Inland aquaculture pond water surface recognition method based on high-resolution remote sensing image data

Abstract

The present invention relates to an inland aquaculture pond water surface identification method based on high-resolution remote sensing image data, a normalized water body index is calculated based on satellite remote sensing image data, an inland water body is extracted using a bimodal threshold method, and the aquaculture pond water surface is identified by combining the geometric characteristics and phenological characteristics of the aquaculture pond. The method of the present invention can achieve high-precision identification of aquaculture ponds using high-resolution remote sensing image data, and the method has certain universality, and has achieved good results in the Taihu Lake Basin and the Luoma Lake Basin, which is helpful to achieve rapid and accurate extraction of the spatial distribution of inland aquaculture ponds, and can provide a data basis for grasping the situation of regional agricultural production activities, and provide a scientific basis and theoretical basis for promoting the sustainable development of the ecological environment.

Description

Technical Field

[0001] The invention belongs to the field of remote sensing technology and relates to a method for identifying water surfaces of inland aquaculture ponds based on high-resolution remote sensing image data. Background Art

[0002] Wetlands are an important component of terrestrial ecosystems, with one-third of the water surface consisting of an area less than 0.01 km 2 Small pond composition [1] Despite their large number and proportion, ponds have received far less attention than large lakes, reservoirs, and rivers in ecosystem and regional water quality research. [2] Moreover, as the main production method of inland freshwater aquaculture, pit ponds are an important part of my country's agricultural production and have become the main growth point of the current rural economy. [3] In recent years, the water surface of ponds has expanded rapidly, and the impact on the surrounding ecological environment has gradually become prominent, becoming an important source of agricultural pollution in my country. [4,5] Therefore, aquaculture ponds, which have been neglected in traditional research, have become an important part of regional economic development and ecological and environmental changes. Rapid and accurate extraction of the spatial distribution of inland aquaculture ponds can provide a scientific basis and theoretical foundation for understanding regional agricultural production activities and promoting sustainable development of the ecological environment.

[0003] The research on extracting water bodies using remote sensing technology started early. Currently, the most commonly used methods for automatic extraction of water body information include pixel-based threshold method and target-based classification method. [6] The threshold method is mainly based on the strong absorption characteristics of water bodies in the near-infrared or mid-infrared bands. It uses single-band and multi-band remote sensing indexes to extract water bodies. Commonly used indexes include the normalized water index NDWI [7] , Modified Normalized Difference Water Index MNDWI (ModifiedNormalizedDifferenceWater Index) [8] .

[0004] Since there are a large number of mixed pixels in medium and low resolution images, it is difficult to accurately classify small water bodies using the threshold method. Therefore, a classification method that uses the rich spatial and texture information of medium and high resolution images to extract small water bodies is proposed. [6,9] For example, Jiang Hao et al.

[10] A method for extracting small rivers based on linear features was established using TM (Thematic Mapper) images; Gasnier et al.

[11] By combining the prior database and SAR (Synthetic Aperture Radar), an automatic extraction method for small rivers was established. The extraction method for inland aquaculture ponds is currently mainly extracted through manual visual interpretation sampling and object-oriented classification methods, such as Luo Ling et al.

[12] , using e Cognition software and object-oriented classification methods to extract the changing characteristics of the water storage area and quantity of lakes and reservoirs / ponds in the western Songnen Plain in 1990, 2000 and 2013, Han Pengpeng et al.

[13] Based on the object-oriented classification method, multi-scale segmentation is performed, and spectral, spatial, texture and other features are used, combined with expert knowledge to extract pond information. Research on automatic extraction of aquaculture water surfaces mainly focuses on coastal aquaculture areas. For example, Duan et al.

[14] Combining spectra with shape rules, using Google Earth Engine to detect spatiotemporal changes in large aquaculture pond areas; Ottinger et al.

[15] A large-scale assessment of coastal aquaculture areas was conducted by combining spectral information with DEM terrain information; Wu Yitian et al.

