Layered water body extraction method for enhancing water body index driving

Abstract

The invention relates to the field of remote sensing science, and discloses a layered water body extraction method for enhancing water body index driving, which comprises the following steps: preprocessing a remote sensing image; calculating a remote sensing index; segmenting ground feature features; performing hierarchical feature fusion and decision voting; and water body space distribution and drawing. According to the method, the enhanced water body index EWI is constructed by using the space-time invariance of ground feature spectral information, and the ground feature features in the scene are subjected to conjoint analysis in combination with remote sensing indexes such as vegetation and cloud, so that accurate quantitative recognition of surface water body distribution is realized. Aiming at the problem that a water body in a remote sensing image and other interference ground objects present a similar trend in spectral characteristics, the method is combined with EWI and other remote sensing indexes to provide a layered water body extraction strategy, so that a spectral signal of the water body can be remarkably enhanced under a thin cloud coverage condition, interference caused by cloud and shadow can be effectively inhibited, and the accuracy of the water body extraction is improved. The accuracy of water body extraction is improved; the method has wide application prospects in the fields of flood disaster emergency response, water resource management and the like.

Description

Technical Field

[0001] The present invention relates to the field of remote sensing science, and in particular to a stratified water body extraction method driven by enhanced water body index. Background Art

[0002] Accurately extracting water body information is crucial for disaster prevention and mitigation, water resource management, and ecological and environmental protection. Dynamic changes in water distribution are not only closely related to natural disasters like floods and droughts, but also directly impact the balance of ecosystems. Therefore, rapid and reliable water body extraction methods are crucial for disaster prevention and early warning, water resource management, and ecological protection.

[0003] However, in practical applications, complex environmental conditions (such as clouds, turbid water, and cloud shadows) pose significant challenges to water extraction, making accurate and real-time acquisition of water distribution extremely difficult. Traditional water extraction methods face significant limitations under these complex conditions, making it difficult to accurately identify water areas. These limitations are primarily manifested in the following aspects: 1. Cloud-water confusion: In remote sensing imagery, clouds typically appear as bright areas, and their reflectivity overlaps with water trends in certain bands, making traditional spectral index methods prone to misidentifying clouds as water. Furthermore, the uneven distribution of cumulus or low-level clouds and the complex spectral information at their edges further complicate cloud-water distinction. 2. Abnormal spectral signatures of thin clouds and turbid water: Thin cloud cover or sediment mixing introduces light scattering and mixing effects, causing the spectral signal of obscured water to surge or become distorted. For example, thin cloud cover significantly increases the spectral signature of water, making it difficult for traditional water index methods to extract the complete water boundary. At the same time, due to the influence of sediment and suspended particles, the spectral reflectance characteristics of turbid water in the near-infrared band may be similar to those of bare soil or vegetation, resulting in frequent missed detections or false detections. 3. Cloud shadow interference problem: Cloud shadows will weaken the reflection signals of objects in the obscured area, making the spectral characteristics of dark target objects such as vegetation and bare soil converge with those of water bodies. For example, in mountainous areas or areas with large terrain undulations, cloud shadows are irregularly distributed and often overlap with actual water areas, increasing the difficulty of accurately extracting water bodies. In addition, the weak spectral signals in the shadow area are easily misidentified as water bodies by traditional methods, thereby reducing the extraction accuracy.

[0004] These complex remote sensing phenomena demonstrate that traditional water extraction methods have significant shortcomings when dealing with complex scenarios. More robust techniques are urgently needed to effectively overcome interference from clouds, thin clouds, turbid water, and cloud shadows, enabling accurate extraction of water information. To address these challenges, developing a hierarchical water extraction method driven by an enhanced water index is particularly necessary. Summary of the Invention

[0005] The present invention aims to provide a stratified water extraction method driven by enhanced water indexes. By leveraging the spatiotemporal invariance of surface feature spectral information and combining it with a stratified water extraction strategy, it can effectively distinguish surface water from interfering features. The research and application of this invention will provide reliable technical support for remote sensing mapping of flood disaster areas, rapid disaster assessment, and emergency disaster response decision-making, thus possessing significant scientific significance and broad practical value.

