A water body anomaly identification method and system
By using the unsupervised isolated forest algorithm and decision rules, the computational complexity and false alarm problems of water anomaly identification in remote sensing images are solved, and efficient and accurate water anomaly monitoring is achieved.
Patent Information
- Application Number
- CN202211507519.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-11-29
- Publication Date
- 2026-01-02
- Estimated Expiration
- 2042-11-29
AI Technical Summary
Existing remote sensing image anomaly detection technologies suffer from high computational complexity, low efficiency, and difficulty in identifying water anomalies with unchanged spectral information in water pollution monitoring, leading to frequent false alarms.
An unsupervised isolated forest algorithm (U-IForest) combined with decision rules is used to identify water anomaly types and avoid misjudgments through preprocessing, isolated forest processing, anomaly degree statistics, and NDVI analysis.
It has improved the accuracy of water anomaly identification, reduced the occurrence of false alarms and misleading alarms, and achieved efficient and accurate water anomaly monitoring.
Smart Images
Figure CN115761367B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of water body anomaly identification, in particular to a water body anomaly identification method and system. BACKGROUND
[0002] Aquatic environment is an important part of the ecosystem. In recent years, with the intensification of human activities, the frequency of water pollution events in nearshore and inland water environments has increased significantly. Many scholars at home and abroad use remote sensing technology to study water environment, including monitoring of water pollution events such as red tide, enteromorpha, black odor, and oil spill. However, existing researches are usually carried out for specific regions, specific data sources and specific time phases, and the methods lack universality.
[0003] Water pollution events caused by human or natural factors can be regarded as water anomaly events, which can be identified by remote sensing anomaly detection technology. Remote sensing image anomaly target detection technology has been developed for a long time, and a large number of algorithms have been proposed, such as anomaly detection algorithms based on deep learning model, support vector, adaptive threshold, subspace, clustering, etc. However, remote sensing image data is large in size, high in dimension and serious in redundancy. Most common algorithms have high computational complexity, and are superior in theory but difficult to meet the needs of real-time remote sensing services in efficiency. Moreover, water anomaly events are various, including organic nitrogen and phosphorus pollution, COD exceeding standard, heavy metal pollution, algae outbreak, oil spill, etc. Some of the anomalies do not change the spectral information of the water body obviously, and it is difficult to monitor effectively using remote sensing means. Therefore, how to realize high-precision and high-efficiency anomaly target detection task has been the direction of researchers' efforts in the development of remote sensing image anomaly target detection technology. SUMMARY
[0004] The purpose of the present application is to provide a water body anomaly identification method and system, which can improve the accuracy of water body anomaly identification and effectively avoid the occurrence of "false alarm" and "false alarm".
[0005] To achieve the above purpose, the present application provides the following scheme:
[0006] A water body anomaly identification method, comprising:
[0007] obtaining an image of a water body to be identified;
[0008] preprocessing the image of the water body to be identified; the preprocessing includes atmospheric correction, geometric correction, water body mask making and NDVI calculation;
[0009] performing unsupervised isolation forest processing on the preprocessed image to determine an anomaly degree value of a pixel of the water body to be identified;
[0010] The abnormal degree values of the water body pixels to be identified are counted to determine the mean value and the standard deviation of the abnormal degree;
[0011] The pixels of the image of the water body to be identified are classified according to the mean value and the standard deviation of the abnormal degree; the classification result includes: a normal water body pixel or an abnormal water body pixel;
[0012] The water body abnormal type of each abnormal water body pixel is determined according to the NDVI of the abnormal water body pixel;
[0013] If a set percentage of abnormal pixels in all abnormal water body pixels do not belong to any water body abnormal type, the corresponding abnormal water body pixel is re-determined as a normal water body pixel;
[0014] The water body abnormal type with the largest proportion in all abnormal water body pixels is determined as the water body abnormal type of the water body to be identified.
[0015] Optionally, the pixels of the image of the water body to be identified are classified according to the mean value and the standard deviation of the abnormal degree, and specifically include:
[0016] The minimum value of the range of the normal water body pixel is determined by the difference between the mean value and the standard deviation, and the maximum value of the range of the normal water body pixel is determined by the sum of the mean value and the standard deviation;
[0017] The pixel in the image of the water body to be identified whose abnormal degree value is not in the range of the normal water body pixel is determined as an abnormal water body pixel.
