A Method, System, Electronic Device and Storage Medium for Automatic Monitoring of Warm Drainage

Abstract

The present invention discloses a method, a system, an electronic device and a storage medium for automatic monitoring of warm wastewater discharge, which relates to the technical field of environmental monitoring. The method includes: obtaining an image of the water area to be monitored according to the Landsat-8 thermal infrared band image corresponding to the water area to be monitored; using the Isolation Forest algorithm to determine the anomaly scores of each water body pixel in the image of the water area to be monitored; obtaining anomaly water body samples and normal water body samples according to the anomaly scores of each water body pixel; judging the spatial adjacency relationship of the anomaly water body samples to obtain warm wastewater discharge samples; training a support vector machine using the warm wastewater discharge samples and the normal water body samples; inputting the image of the water area to be monitored into the trained support vector machine, and initially identifying the water body pixels affected by warm wastewater discharge in the image of the water area to be monitored by using the trained support vector machine, and further combining the judgment of the spatial adjacency relationship to obtain the final spatial distribution map of warm wastewater discharge. The present invention can achieve fast and real-time monitoring of warm wastewater discharge.

Description

Technical Field

[0001] The present invention relates to the technical field of environmental monitoring, and particularly to an automatic monitoring method, system, electronic device and storage medium for warm water discharge. Background Art

[0002] With the increase in energy demand and the development of nuclear energy technology, many nuclear power plants have been built in coastal areas of China to alleviate the problem of power shortage in coastal areas and promote the economic development of coastal areas. However, the power generation efficiency of nuclear power plants is lower than that of thermal power plants, and only 30% - 35% of nuclear energy is converted into electrical energy. Most of the remaining energy is discharged with cooling water in the form of heat, causing the temperature of the surrounding water body to rise rapidly, thus affecting the growth and reproduction of aquatic animals and plants. Algae are particularly sensitive to temperature changes, and the increase in seawater temperature may lead to abnormal outbreaks of algae. Driven by wind and tides, algae will invade the water cooling system of nuclear power plants, affecting the discharge of cooling water and hindering the safe operation of nuclear power plants. Therefore, it is of great significance to study the distribution of warm water discharge in the waters near nuclear power plants.

[0003] The methods for warm water discharge monitoring mainly include numerical simulation method, large-area field measurement method and remote sensing monitoring method. The numerical simulation method can obtain two-dimensional and three-dimensional temperature field distributions with large range and high spatio-temporal resolution, but the accuracy of its simulation results depends on the model used by the researcher. The field measurement method can obtain the most real data, but it can only provide scattered point data about temperature, cannot provide a complete spatial distribution, and consumes a lot of manpower and material resources. In some studies, it is used for the calibration and verification of remote sensing algorithms. Remote sensing technology has become an important method for warm water discharge monitoring due to its advantages such as good synchronization, large observation range, repeatable observation and relatively low cost. At present, warm water discharge monitoring based on remote sensing means is also widely used in the waters near many nuclear power plants in China, such as Daya Bay Nuclear Power Plant, Tianwan Nuclear Power Plant, etc.

[0004] In the early stage, AVHRR data was used for remote sensing monitoring of warm water discharge. However, due to its low spatial resolution, it was unable to accurately obtain the detailed distribution of warm water discharge. Later, with the release of satellite data series such as Landsat, MODIS, HJ-1, and GF-5, this problem was solved to a certain extent. Some scholars also detected warm water discharge through X-band radar data. However, the data that meets the requirements of high spatio-temporal resolution is usually ground-based radar, which is expensive and difficult to be applied on a large scale. Manned aerial remote sensing has better flexibility and spatial resolution than satellite remote sensing. However, it has special requirements for takeoff and landing sites, and the procedures for standby flight are cumbersome and costly, making it difficult to be promoted commercially. In addition, some achievements have been made in using unmanned aerial vehicles carrying thermal infrared sensors to monitor warm water discharge from nuclear power plants. However, its hardware equipment and image processing technology are still in the development process, and its flight range and flight conditions are also restricted to a certain extent. Generally speaking, the extraction of warm water discharge based on satellite remote sensing data is still the most widely used in practical applications.

[0005] At present, the warm water discharge monitoring carried out based on remote sensing means mainly uses remote sensing data containing thermal infrared bands to invert the water surface temperature, and then discriminates the warm water discharge based on statistical methods. The whole process is time-consuming and laborious, and it is difficult to achieve rapid and real-time warm water discharge monitoring. The commonly used algorithms in temperature inversion include the single-channel method, the split-window method, and the multi-channel method. The single-channel algorithm is derived from the surface temperature radiation transfer equation and requires three parameters: atmospheric transmittance, surface emissivity, and atmospheric mean acting temperature. Some of these parameters need to be measured. Although the split-window method and the multi-channel method require less measured data, their model complexity is much higher than that of the single-channel method, and there are certain requirements for the number of thermal infrared bands.

[0006] In summary, how to achieve rapid and real-time warm water discharge monitoring has become an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0007] The purpose of the present invention is to provide a warm water discharge automatic monitoring method, system, electronic device, and storage medium, which can achieve rapid and real-time warm water discharge monitoring.

[0008] To achieve the above purpose, the present invention provides the following solutions:

[0009] A warm water discharge automatic monitoring method, the method includes:

[0010] Obtain the Landsat-8 thermal infrared band image corresponding to the water area to be monitored;

[0011] Obtain the water area image to be monitored according to the Landsat-8 thermal infrared band image;

[0012] Use the isolation forest algorithm to determine the anomaly score of each water body pixel in the water area image to be monitored;

[0013] Two types of samples are obtained according to the anomaly scores of the respective water body pixels; the two types of samples include abnormal water body samples and normal water body samples;

[0014] Judge the spatial adjacency relationship of the abnormal water body samples to obtain thermal discharge samples;

[0015] Use the thermal discharge samples and the normal water body samples to train a support vector machine to obtain a trained support vector machine;

[0016] Substitute the image of the water area to be monitored into the trained support vector machine, and use the trained support vector machine to initially identify the water body pixels affected by thermal discharge in the image of the water area to be monitored, and obtain a preliminary spatial distribution map of thermal discharge;

[0017] Extract the patches of thermal discharge with spatial adjacency relationship to the nuclear power plant drainage outlet from the preliminary spatial distribution map of thermal discharge to obtain the final spatial distribution map of thermal discharge.

[0018] Optionally, obtaining the image of the water area to be monitored according to the Landsat-8 thermal infrared band image specifically includes:

[0019] Use a mask to crop the Landsat-8 thermal infrared band image to obtain an image of the water area to be monitored; the image of the water area to be monitored includes multiple water body pixels.

