Multi-temporal fusion uncontrolled water depth inversion method based on remote sensing image substrate recognition

Abstract

The invention provides a multi-temporal fusion uncontrolled water depth inversion method based on remote sensing image substrate identification, and relates to the technical field of water depth information extraction. Comprising the following steps: S1, collecting and processing data of a multi-temporal remote sensing image and actually measured data; s2, determining an inversion factor, and constructing an MM-P-DLA model, namely a multi-temporal fusion dual-band logarithmic linear model based on physics; s3, extracting feature pixels based on substrate identification of the multi-temporal remote sensing image; s4, performing preliminary water depth inversion based on the MM-P-DLA model to obtain each single-time-phase water depth inversion result; and S5, according to each single-temporal water depth inversion result, performing multi-temporal uncontrolled water depth inversion fusion based on an MM-P-DLA model to obtain a final inversion result. According to the method, efficient inversion of the uncontrolled water depth of the shallow sea can be completed, high inversion precision is achieved, and the problem that the water depth inversion precision is affected due to limitation of an existing shallow sea uncontrolled water depth inversion algorithm is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of water depth information extraction, and in particular to a multi-temporal fusion uncontrolled water depth inversion method based on remote sensing image bottom identification. Background Art

[0002] Shallow water depth information has an important impact on offshore development and utilization, marine economic activities and shipping safety, and is one of the indispensable basic information in seabed topography mapping and marine environmental surveys. Traditional bathymetric methods, such as single-beam bathymetry, multi-beam bathymetry and airborne lidar bathymetry, have high accuracy but are costly and inefficient.

[0003] Therefore, the method of inverting shallow sea water depth based on remote sensing images has become an important means of obtaining shallow sea water depth because it has the advantages of multi-temporal observation, large range and low cost, and can obtain information on islands and reefs in disputed, dangerous or remote sea areas. In recent years, scholars in this field have gradually proposed different remote sensing image water depth inversion algorithms through research, mainly theoretical analytical models, semi-theoretical and semi-empirical models, statistical models, etc. These algorithms all use a certain amount of measured data as input values ​​to solve the model parameters, and apply the model to the bathymetric inversion of a specific study area. There is a limitation of low portability, which makes it impossible to invert the water depth in areas without available water depth control points. Based on this, in order to obtain the water depth information of sea areas without measured control points, scholars at home and abroad have carried out different studies on the theory and methods of uncontrolled water depth inversion, which can be roughly divided into the following three categories: 1) Hyperspectral Optimization Processing Exemplar (HOPE): The remote sensing reflectance of satellite remote sensing images in shallow waters is regarded as a function of variables that vary independently with inherent optical quantities in the water body, including water depth, seabed sediments, and water body. The inherent optical properties and water depth of the water body are inverted using a nonlinear optimization solution method based on the least squares method. This algorithm has been relatively mature after continuous research. For example, the Chinese patent with the publication number CN114758254A uses the HOPE model for water depth inversion. However, this algorithm requires many optical parameters, so unstable optical properties will bring greater instability to the single-phase inversion, which will have a negative impact on the final inversion result. 2) Combination of hyperspectral optimization model with semi-theoretical and semi-empirical model: The hyperspectral optimization model is constrained by a semi-theoretical and semi-empirical model (such as the logarithmic ratio model) to reduce the number of unknown parameters of the hyperspectral optimization model, so that the bottom reflectivity information can be obtained while obtaining the water depth and water body optical properties. This algorithm can break through the limitation of the number of visible light bands in multispectral images, but the single-phase inversion results in deep water areas are underestimated, which can easily cause large errors. 3) Dual-Band Log-Linear Analysis Model Based on Physics (P-DLA): This algorithm is based on the radiation transfer model. It extracts characteristic pixels of different depths and bottom textures from multispectral images to solve the unknown parameters of the model, thereby reducing the input of optical parameters and achieving good accuracy in different water depths. However, the calculation of key model parameters in the algorithm depends on the sampling of spectral pixels in the study area, which makes the inversion of single-phase images uncertain.

[0004] In summary, most current studies use single-phase satellite data, and single-phase images are easily affected by factors such as cloud cover and atmospheric interference, and contain a lot of noise, which will lead to poor image quality or partial information loss, thereby affecting the accuracy of water depth inversion. Therefore, the present invention provides a multi-phase fusion uncontrolled water depth inversion method based on remote sensing image bottom identification. Summary of the invention

[0005] The purpose of the present invention is to provide a multi-temporal fusion uncontrolled water depth inversion method based on remote sensing image bottom identification, so as to solve the problem mentioned in the above background technology that the existing shallow sea uncontrolled point water depth inversion algorithm has limitations, which will affect the water depth inversion accuracy.