[16] The offshore aquaculture areas were automatically extracted by comprehensively utilizing spectral and texture feature information, with an overall accuracy of over 93%.

[0005] At present, the research on automatic extraction of small water bodies by remote sensing is mainly focused on water bodies with slender characteristics such as rivers. Inland aquaculture ponds are often surrounded by rivers, lakes, paddy fields, ditches and other features with similar spectral and geometric texture expressions, resulting in the extraction of inland aquaculture ponds still relying on manual visual interpretation; at the same time, the coastal aquaculture areas are large in scale, concentrated in distribution, and have single spectral and texture features, making their remote sensing extraction methods difficult to apply to smaller and scattered inland aquaculture ponds.

[0006] References:

[0007] [1]VERPOORTER C,KUTSER T,SEEKELL DA,et al.A global inventory of lakes based on high-resolution satellite imagery[J].Geophys Res Lett,2014,41(18):6396-402.

[0008] [2]SCHMADEL NM,HARVEY JW,SCHWARZ GE,et al.Small Ponds in HeadwaterCatchments Are a Dominant Influence on Regional Nutrient and Sediment Budgets[J].Geophys Res Lett,2019,46(16):9669-77.

[0009] [3] Huang Senwen, Wei Chunzhu. Changes in artificial aquaculture bases in coastal cities of Guangdong Province: An empirical study based on Sentinel-1 data from 2015 to 2019[J]. Tropical Geography, 2021, 41(03): 622-34.

[0010] [4] Peng Lingyun, Chuan Chaopu, Li Hengpeng, et al. Estimation of pollution emissions from pond aquaculture in Taihu Lake Basin and its spatial distribution characteristics[J]. Journal of Lake Science, 2020, 32(01): 70-8.

[0011] [5]BROOKS BW,CONKLE J L.Commentary:Perspectives on aquaculture,urbanization and water quality[J].Comp BiochemPhys C, 2019,217(1-4.

[0012] [6] Wang Hang, Qin Fen. A review of water extraction from remote sensing images[J]. Surveying and Mapping Science, 2018, 43(05): 23-32.

[0013] [7]MCFEETERS S K.The use of the normalized difference water index(NDWI) in the delineation of open water features[J].Int J Remote Sens,1996,17(7):1425-32.

[0014] [8] Xu Hanqiu. Research on extracting water body information using improved normalized difference water index (MNDWI) [J]. Journal of Remote Sensing, 2005, 05: 589-95.

[0015] [9] Li Dan, Wu Baosheng, Chen Bowei, et al. Research progress and prospects of water information extraction based on satellite remote sensing[J]. Journal of Tsinghua University (Science and Technology), 2020, 60(02): 147-61.

[0016]

[10] Jiang Hao, Feng Min, Xiao Tong, et al. Small river extraction method from TM remote sensing images based on linear feature enhancement[J]. Acta Geodaetica et Cartographica Sinica, 2014, 43(07):705-10.

[0017]

[11] GASNIER N,DENIS L,FJORTOFT R,et al.Narrow River Extraction FromSAR Images Using Exogenous Information[J].Ieee J-Stars,2021,14:5720-34.

[0018]

[12] Luo Ling, Mao Dehua, Wang Zongming, Zhang Bai, Ren Chunying, Jia Mingming. Dynamic changes and driving factors of lakes and reservoirs / ponds in the western Songnen Plain [J]. Transactions of the Chinese Society of Agricultural Engineering, 2015, 31(22): 285-291.

[0019]

[13] Han Pengpeng, Huang Jinliang, Li Rendong, Wang Lihui, Hu Yanxia, ​​Huang Wei. Remote sensing monitoring and dynamic analysis of ponds based on object-oriented approach[J]. Transactions of the Chinese Society of Agricultural Machinery, 2015, 46(01): 272-277.

