Flood situation detection method based on computer vision technology

Abstract

The invention discloses a flood situation detection method based on a computer vision technology. According to the method, hydrological information, meteorological information, dam multi-angle images and structural displacement data are acquired in real time through a data acquisition module; the data processing module performs filtering, denoising, standardization and semantic segmentation on the data, and extracts pixel-level masks of a water area and a dam crack; the three-dimensional modeling module constructs a dynamic three-dimensional model of a water area and a dam; the data analysis module evaluates the stability of the river embankment in combination with parameters such as deformation gradient and crack amplification rate, and judges a risk level based on a water level deviation degree, a flow velocity abnormal index and a stability coefficient; finally, a three-dimensional dynamic model containing a risk thermodynamic diagram and a crack evolution report and a structured PDF detection report are generated, and water level simulation, history playback and report export are supported through a computer terminal. According to the invention, automatic and visual accurate early warning of flood risk is realized through multi-source data fusion.

Description

Technical Field

[0001] The present invention belongs to the technical field of smart water conservancy monitoring, and specifically discloses a flood situation detection method based on computer vision technology. Background Art

[0002] Traditional flood monitoring methods rely primarily on manual inspections and single-sensor data collection, which have the following limitations: Insufficient efficiency and coverage: Manual inspections take a long time, and there are blind spots at night and in bad weather, making it difficult to detect sudden emergencies in a timely manner. Single data dimension: Existing systems are mostly based on isolated sensors such as water level gauges and rain gauges. They lack quantitative perception of the dam's structural status and are unable to capture the precursory characteristics of dam failures. One-sided risk assessment: Relying on static threshold alarms, such as water level exceeding the limit, without integrating the dynamic coupling relationship between multiple parameters such as weather forecasts, abnormal flow velocity, and structural stability, resulting in a high false alarm rate and difficulty in locating risk areas; Lack of visualization: 2D plane images cannot show the 3D deformation gradient of the dam and the spatial evolution trend of cracks, making it difficult for decision makers to intuitively assess the risk situation; Although computer vision technology has achieved object recognition and 3D reconstruction in other fields, an automated flood analysis system that integrates multi-source heterogeneous data has not yet been established in smart water conservancy monitoring. Therefore, it is necessary to invent a flood detection method based on computer vision technology to solve the above problems. Summary of the Invention

[0003] In order to overcome the above-mentioned defects of the prior art, the present invention provides a flood detection method based on computer vision technology, which obtains hydrological information, meteorological information, multi-angle images of dams and structural displacement data in real time through a data acquisition module; the data processing module filters, denoises, standardizes and semantically segments the data, and extracts pixel-level masks of water areas and dam cracks; the three-dimensional modeling module constructs a dynamic three-dimensional model of the water area and dam; the data analysis module combines parameters such as deformation gradient and crack amplification rate to evaluate the stability of the river embankment, and determines the risk level based on water level deviation, flow velocity anomaly index and stability coefficient; finally, a three-dimensional dynamic model containing a risk heat map and a crack evolution report and a structured PDF detection report are generated, and water level simulation, historical playback and report export are supported through a computer terminal, effectively solving the problems mentioned in the background technology.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a flood detection method based on computer vision technology, comprising a data acquisition module, a data processing module, a three-dimensional modeling module, a data analysis module, and a computer terminal, specifically comprising the following steps: S1, data acquisition module collects hydrological information, meteorological data, image information and dam status data; S2, the data processing module performs preprocessing and image segmentation on the collected information to obtain preprocessed data; S3, the three-dimensional modeling module establishes a three-dimensional dynamic model of the water area and dam based on the pre-processed data; S4, the data analysis module combines pre-processed data with the three-dimensional dynamic model to analyze the stability of the riverbank and determine the risk level, and finally generates a flood detection report; S5. The computer terminal outputs an operational three-dimensional dynamic model and flood detection report.

[0005] Based on the above embodiment, the hydrological data includes real-time water level and flow rate; the meteorological data includes precipitation and wind speed; the image data is multi-angle images of the water area and dam; and the dam status data includes soil pressure data and structural displacement.

[0006] On the basis of the above embodiment, the collected data is pre-processed and image segmented, and the specific analysis method is as follows: Eliminate environmental interference by filtering image data; Standardize hydrological data to a common dimension; A semantic segmentation algorithm is used on the image data to separate the water area, dam structure and background environment, and to extract the pixel-level mask of cracks on the dam surface.