[0020] Optionally, obtaining two types of samples according to the anomaly scores of the respective water body pixels specifically includes:

[0021] Obtain the average value and standard deviation of all the anomaly scores according to the anomaly scores of the respective water body pixels;

[0022] Take all the water body pixels whose anomaly scores are less than or equal to the set threshold as abnormal water body samples; the set threshold is the average value minus 2 times the standard deviation;

[0023] Obtain the respective water body pixels with positive anomaly scores according to the anomaly scores of the respective water body pixels;

[0024] Arrange the respective water body pixels with positive anomaly scores in ascending order of the anomaly scores to obtain all the water body pixels with positive anomaly scores within the range from the minimum anomaly score to the maximum anomaly score;

[0025] Take all the water body pixels within the range from the median anomaly score to the maximum anomaly score among all the water body pixels with positive anomaly scores as normal water body samples.

[0026] Optionally, judge the spatial adjacency relationship of the abnormal water body sample to obtain a warm water discharge sample, specifically including:

[0027] Extract the patches of the abnormal water body sample with a spatial adjacency relationship to the nuclear power plant drainage outlet from the abnormal water body sample as the warm water discharge sample.

[0028] The present invention also provides the following solution:

[0029] A warm water discharge automatic monitoring system, the system includes:

[0030] A thermal infrared band image acquisition module, configured to acquire a Landsat-8 thermal infrared band image corresponding to the water area to be monitored;

[0031] A water area to be monitored image obtaining module, configured to obtain an image of the water area to be monitored according to the Landsat-8 thermal infrared band image;

[0032] An abnormal score determination module, configured to use the isolation forest algorithm to determine the abnormal score of each water body pixel in the image of the water area to be monitored;

[0033] A two-category sample obtaining module, configured to obtain two categories of samples according to the abnormal scores of each water body pixel; the two categories of samples include abnormal water body samples and normal water body samples;

[0034] A first spatial adjacency relationship judgment module, configured to judge the spatial adjacency relationship of the abnormal water body sample to obtain a warm water discharge sample;

[0035] A support vector machine training module, configured to train a support vector machine using the warm water discharge sample and the normal water body sample to obtain a trained support vector machine;

[0036] A warm water discharge preliminary identification module, configured to substitute the image of the water area to be monitored into the trained support vector machine, and use the trained support vector machine to preliminarily identify the water body pixels affected by warm water discharge in the image of the water area to be monitored, to obtain a preliminary warm water discharge spatial distribution map;

[0037] A second spatial adjacency relationship judgment module, configured to extract the patches of warm water discharge with a spatial adjacency relationship to the nuclear power plant drainage outlet from the preliminary warm water discharge spatial distribution map to obtain a final warm water discharge spatial distribution map.

[0038] Optionally, the water area to be monitored image obtaining module specifically includes:

[0039] A cropping unit, configured to crop the Landsat-8 thermal infrared band image using a mask to obtain an image of the water area to be monitored; the image of the water area to be monitored includes a plurality of water body pixels.

[0040] Optionally, the two types of sample obtaining modules specifically include:

[0041] An average value and standard deviation obtaining unit, configured to obtain the average value and standard deviation of all the anomaly scores according to the anomaly scores of each water body pixel;

[0042] An abnormal water body sample obtaining unit, configured to use all the water body pixels whose anomaly scores are less than or equal to a set threshold as abnormal water body samples; the set threshold is the average value minus 2 times the standard deviation;

[0043] A positive anomaly score water body pixel obtaining unit, configured to obtain each water body pixel with a positive anomaly score according to the anomaly scores of each water body pixel;

[0044] A positive anomaly score water body pixel arranging unit, configured to arrange in ascending order of the anomaly scores each water body pixel with a positive anomaly score, to obtain all water body pixels with positive anomaly scores within the range from the minimum anomaly score to the maximum anomaly score;

[0045] A normal water body sample obtaining unit, configured to use all water body pixels with positive anomaly scores within the range from the median anomaly score to the maximum anomaly score as normal water body samples.

[0046] Optionally, the first spatial adjacency relationship determination module specifically includes:

[0047] A warm discharge sample obtaining unit, configured to extract patches of the abnormal water body samples having a spatial adjacency relationship with the nuclear power plant drainage outlet from the abnormal water body samples as warm discharge samples.

[0048] The present invention also provides the following solution:

[0049] An electronic device, including a memory and a processor, where the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the warm discharge automatic monitoring method.

[0050] The present invention also provides the following solution:

[0051] A computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the warm discharge automatic monitoring method is implemented.

[0052] According to the specific embodiments provided by the present invention, the following technical effects are disclosed by the present invention:

[0053] The automatic monitoring method, system, electronic device and storage medium for warm wastewater disclosed in the present invention realize automatic sample extraction based on the isolation forest algorithm and the judgment of spatial adjacency relationship, obtain normal water body samples and warm wastewater samples, train a support vector machine with the normal water body samples and warm wastewater samples, initially identify the water body pixels affected by warm wastewater in the image of the water area to be monitored based on the trained support vector machine, and further combine the judgment of spatial adjacency relationship to obtain the final spatial distribution map of warm wastewater, so as to realize the automatic monitoring of warm wastewater; by comprehensively using the isolation forest algorithm, spatial adjacency relationship determination and support vector machine supervised classification algorithm, the automatic training sample extraction and supervised classification are combined, without temperature inversion, saving the inversion time, and being more capable of quickly and real-time monitoring of warm wastewater compared with traditional methods; in addition, the automatic training sample extraction greatly reduces the manpower, material resources and time investment in sample collection, minimizes the human intervention process, and has less input data, accelerating the algorithm speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required to be used in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings without creative efforts based on these drawings.

[0055] Figure 1 It is a flowchart of the first embodiment of the automatic monitoring method for warm wastewater of the present invention;

[0056] Figure 2 It is a general situation diagram of the study area;

[0057] Figure 3 It is an image map of the water area to be monitored in 4 time phases;

[0058] Figure 4 It is a schematic diagram of the technical route of the present invention;

[0059] Figure 5 It is a spatial distribution map of warm wastewater identified by different methods;

[0060] Figure 6 It is a detection result map when warm wastewater does not break out;

[0061] Figure 7 It is a schematic diagram of the accuracy of warm wastewater extraction based on samples selected with different thresholds. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0062] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0063] The object of the present invention is to provide a method, system, electronic device and storage medium for automatic monitoring of warm wastewater discharge, which can realize fast and real-time monitoring of warm wastewater discharge.