[0018] Optionally, the set percentage is 90%.
[0019] A water body abnormality identification system, comprising:
[0020] An image acquisition module for acquiring an image of a water body to be identified;
[0021] A preprocessing module for preprocessing the image of the water body to be identified; the preprocessing includes: atmospheric correction, geometric correction, water body mask making, and NDVI calculation;
[0022] An abnormal degree value determination module for performing unsupervised isolation forest processing on the preprocessed image to determine the abnormal degree values of the water body pixels to be identified;
[0023] A mean value and standard deviation determination module for counting the abnormal degree values of the water body pixels to be identified to determine the mean value and the standard deviation of the abnormal degree;
[0024] A pixel classification module for classifying the pixels of the image of the water body to be identified according to the mean value and the standard deviation of the abnormal degree; the classification result includes: a normal water body pixel or an abnormal water body pixel;
[0025] The type determination module of the pixel is used for determining the corresponding water body abnormal type according to the NDVI of each abnormal water body pixel;
[0026] The first decision module is used for re-determining the corresponding abnormal water body pixel as a normal water body pixel if a set percentage of abnormal pixels in all abnormal water body pixels do not belong to any water body abnormal type.
[0027] The second decision module is used for determining the water body abnormal type of the water body to be identified as the water body abnormal type with the largest proportion in all abnormal water body pixels.
[0028] A water body abnormal identification system comprises at least one processor, at least one memory and computer program instructions stored in the memory, and the computer program instructions realize the water body abnormal identification method when executed by the processor.
[0029] According to the specific embodiments provided by the application, the following technical effects are disclosed.
[0030] The water body abnormal identification method and system provided by the application determine the abnormal degree value of the water body pixel to be identified by using unsupervised isolation forest, and then perform statistics on the abnormal degree value of the water body pixel to be identified, determine the mean value and standard deviation of the abnormal degree, classify the pixels of the image of the water body to be identified by using the mean value and standard deviation of the abnormal degree, and then determine the corresponding water body abnormal type according to the NDVI of each abnormal water body pixel. If a set percentage of abnormal pixels in all abnormal water body pixels do not belong to any water body abnormal type, the corresponding abnormal water body pixel is re-determined as a normal water body pixel. The water body abnormal type of the water body to be identified is determined as the water body abnormal type with the largest proportion in all abnormal water body pixels, that is, the water body abnormal type is judged by using the decision rule. The universality and stability of the decision rule are used to improve the accuracy of water body abnormal identification, and the occurrence of false alarm and false alarm is effectively avoided. BRIEF DESCRIPTION OF DRAWINGS
[0031] In order to more clearly illustrate the technical solutions of the embodiments of the application or the prior art, the following will briefly introduce the drawings needed in the embodiments. Obviously, the drawings in the following description are only some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.
[0032] Figure 1 A water body abnormal identification method flow chart provided by the application;
[0033] Figure 2 A contrast diagram of reflectivity (a) and NDVI (b) of normal water body and abnormal water body.
[0034] Figure 3 Flowchart for water body anomaly discovery and identification;
[0035] Figure 4 Schematic diagram of abnormal water body identification result based on three types of methods of SVM, S-IForest and U-IForest;
[0036] Figure 5 Schematic diagram of abnormal water body identification result only relying on decision rules;
[0037] Figure 6 Schematic diagram of abnormal water body identification result only using U-IForest. DETAILED DESCRIPTION
[0038] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.
[0039] The purpose of the present application is to provide a water body anomaly identification method and system, which can improve the accuracy of water body anomaly identification, and thus can effectively avoid the occurrence of "false alarm" and "false alarm".
[0040] In order to make the above-mentioned purposes, characteristics and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.
[0041] Figure 1 A flowchart of a water body anomaly identification method provided by the present application is shown in FIG. 1, which comprises the following steps: Figure 1
[0042] S101, acquiring an image of a water body to be identified.
[0043] As a specific embodiment, the image of the water body to be identified is Landsat data and Sentinel data. The Landsat data has good spatial resolution, a large spectral span (from visible light to mid-infrared), and intersects with the spectral settings of most current satellites in orbit. Although the development time of the Sentinel data is not as long as the Landsat series, the highest spatial resolution of 10 meters and the more precise division in the visible light to mid-infrared spectrum make it one of the most popular free data in recent years. Sentinel also provides radar data, and the multi-satellite setting allows it to revisit the target area multiple times within a month, which can meet the needs of many specific work.
