Description

Hydraulic engineering leakage detection method based on computer vision Technical Field The application relates to the technical field of engineering detection, in particular to a hydraulic engineering leakage detection method based on computer vision. Background The traditional hydraulic engineering leakage detection mainly relies on point-like contact sensors such as pressure measuring pipes, osmometers and the like to collect data, and the sensors need to be densely distributed at key positions of a dam body, so that the early installation cost and the later maintenance cost are obviously increased, and particularly for large hydraulic engineering or dam areas with complex terrains, the distribution difficulty and the economic burden are more outstanding, so that an effective means for leakage detection by a remote sensing technology appears; However, although the detection means based on the remote sensing technology can provide a large-scale surface coverage, the detection means still have obvious defects in practical application, for example, the abnormality detection process lacks effective coordination to multiple parameters, the risk assessment link fails to finely classify the geometric features of the abnormal region, so that the early warning result lacks pertinence and practicality, the problems severely restrict the early warning capability of leakage detection, and the prior art needs improvement aiming at the problems. Disclosure of Invention The application aims to provide a hydraulic engineering leakage detection method based on computer vision, which has the advantages of reducing the monitoring cost, realizing global continuous monitoring, improving the accuracy of leakage detection and early warning capability and reducing false alarm. The application aims at realizing the technical scheme that the hydraulic engineering leakage detection method based on computer vision comprises the following steps: Acquiring a historical remote sensing image, regional meteorological data and a current remote sensing image of a target dam region; extracting different types of historical remote sensing parameters according to the historical remote sensing images, and generating parameter fluctuation ranges at different pixels in a target dam area by combining regional meteorological data; Extracting different types of current remote sensing parameters according to the current remote sensing image, and comparing the current remote sensing parameters with the parameter fluctuation range to obtain parameter deviation degrees at different pixels; extracting abnormal pixels in the target dam area according to the parameter deviation degree, and performing spatial clustering on the abnormal pixels to obtain an abnormal pixel area; Acquiring the accumulated time length of the abnormal pixel in the abnormal pixel area within a preset time window, generating a deviation adjusting factor, and adjusting the parameter deviation degree according to the deviation adjusting factor to obtain a corrected deviation degree; Respectively obtaining a morphological risk index and a deviation risk index according to the geometric characteristics and the correction deviation degree of the abnormal pixel region, and obtaining a comprehensive risk index; And obtaining leakage risk grades corresponding to the abnormal pixel areas according to the comprehensive risk indexes, and generating a leakage early warning map of the target dam area by combining a preset dam area map and feeding back the leakage early warning map. In a second aspect, a hydraulic engineering leakage detection system based on computer vision comprises the following modules: The data acquisition module is used for acquiring historical remote sensing images, regional weather data and current remote sensing images of the target dam region; the parameter extraction module is used for extracting different types of historical remote sensing parameters according to the historical remote sensing images and generating parameter fluctuation ranges at different pixels in a target dam area by combining the regional meteorological data; The deviation analysis module is used for extracting different types of current remote sensing parameters according to the current remote sensing image, and comparing the current remote sensing parameters with the parameter fluctuation range to obtain parameter deviation degrees at different pixels; The region extraction module is used for extracting abnormal pixels in the target dam region according to the parameter deviation degree, and performing spatial clustering processing on the abnormal pixels to obtain an abnormal pixel region of the abnormal pixels; the deviation correction module is used for obtaining the accumulated duration of the abnormal pixel in the abnormal pixel area in a preset time window, generating a deviation adjustment factor, and adjusting the parameter deviation degree according to the deviation adjustment factor t