Ultrahigh spatial resolution water depth inversion method fused with panchromatic spectral data

Abstract

The invention discloses an ultrahigh spatial resolution water depth inversion method fused with panchromatic spectral data. The method comprises the following steps: collecting ICESat-2 satellite water depth data and satellite multispectral image and panchromatic image data; performing water depth inversion on the multispectral image with the original spatial resolution to obtain water depth data with the original spatial resolution; performing panchromatic spectrum fusion processing on the multispectral image and panchromatic image data to obtain an ultrahigh spatial resolution multispectral image; carrying out model training by adopting a machine learning algorithm, and constructing a complex nonlinear mapping relation between the water depth and the multispectral reflectivity; and performing pixel-by-pixel prediction on the multispectral image with the ultrahigh spatial resolution by using the trained model, thereby obtaining water depth data with the ultrahigh spatial resolution. According to the method, the nonlinear mapping relation between the spectral reflectivity of the multispectral images with different spatial resolutions and the water depth data is constructed through a machine learning algorithm, and the problem that the inversion precision is reduced due to spectral information loss in the panchromatic spectrum fusion process is effectively solved.

Description

Technical Field

[0001] The present invention relates to a method for ultra-high spatial resolution water depth inversion by fusing panchromatic spectral data, belonging to the field of marine surveying and mapping. Background Art

[0002] Water depth, as an important parameter of the marine environment, is one of the basic elements of marine geospatial information and is of great significance to fields such as maritime transportation, coastal engineering, marine resource development, and benthic environment mapping. Although the traditional shipborne sonar sounding technology has high accuracy, it is costly and has a limited coverage. With the rapid development of remote sensing technology, the satellite optical remote sensing water depth inversion technology has gradually become an important supplementary means to the traditional sounding technology by virtue of its advantages of large-scale synchronous observation and multi-temporal continuous observation. In particular, the successful launch of satellites equipped with active laser altimetry systems in recent years has promoted the development of the active-passive fusion water depth inversion method, providing a new way to obtain shallow water bathymetry over a large area at low cost. Currently, the water depth inversion models based on multi-spectral images are mainly divided into three categories: theoretical analytical models, semi-theoretical semi-empirical models, and statistical models.

[0003] With the continuous deepening of human exploration of the ocean, the requirements for accurate measurement of seabed topography and geomorphology are increasing. Ultra-high spatial resolution bathymetric maps can accurately depict complex seabed topography features. Especially in areas with drastic water depth changes or complex seabed geomorphology, rich seabed topography details can be accurately captured to achieve the accurate expression of seabed micro-topography and geomorphology features. However, the existing water depth inversion mainly relies on multi-spectral images. Limited by the spatial resolution, it is difficult to fully display the subtle undulations of seabed geomorphology features. In contrast, panchromatic images have higher spatial resolution and can provide more detailed seabed topography information. Through the panchromatic spectral fusion technology, while retaining the rich spectral information of multi-spectral images, the spatial resolution can be significantly improved, making the information of the fused remote sensing images more abundant and accurate, thus providing a reliable data basis for obtaining high-precision water depth data.

[0004] Although the panchromatic spectral fusion technology has been widely used in the field of land topographic mapping, due to the inevitable spectral information loss problem in the fusion process, its application in the field of water depth inversion is relatively limited. Existing methods usually directly use the fused images for water depth inversion without fully considering the problem of spectral information loss in the image fusion process. Especially for the fusion of different satellite images, the loss of spectral information is more significant. Although directly performing water depth inversion on the fused images by existing methods improves the spatial resolution, the inversion accuracy is lost to a certain extent. Therefore, there is an urgent need to develop an ultra-high spatial resolution water depth inversion method that can effectively reduce the influence of spectral information loss. Summary of the Invention

[0005] The objective of the present invention is to provide a method for ultra-high spatial resolution water depth inversion that integrates panchromatic spectral data, effectively ensuring the accuracy of water depth inversion while improving the spatial resolution, so as to solve the problems of low acquisition efficiency and accuracy and high cost of ultra-high spatial resolution water depth data in current water depth inversion.

[0006] A method for ultra-high spatial resolution water depth inversion that integrates panchromatic spectral data, comprising the following steps: Step S1: Collect ICESat-2 satellite water depth data, as well as satellite multispectral image and panchromatic image data, and preprocess the acquired original data to ensure the consistency and usability of the data.

