A method and detection system for accurately characterizing groundwater pollution

Abstract

The present invention discloses a method and a detection system for accurately depicting groundwater pollution, and relates to the field of groundwater detection technology. The method comprises the following steps: S1, by opening an observation well in the area to be detected, the detection equipment is submerged into the target water area; S2, the water body information, survey video and location information of the target water area are collected in real time by the detection equipment; S3, based on the water body information and location information, the regional pollution index at different locations in the target water area is analyzed; S4, based on the survey video and location information, the water quality status at different locations in the target water area is identified; S5, the pollution distribution of the target water area is depicted by combining the geographic information system and the pollution distribution model. The present invention provides reliable technical support for the monitoring and control of groundwater pollution, so as to accurately evaluate the pollution degree of the groundwater detection point, especially for non-aqueous phase liquid pollutants, so as to carry out targeted extraction and treatment.

Description

Technical Field

[0001] The present invention relates to the field of groundwater detection technology, and in particular to a method and a detection system for accurately characterizing groundwater pollution. Background Art

[0002] Common groundwater pollution is an environmental problem caused by the accidental leakage of non-aqueous liquid substances such as petroleum. Non-aqueous liquid pollutants are a complex mixture composed of many hydrocarbons and chlorinated hydrocarbon compounds of different molecular sizes. In the process of oil and gas field development and various industrial activities, the accidental leakage of crude oil, hydrocarbons and chlorinated hydrocarbons will cause groundwater pollution. Moreover, under certain conditions, these pollutants will also transfer to the surrounding environment in different ways, further polluting the atmosphere and soil, and ultimately posing a serious threat to human health. Therefore, it is very important to monitor petroleum hydrocarbons in groundwater. High-concentration organic pollutants form a free phase in groundwater. This free phase continues to migrate and diffuse, becoming a long-term source of pollution and difficult to completely remove. In order to effectively deal with this problem, it is of great significance to develop a precise characterization method for groundwater pollution.

[0003] First, accurate characterization of groundwater pollution can provide more accurate means of pollution monitoring and assessment. Traditional groundwater pollution monitoring methods rely on fixed sampling points and laboratory analysis, which is often insufficient when faced with complex and changeable groundwater flow environments.

[0004] Secondly, precise characterization can improve the efficiency and effectiveness of pollution control. In traditional control methods, due to insufficient understanding of pollution sources and pollution diffusion paths, control measures are often difficult to target and the results are poor. However, through precise characterization, pollution sources and pollution ranges can be accurately located, control plans can be scientifically formulated, unnecessary control costs and resource waste can be reduced, and control effects can be improved. For example, in the case of high-concentration organic pollutants forming a free phase, precise characterization can help identify the distribution and migration path of the free phase, so as to carry out targeted extraction and treatment, avoid the continued diffusion and migration of the free phase, and reduce the threat of long-term pollution sources.

[0005] Existing groundwater pollution monitoring technologies mainly include traditional sampling and analysis methods and modern online monitoring technologies. Traditional methods rely on the setting of sampling points and laboratory analysis, and have problems such as poor timeliness, limited coverage, and high costs. Although modern online monitoring technology has made up for these shortcomings to a certain extent, there are still problems such as insufficient monitoring accuracy and equipment being easily affected by the environment. In addition, when faced with complex groundwater flows and the migration and diffusion of pollutants, existing technologies often find it difficult to provide comprehensive and accurate pollution information, making it difficult for control measures to achieve the expected results.

[0006] Currently, no effective solution has been proposed for the problems in the related technologies. Summary of the invention

[0007] 1. Technical issues to be resolved

[0008] In view of the shortcomings of the existing technology, the present invention provides a method and a detection system for accurately characterizing groundwater pollution. It has the advantages of comprehensively utilizing advanced sensing technology, data analysis technology and environmental science theory to achieve real-time, dynamic and accurate monitoring and analysis of groundwater pollution, thereby solving the problem that traditional technology is difficult to provide comprehensive and accurate pollution information when faced with complex groundwater flow and migration and diffusion of pollutants, making it difficult for control measures to achieve the expected results.

[0009] (II) Technical solution

[0010] In order to realize the advantages of comprehensive utilization of advanced sensing technology, data analysis technology and environmental science theory to achieve real-time, dynamic and accurate monitoring and analysis of groundwater pollution, the specific technical solutions adopted by the present invention are as follows:

[0011] In a first aspect, a method for accurately characterizing groundwater pollution is provided, the method comprising the following steps:

[0012] S1. By opening an observation well in the area to be inspected, the inspection equipment is submerged into the target waters, and the initial inspection route of the inspection equipment is set based on the geological and hydrological information of the area to be inspected;

[0013] S2, using detection equipment to collect water body information, survey video and location information of the target water area in real time, and transmit it to a remote ground receiver and a computing control center via a wireless network;

[0014] S3. Based on water body information and location information, analyze the regional pollution index at different locations in the target waters, predict the pollution source and pollution range of the target waters, and establish a pollution distribution model;

[0015] S4. Based on the survey video and location information, identify the water quality status at different locations in the target waters; then combine the identification results to optimize the equipment inspection route and pollution distribution model;

[0016] S5. Combine geographic information system with pollution distribution model to characterize the pollution distribution of target waters.

[0017] Furthermore, based on the water body information and location information, the regional pollution index of different locations in the target water area is analyzed, and the pollution source and pollution range of the target water area are predicted. The establishment of the pollution distribution model includes the following steps:

[0018] S31, obtaining water body information collected at all detection points in the target water area, as well as location information of each detection point, and integrating the water body information and location information of each detection point into a sample data; wherein the water body information includes the pollutant concentration and water body temperature of non-aqueous phase liquid pollutants;

[0019] S32. Integrate all sample data to establish a sample data set, evaluate the pollution load of each detection point, and calculate the regional pollution index of each detection point in combination with regional indicators;

[0020] S33. Based on the migration characteristics of pollutants, predict the pollution sources and pollution range of the target waters, and build a pollution distribution model to simulate the potential regional pollution index of the area through which the target waters flow.

