A dynamic monitoring method for monthly changes of surface water bodies

Abstract

The present invention relates to a method for dynamic monitoring of monthly changes of surface water bodies by integrating optical and SAR images, and belongs to the field of remote sensing application technology. The method includes four parts: (1) data preparation and preprocessing: a monthly continuous SAR data set is constructed through preprocessing such as monthly median synthesis and image filtering. (2) initial water body range extraction: the water body-vegetation element coupling index (AWVI) and water body automatic extraction rules are used to extract the water body range of the full coverage month of the optical image. (3) continuous extraction of water body change range: a multi-level rule is established to extract the monthly change range of water bodies by combining post-classification comparison and CVA analysis methods. (4) continuous inversion of monthly water body range: the initial range of water bodies is extracted by optical remote sensing and the change range is extracted by SAR, and the monthly dynamic monitoring of water bodies is realized through a continuous mapping strategy of bidirectional iterative inversion. The present invention provides technical support for the dynamic monitoring of surface water bodies in a continuous manner in time and space.

Description

Technical Field

[0001] The invention relates to a method for dynamically monitoring monthly changes of surface water bodies, belonging to the technical field of remote sensing geoscience applications. Technical Background

[0002] The change of surface water bodies is a complex pattern under the coupling of climate change and human disturbance. Climate factors not only affect the interannual evolution of water bodies, but also give surface water bodies periodic seasonal fluctuation characteristics. Exploring the periodic seasonal laws of surface water bodies is of great significance to regional water resources management, ecological dynamic balance and sustainable development.

[0003] There is still a lack of research on methods for dynamic monitoring of surface water bodies using remote sensing. The main technical difficulties currently include: (1) Water extraction from optical images is difficult to avoid the influence of uncertain weather conditions, which seriously restricts the construction of temporal stability of data required for dynamic monitoring projects. Especially during the rainy season when water bodies change drastically with seasons, the sparsity of optical data time series hinders the implementation of regular and routine dynamic monitoring. (2) SAR data has the advantages of all-weather and all-time observation capabilities, but due to the influence of low backscatter targets on the land surface, the accuracy of water extraction based on SAR images needs to be improved. Therefore, by utilizing the respective characteristics and advantages of optical and SAR images, a method for dynamic monitoring of monthly changes in surface water bodies is developed, which is conducive to carrying out spatiotemporally continuous dynamic monitoring of surface water bodies and providing spatial quantitative data for exploring the laws of seasonal changes in surface water bodies. Summary of the invention

[0004] The problem to be solved by the present invention is to overcome the restriction of continuous cloud and rain weather on the accurate dynamic monitoring of large-scale surface water bodies, and propose a method for dynamic monitoring of monthly changes of surface water bodies. The method makes full use of the characteristics and advantages of optical and SAR images, and can not only realize the routine and dynamic monitoring of surface water bodies, but also realize the accurate extraction of the spatiotemporal distribution of surface water bodies independently of the weather.

[0005] In order to solve the above technical problems, the present invention proposes a method for dynamically monitoring monthly changes of surface water bodies, comprising the following steps:

[0006] Step 1: Active and passive remote sensing data preparation: prepare a multi-source remote sensing data set in the area of ​​interest and construct a time series data set. The multi-source remote sensing data set includes: Sentinel-1 radar remote sensing images and Sentinel-2 optical remote sensing images;

[0007] Step 2: Data preprocessing: Perform median synthesis processing on the Sentinel-1 radar remote sensing images in each month in pixels to construct a monthly continuous time series SAR data set, and filter each image of the time series SAR data set to remove coherent speckle noise; Perform median synthesis processing on the Sentinel-2 optical remote sensing images with cloud coverage less than 20% in each month in pixels to obtain monthly optical images, and select optical images of several months with effective observation pixels fully covering the area of ​​interest;

[0008] Step 3: Initial optical extraction of water body range: Calculate the water body-vegetation element coupling index AWVI of each pixel in the optical image of the month selected in the second step, obtain the AWVI index characteristic images of several months, and extract the water body-vegetation element coupling index AWVI ≥ threshold τ S2 The pixels of the water bodies in several months are obtained;