[0007] Based on the above embodiment, the analysis of riverbank stability is carried out in the following specific manner: The dam deformation gradient is obtained by combining the dam structure displacement in the 3D dynamic model; Analyze the crack length, width, expansion rate and direction evolution trend based on the extracted crack mask; Substitute the embankment deformation gradient, crack length, width and expansion rate into the embankment stability coefficient calculation formula to calculate the embankment stability coefficient; The stability of riverbanks is assessed based on the riverbank stability coefficient.

[0008] Based on the above embodiment, the calculation formula of the riverbank stability coefficient is: , where: S c is the stability coefficient, ▽D is the deformation gradient, R c is the crack expansion rate, σ is the compressive strength of the dam matrix, P e is the real-time earth pressure anomaly index, L c is the crack length, W c is the average crack width, A c is the critical crack area threshold, λ is the material aging coefficient, and K is the geological calibration constant.

[0009] Based on the above embodiment, the flood risk level is determined by the following specific analysis method: Calculate the percentage of real-time water level deviation from the warning value based on the real-time water level and the water level warning value; Calculate the velocity anomaly index based on the real-time velocity and the historical average velocity; Substitute the percentage of real-time water level deviation from the warning value, the velocity anomaly index, and the riverbank stability coefficient into the flood risk coefficient calculation formula to calculate the flood risk coefficient; Assess flood risk level based on flood risk coefficient.

[0010] Based on the above embodiment, the flood risk coefficient calculation formula is: , in the formula, R f is the flood risk coefficient, H r is the percentage of real-time water level deviation from the warning value, H c is the critical threshold of water level mutation, P f is the predicted value of cumulative precipitation in the next 3 hours, V f is the velocity anomaly index, S c is the normalized riverbank stability coefficient, and η, α, and β are basin characteristic parameters.

[0011] Based on the above embodiment, the flood detection report is generated, specifically including the following structured data: Risk situation map: superimpose a real-time risk coefficient heat map on the 3D dynamic model; Key parameter table: time series changes of parameters such as water level / flow velocity / deformation gradient and over-limit marking; Crack evolution report: marking crack location, size change trend and structural safety impact assessment; Early warning recommendations: Propose key inspection areas and engineering disposal measures based on risk levels.

[0012] Based on the above embodiment, the output of the operational three-dimensional dynamic model and flood detection report is specifically analyzed as follows: The interactive 3D model is rendered on the computer terminal through the WebGL engine, which supports dynamic adjustment of the water level to simulate the flooding range and playback of the model state evolution operation in historical periods. Export risk reports to PDF format with one click.

[0013] Technical effects and advantages of the present invention: 1. Multi-source data fusion: Integrate hydrological, meteorological, visual and structural data to improve monitoring comprehensiveness; 2. Accurate crack identification: semantic segmentation enables pixel-level crack extraction and accurate acquisition of crack information; 3. Dynamic risk visualization: 3D heat maps intuitively display risk distribution and support decision-making; 4. Intelligent early warning: through S c With R f Double coefficient model to predict dam break risk in advance. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The present invention is further described with reference to the accompanying drawings. However, 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 effort.

[0015] Figure 1 It is a schematic diagram of the overall structure of the present invention.

[0016] Figure 2 It is an overall step diagram of the present invention. DETAILED DESCRIPTION

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0018] The present invention provides a flood detection method based on computer vision technology, the structure of which is as follows: Figure 1 As shown, it includes a data acquisition module, a data processing module, a three-dimensional modeling module, a data analysis module and a computer terminal; like Figure 2 As shown, the flood detection method based on computer vision technology specifically includes the following steps: S1, data acquisition module collects hydrological information, meteorological data, image information and dam status data; In a more specific application of the present invention, the data acquisition module collects hydrological data, meteorological data, multi-angle images of the dam and structural displacement in real time through a water level meter, a current meter, a weather station, a camera and an earth pressure sensor; Furthermore, in the above technical solution, the hydrological data includes real-time water level and flow rate; the meteorological data includes precipitation and wind speed; the image data is multi-angle images of the water area and dam; and the dam status data includes soil pressure data and structural displacement.

[0019] S2, the data processing module performs preprocessing and image segmentation on the collected information to obtain preprocessed data; Furthermore, in the above technical solution, the collected data is preprocessed and image segmented, and the specific analysis method is as follows: Eliminate environmental interference by filtering image data; Standardize hydrological data to a common dimension; A semantic segmentation algorithm is used on the image data to separate the water area, dam structure and background environment, and to extract pixel-level masks of cracks on the dam surface; In a more specific application of the present invention, the image data is filtered to eliminate environmental interference, and the specific process is as follows: Select the filtering method to remove noise according to the lighting conditions. If it is rainy or foggy, use guided filtering + wavelet threshold denoising; if it is normal weather, use median filtering; The specific process of standardizing hydrological data is as follows: Data cleaning: Eliminate outliers based on the 3σ principle and outlier verification; Missing value filling: fill missing data based on the linear interpolation algorithm of time series; Water level and flow velocity data were normalized using z-score.