[0020]

[14] DUAN YQ, LI

[0021]

[15] OTTINGER M, CLAUSS K, KUENZER C.Large-Scale Assessment of CoastalAquaculture Ponds with Sentinel-1Time Series Data[J].Remote Sens-Basel,2017,9(5):

[0022]

[16] Wu Yitian, Chen Fu, Ma Yong, et al. Research on automatic extraction method of offshore aquaculture areas based on Landsat8 data[J]. Remote Sensing of Land and Resources, 2018, 30(03):96-105. Summary of the invention

[0023] The purpose of the present invention is to provide a method for identifying the water surface of an inland aquaculture pond based on high-resolution remote sensing image data. To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0024] The inland aquaculture pond water surface identification method based on high-resolution remote sensing image data includes:

[0025] Obtain high-resolution remote sensing images of each season, calculate the normalized water index NDWI after radiometric calibration and atmospheric correction, use the bimodal threshold method to determine the threshold of NDWI, and extract water bodies in the images of each season; the high-resolution remote sensing images of each season cover images of aquaculture ponds in different agricultural production cycles;

[0026] After masking the lakes and reservoirs with the water extraction results, the water patches were screened using the preset area, and the patches with an area less than 11000m were eliminated. 2 of fine water bodies;

[0027] The shape index was calculated for the remaining water patches after removing the fragmented water bodies, and the water bodies with a shape index greater than 8.0 were identified as rivers and ditches and removed;

[0028] The NDWI values ​​of the remaining water patches after removing fragmented water bodies, rivers and ditches were counted in different seasons. When the annual mean value was ≥ 0.2, it was identified as aquaculture ponds.

[0029] As a preferred implementation, the high-resolution remote sensing image refers to multispectral image data with a resolution not exceeding 10 meters; if the resolution of some bands of the selected multispectral image data sensor exceeds 10 meters, a 10-meter resolution image is obtained by resampling.

[0030] As a preferred embodiment, the different agricultural production cycles include the periods of water storage in aquaculture ponds, stocking of aquatic seedlings, cultivation, fishing and fallow and water release.

[0031] As a preferred implementation, the cloud coverage of the entire image data of the high-resolution remote sensing image is less than 15%.

[0032] As a preferred implementation, when using the water body extraction results to mask lakes and reservoirs, the water body extraction results are masked using the reservoir and lake ranges obtained from the national land survey in the most recent year. Lakes and reservoirs do not change significantly in a short period of time, and more accurate results can be obtained by using the reservoir and lake ranges obtained from the national land survey in the most recent year.

[0033] As a preferred implementation, after using the water body extraction results to mask lakes and reservoirs, ponds are pre-screened by grayscale segmentation and then screened by area threshold.

[0034] As a preferred implementation, for ponds whose boundaries with adjacent ponds cannot be identified on remote sensing images, grayscale segmentation is used to pre-screen the ponds. Some aquaculture ponds use new aquaculture technologies, or aquaculture ponds opened by large fishery companies separate adjacent ponds with a film / net, or the water level rises during the rainy season, resulting in the inability to identify the boundaries of adjacent ponds on remote sensing images. In this case, there are large areas of contiguous ponds, and it is impossible to simply distinguish aquaculture ponds and lakes by area. NDWI values ​​can be used for pre-screening. The average depth of lakes / reservoirs is generally 2 meters or more, while the depth of aquaculture ponds is generally about 1 meter, and the water is turbid. Therefore, the water index value of lakes will be higher, and NDWI values ​​can be used for pre-screening.

[0035] As a preferred implementation, the pre-screening of ponds by grayscale segmentation refers to: using the calculated normalized water index NDWI value as the grayscale value to perform threshold segmentation to distinguish ponds from lakes and reservoirs.

[0036] As a preferred embodiment, the method further comprises verifying the accuracy of the finally identified aquaculture pond results.