[0006] To achieve the above object, the present invention provides a method for extracting stratified water bodies driven by enhanced water index, comprising the following steps:

[0007] Step S1: Acquire multispectral images of multiple ground scenes and preprocess the acquired multispectral images. The preprocessing steps include:

[0008] Step S1.1, performing fine correction on the multispectral image, wherein the fine correction includes geometric correction, picking and registering homonymous image points;

[0009] Step S1.2, performing atmospheric correction and radiometric calibration based on the precisely corrected multispectral image to obtain a multispectral image data block of the study area;

[0010] Step S2, based on the multispectral image data block obtained by preprocessing in step S1, respectively calculating the enhanced water index feature, the improved multispectral vegetation index feature, and the cloud shadow index feature;

[0011] Step S3, performing threshold segmentation on the enhanced water index feature, the improved multispectral vegetation index feature, and the cloud shadow index feature calculated in step S2, and extracting the corresponding ground feature distribution in layers;

[0012] Step S4: fusing the multi-layer feature distribution obtained in step S3 by voting, and extracting the water body distribution according to the voting results;

[0013] Step S5: extract and draw a surface water distribution map of the satellite observation area.

[0014] Furthermore, in step S2, the calculation formula of the enhanced water index feature EWI is:

[0015]

[0016] Among them, Green is the reflectance of the ground objects in the green light band in the multispectral image, NIR is the reflectance of the ground objects in the near-infrared band in the multispectral image, and SWIR2 is the reflectance of the ground objects in the second short-wave infrared band in the multispectral image.

[0017] Furthermore, in step S2, the calculation formula of the multispectral vegetation index feature MMSVI is:

[0018]

[0019] Among them, RE1 is the reflectance of the ground objects in the first red edge band of the multispectral image, RE2 is the reflectance of the ground objects in the second red edge band of the multispectral image, Red is the reflectance of the ground objects in the red light band of the multispectral image, and Blue is the reflectance of the ground objects in the blue light band of the multispectral image.

[0020] Furthermore, in step S2, the cloud shadow index feature is calculated as follows:

[0021]

[0022] Among them, NIR is the reflectance of the ground objects in the near-infrared band of the multispectral image, and SWIR1 is the reflectance of the ground objects in the first short-wave infrared band of the multispectral image.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] (1) The enhanced water index method proposed in this paper integrates the advantages of multi-band optical remote sensing images. The constructed spectral enhancement model can significantly improve the spectral feature contrast of thin cloud cover and turbid water bodies, while effectively weakening the influence of interference factors such as clouds, cloud shadows and dark targets, thereby achieving efficient and accurate extraction of water bodies, providing reliable technical support for flood disaster monitoring, water resources management and other fields.

[0025] (2) The present invention's water body layered extraction method provides a new technical solution for the collaborative processing of multi-temporal, multi-source optical images. This method is highly automated and can automatically identify and extract water body distribution in satellite images without human intervention, significantly improving the efficiency and accuracy of water body information extraction.

[0026] (3) The enhanced water index-driven stratified extraction strategy proposed in this invention effectively improves the reliability of water extraction results by constructing a multi-scale spectral enhancement and stratified analysis model. This method fully considers the diversity of water distribution and changes in spectral characteristics in complex scenarios, proposes a new technical approach, and achieves accurate extraction of water information. This strategy not only improves the water extraction model in theory, but also provides solid technical support for flood disaster monitoring, water resources management and other fields in practical applications.

[0027] In addition to the above-described objects, features and advantages, the present invention has other objects, features and advantages. The present invention will be further described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following detailed description, they are used to explain the embodiments of the present invention, but do not constitute a limitation of the embodiments of the present invention. In the accompanying drawings:

[0029] Figure 1 A flow chart of the enhanced water index-driven stratified water body extraction method provided in an embodiment of the present invention;

[0030] Figure 2 (a) shows the spectral curve characteristics of water bodies under thin clouds and vegetation under shadows from the Sentinel-2 satellite;

[0031] Figure 2(b) shows the spectral curve characteristics of water bodies, vegetation, and clouds from the Sentinel-2 satellite;

[0032] Figure 2 (c) shows the spectral curve characteristics of water bodies under thin clouds and vegetation under shadows from the Landsat-8 satellite;

[0033] Figure 2 (d) shows the spectral curve characteristics of water bodies, vegetation, and clouds from the Landsat-8 satellite;

[0034] Figure 2 (e) shows the spectral curve characteristics of water bodies under thin clouds and vegetation under shadows from the Landsat-9 satellite.