[0020] It should be further explained that the specific process of data cleaning is as follows: If the measured value deviates from the historical average data of the same month by more than three standard deviations, it will be marked as suspected abnormal data; Verify the marked suspected abnormal data: If the suspected abnormal data is abnormal water level data, retrieve the data of the adjacent water level meter at the same time for verification. If the difference between the suspected abnormal data and the average value of the adjacent data is less than 0.1m, cancel the mark; otherwise, mark it as abnormal data; If the suspected abnormal data is flow velocity data, retrieve the flow velocity data of the flow meter before and after for verification. If the difference between the suspected abnormal data and the average value of the before and after data is less than 1m / s, cancel the mark; otherwise, mark it as abnormal data; According to the formula Normalize the data, and x in the formula n is the standardized data, x i is the actual data, μ h is the historical mean for the same period, σ h is the historical standard deviation.

[0021] S3, the three-dimensional modeling module establishes a three-dimensional dynamic model of the water area and dam based on the pre-processed data; S4, the data analysis module combines pre-processed data with the three-dimensional dynamic model to analyze the stability of the riverbank and determine the risk level, and finally generates a flood detection report; In a more specific application of the present invention, the analysis of riverbank stability is carried out in the following manner: The dam deformation gradient is obtained by combining the dam structure displacement in the 3D dynamic model; Analyze the crack length, width, expansion rate and direction evolution trend based on the extracted crack mask; Substitute the embankment deformation gradient, crack length, width and expansion rate into the embankment stability coefficient calculation formula to calculate the embankment stability coefficient; The stability of riverbanks is assessed based on the riverbank stability coefficient.

[0022] Furthermore, in the above technical solution, the formula for calculating the riverbank stability coefficient is: , where: S c is the stability coefficient, ▽D is the deformation gradient, R c is the crack expansion rate, σ is the compressive strength of the dam matrix, P e is the real-time earth pressure anomaly index, L c is the crack length, W c is the average width, A c is the critical crack area threshold, λ is the material aging coefficient, and K is the geological calibration constant, which is inverted and fitted through historical dam break data.

[0023] It should be further explained that the stability coefficient S c The lower the value, the more dangerous it is; real-time soil pressure anomaly index P e The calculation formula is: e =|P 实时 P 基线 | / P 基线 , in the formula P 实时 is the measured value of earth pressure, P 基线 is the earth pressure baseline value, which is a dynamic reference value for evaluating whether the real-time earth pressure of the embankment is abnormal; the material aging coefficient λ∈[0.2, 0.8], when the embankment material is concrete, λ=0.2, when the embankment material is soil and rock, λ=0.6+; the critical crack area threshold A c Determined by the type of dam. When the dam material is concrete, A c =2.0, when the embankment material is soil and rock, A c =0.5; the calculation formula of the geological calibration constant K is: , where N is the number of historical data, S c(i) is the stability coefficient corresponding to the i-th historical data, ▽D i is the deformation gradient in the i-th historical data, R c(i) The crack propagation rate in the i-th historical data; It should be further explained that the specific method of evaluating the stability of the riverbank is: Using the dam stability benchmark value S cx The stability coefficient S c Normalize to obtain the normalization coefficient S, that is, S=Sc / S cx , S cx Determined based on dry season data in the first year after the dam is completed; When S ≥ 0.9, the structure is considered stable; When 0.7≤S<0.9, it is determined to be a metastable state; When 0.5≤S<0.7, it is determined to be a local unstable state; When S<0.5, it is determined to be in an overall unstable state; If S c If the decline rate is greater than 10% for three consecutive monitoring cycles, the risk level will be automatically raised to a higher level. When a through crack is detected, that is, the crack length is greater than 50% of the dam cross-section width, it will be forcibly judged as a level above the local instability risk level.

