An intertidal zone space and terrain extraction method based on multi-source remote sensing data

Abstract

The application discloses a kind of intertidal zone space and terrain extraction method based on multi-source remote sensing data, comprising: obtaining the target area after pre-processing Landsat8 ground reflectance image and Sentinel-2 ground reflectance image by remote sensing cloud platform, Landsat8 ground reflectance image is calculated after improved spectral water body index, and then coefficient of variation algorithm is carried out variation detection, and the range of intertidal zone is obtained by threshold extraction;ICESat-2 laser radar data of the target area is obtained by laser radar platform, and after the data reliability screening and the overlap selection of the range of intertidal zone, it is combined with Sentinel-2 ground reflectance image, and the terrain information of intertidal zone is inverted by the random forest model of feature screening.The application solves the difficulty that the previous intertidal zone space and terrain extraction method needs to rely on a large number of tide data, and the precision is better.

Description

Technical Field

[0001] This invention relates to the field of remote sensing image processing technology for coastal zone research and application, and in particular to a method for extracting intertidal space and topography based on multi-source remote sensing data. Background Technology

[0002] The Jiangsu coastline faces unprecedented challenges to its ecological environment due to both natural factors and human economic activities. Monitoring the intertidal zone and understanding its changes is the most fundamental task for its rational utilization and sustainable development.

[0003] Previous methods for extracting intertidal space and topography have primarily employed methods such as extracting waterline data and assigning tidal level information for topographic inversion, or constructing empirical models using measured topographic information for topographic inversion. However, these methods suffer from several drawbacks in acquiring waterline data, tidal level information, simulated tidal level information, and measuring topography: a lack of accurate contour and spatial representation, a shortage of tidal gauge stations, asynchronous water level distribution in complex tidal zones, and difficulties in accessing the intertidal zone or requiring substantial human and material resources. These challenges may lead to a decrease in the accuracy of intertidal topographic inversion. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method for extracting intertidal space and topography based on multi-source remote sensing data, so as to solve the problem of reduced accuracy in intertidal topography extraction based on waterline inversion in the prior art.

[0005] To solve the above-mentioned technical problems, the present invention is implemented using the following solution:

[0006] This invention provides a method for extracting intertidal zone space and topography based on multi-source remote sensing data, including:

[0007] Preprocessed Landsat 8 and Sentinel-2 surface reflectance images of the study area were obtained through a remote sensing image processing cloud platform, and ICESat-2 lidar data of the same image area were obtained through a lidar data processing platform.

[0008] Cloud removal was performed on the acquired Landsat 8 and Sentinel-2 surface reflectance images.

[0009] The improved spectral water index was used to calculate the surface reflectance of Landsat 8 images after cloud removal.

[0010] The image after the improved spectral water index calculation was used to calculate the coefficient of variation algorithm and the intertidal spatial range was extracted using a pre-set threshold.

[0011] The credibility of the ICESat-2 lidar data was screened, and data that overlapped with the extracted intertidal spatial range was selected.

[0012] Topographic information of the intertidal zone was retrieved using selected ICESat-2 lidar data and Sentinel-2 surface reflectance images through a feature-filtered random forest model.

[0013] Furthermore, the remote sensing image processing cloud platform uses the GEE cloud platform, and the lidar data processing platform uses the Python-based PyCharm platform.

[0014] Furthermore, the cloud removal operation involves using the qa_pixel cloud quality control band of the maskL8 function in the remote sensing image processing cloud platform to remove clouds from the preprocessed Landsat8 surface reflectance image, and using the qa_60 cloud quality control band of the maskS2clouds function in the remote sensing image processing cloud platform to remove clouds from the preprocessed Sentinel-2 surface reflectance image.

[0015] Furthermore, it also includes: resampling the declouded Sentinel-2 surface reflectance image to a spatial resolution of 10m using bilinear interpolation.

[0016] Furthermore, the formula for calculating the improved spectral water index is as follows:

[0017] IWI=((ρb2+ρb3-ρb6-ρb7) / (ρb2+ρb3+ρb6+ρb7)) 2

[0018] In the formula, IWI is the improved spectral water index, and ρb2, ρb3, ρb6, and ρb7 are the reflectances of the blue band, green band, shortwave infrared 1 band, and shortwave infrared 2 band, respectively.