[0009] Optical extraction post-processing: Use slope information (slope>20°) to mask uneven terrain areas and remove mountain shadow interference; use watershed relative elevation information (HAND>20m) to remove other shadow effects such as vegetation and structures; use the blue light band (surface reflectance>0.3) to mask the snowy surface to remove the impact of ice and snow noise; use the JRC GHSL global human settlement data from 1975 to 2014 to remove errors caused by building shadows and bright buildings in urban areas;

[0010] Step 4: SAR rough extraction of water range: For each image in the time series SAR data set, the Otsu method is used to calculate the water-land segmentation thresholds of multiple typical areas and average them to obtain the optimal water-land segmentation threshold τ S1 , the backscatter coefficient is less than the optimal land-water segmentation threshold τ S1 The pixel with the value of 1 is water body, and the other pixels are non-water body, and the value of 0 is assigned. The relative elevation information of the watershed (HAND>20m) is used to remove the shadow effects of vegetation, structures, etc.

[0011] Step 5: SAR image continuous change detection: Use the change vector analysis method to perform continuous change detection analysis on the time series SAR data set to obtain the time series SAR change intensity data set;

[0012] Step 6: Extract the monthly range of water body changes: Perform differential processing on the water body extraction results of step 4 to obtain the areas where water-land transformation occurs in adjacent months; extract the water body extraction results in step 5 with a change intensity greater than the threshold τ S1C The high-intensity change area is obtained by calculating the pixels of the water-land conversion area and superimposing the high-intensity change area to obtain the monthly change range of the water body.

[0013] Step 7: Iterative inversion of water body range: Using the monthly water body obtained in step 3 As the base period, combined with the monthly variation range of water bodies extracted in step 6 The water range w of phase i+1 is inverted using the following formula: i+1 :

[0014]

[0015] After obtaining the water body range of phase i+1, continue to use the monthly change range of water bodies extracted in step 6 The water range w of phase i+2 is obtained by inversion i+2 Similarly, the water range of i+3, i+4... continuous phases can be obtained by continuous iterative inversion;

[0016] Step 8. Bidirectional iteration of water body range: A bidirectional strategy is introduced to alleviate the error transmission effect of the continuous iteration process. The calculation expression is:

[0017]

[0018] In the formula, i and j are the two adjacent time phases of the initial extraction of optical water bodies, and n is the number of months between the two adjacent time phases i and j. is the range of water body change detected from the positive change point i, is the range of water body change obtained by reverse change detection from point j as the starting point, The water body range obtained by bidirectional inversion in period i+m;

[0019] Step 9, post-processing: By detecting the unchanged area of ​​the multi-period water body range in the third step, the stable water body range is obtained, and then this is used to fill in the water body leakage caused by the strong coherent speckle noise of urban buildings.

[0020] The main innovations and benefits of the present invention are as follows:

[0021] 1. In the second step, most existing SAR image filtering methods are based on fixed sliding windows, while the spatial patch units of surface water bodies are complex and changeable. Fixed window filtering is prone to lose the edge information of water patches, reducing the accuracy of surface water extraction. The present invention first uses an image segmentation method to aggregate, segment, and construct neighboring pixels with similar texture and backscattering intensity characteristics in SAR images into spatial continuous objects, and then performs filtering processing using the segmented objects as basic units, which can effectively eliminate coherent speckle noise in the image and retain the edge shape characteristics of the surface water body.

[0022] 2. In the seventh step, the existing water body dynamic monitoring methods face problems such as incomplete time series data or low monitoring accuracy in areas where optical data acquisition is difficult. The present invention effectively avoids the limitations of SAR and optical imaging for water body dynamic monitoring through decision-level fusion of active and passive remote sensing data, promotes the complementary advantages of SAR and optical imaging, and realizes large-scale accurate monitoring of surface water bodies.

[0023] 3. In the eighth step, the existing continuous change detection method does not take the error transmission effect into consideration, resulting in many accuracy problems. The present invention effectively alleviates the error transmission of the unidirectional continuous change detection process by establishing continuous change detection paths in both positive and negative directions, thereby improving the detection accuracy.

[0024] The continuous extraction and monthly mapping method of surface water bodies proposed in the present invention integrates the time-series Sentinel-1 and Sentinel-2 data, which can fully utilize the all-weather imaging capability of SAR data and the high accuracy of water extraction from optical data. It has good spatiotemporal extraction accuracy and can support dynamic monitoring of surface water bodies over a large range of scales. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a schematic diagram of the process of the present invention.