[0024] Stability determination calculation example: Parameter settings: Historical data: S c(1) =0.2, ▽D1=0.82, R c(1) =0.018, S c(2) =0.2, ▽D2=0.95, R c(2) =0.025, S c(3) =0.2, ▽D3=1.10, R c(3) =0.032, S cx =4.2, ▽D=0.38, R c =0.004,σ=35,P 实时 =26.5, P 基线 =22,L c =0.8, W c =0.03, A c =0.02,λ=0.25; Calculation process: K=[0.2×0.82×(1+e 0.018 )+0.2×0.95×(1+e 0.025 )+0.2×1.10×(1+e 0.032 )] / 3≈(0.331+0.385+0.447) / 3≈0.39; P e =|26.5 22| / 22=4.5 / 22=0.2045; S c =0.39×ln(1+35 / 0.2045)×[1-0.25×tanh(0.8×0.03 / 0.02)] / [0.38×(1+e 0.004)]=0.39×ln(172.15)×0.7917 / 0.7615≈2.088; S=S c / S cx =2.088 / 4.2≈0.497; Conclusion: The dam is currently in a state of overall instability and requires timely treatment.

[0025] In a more specific application of the present invention, the flood risk level is determined by analyzing the following: Calculate the percentage of real-time water level deviation from the warning value based on the real-time water level and the water level warning value; Calculate the velocity anomaly index based on the real-time velocity and the historical average velocity; Substitute the percentage of real-time water level deviation from the warning value, the velocity anomaly index, and the riverbank stability coefficient into the flood risk coefficient calculation formula to calculate the flood risk coefficient; Assess flood risk level based on flood risk coefficient.

[0026] Furthermore, in the above technical solution, the real-time water level deviates from the warning value percentage H r The calculation formula is: r =[(H 实时 H 警戒 ) / H 警戒 ]×100%, where H 实时 is the real-time water level measurement value, H 警戒 The preset value for the warning water level; The flow rate abnormality index V f The calculation formula is: V f =|V 实时 V 平 ∣ / V 平 , in the formula V 实时 is the real-time flow rate measurement value, V 平 is the average value of historical velocity data for the same period; Furthermore, in the above technical solution, the flood risk coefficient calculation formula is: , in the formula R f is the flood risk coefficient, H r is the percentage of real-time water level deviation from the warning value, H c is the critical threshold of water level mutation, P f is the predicted value of cumulative precipitation in the next 3 hours, V f is the velocity anomaly index, S c is the embankment stability coefficient, η, α and β are basin characteristic parameters; F(x) is the hydrodynamic load term.

[0027] Furthermore, in the above technical solution, the calculation formula of the hydrodynamic load term F(x) is: ; It should be further explained that the flood risk coefficient R f ∈[0, 1], the higher the value, the greater the risk; the default value of the critical threshold of water level mutation is H c =0.15H 警戒 The initial values ​​of the basin characteristic parameters are set based on the basin type: mountainous rivers: η=1.8, α=2.2, β=1.5; plain rivers: η=1.2, α=1.6, β=2.0; tidal river sections: η=2.5, α=1.8, β=1.2; It should be further explained that the specific method of assessing flood risk is as follows: The current flood risk factor R f When ≤0.3, it is judged as low risk; When 0.3<R≤0.7, it is judged as medium risk; When R>0.7, it is judged as high risk; When a sudden change in the local deformation gradient of the dam is detected or the direction of the crack is orthogonal to the direction of the water flow, the risk level will be automatically increased by one level; Flood risk factor calculation example: Parameter setting: S c =0.5, η=1.2, α=1.6, β=2.0, H c =0.15H 警戒 , H 实时 =45.8, H 警戒 =42, P f =60, V 实时 =3.2, V 平 =1.8; V f =|V 实时 V 平 ∣ / V 平 =(3.2-1.8) / 1.8=0.778; F(P f ·V f )=F(0.778×60)=F(46.68)=0.4+0.6×tanh(0.8×45.68)≈1; H c =0.15×42=6.3%; H r =[(45.8-42) / 42]×100%=9.05%; -η×(H r / H c )α =-1.2×(9.05 / 6.3)≈-1.728; R f =[1-e -1.728 ]×1×e -2×0.5 =0.82×1×0.37≈0.30; Conclusion: The current flood risk coefficient is 0.30, which is at the critical point of medium and low risk. It is necessary to strengthen monitoring and pay attention to development trends.

[0028] S5. The computer terminal outputs an operational three-dimensional dynamic model and flood detection report; Furthermore, in the above technical solution, the generated flood detection report specifically includes the following structured data: Risk situation map: superimpose a real-time risk coefficient heat map on the 3D dynamic model; Key parameter table: time series changes of parameters such as water level / flow velocity / deformation gradient and over-limit marking; Crack evolution report: marking crack location, size change trend and structural safety impact assessment; Early warning recommendations: Propose key inspection areas and engineering disposal measures based on risk levels.