[0021] Furthermore, all sample data are integrated to establish a sample data set, the pollution load of each detection point is evaluated, and the regional pollution index of each detection point is calculated by combining regional indicators, including the following steps:

[0022] S321, obtaining the pollutant concentration of the non-aqueous phase liquid pollutants collected at each detection point;

[0023] S322, based on the pollutant concentration at each detection point and in combination with the predicted water volume at the detection point, the product of the pollutant concentration and the predicted water volume is defined as the pollution load;

[0024] S323. Define the groundwater recharge index and estimate the value of the groundwater recharge index by combining the rainfall, infiltration rate and radial coefficient of groundwater flow in the detection area;

[0025] S324, defining a distance index, calculating the distance between each detection point and the observation well by using the Euclidean distance in the geographic information system, and then calculating the value of the distance index of each detection point;

[0026] S325. The exponential function method is used to express the functional relationship between pollution load, groundwater recharge index and distance index, and is defined as the regional pollution index of each detection point, where the expression of the regional pollution index is:

[0027] RPIi=Fi·[exp(GIi)+exp(DIi)]

[0028] Where, RPI i represents the pollution index of the i-th detection point area; F i represents the pollution load of the i-th detection point; GI i Represents the groundwater recharge index of the i-th detection point; DI i Represents the distance index of the i-th detection point.

[0029] Furthermore, based on the pollutant migration characteristics, the pollution source and pollution range of the target waters are predicted, and a pollution distribution model is built to simulate the potential regional pollution index of the area through which the target waters flow, including the following steps:

[0030] S331. Based on the migration characteristics of non-aqueous liquid pollutants, the diffusion process of pollutants in the target waters is defined as a normal distribution, and a pollution diffusion equation is established; the pollution diffusion equation formula is:

[0031]

[0032] Where RPI represents the regional pollution index of the point to be predicted; RPI0 represents the regional pollution index of the pollution source; x, y, z represent the position coordinates of the point to be detected; σ represents the parameter constraining the pollution diffusion transformation.

[0033] S332, select any detection point as a hypothetical pollution source in turn, draw a circular area with the hypothetical pollution source as the center and a preset radius, obtain all other detection points in the circular area, and calculate the theoretical value of the regional pollution index of other detection points in the circular area in combination with the pollution diffusion equation;

[0034] S333. Calculate the effective probability of the assumed pollution source through gradient calculation and correlation analysis, set iteration conditions, and select the optimal iteration result as the pollution source of the target waters;

[0035] S334. Select the distributed hydrological model as the carrier, input the regional pollution index of the pollution source and each detection point, calculate the pollution range through the pollution diffusion equation, establish a pollution distribution model, simulate and display the potential regional pollution index of each undetected point in the target water area.

[0036] Furthermore, by gradient calculation and correlation analysis, the effective probability of the assumed pollution source is calculated, and the iteration conditions are set. The optimal iteration result is selected as the pollution source of the target waters, which includes the following steps:

[0037] S3331. Calculate the gradient of the theoretical value and the measured value of the assumed pollution source for other detection points;

[0038] S3332. Analyze the correlation between the theoretical value gradient and the measured value gradient, and define the value of the correlation as the effective probability of the assumed pollution source being the real pollution source;

[0039] S3333, sequentially selecting different detection points as hypothetical pollution sources, calculating the effective probability of each hypothetical pollution source, and establishing a probability matrix to extract detection points that meet the probability threshold;

[0040] S3334. Define the initial parameters of the pollution source distribution, continuously update the parameters using the iterative formula, calculate the iterative value of each detection point after each iteration, until the convergence condition is reached, stop the iteration, and obtain the optimal iterative result as the final pollution source distribution, where the iterative formula is:

[0041]

[0042] Where ΔD represents the parameter change between the jth iteration and the j-1th iteration; D j represents the parameters of the jth iteration; D j-1 represents the parameters of the j-1th iteration; T j-1 represents the target value of the effective probability of the j-1th iteration; represents the square of the L2 norm; λ represents the regularization factor.

[0043] Furthermore, based on the survey video and location information, the water quality status at different locations in the target waters is identified; and combined with the identification results, the equipment inspection route and the pollution distribution model are optimized, including the following steps:

[0044] S41, obtaining the survey video shot by the detection equipment in real time, setting the sampling interval, regularly extracting key frame images, and matching and aligning each key frame image with the corresponding timestamp information and position information to establish a real-time target image set;

[0045] S42, based on the color correction algorithm, compensating the red channel of the key frame image to correct the image color deviation;

[0046] S43, inputting the color-corrected key frame image into the image enhancement model, and enhancing the image brightness and texture information by introducing a combined loss function to obtain an enhanced underwater image;

[0047] S44, by processing and identifying the enhanced underwater image, the turbidity and water color of the target water area at different detection points are obtained, which are defined as the water quality state of the target water area at different positions;

[0048] S45. Select verification detection points in the target waters, drive the detection equipment to re-detect the verification detection points, and identify underground obstacle information based on enhanced underwater images, optimize the inspection route of the detection equipment, and finally optimize the pollution distribution model based on the detection results.

[0049] Furthermore, based on the color correction algorithm, the red channel of the key frame image is compensated, and the image color deviation is corrected, including the following steps:

[0050] S421, respectively calculating the compensation amount and compensation control item of the red channel;

[0051] S422, calculating the stretching result of the key frame image, applying it to each channel of the key frame image, performing histogram stretching, and covering all pixels in the entire dynamic range of the key frame image;

[0052] S423, based on the compensation amount and the compensation control item, define a red channel compensation function of the key frame image, compensate the red channel of each pixel position in the key frame image, make up for the loss of the red channel, and achieve color correction of the key frame image; the formula of the red channel compensation function is:

[0053]

[0054] In the formula, Represents the red channel value of the compensated key frame image; I R (x, y) represents the red channel value; Temp1 represents the compensation amount; Temp2 represents the compensation control item.

[0055] Furthermore, the color-corrected key frame image is input into the image enhancement model, and the image brightness and texture information are enhanced by introducing a combined loss function. The enhanced underwater image is obtained by the following steps:

[0056] S431, construct generator G and generator F for image enhancement and inverse transformation; then construct discriminator D A and the discriminator D B , used to distinguish the enhanced image from the real image;

[0057] S432, using multi-cycle adversarial loss, cycle consistency loss and image strong edge structure similarity loss, defining a combined loss function, and constructing an image enhancement model;

[0058] S433, training the image enhancement model using the training data, and optimizing the parameters of the generator and the discriminator until the image enhancement model converges;

[0059] S434, input the color-corrected key frame image into the image enhancement model, convert the input image into an enhanced image through the generator G, and perform post-processing to obtain an enhanced underwater image.