[0064] To make the above objects, features and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0065] Embodiment 1

[0066] Figure 1 It is a flowchart of Embodiment 1 of the method for automatic monitoring of warm wastewater discharge of the present invention. As Figure 1 shown, this embodiment provides a method for automatic monitoring of warm wastewater discharge, including the following steps:

[0067] Step 101: Obtain the Landsat-8 thermal infrared band image corresponding to the water area to be monitored.

[0068] Step 102: Obtain the image of the water area to be monitored according to the Landsat-8 thermal infrared band image.

[0069] This step 102 specifically includes:

[0070] Use a mask to crop the Landsat-8 thermal infrared band image to obtain the image of the water area to be monitored; the image of the water area to be monitored includes multiple water pixels.

[0071] Step 103: Use the isolation forest algorithm to determine the anomaly score of each water pixel in the image of the water area to be monitored.

[0072] Step 104: Obtain two types of samples according to the anomaly scores of each water pixel; the two types of samples include abnormal water samples and normal water samples.

[0073] This step 104 specifically includes:

[0074] Obtain the average value and standard deviation of all anomaly scores according to the anomaly scores of each water pixel.

[0075] Take all water pixels with anomaly scores less than or equal to the set threshold as abnormal water samples; the set threshold is the average value minus 2 times the standard deviation.

[0076] Obtain water body pixels with positive anomaly scores based on the anomaly scores of each water body pixel.

[0077] Arrange in ascending order of anomaly scores the water body pixels with positive anomaly scores, obtaining all water body pixels with positive anomaly scores within the range from the minimum anomaly score to the maximum anomaly score.

[0078] Use all water body pixels with positive anomaly scores within the range from the median anomaly score to the maximum anomaly score as normal water body samples.

[0079] Step 105: Judge the spatial adjacency relationship of the abnormal water body samples to obtain warm wastewater samples.

[0080] This step 105 specifically includes:

[0081] Extract from the abnormal water body samples the patches of abnormal water body samples having a spatial adjacency relationship with the nuclear power plant drainage outlet as warm wastewater samples.

[0082] Step 106: Train a support vector machine using the warm wastewater samples and the normal water body samples to obtain the trained support vector machine.

[0083] Step 107: Substitute the image of the water area to be monitored into the trained support vector machine, and use the trained support vector machine to initially identify the water body pixels affected by warm wastewater in the image of the water area to be monitored, obtaining a preliminary spatial distribution map of warm wastewater.

[0084] Step 108: Extract from the preliminary spatial distribution map of warm wastewater the patches of warm wastewater having a spatial adjacency relationship with the nuclear power plant drainage outlet to obtain the final spatial distribution map of warm wastewater.

[0085] The present invention considers that warm wastewater is water discharged into natural water bodies with a temperature higher than that of natural water bodies. The temperature of water body pixels affected by warm wastewater will be significantly higher than the temperature of water body pixels around that are not affected by warm wastewater at the same moment. The water body pixels affected by warm wastewater can be regarded as abnormal pixels among water body pixels, and their values in the thermal infrared band are significantly different from normal pixels. Moreover, the pixels of warm wastewater have an adjacency relationship with the nuclear power plant discharge outlet (drainage outlet) in terms of spatial position. A method for automatic monitoring of warm wastewater that comprehensively uses the isolation forest, spatial adjacency relationship, and support vector machine is proposed, that is, the method for automatic monitoring of warm wastewater of the present invention, which can realize fast and real-time monitoring of warm wastewater and help the development of real-time monitoring services for warm wastewater.

[0086] The following uses a specific embodiment to illustrate the technical solution of the present invention:

[0087] This embodiment selects the Fuqing Nuclear Power Plant in the southern section of Sanshan Town, Fuqing City, Fujian Province as the research target, and sets the surrounding sea area as the research area. Geographically, the Fuqing Nuclear Power Plant is surrounded by the sea on three sides, connected to the land in the northeast, with developed surrounding roads and convenient transportation. The average annual water temperature around the sea area is about 22.3°C. Its geographical location is as Figure 2 shown. Figure 2 Figure 4 shows the general situation of the research area. Figure 2 In part (a) and part (b) of Figure 6 are the Landsat-8 true color composite image (data) and the thermal infrared band image (data) on January 18, 2019 respectively. The Fuqing Nuclear Power Plant is the first nuclear power plant in China to adopt the third-generation nuclear power technology, covering an area of about 18 km 2 , with a total of 6 nuclear power generating units and a total installed capacity of 6,600 megawatts. It is one of the important energy bases in the southeastern coast of China and an important landmark project in the Chinese nuclear power industry.

[0088] This embodiment uses the Landsat-8 satellite data obtained through Google Earth Engine after atmospheric correction and geometric correction. The time series of the data used is shown in Table 1. In the above research area, the water body positions near the sea are extracted through visual interpretation and water body recognition algorithms to make a mask. The mask is used to crop the Landsat-8 thermal infrared band (Landsat-8 thermal infrared band image) of 4 time phases to obtain the final image for monitoring (the image of the water area to be monitored), as Figure 3 shown. Figure 3 Figure 15 shows the experimental data. Figure 3 In part (a), part (b), part (c) and part (d) of Figure 17 respectively represent the final images for monitoring on October 30, 2018, November 15, 2018, December 1, 2018 and January 18, 2019.

[0089] Table 1 List of data used

[0090] Time Data Thermal discharge condition 2018 / 10 / 30 Landsat-8 No thermal discharge 2018 / 11 / 15 Landsat-8 Small amount of thermal discharge 2018 / 12 / 01 Landsat-8 Large amount of thermal discharge 2019 / 01 / 18 Landsat-8 Small amount of thermal discharge

[0091] The principle of the Isolation Forest algorithm is introduced as follows:

[0092] Isolation Forest is a machine learning algorithm suitable for anomaly detection. Isolation Forest (IForest) was first proposed at the 8th IEEE International Conference on Data Mining for outlier detection. In Isolation Forest, outliers are those samples with sparse distribution and far from the high-density groups. Isolation Forest is obtained by constructing n isolated binary trees. Its basic process is:

[0093] (1) Randomly select m samples from the sample space to form a subspace, and put these samples into the root node of the tree.

[0094] Among them, the sample space is the final image obtained by cropping the Landsat-8 thermal infrared band using a mask for monitoring. The sample is the water body pixel.