[0044] S102, pre-processing the image of the water body to be identified; the pre-processing includes: atmospheric correction, geometric correction, water body mask making and NDVI calculation. The water body mask is used to control the image of the water body to be identified. There may be water bodies and other various ground object types (buildings, trees, etc.) on an image. The mask is used to control the processing object to ensure that all processing is for water body pixels.
[0045] The FLAASH module of ENVI is used for atmospheric correction of Landsat data.
[0046] The Sen2Cor plug-in is used for atmospheric correction of Sentinel data.
[0047] Resampling is used to keep the spatial resolution of each band of Landsat-8 at 30 meters and the spatial resolution of each band of Sentinel-2 at 10 meters. The water body boundary is digitized by visual interpretation to extract the water body range.
[0048] S103, performing unsupervised isolated forest processing on the pre-processed image to determine the abnormality value of the water body pixel to be identified.
[0049] The isolated forest algorithm (IForest) was first proposed by Liu et al. at the eighth IEEE International Conference on Data Mining. It is mainly used for outlier detection in data mining. The meaning of isolation is "to separate samples with certain characteristics from other sample data". The characteristics have two meanings: (1) the proportion of the sample in the entire data is very small; (2) the sample has a value that is significantly different from the surrounding sample data. For the problem of anomaly target detection in remote sensing images, the concept of isolation is also applicable, because abnormal pixels often have the characteristics of "few and different", so they are more likely to be isolated. When some pixels in the image have shorter path lengths in the isolated forest composed of several isolated trees, these pixels are abnormal.
[0050] S104, statistics of the abnormal degree value of the water body pixel to be identified is performed to determine the mean value and standard deviation of the abnormal degree.
[0051] S105, the pixels of the image of the water body to be identified are classified according to the mean value and standard deviation of the abnormal degree; the classification result includes: normal water body pixel or abnormal water body pixel.
[0052] S105 specifically includes:
[0053] The minimum value of the range of the normal water body pixel is determined by the difference between the mean value and the standard deviation, and the maximum value of the range of the normal water body pixel is determined by the sum of the mean value and the standard deviation.
[0054] The pixel in the image of the water body to be identified, whose abnormal degree value is not in the range of the normal water body pixel, is determined as an abnormal water body pixel.
[0055] S106, the corresponding water body abnormal type is determined according to the NDVI of each abnormal water body pixel.
[0056] Figure 2 The NDVI value range of the three types of abnormal water bodies is shown. Figure 2 b is the normalized difference vegetation index (NDVI) box plot of the three types of abnormal water bodies and the normal water body. As can be seen from the figure, the high reflectivity of Enteromorpha in the near-infrared band is quite different from the normal water body and the other two abnormal events, and the reflectivity of the oil spill is lower than that of the normal water body and the other two abnormal water bodies. The NDVI of the three types of abnormal water bodies is obviously different. On the basis of determining the abnormal pixel, we identify the type of the abnormal pixel by NDVI. Table 1 summarizes the NDVI value range (mean ± 2 times standard deviation) of the three types of abnormal water bodies. The NDVI value of the abnormal water body pixel to be determined is compared with the value range of the three types of abnormalities in Table 1, and different water body abnormal types correspond to the corresponding interval. After the above judgment, some abnormal water body pixels may not be classified into any abnormal type, and a part of the misclassification may exist. That is, the decision rule shown in S107 is performed.
[0057] Table 1
[0058]
[0059] S107, if there are a set percentage of abnormal pixels in all abnormal water body pixels that do not belong to any type of water body abnormality, the corresponding abnormal water body pixel is re-determined as a normal water body pixel; the set percentage is 90%.
[0060] S108, the water body abnormal type with the largest proportion in all abnormal water body pixels is determined as the water body abnormal type of the water body to be identified.