[0060] Furthermore, verification detection points are selected in the target waters, the detection equipment is driven to re-detect the verification detection points, and underground obstacle information is identified according to the enhanced underwater image, and the inspection route of the detection equipment is optimized. Finally, based on the detection results, the pollution distribution model is optimized, including the following steps:

[0061] S451. According to the pollution distribution model, a plurality of verification detection points are randomly selected in the target waters, and the potential regional pollution index simulated by the verification detection points is obtained;

[0062] S452, identifying and enhancing the underwater image in real time, identifying underground obstacles, combining the underground obstacle information with the current location information of the detection equipment, and generating an underground obstacle distribution map;

[0063] S453, according to the distribution location of the verified detection points, use the path planning algorithm to set the secondary inspection route, and dynamically adjust the movement path according to the real-time underground obstacle identification;

[0064] S454. Obtain the water body information of the verification detection point, calculate the regional pollution index of the verification detection point, and compare it with the potential regional pollution index. If the difference is less than the set threshold, use the regional pollution index to update the regional pollution index of the pollution distribution model at the verification detection point. If the difference is greater than the set threshold, adjust the parameters of the pollution distribution model to optimize the overall model evaluation effect.

[0065] In a second aspect, a groundwater pollution detection system is also provided, the system comprising: a detection device, a ground receiver and a computing control center;

[0066] Among them, the detection equipment is used to dive into the target water area through the observation well, and detect the water body information, location information and survey video inside the target water area in real time through concentration detection, location detection and video shooting functions;

[0067] A ground receiver is used to establish wireless communication with the detection equipment and transmit the information data collected by the detection equipment to the computing and control center;

[0068] The computing control center is used to analyze the regional pollution index at different locations in the target waters, predict the pollution sources and pollution range of the target waters, and establish a pollution distribution model; identify the water quality status at different locations in the target waters; and then combine the identification results to optimize the equipment inspection route and the pollution distribution model; finally, combine the geographic information system with the pollution distribution model to characterize the pollution distribution of the target waters.

[0069] (III) Beneficial effects

[0070] Compared with the prior art, the present invention provides a method and a detection system for accurately characterizing groundwater pollution, which has the following beneficial effects:

[0071] (1) By combining concentration detection, video analysis and geological and hydrological information, it is possible to obtain groundwater pollution information in real time and accurately, establish a detailed pollution distribution model, and optimize the inspection route of the detection equipment; through data fusion and comprehensive analysis, it is possible to accurately characterize the pollution distribution of the target waters, providing reliable technical support for the monitoring and control of groundwater pollution, so as to accurately assess the pollution level of groundwater detection points, especially for non-aqueous liquid pollutants (NAPL), so as to carry out targeted extraction and treatment, avoid the continued diffusion and migration of the free phase, and reduce the threat of long-term pollution sources.

[0072] (2) By acquiring and integrating the water body information and location information of all detection points in the target waters, establishing a sample data set, and evaluating the pollution load of each detection point, we can fully grasp the pollution status of the target waters, and accurately calculate the pollution index of each detection point by combining the pollutant concentration data; through the migration characteristics of pollutants, we can accurately predict the pollution sources and their diffusion range, and establish a detailed pollution distribution model, so as to simulate the diffusion and migration of pollutants in the waters, provide accurate data support, help to scientifically formulate pollution control plans, optimize environmental monitoring and control effects, and effectively protect water resources and the ecological environment.

[0073] (3) By processing the survey video captured by the detection equipment in real time, regularly extracting and correcting key frame images, and using the image enhancement model to improve the brightness and texture information of the image, it is possible to accurately identify and evaluate the water quality status; the color cast and blur of the underwater image are eliminated, providing clear water quality status information; by analyzing the enhanced underwater image, the turbidity and water color of different detection points in the target water area can be accurately identified, and then the water quality status can be defined, making water quality monitoring more intuitive and reliable; and based on the verification of the detection results of the detection points, the pollution distribution model is further optimized, making the pollution assessment more accurate and comprehensive, thereby improving the accuracy and efficiency of water quality monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0075] Figure 1 is a flow chart of a method for accurately characterizing groundwater pollution according to an embodiment of the present invention;

[0076] Figure 2 is a system block diagram of a groundwater pollution detection system according to an embodiment of the present invention;

[0077] Figure 3is a functional structure diagram of a detection device according to an embodiment of the present invention.

[0078] In the figure:

[0079] 1. Detection equipment; 2. Ground receiver; 3. Computing control center. DETAILED DESCRIPTION

[0080] According to an embodiment of the present invention, a method and a detection system for accurately characterizing groundwater pollution are provided.

[0081] like Figure 1 As shown, a method for accurately characterizing groundwater pollution is provided, the method comprising the following steps:

[0082] S1. By opening an observation well in the area to be inspected, the detection equipment is submerged into the target waters, and the initial inspection route of the detection equipment is set based on the geological and hydrological information of the area to be inspected.

[0083] Specifically, it is necessary to give priority to opening observation wells, provide entrances for detection equipment to dive, and provide sampling points for groundwater monitoring. Choose appropriate well locations based on geological and hydrological information, usually in areas with relatively stable groundwater flow, and avoid opening wells near pollution sources or in areas with violent flow. Design the location and depth of the wells based on the direction and velocity of groundwater flow in the area and possible sources of pollution.

[0084] Use a dedicated submersible device (such as a cable, pulley system) to slowly and steadily submerge the detection equipment into the target waters. Ensure that the equipment is not damaged during the dive and remains stable. Debug the equipment in the target waters to ensure that the equipment is working properly, and calibrate the sensor to ensure the accuracy of the data.

[0085] Before the detection process, it is necessary to collect geological and hydrological data of the area to be tested, including groundwater flow direction, flow rate, geological structure, potential pollution sources, etc. Analyze the geological and hydrological data to identify possible pollution sources, migration paths of pollutants, groundwater flow patterns, etc. Based on the geological and hydrological information, plan the initial inspection route of the detection equipment to ensure that the route covers all key areas, especially near the pollution source and the main direction of groundwater flow. Use path optimization algorithms (such as A* algorithm, Dijkstra algorithm) to optimize the inspection route, reduce repeated detection, and ensure effective coverage. Finally, after the path design is completed, set the detection points at the key nodes, and establish the corresponding spatial coordinates in the geographic information system to obtain the coordinate data of each detection point as the location information of the detection equipment during movement.

[0086] S2. Use detection equipment to collect water body information, survey video and location information of the target water area in real time, and transmit it to a remote ground receiver and a computing control center via a wireless network.

[0087] Specifically, the collected water body information, video stream and location information are packaged into data packets to ensure that the data packet structure is reasonable and easy to transmit and analyze. In addition, the data packet is encrypted to ensure that the data is not tampered with or leaked during transmission. According to the geographical environment and communication conditions of the target waters, a suitable wireless network (such as 4G / 5G network, satellite communication, underwater acoustic communication, etc.) is selected for data transmission. The packaged data is transmitted to the ground receiver in real time through the wireless network.