[0095] (2) Randomly select a certain feature in a sample space, and select a certain value z between the maximum value and the minimum value in the feature dataset. Using z as a node, place the data in the feature vector that is less than z on the left side of the current node, and the data greater than z on the right side of the current node, thus forming two new feature subspaces on the left and right.

[0096] (3) In the two new feature subspaces on the left and right respectively, continuously repeat the previous step to construct new child nodes until the data itself only contains one sample or the height of the tree has reached the preset height, then the isolation tree is established. Continuously establish new isolation trees. When n isolation binary trees are constructed, an isolation forest is formed. Calculate the average path length of the test data sample reaching the leaf nodes of each isolation tree, and then calculate the anomaly degree value. The specific calculation is shown in formulas (1) and (2). The shorter the path length, the higher the corresponding anomaly degree value, and the earlier it is isolated. Among them, the test data sample is the pixel of all input raster data. Assuming there are 10,000 samples (water body pixels) in total, 1,000 samples are randomly selected, and an isolation forest is constructed using these 1,000 samples, and then all pixels of the input raster data are judged, which refers to the original pixels.

[0097] Given a dataset containing n samples, the average path length of the tree is:

[0098]

[0099] In formula (1): H(i) is the harmonic number, and this value can be estimated as ln(i) + 0.5772156649. c(n) is the average value of the path length when the given number of samples is n, which is used to standardize the path length h(x) of the sample x.

[0100] The anomaly score s(x,n) of the sample is defined as:

[0101]

[0102] In formula (2): E(h(x)) is the average path length that the sample x spends to find its own position in a batch of isolation trees.

[0103] In the process of implementing Isolation Forest based on Python, when calculating the anomaly score, the degree of anomaly of a data point is estimated according to the local density of the data point. The lower the density of a point, the higher the anomaly score it will get. Since negative numbers can more intuitively represent low density, the scikit-learn library has made adjustments based on the original formula. The adjusted calculation method is shown in formula (3). The smaller the value after adjustment, the higher the corresponding degree of anomaly.

[0104]

[0105] The basic processing flow of the automatic monitoring method for thermal discharge water (automatic monitoring method for thermal discharge water) of the present invention is as follows:

[0106] The temperature of the water body pixels affected by thermal discharge water will be significantly higher than the temperature of the water body pixels in the surrounding area that are not affected by thermal discharge water at the same moment, and there is an adjacency relationship in the spatial position between the pixels of thermal discharge water and the nuclear power plant discharge port. For this reason, the present invention proposes an automatic monitoring method for thermal discharge water that comprehensively uses Isolation Forest, spatial adjacency relationship and Support Vector Machine, that is, an automatic extraction method for thermal discharge water based on Isolation Forest, Support Vector Machine and spatial adjacency relationship. Its basic process includes four steps:

[0107] (1) Use Isolation Forest to calculate the anomaly score of each water body pixel in the monitoring water area for the remote sensing data in the thermal infrared band, that is, use the Isolation Forest algorithm to determine the anomaly score of each water body pixel in the image of the water area to be monitored.

[0108] Among them, according to the remote sensing data in the thermal infrared band, the anomaly score of each water body pixel in the monitoring water area is obtained by using the Isolation Forest algorithm. The specific steps include: referring to formula (2) and formula (3), first, input the training data, construct a tree structure, and obtain a trained Isolation Forest model; second, substitute the samples to be detected and calculate the path length of each sample, see the explanation of formula (2); finally, obtain the final anomaly score based on formula (3).

[0109] (2) Conduct statistics on the anomaly scores, extract the pixels with anomaly scores less than the mean minus 2 times the standard deviation as anomaly water body samples, and extract the sample patches with spatial adjacency relationships with the nuclear power plant drainage outlet from the anomaly samples (anomaly water body samples), and set these sample patches as thermal discharge water samples, that is, obtain the average value and standard deviation of all anomaly scores according to the anomaly scores of each water body pixel, and use all water body pixels with anomaly scores less than or equal to the set threshold (the mean minus 2 times the standard deviation) as anomaly water body samples, and judge the spatial adjacency relationship of the anomaly water body samples to obtain thermal discharge water samples.

[0110] Among them, to determine the sample patches with a spatial adjacency relationship with the drainage outlet from the abnormal samples, the specific steps are as follows: Convert the raster data of the extracted abnormal water body samples into vector data to obtain the patches of the abnormal water body, overlay the point vector data of the nuclear power plant drainage outlet on the patches of the abnormal water body samples, and extract the patches of the abnormal water body samples with an adjacent relationship to the point as the sample patches of the warm water discharge.

[0111] (3) Extract the pixels with a positive abnormal score for statistics, and take the pixels in the interval from the median to the maximum value as the normal water body samples (normal water body samples), that is, obtain the water body pixels with a positive abnormal score according to the abnormal scores of each water body pixel, arrange the water body pixels with a positive abnormal score in ascending order of the abnormal scores, obtain all the water body pixels in the interval from the minimum abnormal score to the maximum abnormal score, and take all the water body pixels in the interval from the median abnormal score to the maximum abnormal score as the normal water body samples.

[0112] (4) Substitute the two types of samples into the support vector machine for supervised classification to obtain a preliminary spatial distribution map of the warm water discharge (that is, substitute the warm water discharge samples and the normal water body samples into the support vector machine for training to obtain the trained support vector machine, and substitute the image of the water area to be monitored into the trained support vector machine to obtain a preliminary spatial distribution map of the warm water discharge), and further extract the water body patches (patches of the warm water discharge) with an adjacency relationship (with a spatial adjacency relationship) with the nuclear power plant drainage outlet from this map to obtain the final spatial distribution map of the warm water discharge.

[0113] Among them, the training data set of the support vector machine includes the warm water discharge samples obtained in step (2) and the normal water body samples obtained in step (3). After the support vector machine is trained, input (substitute) the image of the water area to be monitored into the trained support vector machine. Using the trained support vector machine, the water body pixels affected by the warm water discharge in the image of the water area to be monitored can be initially identified to obtain a preliminary spatial distribution map of the warm water discharge. Further, combined with the spatial adjacency relationship judgment, the final spatial distribution map of the warm water discharge is obtained, that is, the water body patches with an adjacency relationship with the nuclear power plant drainage outlet are further extracted from the preliminary spatial distribution map of the warm water discharge to obtain the final spatial distribution map of the warm water discharge.