[0019] Furthermore, the preset threshold is 0.28.

[0020] Furthermore, the credibility screening was conducted using the credibility attributes provided by the ALT08 data and the intersection method was used to remove data outside the intertidal zone.

[0021] Furthermore, the credibility screening involves selecting ICESat-2 lidar data with an overall error of less than 0.2.

[0022] Furthermore, the feature bands added to the random forest model for feature selection are SDD, NDWI, SSC, and TP, and the calculation formulas are as follows:

[0023]

[0024]

[0025]

[0026]

[0027] In the formula, aerosols, red, blue, red, and nir represent the Sentinel-2 aerosol band, blue band, green band, and near-infrared band, respectively. data1 and b data2 This consists of data from 13 bands of Sentinel-2 surface reflectance imagery, as well as SDD, NDWI, and SSC data acquired at different times within the same band.

[0028] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0029] 1. This invention overcomes the limitations of previous intertidal zone extraction methods, such as the difficulty in obtaining waterline extraction accuracy and tidal level data, and the low accuracy caused by the IWI and CV algorithms.

[0030] 2. This invention does not require image screening when identifying the spatial extent of the intertidal zone; all images can be used. It is a method that starts from dynamic changes and captures the intertidal zone from long-sequence remote sensing image changes.

[0031] 3. This invention uses ICESat-2 lidar data to replace in-situ observation data, which greatly solves the difficulty of obtaining observation data.

[0032] 4. The random forest algorithm based on feature selection in this invention is used to invert intertidal terrain. It is applicable to terrain inversion over a large area. The added features after selection improve the inversion accuracy. Attached Figure Description

[0033] Figure 1 This is a schematic diagram of image CV variation detection according to the present invention;

[0034] Figure 2 This is a diagram showing the intertidal zone extraction results of this invention;

[0035] Figure 3 This is a diagram showing partial ICESat-2 data within the research area of ​​this invention;

[0036] Figure 4 These are images of a small area at different times, as presented in this invention.

[0037] Figure 5 This is a correlation analysis diagram showing the relationship between the model predictions of this invention and the spectral values ​​and observed values ​​of each band;

[0038] Figure 6 This is a ranking diagram of the feature contribution of the random forest model in this invention;

[0039] Figure 7 This is a verification diagram of the random forest model of this invention;

[0040] Figure 8 This is a diagram showing the verification results of the intertidal topography inversion of this invention;

[0041] Figure 9 This is a diagram showing the results of the extraction of intertidal space and topography in this invention. Detailed Implementation

[0042] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0043] This embodiment discloses a method for extracting intertidal spatial and topographic information based on multi-source remote sensing data. Taking the central coastal zone of Jiangsu, China as an example, the method extracts intertidal spatial and topographic information of Jiangsu in 2022, specifically including the following steps:

[0044] Step 1: Select the central coastal area of ​​Jiangsu Province on the remote sensing image processing cloud platform and lidar data processing platform, and acquire pre-processed satellite surface reflectance images and ICESat-2 lidar data.

[0045] The remote sensing image processing cloud platform used was Google Earth Engine (GEE), which selected Landsat 8 and Sentinel-2 surface reflectance images. Preprocessing was completed by the data producers, including image registration, radiometric correction, atmospheric correction of the Landsat 8 surface reflectance image using LaSRC (Land Surface Reflectance Code), and atmospheric correction of the Sentinel-2 surface reflectance image using sen2cor.

[0046] The lidar data processing platform uses the Python-based PyCharm platform. The ICESat-2 lidar data in the study area is obtained by filtering, and the ICESat-2 ALT08 product is used.

[0047] Step 2: Perform cloud removal on the acquired Landsat 8 and Sentinel-2 surface reflectance images.

[0048] The cloud removal operation involves using the qa_pixel cloud quality control band of the maskL8 function in GEE to remove clouds from the preprocessed Landsat 8 surface reflectance image, and using the qa_60 cloud quality control band of the maskS2clouds function in GEE to remove clouds from the preprocessed Sentinel-2 surface reflectance image.

[0049] Step 3: Calculate the surface reflectance of the Landsat 8 image after cloud removal using IWI (Improved Spectral Water Index).