[0061] To illustrate the effectiveness of the proposed method, three types of algorithms are used to extract the abnormal water body information simultaneously: 1) abnormal water body information extraction based on traditional supervised classification method (support vector machine); 2) abnormal water body information extraction based on supervised isolated forest (S-IForest); 3) abnormal water body information extraction based on unsupervised isolated forest (U-IForest) and decision rules. In addition, the influence of different outlier determination methods and whether to use band compression method on the recognition accuracy is further compared in the unsupervised isolated forest. Among them, the outlier determination method uses two methods of quartile range method and standard deviation method. In the quartile range method, the outliers are defined as values less than QL-n times IQR or greater than QU+n times IQR, n generally takes 1.5, and n=1 and n=1.5 are tested. QL is the lower quartile, which means that one-fourth of the data values are less than it. QU is the upper quartile, which means that one-fourth of the data values are greater than it. IQR is the interquartile range, which is the difference between the upper quartile QU and the lower quartile QL, and contains half of all observations. The standard deviation method determines the outliers by calculating the mean and standard deviation of the variable. The standard deviation itself can reflect the dispersion degree of the variable. In the outlier processing process, the distance between the variable value and the mean value can be measured by the mean±n times the standard deviation, n generally takes 2, and n=1 and n=2 are tested. Therefore, the abnormal water body information extraction based on unsupervised isolated forest (U-IForest) includes four methods: U-IForest-IQR, U-IForest-1.5IQR, U-IForest-SD, and U-IForest-2SD, which correspond to 1 times the quartile range method, 1.5 times the quartile range method, 1 times the standard deviation method, and 2 times the standard deviation method, respectively. Among them, U-IForest-SD is the method proposed in the present application. The above algorithms are all based on principal component analysis of the original data, and the first three principal components are selected for processing. In order to compare the influence of using and not using band compression on the accuracy of the method, the effect of using U-IForest-SD algorithm to identify abnormal water body by using all original band data is tested, and this method is recorded as UPCA-U-IForest-SD.
[0062] The basic flow of the experiment is as follows Figure 3The flowchart of the proposed method is shown in Fig. 1. It mainly includes 6 steps: 1) Preparation and pre-processing of case data; 2) Preparation of sample data; 3) Abnormal identification based on traditional supervised classification methods; 4) Abnormal water body information extraction based on supervised Isolation Forest (S-IForest); 5) Abnormal water body information extraction based on unsupervised Isolation Forest (U-IForest) and decision rules, including U-IForest-IQR, U-IForest-1.5IQR, U-IForest-SD, U-IForest-2SD; 6) Precision evaluation and comparison of different detection methods using confusion matrix. The pre-processing mainly includes atmospheric correction and band compression based on principal component analysis, and the first three bands are selected for subsequent processing. The preparation of sample data includes two parts: the preparation of training sample data for steps 3 and 4; the preparation of validation sample data for step 6. The training and validation sample data are obtained by visual interpretation based on previous research results, covering three types of abnormal water bodies and normal water bodies.
[0063] In addition, in order to illustrate the necessity of combining U-IForest and decision rules, the effects of U-IForest-SD and only using decision rules to identify three types of abnormal water bodies are compared. At the same time, considering that previous studies often only test the remote sensing images of water bodies with abnormalities, this is an abnormal detection for water bodies that have already appeared abnormalities. However, in real-time water body monitoring, most of the time we are faced with normal water bodies, and whether the developed algorithm will falsely report abnormalities is also a problem to be considered. Therefore, based on the case of Qingdao enteromorpha, a normal water body image of the area after the enteromorpha ended was selected to test the U-IForest-SD method.
[0064] Figure 4The results of three types of abnormal water body detection obtained by different methods are shown, and the detection accuracy statistics of different methods are shown in Table 2. As can be seen from the table, the method (U-IForest-SD) proposed in the application has lower accuracy than SVM, but is better than other methods. The overall recognition accuracy of U-IForest-SD method for green algae, black and odorous water and oil spill is 0.90, 0.96 and 0.94 respectively, and the Kappa coefficient is 0.80, 0.90 and 0.85 respectively. The threshold selection method has a great influence on the final accuracy, and the accuracy of the standard deviation threshold determination method is higher than that of the four quantile distance threshold determination method. The selection of n in the standard deviation and the four quantile distance has a great influence on the result, and the increase of n will lead to the increase of the error of missed classification, and the producer accuracy will be reduced, which is particularly obvious in the identification of black and odorous water and oil spill. When using 1.5 times IQR, the producer accuracy of black and odorous water is only 0.04, and when using 2 times standard deviation, the producer accuracy of black and odorous water is only 0.08, resulting in a very low kappa coefficient. By comparing the results of U-IForest-SD and UPCA-U-IForest-SD, it can be seen that there is no difference between the two in the identification of green algae, and the overall accuracy is 0.9, and the kappa coefficient is 0.80; the overall accuracy and Kappa coefficient of UPCA-U-IForest-SD are slightly higher than those of U-IForest-SD in black and odorous water, while the overall accuracy of U-IForest-SD is 6% higher than that of UPCA-U-IForest-SD in oil spill identification, and the Kappa coefficient is 0.15 higher. In addition, in the case study, the time consumed by using the original 12 bands of sentinel data for anomaly detection is about 5 times higher than that of using band compression for anomaly detection. In summary, it is a better choice to compress the bands before processing.