[0026] Figure 2 It is the monthly coverage of Sentinel-1 / 2 data in the middle and lower reaches of the Yangtze River Plain.

[0027] Figure 3 It is a water body variation range extraction process that combines the post-classification comparison method and CVA.

[0028] Figure 4 This is a schematic diagram of a water body dynamic monitoring method based on continuous change detection. DETAILED DESCRIPTION

[0029] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby making a clearer and more definite definition of the protection scope of the present invention.

[0030] The study area of ​​the present invention is the middle and lower reaches of the Yangtze River Plain, and the monitoring period is from January 2016 to December 2020. The flow chart of the large-scale dynamic monitoring of surface water bodies based on the fusion of optical and SAR images in this embodiment is as follows: Figure 1 As shown, the following steps are included:

[0031] The first step is to prepare active and passive remote sensing data. Google Earth Engine (GEE) stores the entire pre-processed Sentinel-1GRD and Sentinel-2 data, and allows users to directly access and call, so this embodiment carries out data processing and analysis based on the GEE platform. Sentinel-1 images are screened according to the scope of the middle and lower reaches of the Yangtze River Plain. The images involve six relative orbit numbers of Sentinel-1, namely 11, 113, 40, 142, 69, and 171. The Sentinel-2 images of the study area are screened based on the scope of the middle and lower reaches of the Yangtze River Plain and the cloud coverage rate is less than 20%, and the monthly images are obtained by the median synthesis method.

[0032] The second step is data preprocessing. The Sentinel-1 radar remote sensing images in each month are processed by median synthesis in pixels to construct a monthly continuous time series SAR data set. Taking into account the spatial autocorrelation characteristics of water elements to form continuous patches, a SAR image filtering method based on superpixel segmentation algorithm is used, which can reduce the coherent speckle noise of SAR images, retain the edge information of the patches, and also has the ability to process quickly on a large scale. SAR continuous monthly filtered images that take shape characteristics into account are obtained.

[0033] The monthly optical images are obtained by median synthesis of the Sentinel-2 optical remote sensing images with cloud coverage less than 20% in each month in pixel units. The coverage of the monthly Sentinel-2 images of the middle and lower reaches of the Yangtze River Plain is calculated. The results are as follows: Figure 2 The optical images of several months with effective observation pixels fully covering the area of ​​interest are screened out as follows: September 2016, July 2017, July 2018, August 2019, and August 2020.

[0034] The third step is the initial optical extraction of the water body range. Calculate the water body-vegetation element coupling index AWVI of each pixel in the optical image of the month selected in the second step, obtain the AWVI index characteristic images of several months, and extract the water body-vegetation element coupling index AWVI ≥ threshold τ S2 The pixels of the water bodies in several months are obtained, where the threshold τ S2 It is 0.006.

[0035] Optical extraction post-processing: Use slope information (slope>20°) to mask uneven terrain areas and remove mountain shadow interference; use watershed relative elevation information (HAND>20m) to remove other shadow effects such as vegetation and structures; use the blue light band (surface reflectance>0.3) to mask the snowy surface to remove the impact of ice and snow noise; use the JRC GHSL global human settlement data from 1975 to 2014 to remove errors caused by building shadows and bright buildings in urban areas.

[0036] The fourth step is SAR rough extraction of water body range. The surface of the water body is smooth and homogeneous, the noise level in the cross-polarization image is low, and it has a small intra-class variance and high water-land separability. Therefore, VH cross-polarization images are used to extract surface water body pixels. For each image in the time series SAR data set, the Otsu method is used to calculate the water-land segmentation thresholds of multiple typical areas and average them to obtain the optimal water-land segmentation threshold τ S1 , threshold τ S1 The monthly continuous time-series SAR data set obtained in the second step is then binarized to extract the rough monthly water range containing low-scattering targets on the land surface. That is, the backscatter coefficient is less than the optimal land-water segmentation threshold τ S1 The pixel with the value of 1 is water body, and the other pixels are non-water body, and the value of 0 is used. The relative elevation information of the watershed (HAND>20m) is used to remove the shadow effects of vegetation, structures, etc.