[0037] As a preferred implementation, a confusion matrix is ​​used to compare and verify the recognition results through field survey verification, drone images of the corresponding period, and national land survey data. The confusion matrix includes six elements: overall classification accuracy, Kappa coefficient, misclassification error, omission error, mapping accuracy, and user accuracy.

[0038] As a preferred implementation, the method of determining the NDWI threshold using the bimodal threshold method is as follows: statistically calculate the NDWI results of images of each season, make a distribution histogram, use the bimodal threshold method to select the lowest valley between the two peaks as the threshold for image segmentation, identify pixels above the threshold as water bodies, assign a value of 1, and identify pixels below the threshold as land, assign a value of 0.

[0039] Breeding ponds, as the main production mode of inland freshwater aquaculture, are an important part of my country's agricultural production, and their area has increased rapidly in recent years. The remote sensing identification method of inland breeding pond water bodies of the present invention takes into account that breeding ponds, as typical agricultural land, have an agricultural production cycle of water storage, aquatic seedling release, cultivation, fishing and fallow release, so the phenological characteristics of significant spatiotemporal changes within the year, and utilizes the phenological characteristics to assist in the identification of breeding ponds. Based on high-resolution satellite remote sensing images, the normalized water body index NDWI is calculated, and the inland water body is extracted using the bimodal threshold method. The water surface of the breeding pond is extracted by screening through area, shape and phenological indicators, and the inland breeding ponds are extracted with high spatiotemporal accuracy. The algorithm performance has been verified by wetland surveys, drone images and national land survey data, and can be used for long-term and large-scale breeding pond identification, filling the neglect of breeding ponds in the remote sensing extraction research of small water bodies. It not only prepares the technical foundation for the future agricultural development trend in advance, but also provides a scientific basis and theoretical basis for the sustainable development of the ecological environment.

[0040] It should be understood that all combinations of the aforementioned concepts and the additional concepts described in more detail below can be considered as part of the inventive subject matter of the present disclosure as long as such concepts are not mutually inconsistent. In addition, all combinations of the claimed subject matter are considered as part of the inventive subject matter of the present disclosure.

[0041] The foregoing and other aspects, embodiments and features of the present invention can be more fully understood from the following description in conjunction with the accompanying drawings. Other additional aspects of the present invention, such as the features and / or beneficial effects of the exemplary embodiments, will be apparent from the following description or learned from the practice of the specific embodiments according to the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The drawings are not intended to be drawn to scale. In the drawings, each identical or nearly identical component shown in various figures may be represented by the same reference numeral. For clarity, not every component is labeled in every figure. Embodiments of various aspects of the present invention will now be described by way of example and with reference to the accompanying drawings, in which:

[0043] Figure 1 It is a technical flow chart of the remote sensing identification method of inland aquaculture ponds.

[0044] Figure 2 It is the estimation result and threshold selection of extraction index of different water bodies in Shezhu Town, Liyang City.

[0045] Figure 3 This is the result of water surface extraction in Shezhu Town from 2017 to 2018.

[0046] Figure 4 It is the extraction result from aquaculture ponds in Shezhu Town, Liyang City.

[0047] Figure 5 It is a verification point.

[0048] Figure 6 It is the application of remote sensing extraction method of aquaculture ponds in Siyang County, Suqian City.

[0049] The aforementioned Figure 1-6 In the example, each coordinate, symbol or other representation expressed in English is well known in the art and will not be described in detail in this example. DETAILED DESCRIPTION

[0050] In order to better understand the technical content of the present invention, specific embodiments are given and described as follows in conjunction with the accompanying drawings.

[0051] Various aspects of the present invention are described in this disclosure with reference to the accompanying drawings, in which many illustrative embodiments are shown. The embodiments of the present disclosure are not necessarily intended to include all aspects of the present invention. It should be understood that the various concepts and embodiments introduced above, as well as those described in more detail below, can be implemented in any of many ways, and this is because the concepts and embodiments disclosed by the present invention are not limited to any implementation. In addition, some aspects disclosed by the present invention can be used alone or in any appropriate combination with other aspects disclosed by the present invention.