[0029] The output of the operational three-dimensional dynamic model and flood detection report is specifically analyzed as follows: The interactive 3D model is rendered on the computer terminal through the WebGL engine, which supports dynamic adjustment of the water level to simulate the flooding range and playback of the model state evolution operation in historical periods. Export risk reports to PDF format with one click.

[0030] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A flood detection method based on computer vision technology, characterized in that: It includes a data acquisition module, a data processing module, a 3D modeling module, a data analysis module and a computer terminal, and specifically includes the following steps: S1, data acquisition module collects hydrological information, meteorological data, image information and dam status data; S2, the data processing module performs preprocessing and image segmentation on the collected information to obtain preprocessed data; S3, the three-dimensional modeling module establishes a three-dimensional dynamic model of the water area and dam based on the pre-processed data; S4, the data analysis module combines pre-processed data with the three-dimensional dynamic model to analyze the stability of the riverbank and determine the risk level, and finally generates a flood detection report; S5. The computer terminal outputs an operational three-dimensional dynamic model and flood detection report.

2. The flood detection method based on computer vision technology according to claim 1, characterized in that: The hydrological data includes real-time water level and flow rate; the meteorological data includes precipitation and wind speed; the image data is multi-angle images of the water area and dam; and the dam status data includes soil pressure data and structural displacement.

3. The flood detection method based on computer vision technology according to claim 1, characterized in that: The collected data is preprocessed and image segmented, and the specific analysis method is as follows: Eliminate environmental interference by filtering image data; Standardize hydrological data to a common dimension; A semantic segmentation algorithm is used on the image data to separate the water area, dam structure and background environment, and to extract the pixel-level mask of cracks on the dam surface.

4. The flood detection method based on computer vision technology according to claim 1, characterized in that: The specific analysis method for analyzing the stability of the riverbank is as follows: The dam deformation gradient is obtained by combining the dam structure displacement in the 3D dynamic model; Analyze the crack length, width, expansion rate and direction evolution trend based on the extracted crack mask; Substitute the embankment deformation gradient, crack length, width and expansion rate into the embankment stability coefficient calculation formula to calculate the embankment stability coefficient; The stability of riverbanks is assessed based on the riverbank stability coefficient.

5. The flood detection method based on computer vision technology according to claim 4, characterized in that: The calculation formula of the riverbank stability coefficient is: , where: S c is the stability coefficient, ▽D is the deformation gradient, R c is the crack expansion rate, σ is the compressive strength of the dam matrix, P e is the real-time earth pressure anomaly index, L c is the crack length, W c is the average width, A c is the critical crack area threshold, λ is the material aging coefficient, and K is the geological calibration constant.

6. The flood detection method based on computer vision technology according to claim 1, characterized in that: The specific analysis method for determining the flood risk level is as follows: Calculate the percentage of real-time water level deviation from the warning value based on the real-time water level and the water level warning value; Calculate the velocity anomaly index based on the real-time velocity and the historical average velocity; Substitute the percentage of real-time water level deviation from the warning value, the velocity anomaly index, and the riverbank stability coefficient into the flood risk coefficient calculation formula to calculate the flood risk coefficient; Assess flood risk level based on flood risk coefficient.

7. The flood detection method based on computer vision technology according to claim 1, characterized in that: The calculation formula for the flood risk coefficient is: , in the formula R f is the flood risk coefficient, H r is the percentage of real-time water level deviation from the warning value, H c is the critical threshold of water level mutation, P f is the predicted value of cumulative precipitation in the next 3 hours, V f is the velocity anomaly index, S c is the riverbank stability coefficient, η, α and β are the basin characteristic parameters.

8. The flood detection method based on computer vision technology according to claim 1, characterized in that: The generated flood detection report specifically includes the following structured data: Risk situation map: superimpose a real-time risk coefficient heat map on the 3D dynamic model; Key parameter table: time series changes of parameters such as water level / flow velocity / deformation gradient and over-limit marking; Crack evolution report: marking crack location, size change trend and structural safety impact assessment; Early warning recommendations: Propose key inspection areas and engineering disposal measures based on risk levels.

9. The flood detection method based on computer vision technology according to claim 1, characterized in that: The output of the operational three-dimensional dynamic model and flood detection report is specifically analyzed as follows: The interactive 3D model is rendered on the computer terminal through the WebGL engine, which supports dynamic adjustment of the water level to simulate the flooding range and playback of the model state evolution operation in historical periods. Export risk reports to PDF format with one click.

Images