[0035] Figure 2 (f) shows the spectral curve characteristics of water bodies, vegetation, and clouds from the Landsat-9 satellite;

[0036] Figure 3 (a) is a false color composite image of Sentinel-2 (blue channel: Green, green channel: Red, red channel: NIR);

[0037] Figure 3(b) shows the EWI index results obtained by Sentinel-2, with the numerical display range being (0-0.2);

[0038] Figure 3(c) shows the MNDWI index results calculated by Sentinel-2, with the numerical display range being (0.2-0.5);

[0039] Figure 3(d) shows the MNDWI index results calculated by Sentinel-2, with the numerical display range being (0.3-0.5);

[0040] Figure 3(e) shows the MNDWI index results calculated by Sentinel-2, with the numerical display range being (0.4-0.5);

[0041] Figure 3 (f) is a Landsat-9 false color composite image (blue channel: Green, green channel: Red, red channel: NIR);

[0042] Figure 3 (g) shows the EWI index results calculated by Landsat-9, with the numerical display range being (0-0.2);

[0043] Figure 3 (h) shows the MNDWI index results calculated by Landsat-9, with the numerical display range being (0.2-0.5);

[0044] Figure 3 (i) shows the MNDWI index results calculated by Landsat-9, with the numerical display range being (0.3-0.5);

[0045] Figure 3 (j) shows the MNDWI index results calculated by Landsat-9, with the numerical display range being (0.4-0.5);

[0046] Figure 3 (k) is a Landsat-8 false color composite image (blue channel: Green, green channel: Red, red channel: NIR);

[0047] Figure 3 (l) shows the EWI index results calculated by Landsat-8, with the numerical display range being (0-0.2);

[0048] Figure 3 (m) shows the MNDWI index results calculated by Landsat-8, with the numerical display range being (0.2-0.5);

[0049] Figure 3 (n) shows the MNDWI index results calculated by Landsat-8, with the numerical display range being (0.3-0.5);

[0050] Figure 3 (o) shows the MNDWI index results calculated by Landsat-8, with the numerical display range being (0.4-0.5). DETAILED DESCRIPTION

[0051] The present invention will be described in detail below with reference to the various embodiments shown in the accompanying drawings, but it should be noted that these embodiments are not limitations of the present invention, and any equivalent transformations or substitutions in functions, methods, or structures made by ordinary technicians in this field based on these embodiments are all within the scope of protection of the present invention.

[0052] By employing the Enhanced Water Index (EWI), this invention develops a technical solution specifically for layered water extraction in complex scenes. Its key advantages lie in the following three aspects: 1. Effectively addressing cloud-water confusion: Because clouds and water exhibit similar spectral characteristics in certain bands, traditional methods are prone to misidentification. By integrating multi-band information to construct an efficient spectral difference model, this invention significantly improves the ability to distinguish between clouds and water, thereby enabling precise extraction of water areas. 2. Enhancement of spectral information for thin cloud cover and turbid water: To address the weakening of water spectral signals under thin cloud cover, this invention employs the EWI to significantly enhance the spectral characteristics of water within thin cloud cover, highlighting the spectral differences between water and other ground objects. Furthermore, for water that appears turbid due to sediment and suspended particles, the EWI enhances its spectral characteristics, enabling the effective identification of more water areas. 3. Effectively suppressing cloud shadow interference: Cloud shadows tend to weaken the spectral signals of objects within the obscured area, causing dark objects such as vegetation to have similar spectral characteristics to those of water. The present invention implements targeted compensation for cloud shadow effect by jointly improving the multispectral vegetation index and the cloud shadow index, reducing its interference on water body extraction results, thereby further improving extraction accuracy.

[0053] like Figure 1 As shown, an embodiment of the present invention provides a method for extracting stratified water bodies (including floods, rivers, lakes, etc.) driven by enhanced water body index, comprising the following steps:

[0054] Step S1: After obtaining multi-view multispectral images of the surface in the study area, pre-process the images. Specifically:

[0055] Step S1.1: Fine image calibration: Perform fine registration of the time-series multispectral images. This fine registration includes the selection of homonymous image points and image calibration. ENVI 5.3 software is used to effectively select homonymous image points and perform fine registration of heterogeneous multispectral images using a polynomial image calibration method.

[0056] Step S1.2: Based on the precisely registered multispectral image, ENVI5.3 software is used to perform atmospheric correction and radiometric calibration to obtain the multispectral image data block of the study area.

[0057] Step S2: Based on the multispectral image data block obtained by preprocessing in step S1, the enhanced water index feature, the improved multispectral vegetation index feature, and the cloud shadow index feature are calculated respectively.