[0050] The formula for calculating IWI is as follows:

[0051] IWI=((ρb2+ρb3-ρb6-ρb7) / (ρb2+ρb3+ρb6+ρb7)) 2

[0052] In the formula, ρb2, ρb3, ρb6, and ρb7 are the reflectivities of the blue band, green band, shortwave infrared 1 band, and shortwave infrared 2 band, respectively.

[0053] The IWI formula can better increase the inter-class variance of sediment-laden water bodies and silty mudflats, and is more adaptable to any tidal conditions and various types of coastlines. It is superior to NDWI (Normalized Difference Water Index), MNDWI (Modified Normalized Difference Water Index), RNDWI (Revised Normalized Different Water Index), AWEI (Automated Water Extraction Index), and EWI (Enhanced Water Index).

[0054] Step 4: Calculate the intertidal spatial range using the coefficient of variation algorithm on the image after the improved spectral water index calculation, and apply a pre-set threshold.

[0055] Using CV (coefficient of variation) for change detection makes it easier to extract areas of change in the image (see...). Figure 1 ), and through experiments, thresholds were set to extract areas with large variations, which were then identified as intertidal zones;

[0056] The formula for calculating CV is as follows:

[0057]

[0058] In the formula, sd and mean are the standard deviation and mean of a set of data, respectively;

[0059] In this embodiment, the threshold obtained through multiple experiments was 0.28, and the extracted intertidal region was as follows: Figure 2 As shown.

[0060] Step 5: Perform credibility screening on the ICESat-2 lidar data and select data that overlaps with the extracted intertidal spatial range.

[0061] The credibility attribute is a test of the accuracy of the generated product using ALT08 data. The smaller the overall error of the data, the higher the credibility and the more reliable the data quality. In this example, ICESat-2 lidar data with an overall error of less than 0.2 is selected to eliminate data with low credibility. Then, intersection is used to determine whether the data is located within the intertidal zone extracted in step four, and data outside the intertidal zone is eliminated (see...). Figure 3 ).

[0062] Step Six: Using the ICESat-2 lidar data and Sentinel-2 surface reflectance imagery (spectral band information) selected in Step Five, invert the intertidal zone topographic information using a feature-filtered random forest model.

[0063] After cloud removal in step three, all Sentinel-2 surface reflectance images (spectral band information) were resampled to 10m spatial resolution via bilinear interpolation.

[0064] The random forest model for feature selection adds the feature bands SDD, NDWI, SSC, and TP, and the calculation formulas are as follows:

[0065]

[0066]

[0067]

[0068]

[0069] In the formula, aerosols, red, blue, red, and nir represent the Sentinel-2 aerosol band, blue band, green band, and near-infrared band, respectively. data1 and b data2 This consists of data from 13 bands of Sentinel-2 surface reflectance imagery, as well as SDD, NDWI, and SSC data acquired at different times within the same band.

[0070] In this embodiment, the study area requires the stitching together of multiple images. The experiment first constructed the area with the most dramatic dynamic changes in the intertidal zone (radiating sandbar area, see...). Figure 4 The stochastic forest topography inversion model was tested through multiple image combination experiments (see Table 1) and data correlation analysis (see...). Figure 5 The results show that: (1) the larger the area of ​​water covering the tidal flats, the higher the accuracy of topography inversion using spectral information, and the impact on the accuracy of topography inversion is positively correlated; (2) the bands with lower water reflectance have a higher correlation with the model prediction, that is, the correlation between the bands and the predicted values ​​gradually increases with the increase of spectral wavelength.

[0071] Table 1 Figure 4 Training results of different grouped image models

[0072]

[0073] Based on the above conclusions, the model incorporates SDD, NDWI, and SSC features. Considering the impact of image temporal variations, TP features are also added. These features, along with the original spectral bands (the aforementioned 13 bands), are used to construct the inversion model. The inversion results show that R... 2 The accuracy was 0.89, RMSE (root mean square error) was 0.31m, MAE (mean absolute error) was 0.22m, and the sample size N = 10530. Compared with the original model, the accuracy was improved. Further analysis of feature contribution (see...) Figure 6 This indicates that the newly added features significantly improve model accuracy. Based on the model constructed above, topographic inversion was performed on the entire intertidal region of central Jiangsu. The model validation results are shown in [link to relevant documentation]. Figure 7 .