[0037] The fifth step is the continuous change detection of SAR images. The change vector analysis (CVA) method is used to perform change detection analysis on the SAR data sets of adjacent months. The calculation formula for the change intensity is:

[0038]

[0039] Where ||ΔG p ||2 is the change intensity of pixel p between the two time phases h and g. Through continuous change detection, the SAR monthly change intensity image from January 2016 to December 2020 can be obtained.

[0040] Step 6: Extract the monthly range of water body changes. Combine the post-classification comparison method and CVA change detection to establish the inversion water body change range extraction rules. The process is as follows: Figure 3 As shown in Figure 2, the constraints of the rule are: 1. The pixel undergoes water-land transformation in adjacent months; 2. The intensity of the change is greater than the threshold τ S1C , threshold τ S1C is 7dB.

[0041] Specifically, the water body extraction results in the fourth step are differentially processed to obtain the areas where water-land transformation occurs in adjacent months; the areas where the change intensity in the fifth step is greater than the threshold τ S1C The high-intensity change area is obtained by combining the water-land conversion area with the high-intensity change area to obtain the monthly change range of the water body. Using the results of the fourth and fifth steps, the monthly change range of the water body from January 2016 to December 2020 is inverted.

[0042] Step 7: Iterative inversion of water range. Using the results of optical water extraction in step 3 As the base period, combined with the monthly variation range of water bodies extracted in step 6 The water range w of phase i+1 is inverted using the following formula: i+1 :

[0043]

[0044] After obtaining the water range of phase i+1, continue to use the water extracted in step 6 The water range w of phase i+2 can be inverted i+2 Similarly, the water range of continuous phases i+3, i+4, etc. can be obtained by continuous iterative inversion. The technical process is as follows Figure 4 shown.

[0045] Step 8: Bidirectional iteration of water body range. The error transmission effect of the continuous iteration process will affect the water body inversion accuracy. The bidirectional strategy is introduced to alleviate the error. The calculation expression is:

[0046]

[0047] In the formula, i and j are two adjacent time phases of the initial extraction of optical water bodies, and n is the number of months between i and j. is the range of water body change detected from the positive change point i, is the range of water body change obtained by reverse change detection from point j as the starting point, The range of water bodies obtained by bidirectional inversion in the i+m period. Through the bidirectional iterative algorithm, the monthly range of water bodies in the middle and lower reaches of the Yangtze River Plain from January 2016 to December 2020 was preliminarily obtained.

[0048] Step 9: Post-processing of results: By detecting the unchanged area of ​​the multi-period water body range in the third step, the stable water body range is obtained, and then the water body leakage caused by the strong coherent speckle noise of urban buildings is supplemented.

[0049] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

[0050] In addition to the above embodiments, the present invention may also have other implementation modes. Any technical solution formed by equivalent replacement or equivalent transformation falls within the protection scope required by the present invention.