[0088] The ground receiver is equipped with a corresponding communication module that can match the communication module of the detection equipment to ensure that the data can be received smoothly. After receiving the data packet, the receiver decrypts and analyzes the data packet to extract water body information, video stream and location information. The decrypted and analyzed data is then transmitted to the computing control center via a wired or wireless network. A database system is established in the computing control center to store all received data to ensure data security and integrity.

[0089] S3. Based on water body information and location information, analyze the regional pollution index at different locations in the target water area, predict the pollution source and pollution range of the target water area, and establish a pollution distribution model.

[0090] In the description of the present invention, based on water body information and location information, the regional pollution index of different locations in the target water area is analyzed, and the pollution source and pollution range of the target water area are predicted. Establishing a pollution distribution model includes the following steps:

[0091] S31, obtaining water body information collected at all detection points in the target water area and location information of each detection point, and integrating the water body information and location information of each detection point into a sample data.

[0092] Among them, water body information includes the pollutant concentration of non-aqueous phase liquid (NAPL) and water body temperature. The pollutant concentration is detected by a PID (photoionization detector), which is a tool used to detect volatile organic compounds (VOCs) and other gaseous pollutants. It irradiates the sample with ultraviolet light to ionize the molecules in it, thereby detecting the presence and concentration of pollutants.

[0093] S32. Integrate all sample data to establish a sample data set, evaluate the pollution load of each detection point, and calculate the regional pollution index of each detection point in combination with regional indicators.

[0094] In the description of the present invention, integrating all sample data to establish a sample data set, evaluating the pollution load of each detection point, and combining regional indicators to calculate the regional pollution index of each detection point includes the following steps:

[0095] S321. Obtain the pollutant concentration of the non-aqueous liquid pollutants collected at each detection point.

[0096] Specifically, the concentration data of non-aqueous phase liquid pollutants (NAPL) at each detection point are collected, and the data are cleaned to remove outliers and noise.

[0097] S322. Based on the pollutant concentration at each detection point and in combination with the predicted water volume at the detection point, the product of the pollutant concentration and the predicted water volume is defined as the pollution load.

[0098] S323. Define the groundwater recharge index and estimate the value of the groundwater recharge index based on the rainfall, infiltration rate and radial coefficient of groundwater flow in the test area.

[0099] Specifically, the groundwater recharge index is a comprehensive indicator that represents the groundwater recharge capacity, taking into account rainfall, infiltration rate, and radial coefficient of groundwater flow.

[0100] The rainfall data, infiltration rate data and groundwater flow radial coefficient of the detection area are collected, and the calculation formula is: groundwater recharge index = rainfall × infiltration rate × groundwater flow radial coefficient.

[0101] S324. Define a distance index, calculate the distance between each detection point and the observation well by using the Euclidean distance in the Geographic Information System (GIS), and then calculate the value of the distance index of each detection point.

[0102] S325. The exponential function method is used to express the functional relationship between pollution load, groundwater recharge index and distance index, and is defined as the regional pollution index of each detection point, where the expression of the regional pollution index is:

[0103] RPI i =F i ·[exp(GI i )+exp(DI i )]

[0104] Where, RPI i represents the pollution index of the i-th detection point area; F i represents the pollution load of the i-th detection point; GI i Represents the groundwater recharge index of the i-th detection point; DI i Represents the distance index of the i-th detection point.

[0105] S33. Based on the migration characteristics of pollutants, predict the pollution sources and pollution range of the target waters, and build a pollution distribution model to simulate the potential regional pollution index of the area through which the target waters flow.

[0106] In the description of the present invention, based on the pollutant migration characteristics, the pollution source and pollution range of the target water area are predicted, and a pollution distribution model is built to simulate the potential regional pollution index of the area through which the target water area flows, including the following steps:

[0107] S331. Based on the migration characteristics of non-aqueous liquid pollutants (NAPL), the diffusion process of pollutants in the target waters is defined as a normal distribution, and a pollution diffusion equation is established; the pollution diffusion equation formula is:

[0108]

[0109] Where RPI represents the regional pollution index of the point to be predicted; RPI0 represents the regional pollution index of the pollution source; x, y, z represent the position coordinates of the point to be detected; σ represents the parameter constraining the pollution diffusion transformation.

[0110] S332. Select any detection point as a hypothetical pollution source in turn, draw a circular area with the hypothetical pollution source as the center and a preset radius, obtain all other detection points in the circular area, and calculate the theoretical value of the regional pollution index of other detection points in the circular area in combination with the pollution diffusion equation.

[0111] S333. Calculate the effective probability of the assumed pollution source through gradient calculation and correlation analysis, set the iteration conditions, and select the optimal iteration result as the pollution source of the target waters.

[0112] In the description of the present invention, the effective probability of the assumed pollution source is calculated through gradient calculation and correlation analysis, and the iteration conditions are set. The optimal iteration result is selected as the pollution source of the target waters, which includes the following steps:

[0113] S3331. Calculate the gradient of the theoretical value and the measured value of the assumed pollution source for other detection points; wherein the gradient calculation formula of the theoretical value and the measured value is:

[0114]

[0115] In the formula, represents the theoretical value gradient; Represents the measured value gradient; RPI0 represents the regional pollution index of the pollution source; RPI l Represents the theoretical regional pollution index of other detection points; RPI s Indicates the measured regional pollution index of other detection points;

[0116] S3332. Analyze the correlation between the theoretical value gradient and the measured value gradient, and define the value of the correlation as the effective probability that the assumed pollution source is the real pollution source; wherein the calculation formula for the effective probability is:

[0117]

[0118] In the formula, Indicates the effective probability of a pollution source being a real pollution source (correlation between the theoretical value gradient and the measured value gradient); Represents the covariance of the theoretical value gradient and the measured value gradient; represents the variance of the theoretical value gradient; represents the variance of the measured value gradient;

[0119] S3333. Select different detection points as hypothetical pollution sources in turn, calculate the effective probability of each hypothetical pollution source, establish a probability matrix, and extract detection points that meet the probability threshold.

[0120] S3334. Define the initial parameters of the pollution source distribution, continuously update the parameters using the iterative formula, calculate the iterative value of each detection point after each iteration, until the convergence condition is reached, stop the iteration, and obtain the optimal iterative result as the final pollution source distribution, where the iterative formula is:

[0121]

[0122] Where ΔD represents the parameter change between the jth iteration and the j-1th iteration; D j represents the parameters of the jth iteration; D j-1 represents the parameters of the j-1th iteration; T j-1 represents the target value of the effective probability of the j-1th iteration; represents the square of the L2 norm; λ represents the regularization factor.