[0052] Example 1

[0053] This embodiment takes Shezhu Town, Liyang City as an example to further describe the technical solution of the present invention.

[0054] Shezhu Town in Liyang City has now developed into China's largest shrimp farming town, with more than 65,000 mu of aquaculture water surface. With the continuous development of the aquaculture industry, Shezhu has gradually developed from a small-scale aquaculture to a large-scale, organized aquaculture industry. Therefore, Shezhu Town is an ideal area for applying the method of the present invention. The present invention is based on high-resolution satellite remote sensing image data, calculates the normalized water index, uses the bimodal threshold method to extract inland water bodies, combines the geometric characteristics and phenological characteristics of aquaculture ponds, identifies the aquaculture pond water surface, and finally applies and evaluates it in typical aquaculture areas in the Taihu Lake Basin and Luomahu Lake Basin, confirming the accuracy of the method and its wide applicability.

[0055] As an exemplary description, the implementation of the above method is specifically described below in conjunction with the accompanying drawings.

[0056] Step 1: Based on the high-resolution remote sensing images of different seasons, after atmospheric correction and format conversion, the normalized difference water index NDWI is calculated to obtain the spatial distribution of NDWI in different seasons.

[0057] High-resolution remote sensing images refer to multispectral image data with a resolution not exceeding 10 meters. If the resolution of some bands of the sensor exceeds 10 meters, 10-meter resolution images can be obtained by resampling. This embodiment uses the most widely used optical image data, Sentinel-2MSI, with a revisit period of 5 days and a spatial resolution of up to 10 meters, which can meet the requirements for high-resolution and fine-grained extraction of regional pond water bodies at high temporal and spatial resolution. Table 1 shows the bandwidth and resolution of each band of MSI. Since the resolution of the GREEN band and NIR band required by NDWI is 10 meters, no resampling is required. MSI image data was downloaded from the European Space Agency https: / / scihub.copernicus.eu / dhus / # / home. Images with cloud cover less than 15% in spring (March-May), summer (June-August), autumn (September-November) and winter (December-February) from 2017 to 2018 were selected, and 8 images were finally obtained, including 2017 / 02 / 28, 2017 / 05 / 29, 2017 / 08 / 27, 2017 / 12 / 18, 2018 / 01 / 09, 2018 / 04 / 19, 2018 / 07 / 13 and 2018 / 12 / 18. The images cover the agricultural production cycle of aquaculture ponds from water storage preparation, aquatic seedlings, cultivation, fishing and fallow release. The preprocessing of MSI data first uses the plug-in Sen2Cor-2.4.0 in the SNAP software to perform radiation calibration and atmospheric correction on the L1C level data to obtain the atmospherically corrected data - L2A level data. Then, the L2A level data obtained after preprocessing is converted into the envi (.dat) format for subsequent processing.

[0058] Table 1 Band settings of Sentinel-2MSI

[0059] Serial number Band Wavewidth(nm) Resolution(m) describe 1 Coastaeroso 433-453 60 Coastal / Aerosol Band 2 Blue 458-523 10 Visible light band 3 Green 543-578 10 Visible light band 4 Red 650-680 10 Visible light band 5 VegetationRedEdge 698-713 20 Band within the red edge range 6 VegetationRedEdge 733-748 20 Band within the red edge range 7 NIR 773-793 10 Near infrared band (wide) 8 NarrowNIR 785-900 20 Near infrared band (narrow) 9 Watervapour 935-955 60 Water vapor band 10 SWIR–Cirrus 1360-1390 60 Shortwave infrared band 11 SWIR 1565-1655 20 Shortwave infrared band 12 SWIR 2100-2280 20 Shortwave infrared band