[0065] Table 2
[0066]
[0067]
[0068] Figure 5 The results of three types of abnormal water body detection obtained by only using decision rules are shown, and from the figure it can be seen that if only decision rules are used, a large number of mixed classification will occur, that is, false alarm, and the same water area will be identified as many different types of abnormal water bodies. However, the method of the application can effectively avoid the occurrence of false alarm and false alarm. Figure 6 The abnormal water bodies identified by U-IForest based on the image of normal water bodies are shown. 6.98% of the normal water bodies are wrongly identified as abnormal water bodies, but after combining the decision rules of abnormal types, all the abnormal water bodies are corrected to normal water bodies, avoiding the occurrence of false alarm. As can be seen, the method of U-IForest+ decision rules proposed in the application can effectively avoid the occurrence of false alarm and false alarm.
[0069] As another specific embodiment, the application also provides a water body anomaly identification system, comprising:
[0070] An image acquisition module for acquiring an image of a water body to be identified.
[0071] A preprocessing module for preprocessing the image of the water body to be identified; the preprocessing includes atmospheric correction, geometric correction, water body mask making, and NDVI calculation.
[0072] An anomaly degree value determination module for performing unsupervised isolation forest processing on the preprocessed image to determine an anomaly degree value of a water body pixel to be identified.
[0073] A mean value and standard deviation determination module for performing statistics on the anomaly degree value of the water body pixel to be identified to determine a mean value and a standard deviation of the anomaly degree.
[0074] A pixel classification module for classifying pixels of the image of the water body to be identified according to the mean value and the standard deviation of the anomaly degree; the classification result includes a normal water body pixel or an abnormal water body pixel.
[0075] A pixel type determination module for determining a water body anomaly type corresponding to each abnormal water body pixel according to the NDVI of the abnormal water body pixel.
[0076] A first decision module for re-determining a corresponding abnormal water body pixel as a normal water body pixel if a set percentage of abnormal pixels in all abnormal water body pixels do not belong to any water body anomaly type.
[0077] A second decision module for determining a water body anomaly type with the largest proportion in all abnormal water body pixels as a water body anomaly type of the water body to be identified.
[0078] In order to implement the method corresponding to the above-mentioned embodiment one to achieve the corresponding functions and technical effects, the application also provides a water body anomaly identification system, comprising: at least one processor, at least one memory, and computer program instructions stored in the memory, which, when executed by the processor, implement the water body anomaly identification method.
[0079] The application has the following advantages:
[0080] Firstly, the entire algorithm only inputs single-phase data, while the traditional method based on change detection to discover and identify anomalies often requires multi-phase data, has higher requirements for data quality, more processing steps, and more cumbersome, which reduces the instantaneity of services to some extent.
[0081] Secondly, unsupervised does not need training samples, which improves the portability of the algorithm. Supervised classification needs to establish a sample set, which is often based on specific cases and specific data sets in a specific area. Such sample data is difficult to reuse in different areas and different remote sensing data. For example, a sample set based on 7 multispectral band data of Landsat-8 cannot be well applied to Sentinel-2 data with 12 bands, and if a large number of different satellite data are introduced to make samples, a large amount of manpower, material resources, time and funds need to be invested.
[0082] Thirdly, when formulating the decision rule, the differences in the settings of different sensors are taken into account. Even for the same type of band (such as the red band), the specific wavelength range of the band is not completely consistent, but most sensors have near-infrared and red bands, which can enhance the comparability between different sensors after conversion into NDVI, so that the decision rule is more universal and stable.