[0123] Specifically, the initial parameter represents the initial distribution of pollution sources, which has no direct relationship with the detection point and regional pollution index, but affects the starting point of the inversion. The size of λ affects the stability and convergence speed of the inversion process, which has no direct relationship with the detection point and regional pollution index, but should be adjusted according to the characteristics of the data.

[0124] Example: Assume there are three detection points, the observed pollution indexes are b = [10, 20, 15], the initial guess of the pollution source distribution is D0 = [0, 0, 0], and the proportional factor λ is set to 0.1. In each iteration:

[0125] 1. Calculate the simulated pollution index (target value T j-1 ).

[0126] 2. Calculate the residual between the current parameter and the target value.

[0127] 3. Calculate the update amount ΔD and update the parameters.

[0128] 4. Repeat the iteration until convergence.

[0129] S334. Select the distributed hydrological model as the carrier, input the regional pollution index of the pollution source and each detection point, calculate the pollution range through the pollution diffusion equation, establish a pollution distribution model, simulate and display the potential regional pollution index of each undetected point in the target water area.

[0130] Specifically, select a suitable distributed hydrological model (such as the SWAT model), which needs to be able to handle groundwater flow, pollutant diffusion and hydrological parameters. Input the location information and pollutant concentration of known pollution sources into the distributed hydrological model. Collect the regional pollution index of each detection point and use it as an input parameter, as well as geohydrological data, including topography, soil type, rainfall, infiltration rate, groundwater flow rate and other parameters.

[0131] Combined with the pollution diffusion equation of pollutants in groundwater, the boundary conditions (such as the exchange of surface water and groundwater) and initial conditions (such as the initial pollutant concentration distribution) of the model are set. The location and related data of the known pollution source are input into the model, and then the regional pollution index of each detection point is input into the model as calibration data. By running the hydrological model, the pollution diffusion equation is used to calculate the diffusion range and concentration distribution of pollutants in groundwater, simulating the diffusion of pollutants from the pollution source, passing through each detection point, and spreading to the entire target waters.

[0132] A pollution distribution model is established based on the data output by the distributed hydrological model. The model shows the potential regional pollution index of each undetected point in the target water area. The pollution distribution model can be displayed through a map visualization tool. The model is then calibrated and verified using known detection point data to ensure the accuracy and reliability of the model. If the pollution index predicted by the model deviates significantly from the actual detection data, the model parameters need to be adjusted for recalibration.

[0133] Finally, the pollution distribution results simulated by the model are exported and displayed through GIS system or other visualization tools. The display content includes: location of pollution source, diffusion path of pollutants, regional pollution index of each detection point and potential pollution index of undetected points, etc.

[0134] S4. Based on the survey video and location information, identify the water quality status at different locations in the target waters. Combined with the identification results, optimize the equipment inspection route and pollution distribution model.

[0135] In the description of the present invention, based on the survey video and location information, the water quality status of different locations in the target water area is identified. Combined with the identification results, the optimization of the equipment inspection route and the pollution distribution model includes the following steps:

[0136] S41. Obtain the survey video shot in real time by the detection equipment, set the sampling interval, extract key frame images regularly, and align each key frame image with the corresponding timestamp information and location information to establish a real-time target image set.

[0137] S42. Based on a color correction algorithm, the red channel of the key frame image is compensated to correct the image color cast.

[0138] In groundwater with weak light, only supplementary lighting can be relied on. However, since the photon density is relatively concentrated in the illuminated area, the uneven illumination caused by the gradual decrease in the surroundings and the change in light transmittance caused by the change in imaging distance will cause color distortion in the non-illuminated area. In order to correct the image color cast, it is necessary to design a red channel compensation function to eliminate the influence of the overexposure area.

[0139] In the description of the present invention, based on the color correction algorithm, compensating the red channel of the key frame image and correcting the image color deviation include the following steps:

[0140] S421, respectively calculating the compensation amount and compensation control item of the red channel, the calculation formula of the compensation amount and the compensation control item is:

[0141]

[0142] Where Temp1 represents the compensation amount; Temp2 represents the compensation control item; α represents the lighting adaptation parameter; I R (x, y) represents the red channel value; I G (x, y) represents the green channel value; I B (x, y) represents the blue channel value; Represents the average value of the green channel within the filter window centered at pixel (x, y); Represents the average value of the blue channel within the filter window centered at pixel (x, y).

[0143] S422, calculating the stretching result of the key frame image, applying it to each channel of the key frame image, performing histogram stretching, and covering all pixels in the entire dynamic range of the key frame image; the calculation formula of the stretching result is:

[0144]

[0145] In the formula, I cc Indicates the stretching result; i in Represents the red channel compensated image; o min Indicates the minimum and maximum values ​​of the output image; i min and i max Indicates the minimum and maximum values ​​of the key frame image;

[0146] S423, based on the compensation amount and the compensation control item, define a red channel compensation function of the key frame image, compensate the red channel of each pixel position in the key frame image, make up for the loss of the red channel, and achieve color correction of the key frame image; the formula of the red channel compensation function is:

[0147]

[0148] In the formula, Represents the red channel value of the compensated key frame image; I R (x, y) represents the red channel value; Temp1 represents the compensation amount; Temp2 represents the compensation control item.

[0149] S43, inputting the color-corrected key frame image into the image enhancement model, and enhancing the image brightness and texture information by introducing a combined loss function to obtain an enhanced underwater image.

[0150] Specifically, the image enhancement model adopts the MCA-CycleGAN model. The MCA-CycleGAN model (Multi-scaleCycleGAN) is an improved CycleGAN (Cycle-Consistent Adversarial Networks) model, which enhances the image conversion effect by introducing a multi-scale cycle adversarial loss (Multi-scale Cycle Adversarial Loss), which is particularly suitable for image enhancement tasks, such as underwater image enhancement.

[0151] The MCA-CycleGAN model is improved on the basis of the original CycleGAN, specifically including the following parts:

[0152] 1. Generator: Responsible for converting the input image from one domain (such as low-quality images) to another domain (such as high-quality images). Use residual blocks and multi-scale feature extraction modules to enhance image details and texture information.

[0153] 2. Discriminators: Responsible for distinguishing generated images from real images. Using the PatchGAN structure, multiple scale discriminators are used to capture image details at different scales.