[0058] Step S2.1: Calculate the enhanced water index features for the multispectral image data block. As shown in Figures 2(a) to 2(f), the principle diagrams for constructing the enhanced water index provided in this embodiment are as follows:

[0059] As shown in Figure 2(a), among non-water objects, the Green band of shaded vegetation and clouds is significantly higher than that of other bands, closely resembling the spectral characteristics of water. This makes it difficult for traditional water indices to remove such object interference when calculating index features. However, by superimposing the reflectance of objects in the SWIR2 band with that in the NIR band, it is clear that the reflectance of water beneath thin clouds is significantly higher than that of vegetation beneath shadows. This also makes the superimposed reflectance of non-water objects significantly lower than their own Green band, effectively enhancing water information while suppressing the spectral information of non-water objects, as shown by the dashed line in Figure 2(a). Then, by subtracting the reflectance of the Green band from the superimposed band, water and non-water bodies can be effectively distinguished, as illustrated by arrows ① and ② in the figure. As shown by arrows ③ and ④ in Figure 2(a), the subtraction results in a positive value for water (arrow ④), while a negative value for non-water (arrow ③). To further enhance the robustness of the index, a normalized difference index structure was introduced, and the Green and NIR bands, which are sensitive to water bodies, were selected as the denominator to construct the index. The calculation formula for the enhanced water index constructed based on this principle is:

[0060]

[0061] Among them, Green is the reflectance of the ground objects in the green light band in the multispectral image, NIR is the reflectance of the ground objects in the near-infrared band in the multispectral image, and SWIR2 is the reflectance of the ground objects in the second short-wave infrared band in the multispectral image.

[0062] Applying the EWI construction process to other land features, such as water bodies, vegetation, and clouds, the feature construction process remains effective, effectively distinguishing between water and non-water. The detailed process is shown in Figure 2(b). Similarly, applying this principle from the Sentinel-2 satellite to the Landsat-8 satellite (Figures 2(c) and (d)) and Landsat-9 satellite (Figures 2(e) and (f)) also works.

[0063] Step S2.2: Calculate the improved multispectral vegetation index feature for the multispectral image data block. The calculation formula is:

[0064]

[0065] RE2 is the reflectance of the second red-edge band in the multispectral image, RE1 is the reflectance of the first red-edge band in the multispectral image, Red is the reflectance of the red band in the multispectral image, and Blue is the reflectance of the blue band in the multispectral image. If the sensor does not have a red-edge band, both RE2 and RE1 parameters can be replaced by the near-infrared band.

[0066] Step S2.3: Calculate the Cloud Shadow Index (CSI) for the multispectral image data block. The calculation formula is:

[0067]

[0068] Among them, NIR is the reflectance of the ground objects in the near-infrared band of the multispectral image, and SWIR1 is the reflectance of the ground objects in the first short-wave infrared band of the multispectral image.

[0069] Step S3: Threshold segmentation is performed on the enhanced water index feature, improved multispectral vegetation index feature, and cloud shadow index feature calculated in step S2, and the corresponding ground feature distribution is extracted in layers. The specific steps are:

[0070] Step S3.1: Use threshold segmentation to obtain water body distribution calculation for the enhanced water body index feature. The calculation formula is:

[0071]

[0072] Among them, Water is the spatial distribution of water layer characteristics.

[0073] Step S3.2: Use threshold segmentation to obtain vegetation distribution calculation for the improved multispectral vegetation index feature. The calculation formula is:

[0074]

[0075] Among them, Vegetation is the spatial distribution of vegetation layer characteristics.

[0076] Step S3.3: Use threshold segmentation to obtain cloud and cloud shadow distribution calculation based on cloud shadow index features. The calculation formula is:

[0077]

[0078] Among them, Cloud is the spatial distribution of cloud and cloud shadow layer characteristics.

[0079] Step S4: The multi-layer feature distribution obtained in step S3 is fused by voting, and the water body distribution is extracted according to the voting result threshold. The final water body distribution is calculated by fusion of the multi-layer feature distribution by voting. The calculation formula is:

[0080]

[0081] Among them, FinalWater is the final spatial distribution of water bodies in the algorithm process.

[0082] Step S5: Obtain the water distribution and map the flooded area.