[0007] Step S2: Perform spatial matching on the spectral reflectance of the ICESat-2 water depth data and the multispectral image with the original spatial resolution, construct a water depth inversion data set, select a water depth inversion model, perform water depth inversion on the multispectral image with the original spatial resolution, and obtain water depth data with the original spatial resolution, providing reliable basic data for ultra-high spatial resolution water depth inversion.

[0008] Step S3: Perform panchromatic spectral fusion processing on the preprocessed multispectral image and panchromatic image data to obtain an ultra-high spatial resolution multispectral image.

[0009] Step S4: Perform pixel matching between the water depth data with the original spatial resolution and the ultra-high spatial resolution multispectral image, extract the spectral reflectance of each band in the ultra-high spatial resolution multispectral image, and combine with the water depth data at the corresponding position to construct a training sample set; based on this sample set, use a machine learning algorithm to perform model training and construct a complex non-linear mapping relationship between water depth and multispectral reflectance.

[0010] During the training process, ensure the generalization ability of the model through cross-validation and hyperparameter optimization.

[0011] Step S5: Use the trained model to perform pixel-by-pixel prediction on the ultra-high spatial resolution multispectral image, and finally invert to obtain ultra-high spatial resolution water depth data.

[0012] The ultra-high spatial resolution water depth inversion method integrating panchromatic spectral data constructs a non-linear mapping relationship between the spectral reflectance of multi-spectral images with different spatial resolutions and water depth data through machine learning algorithms, effectively solving the problem of decreased inversion accuracy caused by spectral information loss during the panchromatic spectral fusion process, ensuring the inversion accuracy while improving the spatial resolution. The ultra-high spatial resolution water depth data obtained by this method can more accurately depict the submarine micro-topography and geomorphology features, and the inversion accuracy is significantly better than that of traditional methods, providing an efficient and accurate technical means for obtaining ultra-high spatial resolution water depth data on a large scale and with high efficiency. This method can provide accurate and efficient data support for fields such as the analysis of water depth topographic changes around islands and reefs and the monitoring of underwater targets. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 is a schematic flow chart of the ultra-high spatial resolution water depth inversion method integrating panchromatic data of the present invention.

[0014] Figure 2 is a schematic diagram of the experimental area in an embodiment of the present invention, where (a) is the satellite image of the experimental area, the red line is the distribution of ICESat-2 survey lines, and (b) is the distribution of verification points in the local area of (a).

[0015] Figure 3 is an example diagram of the result obtained by inversion in the Ganquan Island area in an embodiment of the present invention (water depth raster map), where (a) is the inversion of the original multi-spectral image, (b) is the direct inversion of the panchromatic spectral fusion super-resolution image, and (c) is the inversion of the panchromatic spectral fusion super-resolution image combined with random forest.

[0016] Figure 4 is a partial detail enlarged view of the water depth result obtained by inversion in an embodiment of the present invention, where the first row is the northwest side edge area of the reef flat, the second row is the northeast side edge area of the reef flat, and the third row is the reef convex area in the southern part of the research area. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0018] The present invention provides an ultra-high spatial resolution water depth inversion method integrating panchromatic spectral data to obtain ultra-high spatial resolution shallow water depth data.

[0019] To more accurately evaluate the effectiveness of the ultra-high spatial resolution water depth inversion method integrating panchromatic spectral data, the present invention selects the sea area around Ganquan Island as the research area for experiments. The scope of the research area is shown in Figure 2, the airborne lidar bathymetric data in this area was synchronously collected for accuracy evaluation. The accuracy of the bathymetric data obtained by the present invention was compared and analyzed with the bathymetric data directly retrieved by fusing panchromatic images, further intuitively demonstrating the effectiveness of the method of the present invention.

[0020] Referring to Figure 1 , the method includes the following steps: Step S1: Collect and process ICESat-2 satellite bathymetric data, GF-2 satellite multispectral images and panchromatic image data. For effect verification, on-site Optech Aquarius airborne lidar bathymetric data was collected simultaneously. Specifically: Perform data preprocessing such as radiometric calibration, atmospheric correction, orthorectification, and flare correction on the obtained GF-2 satellite multispectral images with a spatial resolution of 4m, reduce the influence of various environmental factors including the atmosphere and seawater on the image brightness value, highlight water body information, and ensure the accuracy of the inversion result.

[0021] Perform radiometric calibration, orthorectification, and image registration on the GF-2 satellite panchromatic images with a spatial resolution of 1m to facilitate subsequent panchromatic spectral fusion processing.