[0154] 3. Multi-scale Cycle Confrontation Loss (MCA Loss): Through adversarial training of generators and discriminators at multiple scales, the loss of images at different scales is calculated. Combined with Cycle Consistency Loss at multiple scales, the consistency of image conversion and detail retention are improved.

[0155] By introducing multi-scale cyclic adversarial loss, the MCA-CycleGAN model can better preserve the details and texture information of the image and enhance the brightness and contrast of the image. Especially in the underwater image enhancement task, the MCA-CycleGAN model can effectively remove blur and noise in the underwater environment and improve the clarity and visibility of the image.

[0156] In the description of the present invention, the color-corrected key frame image is input into the image enhancement model, and the image brightness and texture information are enhanced by introducing a combined loss function to obtain an enhanced underwater image, which includes the following steps:

[0157] S431, construct generator G and generator F for image enhancement and inverse transformation. Then construct discriminator D A and the discriminator D B , which is used to distinguish the enhanced image from the real image.

[0158] Specifically, the generator is based on U-Net (U-shaped structure deep learning model), and adds multi-scale convolution and residual hole convolution, and combines the brightness equalization module of full convolution. Most underwater image enhancement algorithms enhance the brightness of the entire image, so there is a high possibility of overexposure. The brightness equalization module is used to solve this problem. This module is composed of a full convolution network with 32 channels. It is used as an auxiliary network to locate the low-light area of ​​the underwater image, so as to achieve uniform brightness distribution of the enhanced image.

[0159] The dual discriminator consists of the discriminator D A and the discriminator D B It consists of two identical sub-discriminators. Except for the last layer of the sub-discriminator, the convolution step size is set to 1, and the rest of the layers are all 2. The dual discriminator adopts a full convolutional network structure, which calculates the authenticity of the local area and then averages it as a standard for measuring the authenticity of the entire image.

[0160] S432. Multi-cycle adversarial loss, cycle consistency loss and image strong edge structure similarity loss are used to define a combined loss function and construct an image enhancement model.

[0161] Specifically, the loss function is the optimization criterion of the convolutional neural network. The total loss function of the network model consists of three losses: multi-scale cyclic adversarial, cyclic consistency, and image strong edge structure similarity. The expression of the combined loss function is:

[0162] Loss = L GAN (G,D Y ,x,y)+L GAN (F,D Y ,x,y)+αL cyc(G,F)+βL SESS

[0163] Where, L GAN (G,D Y ,x,y) represents the distribution gap between the key frame image x1 and the real dehydrated image y in the multi-scale cyclic adversarial loss; L GAN (F,D Y ,x,y) represents the loss between the key frame image y1 and the real dehydrated image x in the multi-scale cyclic adversarial loss; L cyc (G, F) represents the cycle consistency loss, that is, the difference between the image generated by generator G and the image restored by generator F; L SESS It represents the similarity loss of the strong edge structure of the image, and α and β represent the weight values ​​between the loss functions.

[0164] S433. Train the image enhancement model using the training data, and optimize the parameters of the generator and the discriminator until the image enhancement model converges.

[0165] S434, input the color-corrected key frame image into the image enhancement model, convert the input image into an enhanced image through the generator G, and perform post-processing to obtain an enhanced underwater image.

[0166] S44. By processing and identifying the enhanced underwater image, the turbidity and water color of the target water area at different detection points are obtained, which are defined as the water quality status of the target water area at different locations.

[0167] S45. Select verification detection points in the target waters, drive the detection equipment to re-detect the verification detection points, and identify underground obstacle information based on enhanced underwater images, optimize the inspection route of the detection equipment, and finally optimize the pollution distribution model based on the detection results.

[0168] In the description of the present invention, a verification detection point is selected in the target water area, the detection equipment is driven to re-detect the verification detection point, and the underground obstacle information is recognized according to the enhanced underwater image, and the inspection route of the detection equipment is optimized. Finally, based on the detection results, the pollution distribution model is optimized, which includes the following steps:

[0169] S451. According to the pollution distribution model, a plurality of verification and detection points are randomly selected in the target waters, and the potential regional pollution index simulated by the verification and detection points is obtained.

[0170] S452, real-time recognition and enhancement of underwater images, identification of underground obstacles, combining underground obstacle information with current location information of the detection equipment, and generating an underground obstacle distribution map.

[0171] Specifically, the image enhancement model is used to process the acquired underwater images to improve the brightness and texture information of the images, making underground obstacles more visible. Underground obstacles in underwater images are identified through image processing and computer vision algorithms (such as target detection algorithms), including YOLO, Faster R-CNN, etc. The identified underground obstacle information is combined with the current position of the detection equipment to generate an underground obstacle distribution map, showing the specific location of each obstacle and its impact on the movement path of the detection equipment.

[0172] S453. According to the distribution location of the verified detection points, a secondary inspection route is set using a path planning algorithm, and the moving path is dynamically adjusted based on real-time underground obstacle identification.

[0173] Specifically, based on the distribution of the verified detection points and the current underground obstacle distribution map, a path planning algorithm (such as A* algorithm, Dijkstra algorithm, etc.) is used to set the optimized secondary inspection route. Path planning should take into account the location of obstacles, the priority of detection points, and the mobility of the equipment. According to the real-time obstacle identification results, the inspection route is dynamically adjusted. If new obstacles or changing environmental conditions are found during the inspection process, the path planning algorithm should update the movement path of the equipment in real time to avoid obstacles and optimize inspection efficiency.

[0174] S454. Obtain the water body information of the verification detection point, calculate the regional pollution index of the verification detection point, and compare it with the potential regional pollution index. If the difference is less than the set threshold, use the regional pollution index to update the regional pollution index of the pollution distribution model at the verification detection point. If the difference is greater than the set threshold, adjust the parameters of the pollution distribution model to optimize the overall model evaluation effect.

[0175] Specifically, a secondary test is conducted on the verification test point to obtain the actual water body information at that point, including new NAPL pollutant concentration data. The regional pollution index of the actual test result is then calculated and compared with the potential regional pollution index obtained from the previous simulation. If the difference between the actual test result and the potential pollution index is less than the set threshold, the new regional pollution index is used to update the pollution value of the pollution distribution model at the verification test point. If the difference is greater than the set threshold, it means that there is a large deviation in the model and the parameters of the pollution distribution model need to be adjusted. The model parameters are adjusted based on the test results and the difference to optimize the entire pollution distribution model and improve its prediction accuracy. The overall optimization includes recalibrating the model parameters, updating the model formula and improving the prediction algorithm.

[0176] S5. Combine geographic information system with pollution distribution model to characterize the pollution distribution of target waters.