[0083] Finally, in real-time water body anomaly monitoring, it is unknown whether an anomaly occurs (the location of the anomaly) and the type of the anomaly (Enteromorpha, black and odorous water and oil spill), but based on past experience, the type of water body anomaly event that may occur in a specific water area can be predicted under certain circumstances. Therefore, there are two business scenarios: 1) for a specific area with prior knowledge, such as Enteromorpha in Qingdao, which occurs from time to time in recent years, the specified type (Enteromorpha) of abnormal water body monitoring can be carried out in the specific water area; 2) for an area without prior knowledge, various types of abnormal water bodies (specifically Enteromorpha, black and odorous water and oil spill in the present application) need to be monitored. No matter which business scenario the algorithm is suitable for, the entire algorithm does not require human intervention. After the monitoring area is determined, as long as the image of the monitoring area is obtained, the process can be processed according to the process designed in the present application, which improves the degree of automation.
[0084] The embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. 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 are described in the method part.
[0085] In the present application, specific examples are used to illustrate the principles and implementation methods of the present application. The above embodiment description is only used to help understand the method and core idea of the present application; at the same time, for those skilled in the art, according to the idea of the present application, the specific implementation method and application range will be changed. In view of the above, the content of the specification should not be understood as a limitation of the present application.
Claims
1. A water body anomaly identification method, characterized in that, The method comprises the following steps: acquiring an image of a water body to be identified; preprocessing the image of the water body to be identified; the preprocessing comprises atmospheric correction, geometric correction, water body mask making, and NDVI calculation; performing unsupervised isolated forest processing on the preprocessed image to determine an abnormality degree value of a water body pixel to be identified; statistically determining the abnormality degree value of the water body pixel to be identified to determine a mean value and a standard deviation of the abnormality degree; classifying the pixels of the image of the water body to be identified according to the mean value and the standard deviation of the abnormality degree; the classification result comprises a normal water body pixel or an abnormal water body pixel; determining a corresponding water body abnormality type according to the NDVI of each abnormal water body pixel; if a set percentage of abnormal pixels in all abnormal water body pixels do not belong to any water body abnormality type, re-determining the corresponding abnormal water body pixel as a normal water body pixel; determining the water body abnormality type with the largest proportion in all abnormal water body pixels as the water body abnormality type of the water body to be identified.
2. The method according to claim 1, characterized in that, The step of classifying the pixels of the image of the water body to be identified according to the mean value and the standard deviation of the abnormality degree comprises the following steps: determining the minimum value of the range of the normal water body pixel by using the difference between the mean value and the standard deviation, and determining the maximum value of the range of the normal water body pixel by using the sum of the mean value and the standard deviation; determining the abnormal water body pixel by using the pixels in the image of the water body to be identified whose abnormality degree value is not in the range of the normal water body pixel.
3. The method of claim 1, wherein, The set percentage is 90%.
4. A water body anomaly identification system characterized by, The method comprises the following steps: an image acquisition module for acquiring an image of a water body to be identified; a preprocessing module for preprocessing the image of the water body to be identified; the preprocessing comprises atmospheric correction, geometric correction, water body mask making, and NDVI calculation; an abnormality degree value determination module for performing unsupervised isolated forest processing on the preprocessed image to determine an abnormality degree value of a water body pixel to be identified; a mean value and a standard deviation determination module for statistically determining the abnormality degree value of the water body pixel to be identified to determine a mean value and a standard deviation of the abnormality degree; a pixel classification module for classifying the pixels of the image of the water body to be identified according to the mean value and the standard deviation of the abnormality degree; the classification result comprises a normal water body pixel or an abnormal water body pixel; a pixel type determination module for determining a corresponding water body abnormality type according to the NDVI of each abnormal water body pixel; a first decision module for re-determining the corresponding abnormal water body pixel as a normal water body pixel if a set percentage of abnormal pixels in all abnormal water body pixels do not belong to any water body abnormality type; a second decision module for determining the water body abnormality type with the largest proportion in all abnormal water body pixels as the water body abnormality type of the water body to be identified.
5. A water body anomaly identification system, characterized by, The method comprises the following steps: at least one processor, at least one memory, and computer program instructions stored in the memory, when the computer program instructions are executed by the processor, realizing a water body abnormality identification method according to any one of claims 1-3.