[0022] Perform preprocessing such as point cloud denoising, refraction correction, and tidal correction on ICESat-2 data to obtain high-precision bathymetric data for calibrating the inversion model. There is too little survey line data of ICESat-2 directly passing through the study area. In this paper, a total of 6 survey lines passing through the Ganquan Island area and nearby reefs are selected. The positions of the survey lines are shown in Figure 2 (a). Since part of the northeast region of the GF-2 satellite image for inversion is covered by clouds, the ICESat-2 data passing through the cloud-covered area was removed to reduce the influence of cloud cover on the accuracy of the bathymetric inversion result.

[0023] Perform tidal correction on the collected measured data to unify the datum plane with the bathymetric data extracted by ICESat-2 for subsequent accuracy evaluation. To improve the accuracy evaluation efficiency, the measured data was thinned before use, and 1334 points evenly distributed in the study area were randomly selected. The distribution positions are shown in Figure 2 (b).

[0024] Step S2: Use the spectral reflectance of ICESat-2 bathymetric data and the 4m spatial resolution multispectral images for spatial matching, construct a bathymetric inversion data set, select the logarithmic ratio inversion model, perform bathymetric inversion on the 4m spatial resolution multispectral images, and obtain 4m spatial resolution bathymetric data. Specifically: In this step, first, spatially match the preprocessed ICESat-2 photon point water depth data with the spectral reflectance of the multispectral image with a 4m spatial resolution to construct a water depth inversion dataset. Subsequently, select the spectral reflectances of the blue and green bands with the highest correlation with water depth in the image as input variables, use the processed ICESat-2 photon point water depth data as the dependent variable, and adopt the logarithmic ratio water depth inversion model to perform water depth inversion to obtain water depth data with a 1m spatial resolution, providing reliable basic data for subsequent ultra-high spatial resolution water depth inversion. The logarithmic ratio model can, to a certain extent, weaken the influence caused by the differences in seabed substrates during the water depth inversion process, improving the accuracy and reliability of the water depth inversion results. This model only contains two key parameters, avoiding the problem of numerous and complex parameters in the traditional water depth inversion process.

[0025] Step S3: Select an appropriate algorithm to perform panchromatic spectral fusion processing on the 1m spatial resolution panchromatic data and the 4m spatial resolution multispectral data to obtain a 1m spatial resolution multispectral image. Specifically: In this step, use the Gram-Schmidt fusion algorithm to fuse the 4m spatial resolution multispectral image and the 1m spatial resolution panchromatic image of the GF-2 satellite into a 1m spatial resolution multispectral image. The quality and visual clarity of the remotely sensed image after fusion by the Gram-Schmidt algorithm are significantly improved, the texture details are more abundant, the spectral characteristics of the original multispectral image are retained to the greatest extent, and the loss of spectral information is reduced.

[0026] Step S4: Establish a non-linear mapping relationship between the 4m spatial resolution water depth data and the spectral reflectance of the 1m spatial resolution multispectral image, and then invert to obtain 1m spatial resolution water depth data. Specifically: In this step, aiming at the problem that the difference in spatial resolution will cause the 4m spatial resolution water depth data to correspond to multiple pixels of the 1m spatial resolution multispectral image, select the pixels of the 1m spatial resolution multispectral image at the middle position within 4m to match the water depth data, ensuring that each 4m spatial resolution water depth data corresponds to a pixel of the 1m spatial resolution multispectral image. Extract the spectral reflectances of each band in the 1m spatial resolution multispectral image, and combine the corresponding water depth data to construct a training sample set.

[0027] A non - linear mapping relationship between the spectral reflectance of 1m spatial resolution multispectral images and the water depth data of 4m spatial resolution is established using the random forest algorithm. Based on the constructed training sample set, the feature data set composed of the spectral reflectances of the red, green, blue, and near - infrared bands of the 1m spatial resolution multispectral image is used as the input variable, and the water depth data is used as the target variable. The random forest algorithm is used to train the model and predict the water depth value, and finally, the water depth data with 1m ultra - high spatial resolution is obtained. The construction process of the random forest model is simple and efficient, and is suitable for solving non - linear complex problems.

[0028] During the training process, the performance of the model is optimized by changing the number of decision trees and the number of random features of a single decision tree.

[0029] Step S5: Use the trained model to predict each pixel of the 1m spatial resolution multispectral image, and finally invert to obtain the water depth data with 1m spatial resolution.