[0114] To determine the water body patches with a spatial adjacency relationship with the nuclear power plant drainage outlet from the preliminary spatial distribution map of the warm water discharge, the specific steps are as follows: Convert the preliminary spatial distribution map of the warm water discharge into vector data, then overlay the point vector data of the nuclear power plant drainage outlet on the preliminary warm water discharge vector data, and extract the patches of the warm water discharge with an adjacent relationship to the point as the final extraction result of the warm water discharge.

[0115] Compared with the final image for monitoring obtained by cropping the thermal infrared band of Landsat-8 using a mask, the spatial distribution map of thermal discharge obtained in the (4)th step identifies the water body pixels affected by thermal discharge that are adjacent to the nuclear power plant drainage outlet (i.e., the abnormal pixels in the water body pixels). That is, the final spatial distribution map of thermal discharge is a map that identifies the water body pixels affected by thermal discharge that are adjacent to the nuclear power plant drainage outlet.

[0116] The technical route and experimental design of this embodiment are as follows:

[0117] To test the effectiveness of the proposed automated thermal discharge monitoring method of the present invention in real-time monitoring of thermal discharge, a technical route as shown in Figure 4 is designed, and the specific steps are as follows:

[0118] (1) Vectorize the perennial coastal boundary based on the data on January 18, 2019, and then identify the water body boundary of each image based on the Automated Water Extraction Index (AWEI) as shown in formula (4) to obtain the water body boundary data. Combine the two boundaries to extract the monitored water body range at each time point as shown in Figure 3 .

[0119] AWEI = 4×(B2 - B5) - 0.25×B4 + 2.75×B7 (4)

[0120] In formula (4), B2 is the green band, B4 is the near-infrared band, and B5 and B7 are two shortwave infrared bands respectively.

[0121] (2) Calibrate the positions of the nuclear power plant and its inlet and outlet through visual interpretation.

[0122] (3) Use the proposed automated thermal discharge monitoring method of the present invention to extract the spatial distribution of thermal discharge from the thermal infrared band satellite data of Landsat-8 at different time phases (November 15, 2018, December 1, 2018, and January 18, 2019).

[0123] (4) Through visual interpretation, determine the thermal discharge area on each image. Using this as the true value sample, calculate the user accuracy and producer accuracy of thermal discharge extraction using formula (5) and formula (6).

[0124] P User = Count R / (Count R + Count W ) (5)

[0125] P Prod = Count R / (Count R+Count A ) (6)

[0126] Among them, P User is the user accuracy, and P Prod is the producer accuracy. Count R is the number of correctly classified pixels, and Count W is the number of pixels of normal water bodies misclassified as thermal discharges, and Count A is the number of pixels that are actually thermal discharges but misclassified as normal water bodies. The user accuracy corresponds to the misclassification error. The higher the user accuracy, the smaller the misclassification error; the producer accuracy corresponds to the omission error. The higher the producer accuracy, the smaller the omission error.

[0127] In order to explore the role of adjacency relationship judgment in the automated monitoring method for thermal discharges proposed in the present invention (abbreviated as Method I), three groups of control experiments were also carried out in this embodiment:

[0128] (1) On the basis of extracting abnormal water body pixels using Isolation Forest, extract the pixels having an adjacency relationship with the nuclear power plant drainage outlet as the training samples for thermal discharges, and substitute the samples into the supervised classifier of the support vector machine to obtain the thermal discharge extraction result, that is, Isolation Forest + adjacency relationship judgment + SVM, abbreviated as Method II.

[0129] (2) Substitute the samples extracted by Isolation Forest directly into the supervised classifier of the support vector machine to obtain the thermal discharge extraction result, that is, Isolation Forest + no adjacency relationship judgment + SVM, abbreviated as Method III.

[0130] (3) On the basis of the comparative experiment (2), after the supervised classification is completed, extract the patches having an adjacency relationship with the nuclear power plant drainage outlet as the final result, that is, Isolation Forest + no adjacency relationship judgment + SVM + adjacency relationship judgment, abbreviated as Method IV.

[0131] The results of this embodiment are introduced below:

[0132] Figure 5 shows the thermal discharges of three time phases extracted by different methods. Among them, the spatial distribution of the thermal discharge identified by Method I on November 15, 2018 is as Figure 5 shown in part (a) therein, the spatial distribution of the thermal discharge identified by Method II on November 15, 2018 is as Figure 5 shown in part (b) therein, the spatial distribution of the thermal discharge identified by Method III on November 15, 2018 is as Figure 5 shown in part (c) therein, the spatial distribution of the thermal discharge identified by Method IV on November 15, 2018 is as Figure 5 shown in part (d) therein, and the spatial distribution of the thermal discharge identified by Method I on December 1, 2018 is as Figure 5As shown in part (e), the spatial distribution of the warm wastewater discharged on December 1, 2018 identified by Method Ⅱ is as Figure 5 As shown in part (f), the spatial distribution of the warm wastewater discharged on December 1, 2018 identified by Method Ⅲ is as Figure 5 As shown in part (g), the spatial distribution of the warm wastewater discharged on December 1, 2018 identified by Method Ⅳ is as Figure 5 As shown in part (h). The spatial distribution of the warm wastewater discharged on January 18, 2019 identified by Method Ⅰ is as Figure 5 As shown in part (i), the spatial distribution of the warm wastewater discharged on January 18, 2019 identified by Method Ⅱ is as Figure 5 As shown in part (j), the spatial distribution of the warm wastewater discharged on January 18, 2019 identified by Method Ⅲ is as Figure 5 As shown in part (k), the spatial distribution of the warm wastewater discharged on January 18, 2019 identified by Method Ⅳ is as Figure 5 As shown in part (l). According to the results of visual interpretation, the areas of the warm wastewater regions were calculated using ArcGIS, and the total areas of the warm wastewater discharged on November 15, 2018, December 1, 2018, and January 18, 2019 were 0.76 km 2 , 8.30 km 2 and 2.25 km 2 . The warm wastewater monitored by Method Ⅰ, Method Ⅱ, Method Ⅲ, and Method Ⅳ was verified, and the results are shown in Table 2.