[0060] At present, some high-resolution sensors only have visible light to near-infrared bands. Because the electron wave energy of short-wave infrared and longer bands is low, it is difficult for high-resolution sensors to capture, so some high-resolution sensors do not include bands outside the near-infrared. Although the bands of MSI serial numbers 9-12 belong to bands outside the near-infrared range, spectral information can only be obtained by reducing the resolution or increasing the bandwidth (Table 1). At present, most water body extraction indexes mainly obtain water body information through bands other than red light. In order to adapt to the characteristics of high-resolution sensors, the present invention uses the normalized water body index NDWI to extract water bodies. This index only requires green light and near-infrared bands to obtain, and the calculation method is:

[0061] NDWI=(GREEN-NIR) / (GREEN+NIR) (1)

[0062] Where GREEN is the green light band; NIR is the near infrared band.

[0063] In order to further illustrate that the NDWI index is the preferred embodiment of the present invention, this example adds two commonly used water body indices, AWEI and MNDWI, for comparison. The results are as follows: Figure 2 As shown in the frequency diagram of pixel distribution, it can be seen that the peak of water body of NWDI is more obvious among the three water body indices, which is conducive to the identification of small water body pixels.

[0064] Step 2: Use the bimodal threshold method to determine the threshold of NDWI and extract the water surface distribution in each season.

[0065] Based on the NDWI calculation results of each season, the distribution of NDWI is statistically analyzed to make a distribution histogram. The threshold for water body extraction is determined by the adaptive bimodal threshold method. Pixels above the threshold are identified as water bodies and assigned a value of 1, and pixels below the threshold are identified as land and assigned a value of 0. The water body extraction results for each scene are shown in the figure below. Figure 3 As shown, the water area of ​​Shezhu Town can reach 60km 2 , the water body is mainly distributed in the southwest of Shezhu Town.

[0066] Step 3: After masking the lakes and reservoirs, use the area to eliminate small water bodies, use the shape index to eliminate rivers and ditches, and use the phenological characteristics to screen out the aquaculture pond water surface.

[0067] Since lakes and reservoirs do not change significantly in a short period of time, the ranges of reservoirs and lakes obtained from the national land survey in recent years are used to mask the water body extraction results. The third national land survey was launched in 2017, and the results of various places were summarized and released in 2018. Therefore, the results of the third national land survey are used to mask the lakes and reservoirs in Shezhu Town.

[0068] After using the water body extraction results to mask lakes and reservoirs, aquaculture ponds were screened through area, shape and phenological indicators.

[0069] In this embodiment, the aquaculture ponds are not contiguous, so the first step of screening can be performed directly by the area threshold. In the case of contiguous ponds or simultaneous identification of multiple ponds, grayscale segmentation can be used for pre-screening to avoid unidentified contiguous ponds.

[0070] (1) Use the area index to eliminate small water bodies such as puddles and reservoirs, and calculate the area of ​​each patch. When the area is less than 11,000 m 2 When it is detected, it is identified as a fine water body;

[0071] (2) Calculate the shape index of the remaining patches: Where C represents the perimeter and S represents the area. When the shape index is greater than 8.00, it is identified as a river or ditch.

[0072] (3) Phenological characteristics refer to the changes in aquaculture ponds during the agricultural production cycle of water storage, aquatic seedlings, cultivation, fishing, and fallow water release. The NDWI of the remaining water patches in different seasons is counted. When the annual mean value is ≥-0.2, it is identified as an aquaculture pond. The extraction results are as follows Figure 5 As shown in Figure 1, the ponds in Shezhu Town are mainly distributed in the southwest of Shezhu Town. From 2017 to 2018, the government implemented the policy of “returning ponds to farmland”, and the area of ​​ponds showed a shrinking trend, from 41.29 km in 2017 to 2 Reached 34.10km in 2018 2 .