[0177] Specifically, collect geographic information data of the target water area and its surrounding areas from geographic information system (GIS) databases, satellite images or remote sensing data, including topography, soil type, meteorological data, etc. Convert various data formats into formats that can be processed by the GIS system, such as vector data (points, lines, surfaces) and raster data (raster images). Ensure that pollution data and geographic information data are spatially aligned, such as through coordinate system conversion or data resampling.

[0178] Generate pollution distribution maps of the target water area and its surroundings in the geographic information system, showing pollution areas with different concentration levels. Generate pollution heat maps to highlight areas with higher pollutant concentrations. Users can use software map services to display pollution data, allowing users to query pollution information at specific locations. Generate detailed reports on pollution distribution, including charts, maps, and data analysis results, to facilitate decision-making and public communication.

[0179] See also Figure 2-Figure 3 , also provides a groundwater pollution detection system, the system includes: a detection device 1, a ground receiver 2 and a computing control center 3.

[0180] Among them, the detection equipment 1 is used to dive into the target water area through the observation well, and detect the water body information, position information and survey video inside the target water area in real time through concentration detection, position detection and video shooting functions.

[0181] The ground receiver 2 is used to establish wireless communication with the detection equipment and transmit the information data collected by the detection equipment to the computing and control center.

[0182] The computing control center 3 is used to analyze the regional pollution index at different locations in the target waters, predict the pollution source and pollution range of the target waters, and establish a pollution distribution model. The water quality status at different locations in the target waters is identified. Based on the identification results, the equipment inspection route and the pollution distribution model are optimized. Finally, the geographic information system and the pollution distribution model are combined to characterize the pollution distribution of the target waters.

[0183] In summary, with the help of the above technical scheme of the present invention, by combining concentration detection, video analysis and geological and hydrological information, it is possible to obtain the pollution information of groundwater bodies in real time and accurately, establish a detailed pollution distribution model, and optimize the inspection route of the detection equipment; through data fusion and comprehensive analysis, the accurate characterization of the pollution distribution of the target water area is achieved, and reliable technical support is provided for the monitoring and control of groundwater pollution, so as to accurately evaluate the pollution degree of the groundwater detection point, especially for the pollution of NAPL substances, so as to carry out targeted extraction and treatment, avoid the continued diffusion and migration of the free phase, and reduce the threat of long-term pollution sources. By acquiring and integrating the water body information and location information of all detection points in the target water area, establishing a sample data set, and evaluating the pollution load of each detection point, it is possible to fully grasp the pollution status of the target water area, and accurately calculate the pollution index of each detection point by combining the volatile pollutant concentration data; through the migration characteristics of pollutants, the pollution source and its diffusion range are accurately predicted, and a detailed pollution distribution model is established, so that the diffusion and migration of pollutants in the water area can be simulated, providing accurate data support, which is helpful to scientifically formulate pollution control plans, optimize environmental monitoring and control effects, and effectively protect water resources and ecological environment. By processing the survey video captured by the detection equipment in real time, regularly extracting and correcting key frame images, and using the image enhancement model to improve the brightness and texture information of the image, it is possible to accurately identify and evaluate the water quality status; it eliminates the color cast and blur of underwater images, and provides clear water quality status information; by analyzing the enhanced underwater images, the turbidity and water color of different detection points in the target waters can be accurately identified, and then the water quality status can be defined, making water quality monitoring more intuitive and reliable; and based on the verification of the detection results of the detection points, the pollution distribution model is further optimized, making the pollution assessment more accurate and comprehensive, thereby improving the accuracy and efficiency of water quality monitoring.

[0184] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for accurately characterizing groundwater pollution, characterized in that: The method comprises the following steps: S1. By opening an observation well in the area to be inspected, the inspection equipment is submerged into the target waters, and the initial inspection route of the inspection equipment is set based on the geological and hydrological information of the area to be inspected; S2, using detection equipment to collect water body information, survey video and location information of the target water area in real time, and transmit it to a remote ground receiver and a computing control center via a wireless network; S3. Based on water body information and location information, analyze the regional pollution index at different locations in the target waters, predict the pollution source and pollution range of the target waters, and establish a pollution distribution model; Building a pollution distribution model includes the following steps: S31, obtaining water body information collected at all detection points in the target water area, as well as the location information of each detection point, and integrating the water body information and location information of each detection point into a sample data; wherein the water body information includes the pollutant concentration and water body temperature of non-aqueous phase liquid pollutants; S32. Integrate all sample data to establish a sample data set, evaluate the pollution load of each detection point, and calculate the regional pollution index of each detection point in combination with regional indicators; S33. Based on the migration characteristics of pollutants, predict the pollution sources and pollution range of the target waters, and build a pollution distribution model to simulate the potential regional pollution index of the area through which the target waters flow; S4. Based on the survey video and location information, identify the water quality status at different locations in the target waters; then combine the identification results to optimize the equipment inspection route and pollution distribution model; Optimizing equipment inspection routes and pollution distribution models includes the following steps: S41, obtaining the survey video shot by the detection equipment in real time, setting the sampling interval, regularly extracting key frame images, and matching and aligning each key frame image with the corresponding timestamp information and position information to establish a real-time target image set; S42, based on the color correction algorithm, compensating the red channel of the key frame image to correct the image color deviation; S43, inputting the color-corrected key frame image into the image enhancement model, and enhancing the image brightness and texture information by introducing a combined loss function to obtain an enhanced underwater image; S44, by processing and identifying the enhanced underwater image, the turbidity and water color of the target water area at different detection points are obtained, which are defined as the water quality state of the target water area at different positions; S45, selecting verification detection points in the target waters, driving the detection equipment to re-detect the verification detection points, and identifying underground obstacle information based on the enhanced underwater image, optimizing the inspection route of the detection equipment, and finally optimizing the pollution distribution model based on the detection results; Optimizing the pollution distribution model includes the following steps: S451. According to the pollution distribution model, a plurality of verification detection points are randomly selected in the target waters, and potential regional pollution indexes simulated by the verification detection points are obtained; S452, identifying and enhancing the underwater image in real time, identifying underground obstacles, combining the underground obstacle information with the current location information of the detection equipment, and generating an underground obstacle distribution map; S453, according to the distribution location of the verified detection points, use the path planning algorithm to set the secondary inspection route, and dynamically adjust the movement path according to the real-time underground obstacle identification; S454, obtaining water body information of the verification detection point, calculating the regional pollution index of the verification detection point, and comparing it with the potential regional pollution index. If the difference is less than the set threshold, the regional pollution index is used to update the regional pollution index of the pollution distribution model at the verification detection point. If the difference is greater than the set threshold, the parameters of the pollution distribution model are adjusted to optimize the evaluation effect of the model as a whole; S5. Combine geographic information system with pollution distribution model to characterize the pollution distribution of target waters.