[0133] Table 2 Precision Evaluation Table

[0134]

[0135] In Table 2: Ⅰ represents the method of Isolation Forest + Adjacency Relationship Judgment + SVM + Adjacency Relationship Judgment (Method Ⅰ), that is, the automated monitoring method for warm wastewater proposed in the present invention; Ⅱ represents the method of Isolation Forest + Adjacency Relationship Judgment + SVM (Method Ⅱ); Ⅲ represents the method of Isolation Forest + No Adjacency Relationship Judgment + SVM (Method Ⅲ); Ⅳ represents the method of Isolation Forest + No Adjacency Relationship Judgment + SVM + Adjacency Relationship Judgment (Method Ⅳ). Combining Figure 5 and Table 2, it can be seen that:

[0136] (1) Regardless of Method Ⅰ, Method Ⅱ, Method Ⅲ, or Method Ⅳ, the producer precision of the warm wastewater extracted in the three periods is significantly higher than the user precision. The user precision is the ratio of the number of correctly classified warm wastewater pixels to the total number of warm wastewater pixels obtained by classification. The higher the user precision, the lower the misclassification error. The producer precision is the ratio of the number of correctly classified warm wastewater pixels to the actual number of warm wastewater pixels. The higher the producer precision, the lower the omission error. Therefore, generally speaking, the omission errors of Method Ⅰ, Method Ⅱ, Method Ⅲ, and Method Ⅳ are all lower than the misclassification errors.

[0137] (2) Among Method I, Method II, Method III, and Method IV, the temperature discharge automatic monitoring method proposed by the present invention has the highest accuracy in detecting temperature discharge in three periods. The average values of the user accuracy and producer accuracy of the temperature discharge detected on the three-phase images on November 15, 2018, December 1, 2018, and January 18, 2019 are 89.69%, 94.97%, and 90.04% respectively.

[0138] (3) By comparing the methods with and without adding adjacency relationships, it can be seen that generally speaking, adding adjacency relationships twice can greatly improve the user accuracy and reduce the misclassification error. For example, on November 15, 2018, the user accuracy of adding adjacency relationships twice reached 99.82%, which was 79.27% higher than that of Method II which only added adjacency relationships after the supervised classification was completed. In the use of adjacency relationships twice, the importance of the first adjacency relationship is higher than that of the second. The methods that consider spatial adjacency relationships (Method I and Method II) in preparing samples are better than those that do not consider, and it is particularly obvious in the case of only a small amount of temperature discharge outbreaks (November 15, 2018 and January 18, 2019). Incorrect sample selection can even lead to the inability to detect temperature discharge, such as the results obtained by Method IV on November 15, 2018 and January 18, 2019.

[0139] The advantages of the temperature discharge automatic monitoring method proposed by the present invention are as follows:

[0140] First of all, the temperature discharge automatic monitoring method proposed by the present invention does not require water temperature inversion. Traditional temperature discharge monitoring methods need to perform a series of pre-treatments such as radiometric calibration and atmospheric correction, then perform the inversion of the water surface temperature, and then perform statistical analysis to screen out abnormal positions. The whole process involves many methods and is relatively cumbersome. Especially for the inversion of temperature, other factors outside the remote sensing data may need to be input. The temperature discharge automatic monitoring method proposed by the present invention does not require temperature inversion, saving the inversion time and being more capable of real-time monitoring compared with traditional methods.

[0141] Secondly, the temperature discharge automatic monitoring method proposed by the present invention has good portability and high automation. The temperature discharge automatic monitoring method proposed by the present invention is essentially a method that combines automatic sample extraction and supervised classification. By calculating the anomaly scores of the water body pixels in the monitoring area through the Isolation Forest, and combining the mean standard deviation method to determine the threshold to extract the temperature discharge samples. The anomaly scores and temperature discharge thresholds calculated for each area or different times in the same area are different, that is, the thresholds calculated according to the temperature discharge automatic monitoring method of the present invention are dynamic and adaptive, ensuring the portability of the method. In addition, the automatic extraction of training samples greatly reduces the input of manpower, material resources and time in sample collection.

[0142] Finally, the automated monitoring method for thermal wastewater discharge proposed by the present invention can perform real-time monitoring of thermal wastewater discharge. To achieve real-time monitoring of thermal wastewater discharge, it is necessary to continuously track the monitored water area. The algorithm used should not produce false alarms when there is no thermal wastewater discharge (i.e., falsely detecting thermal wastewater discharge when there is actually none), and when there is thermal wastewater discharge, it should extract the range of thermal wastewater discharge with high precision as much as possible and reduce misreporting. Previous studies often used images during the occurrence of thermal wastewater discharge to test the effectiveness of algorithms. Such algorithms are often "after-the-fact" and not timely, and are not suitable for real-time monitoring business scenarios. To further verify the effectiveness of the automated monitoring method for thermal wastewater discharge proposed by the present invention in real-time monitoring of thermal wastewater discharge, this embodiment not only conducted experiments on the extraction of thermal wastewater discharge from images with thermal wastewater discharge on November 15, 2018, December 1, 2018, and January 18, 2019. The results are as Figure 5 shown. The same processing was also carried out on the thermal infrared band of Landsat-8 on October 30, 2018, when there was no thermal wastewater discharge, to observe whether there were "false alarms". The detection results when there was no thermal wastewater discharge outbreak are as Figure 6 shown. When there is no thermal wastewater discharge outbreak, the automated monitoring method for thermal wastewater discharge proposed by the present invention will not produce false alarms. Combining ​ it can be seen that the automated monitoring method for thermal wastewater discharge proposed by the present invention is applicable when there is no thermal wastewater discharge, a small amount of thermal wastewater discharge, and a large amount of thermal wastewater discharge.

[0143] Next, the factors affecting the accuracy of the automated monitoring method for thermal wastewater discharge proposed by the present invention will be introduced:

[0144] First, when preparing samples based on the Isolation Forest, the selection of the standard deviation multiple has a certain impact on the sample preparation. ​ Schematic diagram of the accuracy of thermal wastewater discharge extraction for samples selected based on different thresholds; ​ shows the changes in the producer accuracy and user accuracy of the thermal wastewater discharge sample extraction results as the standard deviation multiple increases (the standard deviation multiple is set to change in steps of 0.1 between 1 and 2); ​ Part (a) in it shows the changes in the producer accuracy and user accuracy of the thermal wastewater discharge sample extraction results on November 15, 2018, as the standard deviation multiple increases, ​ Part (b) in it shows the changes in the producer accuracy and user accuracy of the thermal wastewater discharge sample extraction results on December 1, 2018, as the standard deviation multiple increases, ​Part (c) shows the variation of producer accuracy and user accuracy of the warm discharge water sample extraction results on January 18, 2019 with the increase of the multiple of the standard deviation. Generally speaking, with the increase of the multiple of the standard deviation, the producer accuracy does not change much, and the lowest is above 0.98, while the user accuracy shows an obvious improvement. This indicates that the samples obtained by the isolation forest plus the adjacency relationship can basically cover all the warm discharge water pixels of the nuclear power plant, and there are very few missed samples; however, when the multiple of the standard deviation is too small, there are many misclassified warm discharge water samples. As the multiple of the standard deviation increases, the user accuracy increases and the misclassified warm discharge water samples decrease. In addition, the rate of increase of the user accuracy with the increase of the multiple of the standard deviation is different in data of different time phases. Generally speaking, when approaching 2 times the standard deviation, the rate of increase of the user accuracy under the three time phases significantly decreases.