[0073] The extraction results were verified using field survey data, drone images, and the third national land survey data to evaluate the extraction accuracy. The distribution of verification points is as follows: Figure 6 As shown in the figure, there are 247 data points for field survey, and the data of the third national land survey is used as an auxiliary for verification. The verification method uses a confusion matrix, including six elements: overall classification accuracy, Kappa coefficient, misclassification error, omission error, mapping accuracy, and user accuracy. The verification results are shown in Table 1, and the overall accuracy reaches 88.98%. The Kappa coefficient reaches 0.7688. It shows that the method of the present invention has high accuracy, can realize the accurate extraction of aquaculture ponds in different years, and has good scalability on a time scale.

[0074] Table 2 Remote sensing extraction accuracy of aquaculture ponds in Shezhu Town, Taihu Lake Basin

[0075] index Overall classification accuracy Kappa coefficient Misclassification error Missing error Cartographic accuracy User Precision Accuracy 88.98％ 0.7688 11.63％ 16.41％ 83.59％ 88.37％

[0076] In order to further evaluate the performance of the method of the present invention, Siyang County, a typical aquaculture area around Luoma Lake, was selected to evaluate the method. Based on Sentinel-2MSI remote sensing images, image data with cloud cover less than 15% were screened. The extraction results are as follows: Figure 6 As shown, the verification accuracy reaches 81.22%.

[0077] Although the present invention has been disclosed as above with preferred embodiments, it is not intended to limit the present invention. A person skilled in the art of the present invention may make various modifications and alterations without departing from the spirit and scope of the present invention.

Claims

1. A method for identifying water surface of inland aquaculture ponds based on high-resolution remote sensing image data, characterized in that: include: Obtain high-resolution remote sensing images of each season, calculate the normalized water index NDWI after radiometric calibration and atmospheric correction, use the bimodal threshold method to determine the threshold of NDWI, and extract water bodies in the images of each season; the high-resolution remote sensing images of each season cover images of aquaculture ponds in different agricultural production cycles; After masking lakes and reservoirs with the water body extraction results, the water body patches were screened using the preset area threshold, and the patches with an area less than 11000 m were eliminated. 2 of fine water bodies; The shape index was calculated for the remaining water patches after removing the fragmented water bodies, and the water bodies with a shape index greater than 8.0 were identified as rivers and ditches and removed; The NDWI values ​​of the remaining water patches after removing fragmented water bodies, rivers and ditches were counted in different seasons. When the annual mean value was ≥ 0.2, it was identified as aquaculture ponds.

2. The method according to claim 1, characterized in that The high-resolution remote sensing image refers to multispectral image data with a resolution not exceeding 10 meters; if the resolution of some bands of the selected multispectral image data sensor exceeds 10 meters, a 10-meter resolution image is obtained by resampling.

3. The method according to claim 1, characterized in that The different agricultural production cycles include the periods of water storage in aquaculture ponds, stocking of aquatic seedlings, cultivation, fishing and fallow and water release.

4. The method according to claim 1, characterized in that The cloud coverage of the entire image data of the high-resolution remote sensing image is less than 15%.

5. The method according to claim 1, characterized in that When using the water body extraction results to mask lakes and reservoirs, the water body extraction results are masked using the reservoir and lake ranges obtained from the national land survey in recent years.

6. The method according to claim 1, characterized in that After using the water body extraction results to mask lakes and reservoirs, ponds are pre-screened through grayscale segmentation and then screened through area thresholds.

7. The method according to claim 6, characterized in that For ponds whose boundaries with adjacent ponds cannot be identified in remote sensing images, grayscale segmentation is used to pre-screen the ponds.

8. The method according to claim 6, characterized in that The pre-screening of ponds by grayscale segmentation refers to: using the calculated normalized water index NDWI value as the grayscale value to perform threshold segmentation to distinguish ponds from lakes and reservoirs.

9. The method according to claim 1, characterized in that: It also includes accuracy verification of the final identified aquaculture pond results.

10. The method according to claim 9, characterized in that The confusion matrix was used to verify the recognition results through field surveys, drone images of the corresponding period, and national land survey data.

Images