[0030] Embodiment Comparing the inversion effect of this article with the inversion effect of the existing method that directly uses the fused image for water depth inversion can further intuitively illustrate the effectiveness of this method.

[0031] The water depth inversion results in the Ganquan Island area are shown in Figure 3 , including the water depth raster map of the Ganquan Island area with 4m spatial resolution ( Figure 3 a), the water depth raster map of the Ganquan Island area with 1m spatial resolution obtained by the traditional method ( Figure 3 b), and the water depth raster map of the Ganquan Island area with 1m spatial resolution obtained by the method proposed in this invention ( Figure 3 c). Overall, the overall change trends of the water depths of the 1m resolution water depth data and the 4m resolution water depth data obtained by different methods are basically the same. In order to more accurately compare the ability of the water depth data to depict the seabed details after the spatial resolution is improved, three local areas with obvious topographic features are selected in the study area for enlarged display, and their water depth inversion results are compared (see Figure 4 ). In the reef bulge area in the south of the study area, the pan - chromatic spectral fusion super - resolution method combined with the random forest proposed in this article can more clearly display the distribution details of the reefs. The obtained reef boundary is clear, and the seabed topography details are depicted more accurately.

[0032] To quantitatively compare the inversion effects of different methods after improving the spatial resolution of image data, this article selects three indicators, the coefficient of determination R2, the mean absolute error MAE, and the root mean square error RMSE, to evaluate the inversion accuracy of several different methods. The calculation formulas are as follows: (1) (2) (3) Among them, is the measured water depth value of the test point, is the inverted water depth value of the check point; is the average water depth value of the test point; n is the number of test points. The larger the R 2 value, the larger the MAE and RMSE values, indicating that the method is more accurate and the fitting effect is better.

[0033] Table 1 Comparison of inversion accuracies after improving spatial resolution by different methods Method Spatial resolution <![CDATA[R 2 > MAE / m RMSE / m Inversion of original multispectral image 4m 0.93 1.26 1.52 Direct inversion of panchromatic spectral fusion super-resolution image 1m 0.87 1.66 2.00 Inversion of panchromatic spectral fusion super-resolution image combined with random forest 1m 0.96 1.18 1.40

[0035] The comparison results show that the method with the highest accuracy is the panchromatic spectral fusion super-resolution method combined with the random forest algorithm proposed in this paper. The R2 of the water depth data with a spatial resolution of 1m obtained by this method is 0.96, the MAE is 1.18m, and the RMSE is 1.40. By constructing a non-linear mapping relationship between the spectral reflectance and water depth data of multi-spectral images with different spatial resolutions before and after image fusion, this method not only makes full use of the rich spectral information of the original multi-spectral image before fusion, effectively solves the problem of spectral information loss in image fusion, but also based on the random forest algorithm, synthesizes the effective spectral data of the blue, green, red, and near-infrared four bands, while improving the spatial resolution, ensuring the inversion accuracy of the water depth data. Compared with directly using the panchromatic spectral fusion super-resolution method, the MAE of the water depth data inverted by the method proposed in this paper decreases by 0.48m, and the RMSE decreases by 0.60m.

Claims

1. An ultra-high spatial resolution water depth inversion method integrating panchromatic spectral data, characterized by The following steps are involved: Step S1: Collect ICESat-2 satellite water depth data and satellite multispectral image and panchromatic image data, and pre-process the acquired raw data to ensure data consistency and availability; Step S2: perform spatial matching between ICESat-2 water depth data and the spectral reflectance of the multispectral image with original spatial resolution, construct a water depth inversion dataset, select a water depth inversion model, perform water depth inversion on the multispectral image with original spatial resolution, obtain water depth data with original spatial resolution, and provide reliable basic data for water depth inversion with ultra-high spatial resolution; Step S3: performing panchromatic spectral fusion processing on the pre-processed multispectral image and panchromatic image data to obtain a multispectral image with ultra-high spatial resolution; Step S4: Pixel matching is performed between the water depth data of the original spatial resolution and the multispectral image of the ultra-high spatial resolution, the spectral reflectance of each band in the multispectral image of the ultra-high spatial resolution is extracted, and the water depth data of the corresponding position is combined to construct a training sample set; based on the sample set, a machine learning algorithm is used to perform model training to construct a complex nonlinear mapping relationship between water depth and multispectral reflectance; Step S5: Use the trained model to perform pixel-by-pixel prediction on the ultra-high spatial resolution multispectral image, and finally invert to obtain ultra-high spatial resolution water depth data.