[0145] Secondly, the choice of the supervised classification algorithm will affect the final result. The present invention selects the combination of the commonly used SVM method and the automated sample extraction. The core of the warm discharge water automated monitoring method proposed by the present invention is the automated sample extraction and the determination of the adjacency relationship. The SVM supervised classifier can also be replaced by other classifiers. Different classifiers have different requirements for samples, and there will be differences in the classification accuracy obtained from the same samples. In the future, combinations with other supervised classifiers can be tried, and classifiers with high and stable accuracy can be preferably selected.

[0146] Based on the low automation level and poor timeliness of existing remote sensing algorithms for detecting warm water discharges, the present invention proposes an automated extraction method for warm water discharges (automated monitoring method for warm water discharges) based on single-temporal thermal infrared bands. Taking the detection of warm water discharges at different times in Fujian Fuqing Nuclear Power Plant as an example, the accuracy of the algorithm was tested. Specifically, 4 periods of Landsat-8 remote sensing images from October 2018 to January 2019 were used to test and verify the automated monitoring method for warm water discharges proposed in the present invention. This method first uses the Isolation Forest to automatically extract warm water discharge and normal water body samples, then purifies the warm water discharge samples through the spatial adjacency relationship between warm water discharges and discharge outlets, and then uses Support Vector Machine supervised classification to extract warm water discharge pixels. Finally, misjudged pixels are removed through the spatial adjacency relationship between warm water discharges and discharge outlets to obtain the final spatial distribution of warm water discharges. This method reduces the human intervention process to the greatest extent. After the algorithm becomes automated, the speed can naturally be increased. In addition, with less input data, the algorithm speed is also accelerated. The test results show that the producer accuracies of warm water discharge detection on November 15, 2018, December 1, 2018, and January 18, 2019 are 79.56%, 95.17%, and 98.46% respectively, and the user accuracies are 99.82%, 94.77%, and 81.62% respectively. The average values of the user accuracy and producer accuracy of the warm water discharges detected by the automated monitoring method for warm water discharges proposed in the present invention on the three-phase images of November 15, 2018, December 1, 2018, and January 18, 2019 are 89.69%, 94.97%, and 90.04% respectively. Adding the determination of the adjacency relationship between warm water discharge pixels and discharge outlets greatly improves the accuracy of warm water discharge identification. It is particularly important to add the adjacency relationship judgment in the sample extraction stage. The accuracy of the classification methods (Methods III and IV) without adding the adjacency relationship in the sample extraction stage is even as low as 0%. In the phase without warm water discharge (October 30, 2018), the automated monitoring method for warm water discharges proposed in the present invention did not cause false detection, that is, there was no false detection on the image without warm water discharge on October 30, 2018, effectively avoiding the occurrence of "false alarms". This algorithm (the automated monitoring method for warm water discharges proposed in the present invention) only needs to input the thermal infrared band of the remote sensing image and does not need to substitute other parameters additionally. It has the advantages of strong portability (good portability), strong universality, and high automation level, and has certain reference value for the real-time monitoring of warm water discharges, and has good application prospects in the real-time monitoring and rapid discovery of warm water discharges.

[0147] Embodiment 2

[0148] In order to execute the method corresponding to the above Embodiment 1 to achieve the corresponding functions and technical effects, an automated monitoring system for warm water discharges is provided below. The system includes the following modules:

[0149] The thermal infrared band image acquisition module is used to acquire the Landsat-8 thermal infrared band image corresponding to the water area to be monitored.

[0150] The water area image to be monitored obtaining module is used to obtain the water area image to be monitored according to the Landsat-8 thermal infrared band image.

[0151] The abnormal score determination module is used to determine the abnormal scores of each water body pixel in the water area image to be monitored by using the isolation forest algorithm.

[0152] The two types of sample obtaining module is used to obtain two types of samples according to the abnormal scores of each water body pixel; the two types of samples include abnormal water body samples and normal water body samples.

[0153] [[ID=1,2]]

[0154] The first spatial adjacency relationship judgment module is used to judge the spatial adjacency relationship of the abnormal water body samples to obtain the warm water discharge samples.

[0155] The support vector machine training module is used to train the support vector machine by using the warm water discharge samples and the normal water body samples to obtain the trained support vector machine.

[0155]

[0156] The warm water discharge preliminary identification module is used to substitute the water area image to be monitored into the trained support vector machine, and use the trained support vector machine to preliminarily identify the water body pixels affected by the warm water discharge in the water area image to be monitored to obtain the preliminary warm water discharge spatial distribution map.

[0157] Among them, the water area image to be monitored obtaining module specifically includes:

[0158] The cropping unit is used to crop the Landsat-8 thermal infrared band image by using the mask to obtain the water area image to be monitored; the water area image to be monitored includes a plurality of water body pixels.

[0159] The two types of sample obtaining module specifically includes:

[0160] The average value and standard deviation obtaining unit is used to obtain the average value and standard deviation of all abnormal scores according to the abnormal scores of each water body pixel.

[0161] The abnormal water body sample obtaining unit is used to use all water body pixels with abnormal scores less than or equal to the set threshold as the abnormal water body samples; the set threshold is the average value minus 2 times the standard deviation.

[0162] The positive abnormal score water body pixel obtaining unit is used to obtain each water body pixel with a positive abnormal score according to the abnormal scores of each water body pixel.

[0163] A positive abnormal score water body pixel arrangement unit is used to arrange each water body pixel with a positive abnormal score in ascending order of the abnormal score, so as to obtain all water body pixels with a positive abnormal score within the range from the minimum abnormal score to the maximum abnormal score.

[0164] A normal water body sample obtaining unit is used to take all water body pixels within the range from the median abnormal score to the maximum abnormal score among all water body pixels with a positive abnormal score as normal water body samples.

[0165] The first spatial adjacency relationship judgment module specifically includes:

[0166] A warm wastewater sample obtaining unit is used to extract the patches of abnormal water body samples having a spatial adjacency relationship with the nuclear power plant drainage outlet from the abnormal water body samples as warm wastewater samples.