2. A method for accurately characterizing groundwater pollution according to claim 1, characterized in that: The steps of integrating all sample data to establish a sample data set, evaluating the pollution load of each detection point, and calculating the regional pollution index of each detection point in combination with regional indicators include the following steps: S321, obtaining the pollutant concentration of the non-aqueous phase liquid pollutants collected at each detection point; S322, based on the pollutant concentration at each detection point and in combination with the predicted water volume at the detection point, the product of the pollutant concentration and the predicted water volume is defined as the pollution load; S323. Define the groundwater recharge index and estimate the value of the groundwater recharge index by combining the rainfall, infiltration rate and radial coefficient of groundwater flow in the detection area; S324, defining a distance index, calculating the distance between each detection point and the observation well by using the Euclidean distance in the geographic information system, and then calculating the value of the distance index of each detection point; S325. The exponential function method is used to express the functional relationship between pollution load, groundwater recharge index and distance index, and is defined as the regional pollution index of each detection point, where the expression of the regional pollution index is: In the formula, RPI i Indicates i Pollution index of each detection point area; F i Indicates i Pollution load at each detection point; GI i Indicates i Groundwater recharge indicators at each monitoring point; DI i Indicates i The distance indicator of the detection points.

3. A method for accurately characterizing groundwater pollution according to claim 2, characterized in that: The method of predicting the pollution source and pollution range of the target water area based on the pollutant migration characteristics, building a pollution distribution model, and simulating the potential regional pollution index of the area through which the target water area flows includes the following steps: S331. Based on the migration characteristics of non-aqueous liquid pollutants, the diffusion process of pollutants in the target waters is defined as a normal distribution, and a pollution diffusion equation is established; the pollution diffusion equation formula is: In the formula, RPI Represents the regional pollution index of the point to be predicted; RPI 0 A regional pollution index indicating pollution sources; x , y , z Indicates the position coordinates of the point to be detected; represents the parameters of the constrained pollution diffusion transformation; S332, select any detection point as a hypothetical pollution source in turn, draw a circular area with the hypothetical pollution source as the center and a preset radius, obtain all other detection points in the circular area, and calculate the theoretical value of the regional pollution index of other detection points in the circular area in combination with the pollution diffusion equation; S333. Calculate the effective probability of the assumed pollution source through gradient calculation and correlation analysis, set iteration conditions, and select the optimal iteration result as the pollution source of the target waters; S334. Select the distributed hydrological model as the carrier, input the regional pollution index of the pollution source and each detection point, calculate the pollution range through the pollution diffusion equation, establish a pollution distribution model, simulate and display the potential regional pollution index of each undetected point in the target water area.

4. A method for accurately characterizing groundwater pollution according to claim 3, characterized in that: The method of calculating the effective probability of the assumed pollution source through gradient calculation and correlation analysis, setting iteration conditions, and selecting the optimal iteration result as the pollution source of the target waters includes the following steps: S3331. Calculate the gradient of the theoretical value and the measured value of the assumed pollution source for other detection points; S3332. Analyze the correlation between the theoretical value gradient and the measured value gradient, and define the value of the correlation as the effective probability of the assumed pollution source being the real pollution source; S3333, sequentially selecting different detection points as hypothetical pollution sources, calculating the effective probability of each hypothetical pollution source, and establishing a probability matrix to extract detection points that meet the probability threshold; S3334. Define the initial parameters of the pollution source distribution, continuously update the parameters using the iterative formula, calculate the iterative value of each detection point after each iteration, until the convergence condition is reached, stop the iteration, and obtain the optimal iterative result as the final pollution source distribution, where the iterative formula is: In the formula, Indicates j The iteration and j -1 parameter change between iterations; D j Indicates j The parameters of the iterations; D j-1 Indicates j -1 parameters for iteration; T j-1 Indicates j -1 target value of the effective probability for iteration; represents the square of L2 norm; λ represents the regularization factor.

5. A method for accurately characterizing groundwater pollution according to claim 1, characterized in that: The method of compensating the red channel of the key frame image and correcting the image color deviation based on the color correction algorithm includes the following steps: S421, respectively calculating the compensation amount and compensation control item of the red channel; S422, calculating the stretching result of the key frame image, applying it to each channel of the key frame image, performing histogram stretching, and covering all pixels in the entire dynamic range of the key frame image; S423, based on the compensation amount and the compensation control item, define a red channel compensation function of the key frame image, compensate the red channel of each pixel position in the key frame image, make up for the loss of the red channel, and achieve color correction of the key frame image; wherein the red channel compensation function formula is: In the formula, Represents the red channel value of the compensated key frame image; I R ( x , y ) represents the red channel value; Temp 1 indicates the compensation amount; Temp 2 indicates the compensation control item.

6. A method for accurately characterizing groundwater pollution according to claim 5, characterized in that: The step of inputting the color-corrected key frame image into the image enhancement model and enhancing the image brightness and texture information by introducing a combined loss function to obtain an enhanced underwater image comprises the following steps: S431, construct generator G and generator F for image enhancement and inverse transformation; then construct discriminator D A and the discriminator D B , used to distinguish the enhanced image from the real image; S432, using multi-cycle adversarial loss, cycle consistency loss and image strong edge structure similarity loss, defining a combined loss function, and constructing an image enhancement model; S433, training the image enhancement model using the training data, and optimizing the parameters of the generator and the discriminator until the image enhancement model converges; S434, input the color-corrected key frame image into the image enhancement model, convert the input image into an enhanced image through the generator G, and perform post-processing to obtain an enhanced underwater image.

7. A groundwater pollution detection system, used to implement the detection of a groundwater pollution accurate characterization method as described in any one of claims 1 to 6, characterized in that: The system includes: detection equipment, ground receiver and computing control center; The detection equipment is used to dive into the target water area through the observation well, and detect the water body information, position information and survey video inside the target water area in real time through concentration detection, position detection and video shooting functions; The ground receiver is used to establish wireless communication with the detection equipment and transmit the information data collected by the detection equipment to the computing control center; The computing control center is used to analyze the regional pollution index at different locations in the target waters, predict the pollution sources and pollution range of the target waters, and establish a pollution distribution model; identify the water quality status at different locations in the target waters; and then combine the identification results to optimize the equipment inspection route and the pollution distribution model; and finally combine the geographic information system with the pollution distribution model to characterize the pollution distribution of the target waters.

Images