[0074] The verification results of the intertidal topography inversion in Jiangsu are shown below. Figure 8 The RMSE for the intertidal mudflat area is 0.20m, indicating high accuracy. However, the spectral density is insufficient to penetrate the water surface to obtain underwater topography, and also insufficient to penetrate vegetation to obtain subvegetated topography, leading to errors. The spatial and topographic extraction results for the intertidal zone along the Jiangsu coast are shown below. Figure 9 .

[0075] In summary, this invention, based on IWI and CV algorithms, overcomes the limitations of previous intertidal zone extraction methods, which suffered from low accuracy due to difficulties in obtaining waterline extraction precision and tidal level data. It eliminates the need for image screening, allowing the use of all available images, and represents a method that captures the intertidal zone from long-sequence remote sensing image changes by focusing on dynamic variations. Furthermore, it utilizes ICESat-2 lidar data instead of in-situ observation data, significantly improving the difficulty of acquiring observation data. The random forest algorithm based on feature screening is suitable for topographic inversion over large areas, and the added features after screening enhance the inversion accuracy.

[0076] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for extracting intertidal space and topography based on multi-source remote sensing data, characterized in that, include: Preprocessed Landsat 8 and Sentinel-2 surface reflectance images of the study area were obtained through a remote sensing image processing cloud platform, and ICESat-2 lidar data of the same image area were obtained through a lidar data processing platform. Cloud removal was performed on the acquired Landsat 8 and Sentinel-2 surface reflectance images. The improved spectral water index was used to calculate the surface reflectance of Landsat 8 images after cloud removal. The image after the improved spectral water index calculation was used to calculate the coefficient of variation algorithm and the intertidal spatial range was extracted using a pre-set threshold. The credibility of the ICESat-2 lidar data was screened, and data that overlapped with the extracted intertidal spatial range was selected. Topographic information of the intertidal zone was retrieved using selected ICESat-2 lidar data and Sentinel-2 surface reflectance images through a feature-selected random forest model. The formula for calculating the improved spectral water index is as follows: ； In the formula, IWI is the improved spectral water index. The reflectances are for the blue band, green band, shortwave infrared band 1, and shortwave infrared band 2, respectively. The random forest model for feature selection adds the feature bands SDD, NDWI, SSC, and TP, and the calculation formulas are as follows: ； ； ； ； In the formula, aerosols, red, blue, green, and nir represent the Sentinel-2 aerosol band, red band, blue band, green band, and near-infrared band, respectively. and This data represents 13 bands of Sentinel-2 surface reflectance imagery, as well as SDD, NDWI, and SSC data acquired at different times within the same band.

2. The method for extracting intertidal space and topography based on multi-source remote sensing data according to claim 1, characterized in that, The remote sensing image processing cloud platform uses the GEE cloud platform, and the lidar data processing platform uses the Python-based PyCharm platform.

3. The method for extracting intertidal space and topography based on multi-source remote sensing data according to claim 1, characterized in that, The cloud removal operation involves using the qa_pixel cloud quality control band of the maskL8 function in the remote sensing image processing cloud platform to remove clouds from the preprocessed Landsat8 surface reflectance image, and using the qa_60 cloud quality control band of the maskS2clouds function in the remote sensing image processing cloud platform to remove clouds from the preprocessed Sentinel-2 surface reflectance image.

4. The method for extracting intertidal space and topography based on multi-source remote sensing data according to claim 1 further includes: The declouded Sentinel-2 surface reflectance image was resampled to a spatial resolution of 10m using bilinear interpolation.

5. The method for extracting intertidal space and topography based on multi-source remote sensing data according to claim 1, characterized in that, The formula for calculating the coefficient of variation is as follows: ； In the formula, sd and mean are the standard deviation and mean of a set of data, respectively.

6. The method for extracting intertidal space and topography based on multi-source remote sensing data according to claim 1, characterized in that, The preset threshold is 0.

28.

7. The method for extracting intertidal space and topography based on multi-source remote sensing data according to claim 1, characterized in that, The credibility screening was performed using the credibility attributes provided by the ALT08 data and the intersection method was used to remove data outside the intertidal zone.

8. The method for extracting intertidal space and topography based on multi-source remote sensing data according to claim 7, characterized in that, The credibility screening method selects ICESat-2 lidar data with an overall error of less than 0.2.

Images