Claims

1. A method for dynamically monitoring monthly changes in surface water bodies, comprising the following steps: Step 1: Active and passive remote sensing data preparation: prepare a multi-source remote sensing data set in the area of ​​interest and construct a time series data set. The multi-source remote sensing data set includes: Sentinel-1 radar remote sensing images and Sentinel-2 optical remote sensing images; Step 2: Data preprocessing: Perform median synthesis processing on the Sentinel-1 radar remote sensing images in each month in pixels to construct a monthly continuous time series SAR data set, and filter each image of the time series SAR data set to remove coherent speckle noise; Perform median synthesis processing on the Sentinel-2 optical remote sensing images with cloud coverage less than 20% in each month in pixels to obtain monthly optical images, and select optical images of several months with effective observation pixels fully covering the area of ​​interest; Step 3: Initial optical extraction of water body range: Calculate the water body-vegetation element coupling index AWVI of each pixel in the optical image of the month selected in the second step, obtain the AWVI index characteristic images of several months, and extract the water body-vegetation element coupling index AWVI ≥ threshold τ S2 The pixels of the water bodies in several months are obtained; Step 4: SAR rough extraction of water range: For each image in the time series SAR data set, the Otsu method is used to calculate the water-land segmentation thresholds of multiple typical areas and average them to obtain the optimal water-land segmentation threshold τ S1 , the backscatter coefficient is less than the optimal land-water segmentation threshold τ S1 The pixel with is water body and is assigned a value of 1, and the other pixels are non-water bodies and are assigned a value of 0; Step 5: SAR image continuous change detection: Use the change vector analysis method to perform continuous change detection analysis on the time series SAR data set to obtain the time series SAR change intensity data set; Step 6: Extract the monthly range of water body changes: perform differential processing on the water body extraction results of step 4 to obtain the areas where water-land transformation occurs in adjacent months; extract the water body extraction results in step 5 with a change intensity greater than the threshold τ S1C The high-intensity change area is obtained by calculating the pixels of the water-land conversion area and superimposing the high-intensity change area to obtain the monthly change range of the water body. Step 7: Iterative inversion of water body range: Using the monthly water body obtained in step 3 As the base period, combined with the monthly variation range of water bodies extracted in step 6 The water range w of phase i+1 is inverted using the following formula: i+1 : After obtaining the water body range of phase i+1, continue to use the monthly change range of water bodies extracted in step 6 The water range w of phase i+2 is obtained by inversion i+2 Similarly, the water range of i+3, i+4... continuous phases can be obtained by continuous iterative inversion; Step 8: Bidirectional iteration of water body range: A bidirectional strategy is introduced to alleviate the error transmission effect of the continuous iteration process. The calculation expression is: In the formula, i and j are the two adjacent time phases of the initial extraction of optical water bodies, and n is the number of months between the two adjacent time phases i and j. is the range of water body change detected from the positive change point i, is the range of water body change obtained by reverse change detection from point j as the starting point, The water body range obtained by bidirectional inversion in period i+m; Step 9, post-processing: By detecting the unchanged area of ​​the multi-period water body range in the third step, the stable water body range is obtained, and then this is used to fill in the water body omissions caused by the strong coherent speckle noise of urban buildings.

2. The method for dynamic monitoring of monthly changes of surface water bodies according to claim 1, characterized in that: Sentinel-1 radar remote sensing data is ground range multi-view image GRD in interferometric wide-band mode IW, with a pixel size of 10×10m.

3. The method for dynamic monitoring of monthly changes of surface water bodies according to claim 1, characterized in that: SAR image filtering is object-level filtering. First, neighboring pixels with similar texture and backscattering intensity characteristics in the SAR image are clustered, segmented, and constructed into spatially continuous objects, and then filtering is performed using the segmented objects as basic units.

4. The method for dynamic monitoring of monthly changes of surface water bodies according to claim 1, characterized in that: Sentinel-2 optical remote sensing data refers to the surface reflectance product L2A that has been processed with orthorectification, geometric precision correction and atmospheric correction. The spatial resolution of its visible light and near-infrared bands is 10m, the spatial resolution of the red edge band and short-wave infrared band is 20m, and the spatial resolution of the aerosol and water vapor band is 60m.

5. The method for dynamic monitoring of monthly changes of surface water bodies according to claim 1, characterized in that: Threshold τ S1 , τ S2 , τ S1C The values ​​are -23dB, 0.006, and 7dB respectively.

6. The method for dynamic monitoring of monthly changes of surface water bodies according to claim 1, characterized in that: The calculation formula of water body-vegetation element coupling index AWVI is: AWVI=AWEI-EVI AWEI=p Blue +2.5×r Green -1.5×(ρ NIR +r SWIR1 )-0.25×ρ SWIR2 7. The method for dynamic monitoring of monthly changes of surface water bodies according to claim 1, characterized in that: The water-land transformation in adjacent months in the dual constraint conditions includes both the transformation of pixels from land to water caused by water body expansion and the transformation of pixels from water to land caused by water body retreat.

8. The method for dynamic monitoring of monthly changes of surface water bodies according to claim 1, characterized in that: Optical extraction post-processing was also performed in step three: the slope information slope>20° was used to mask the uneven terrain areas and remove the interference of mountain shadows; the watershed relative elevation information HAND>20m was used to remove the shadows of vegetation and structures; the surface reflectance of the blue light band>0.3 was used to mask the snowy surface to remove the influence of ice and snow noise; the JRC GHSL global human settlement data from 1975 to 2014 was used to remove the errors caused by building shadows and bright buildings in urban areas.

9. The method for dynamic monitoring of monthly changes of surface water bodies according to claim 1, characterized in that: Step 4 also performs the following processing: using the basin relative elevation information HAND>20m to remove the shadow effects of vegetation and structures.

Images