2. The method for water depth inversion with ultra-high spatial resolution by integrating panchromatic spectral data as claimed in claim 1, characterized in that The satellite multispectral image and panchromatic image data adopt GF-2 satellite multispectral image and panchromatic image data.

3. The method for water depth inversion with ultra-high spatial resolution by integrating panchromatic spectral data as claimed in claim 1, characterized in that The ICESat-2 data are preprocessed, including point cloud denoising, refraction correction, and tide correction, to obtain high-precision water depth data for calibrating the inversion model.

4. The method for water depth inversion with ultra-high spatial resolution by integrating panchromatic spectral data as claimed in claim 2, The method is characterized in that the satellite multispectral image and panchromatic image data are preprocessed as described in step 1, including performing radiation calibration, atmospheric correction, orthorectification and flare correction on the acquired GF-2 satellite 4m spatial resolution multispectral image; Radiometric calibration, orthorectification and image registration were performed on the GF-2 satellite 1m spatial resolution panchromatic image.

5. The method for water depth inversion with ultra-high spatial resolution by integrating panchromatic spectral data as claimed in claim 2, characterized in that The step 2 is specifically as follows: Firstly, the preprocessed ICESat-2 photon point water depth data are spatially matched with the spectral reflectance of the 4m spatial resolution multispectral image to construct a water depth inversion dataset. Then, the spectral reflectance of the blue and green bands in the image with the highest correlation with water depth is selected as the input variable, and the processed ICESat-2 photon point water depth data is used as the dependent variable. The logarithmic ratio water depth inversion model is used to perform water depth inversion, and the water depth data with 4m spatial resolution is obtained, thus providing reliable basic data for the subsequent ultra-high spatial resolution water depth inversion.

6. The method for water depth inversion with ultra-high spatial resolution by fusing panchromatic spectral data as claimed in claim 2, characterized in that In step 3, the Gram-Schmidt fusion algorithm is used to fuse the GF-2 satellite multispectral image with a spatial resolution of 4 m and the panchromatic image with a spatial resolution of 1 m into a multispectral image with a spatial resolution of 1 m.

7. The method for water depth inversion with ultra-high spatial resolution by integrating panchromatic spectral data as claimed in claim 2, characterized in that In step S4, the training sample set is constructed as follows: In order to solve the problem that the difference in spatial resolution will cause the 4m spatial resolution water depth data to correspond to multiple pixels of the 1m spatial resolution multispectral image, the pixel of the 1m spatial resolution multispectral image in the middle position is selected within the 4m range to match the water depth data, ensuring that each 4m spatial resolution water depth data corresponds to a 1m spatial resolution multispectral image pixel; the spectral reflectance of each band in the 1m spatial resolution multispectral image is extracted, and combined with the corresponding water depth data, a training sample set is constructed.

8. The method for water depth inversion with ultra-high spatial resolution by fusing panchromatic spectral data as claimed in claim 7, characterized in that In step S4, a nonlinear mapping relationship is established between the water depth data with a spatial resolution of 4 m and the spectral reflectance of the multispectral image with a spatial resolution of 1 m, and then the water depth data with a spatial resolution of 1 m is inverted, as follows: Based on the constructed training sample set, the feature data set composed of the spectral reflectance of the four bands of red light, green light, blue light and near-infrared of the 1m spatial resolution multispectral image is taken as the input variable, and the water depth data is taken as the target variable. A nonlinear mapping relationship between the spectral reflectance of the 1m spatial resolution multispectral image and the 4m spatial resolution water depth data is established. The random forest algorithm is used to train the model and predict the water depth value, and finally the water depth data with 1m ultra-high spatial resolution is obtained.

9. The method for water depth inversion with ultra-high spatial resolution by fusing panchromatic spectral data as claimed in claim 2, characterized in that In step S5, the trained model is used to perform pixel-by-pixel prediction on the multispectral image with a spatial resolution of 1 m, and finally the water depth data with a spatial resolution of 1 m is inverted.

10. The method for ultra-high spatial resolution water depth inversion integrating panchromatic spectral data as claimed in claim 1 is characterized in that in step S4, a machine learning algorithm is used to perform model training. During the training process of the model, the generalization ability of the model is ensured through cross-validation and hyperparameter optimization.

Images