[0083] Figures 3(a)-3(o) show comparisons of flood detection results provided by this embodiment. Based on the comparisons of EWI and other indices such as MNDWI shown in Figures 3(a)-3(o), the following conclusions can be drawn: EWI demonstrates significant advantages in water extraction, significantly outperforming existing index methods. Specifically,

[0084] As shown in Figures 3(a)–3(e), in the Sentinel-2 imagery and index extraction results, EWI not only effectively suppresses the interference of clouds and cloud shadows, but also accurately extracts water bodies within areas covered by thin clouds. Figure 3(a) clearly shows the distribution of floods, thick clouds, thin clouds, and cloud shadows. Comparison with the EWI index intensity distribution shown in Figure 3(b) demonstrates that EWI significantly reduces the impact of clouds and cloud shadows on water extraction. Figure 3(c), however, shows that traditional MNDWI tends to misclassify vegetation and thick clouds under cloud shadows as water bodies. Furthermore, Figures 3(c)–3(e) demonstrate that while cloud interference is reduced with iterations of the MNDWI threshold, water extraction performance continues to decline, particularly in areas covered by thin clouds and cloud shadows, where water information is significantly weakened. Comparing Figures 3(b) and 3(c), it is clear that EWI maintains water extraction accuracy while better suppressing the influence of clouds and cloud shadows.

[0085] Similarly, as shown in Figures 3(f)–3(j), in the Landsat-9 imagery and index extraction results, EWI is still able to extract surface water distribution to the greatest extent and suppress cloud interference. However, as shown in Figures 3(h)–3(j), as the threshold value of MNDWI is continuously optimized, although cloud interference is eliminated, the water extraction effect gradually deteriorates. Further comparison is verified in Landsat-8 imagery (Figures 3(k)–3(o)): The EWI results in Figure 3(l) are consistent with the aforementioned imagery, successfully eliminating cloud interference, while Figures 3(m)–3(o) show that even with continuous adjustment of the threshold, MNDWI still cannot completely eliminate cloud interference, resulting in poor water extraction results.

[0086] In addition, in the consistency verification of multi-source image platforms, the comprehensive analysis results of multi-source images such as Sentinel-2, Landsat-9 and Landsat-8 show that EWI shows excellent water body extraction performance under different satellite platforms, especially under conditions of complex cloud layers, thin clouds and cloud shadows, which can significantly reduce the impact of clouds and ensure the accuracy and robustness of the extraction results.

[0087] In summary, the enhanced water index-driven stratified water extraction strategy proposed in this paper not only effectively addresses the accuracy issues of traditional methods in the presence of thin clouds, turbid water, and cloud shadows, but also achieves precise extraction of different water types by constructing a multi-order spectral feature enhancement and stratified discrimination model. This method provides reliable water information support for flood disaster warning, water resource monitoring, and ecological and environmental protection, and has broad application value.

[0088] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A method for extracting stratified water bodies driven by enhanced water index, characterized in that: The following steps are involved: S1. Acquire multispectral images of multiple ground surfaces within the study area and preprocess the acquired multispectral images. The preprocessing is specifically as follows: S1.

1. Performing fine correction on the multispectral image, wherein the fine correction includes geometric correction, picking and registering homonymous image points; S1.

2. Perform atmospheric correction and radiometric calibration based on the precisely calibrated multispectral image to obtain multispectral image data blocks for the study area; S2. Based on the multispectral image data block obtained by preprocessing in step S1, respectively calculating the enhanced water index feature, the improved multispectral vegetation index feature, and the cloud shadow index feature; S3, performing threshold segmentation on the enhanced water index feature, the improved multispectral vegetation index feature, and the cloud shadow index feature calculated in step S2, and extracting the corresponding ground feature distribution in layers; S4, using voting to fuse the multi-layer feature distribution obtained in step S3, and extracting the water body distribution according to the voting results; S5. Draw a surface water distribution map of the study area based on the water distribution information extracted in step S4.

2. The method for extracting stratified water according to claim 1, wherein: In step S2, the calculation formula of the enhanced water index feature EWI is: Among them, Green is the reflectance of the ground objects in the green light band in the multispectral image, NIR is the reflectance of the ground objects in the near-infrared band in the multispectral image, and SWIR2 is the reflectance of the ground objects in the second short-wave infrared band in the multispectral image.

3. The method for extracting stratified water according to claim 1, wherein: In step S2, the calculation formula of the multispectral vegetation index feature MMSVI is: Among them, RE1 is the reflectance of the ground objects in the first red edge band of the multispectral image, RE2 is the reflectance of the ground objects in the second red edge band of the multispectral image, Red is the reflectance of the ground objects in the red light band of the multispectral image, and Blue is the reflectance of the ground objects in the blue light band of the multispectral image.

4. The method for extracting stratified water according to claim 1, wherein: In step S2, the cloud shadow index feature is calculated as follows: Among them, NIR is the reflectance of the ground objects in the near-infrared band of the multispectral image, and SWIR1 is the reflectance of the ground objects in the first short-wave infrared band of the multispectral image.

Images