[0167] Embodiment III

[0168] Embodiment III of the present invention provides an electronic device, including a memory and a processor. The memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the warm wastewater automatic monitoring method of Embodiment I.

[0169] The above-mentioned electronic device may be a server.

[0170] Embodiment IV

[0171] Embodiment IV of the present invention provides a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, it implements the warm wastewater automatic monitoring method of Embodiment I.

[0172] In this specification, each embodiment is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other. For the system disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0173] In this article, specific examples are used to elaborate the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. An automated monitoring method for warm wastewater discharge, characterized in that, The method includes: Obtaining a Landsat-8 thermal infrared band image corresponding to the water area to be monitored; Obtaining an image of the water area to be monitored based on the Landsat-8 thermal infrared band image; Using the Isolation Forest algorithm to determine the anomaly scores of each water body pixel in the image of the water area to be monitored; Obtaining two types of samples based on the anomaly scores of each water body pixel; the two types of samples include abnormal water body samples and normal water body samples; Judging the spatial adjacency relationship of the abnormal water body samples to obtain warm water discharge samples; Training a support vector machine using the warm water discharge samples and the normal water body samples to obtain a trained support vector machine; Substituting the image of the water area to be monitored into the trained support vector machine, and using the trained support vector machine to initially identify the water body pixels affected by warm water discharge in the image of the water area to be monitored, to obtain a preliminary spatial distribution map of warm water discharge; Extracting the patches of warm water discharge having a spatial adjacency relationship with the nuclear power plant drainage outlet from the preliminary spatial distribution map of warm water discharge to obtain the final spatial distribution map of warm water discharge.

2. The automated monitoring method for warm wastewater according to claim 1, wherein Obtaining an image of the water area to be monitored based on the Landsat-8 thermal infrared band image, specifically including: Cropping the Landsat-8 thermal infrared band image using a mask to obtain an image of the water area to be monitored; the image of the water area to be monitored includes multiple water body pixels.

3. The automated monitoring method for warm wastewater according to claim 1, wherein Obtaining two types of samples based on the anomaly scores of each water body pixel, specifically including: Obtaining the average value and standard deviation of all the anomaly scores based on the anomaly scores of each water body pixel; Taking all the water body pixels whose anomaly scores are less than or equal to the set threshold as abnormal water body samples; the set threshold is the average value minus 2 times the standard deviation; Obtaining each water body pixel with a positive anomaly score based on the anomaly scores of each water body pixel; Arranging in ascending order of the anomaly scores each water body pixel with a positive anomaly score, to obtain all water body pixels with positive anomaly scores within the range from the minimum anomaly score to the maximum anomaly score; Taking all the water body pixels within the range from the median anomaly score to the maximum anomaly score among all the water body pixels with positive anomaly scores as normal water body samples.

4. The automated monitoring method for warm wastewater according to claim 1, characterized in that, Judging the spatial adjacency relationship of the abnormal water body samples to obtain warm water discharge samples, specifically including: Extracting the patches of the abnormal water body samples having a spatial adjacency relationship with the nuclear power plant drainage outlet from the abnormal water body samples as warm water discharge samples.

5. An automated monitoring system for warm wastewater, characterized in that, The system includes: A thermal infrared band image acquisition module, configured to obtain a Landsat-8 thermal infrared band image corresponding to the water area to be monitored; An image acquisition module of the water area to be monitored, configured to obtain an image of the water area to be monitored based on the Landsat-8 thermal infrared band image; An anomaly score determination module, configured to use the Isolation Forest algorithm to determine the anomaly scores of each water body pixel in the image of the water area to be monitored; A two-type sample acquisition module, configured to obtain two types of samples based on the anomaly scores of each water body pixel; the two types of samples include abnormal water body samples and normal water body samples; The first spatial adjacency relationship judgment module is used to judge the spatial adjacency relationship of the abnormal water body sample to obtain a warm water discharge sample; The support vector machine training module is used to train a support vector machine using the warm water discharge sample and the normal water body sample to obtain a trained support vector machine; The warm water discharge preliminary identification module is used to substitute the image of the water area to be monitored into the trained support vector machine, and use the trained support vector machine to preliminarily identify the water body pixels affected by warm water discharge in the image of the water area to be monitored to obtain a preliminary spatial distribution map of warm water discharge; The second spatial adjacency relationship judgment module is used to extract the patches of warm water discharge having a spatial adjacency relationship with the nuclear power plant drain outlet from the preliminary spatial distribution map of warm water discharge to obtain a final spatial distribution map of warm water discharge.

6. The automated monitoring system for warm wastewater discharge according to claim 5, characterized in that The module for obtaining the image of the water area to be monitored specifically includes: The cropping unit is used to crop the Landsat-8 thermal infrared band image using a mask to obtain an image of the water area to be monitored; the image of the water area to be monitored includes a plurality of water body pixels.

7. The automated monitoring system for warm wastewater discharge according to claim 5, characterized in that The module for obtaining the two types of samples specifically includes: The unit for obtaining the average value and standard deviation is used to obtain the average value and standard deviation of all the abnormal scores according to the abnormal scores of the water body pixels; The unit for obtaining the abnormal water body sample is used to use all the water body pixels whose abnormal scores are less than or equal to the set threshold as the abnormal water body sample; the set threshold is the average value minus twice the standard deviation; The unit for obtaining the water body pixels with positive abnormal scores is used to obtain each of the water body pixels with positive abnormal scores according to the abnormal scores of the water body pixels; The arranging unit for the water body pixels with positive abnormal scores is used to arrange in ascending order of the abnormal scores each of the water body pixels with positive abnormal scores to obtain all the water body pixels with positive abnormal scores within the range from the minimum abnormal score to the maximum abnormal score; The unit for obtaining the normal water body sample is used to use all the water body pixels within the range from the median abnormal score to the maximum abnormal score among all the water body pixels with positive abnormal scores as the normal water body sample.

8. The automated monitoring system for warm wastewater discharge according to claim 5, wherein, The first spatial adjacency relationship judgment module specifically includes: The unit for obtaining the warm water discharge sample is used to extract the patch of the abnormal water body sample having a spatial adjacency relationship with the nuclear power plant drain outlet from the abnormal water body sample as the warm water discharge sample.

9. An electronic device, characterized in that, It includes a memory and a processor. The memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the warm water discharge automatic monitoring method according to any one of claims 1-4.

10. A computer-readable storage medium, characterized in that, It stores a computer program, and when the computer program is executed by the processor, it implements the warm water discharge automatic monitoring method according to any one of claims 1-4.

Images