[0006] The present invention is achieved by adopting the following technical solutions: A multi-temporal fusion uncontrolled water depth inversion method based on remote sensing image bottom identification includes the following steps: Step S1: Collect and process multi-temporal remote sensing image data and measured data; Step S2: Determine the inversion factor and construct the MM-P-DLA model, i.e., the Dual-Band Log-Linear Analysis Model Based on Physics-basedMulti-temporal Merge; Step S3: extracting characteristic pixels based on bottom soil identification of multi-temporal remote sensing images; Step S4: Perform preliminary water depth inversion based on the MM-P-DLA model to obtain water depth inversion results for each single time phase; Step S5: Based on the single-phase water depth inversion results, multi-phase uncontrolled water depth inversion fusion is performed based on the MM-P-DLA model to obtain the final inversion result.

[0007] Furthermore, in the step S1: pre-processing the data of multi-temporal remote sensing images for water depth inversion, and standardizing the measured data for accuracy verification.

[0008] Furthermore, in step S2: correlation analysis is performed on the reflectivity of each band and the water depth in the multi-temporal remote sensing image to obtain the correlation coefficient between the reflectivity of each band and the water depth, and the blue and green bands in the correlation coefficient are used as inversion factors to construct the MM-P-DLA model.

[0009] Furthermore, in step S3: different bottom materials are distinguished according to the brightness of the multi-temporal remote sensing images in shallow water areas, and adjacent pixel pairs of the same depth but different bottom materials, pixels of the same bottom material but different depths, deep water area pixels and waterline pixels are extracted from the processed multi-temporal remote sensing images.

[0010] Furthermore, in step S4: the water depth inversion results of each single time phase are obtained according to the preliminary inversion algorithm in the MM-P-DLA model. The expression of the preliminary inversion algorithm is as follows: , In the formula, for water depth; is the sum of the diffuse attenuation coefficients of the green band, is the ratio of the attenuation coefficients of the blue and green bands; , , is the remote sensing reflectivity of the secondary water surface, is the remote sensing reflectance of the sub-water surface in the optical deep water area; 1 and 2 represent the blue band and the green band respectively. , are the weight feature vectors of the blue and green bands respectively, It is the bottom parameter.

[0011] Furthermore, in step S5: multi-temporal uncontrolled water depth inversion fusion is performed according to the fusion processing algorithm in the MM-P-DLA model, and the expression of the fusion processing algorithm is as follows: , , In the formula, To minimize the function, is the value of the final inversion result.

[0012] Furthermore, the preprocessing includes geometric correction, radiation calibration, atmospheric correction, flare correction and water-land separation.

[0013] Furthermore, for the expression of the preliminary inversion algorithm, it is necessary to solve the optimal band rotation coefficient unit vector , the solution formula is as follows: , In the formula, To minimize the function, Represents a pair of adjacent pixels of the same depth but different substrates. and They correspond to two different types of substrate in a pair of adjacent pixels. is the number of adjacent pixel pairs.

[0014] Furthermore, for the expression of the preliminary inversion algorithm, it is necessary to solve the bottom parameters , when solving, extract the pixels of different bottom types at the waterline and calculate according to the reflectivity And take the average value, where The substrate type of the pixel.

[0015] The beneficial effects achieved by the present invention are: A multi-temporal fusion uncontrolled water depth inversion method based on remote sensing image bottom identification is provided. By sequentially collecting and processing data, building an MM-P-DLA model, extracting characteristic pixels, performing preliminary water depth inversion, and fusion of multi-temporal uncontrolled water depth inversion, the efficient inversion of shallow sea uncontrolled water depth can be completed with high inversion accuracy. This multi-temporal fusion uncontrolled water depth inversion method based on remote sensing image bottom identification is suitable for different satellites and different sea areas. The water depth information obtained based on this method can be directly used in shallow sea navigation, island and reef development, ecological restoration and other projects. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a flow chart of a multi-temporal fusion uncontrolled water depth inversion method based on remote sensing image bottom identification according to an embodiment of the present invention; Figure 2 This is an example of feature pixel extraction in the multi-temporal fusion uncontrolled water depth inversion method based on remote sensing image bottom identification according to the embodiment of the present invention. Figure Ⅰ ; Figure 3 This is an example of feature pixel extraction in the multi-temporal fusion uncontrolled water depth inversion method based on remote sensing image bottom identification according to the embodiment of the present invention. Figure II ; Figure 4 This is an example of the result after preliminary water depth inversion in the multi-temporal fusion uncontrolled water depth inversion method based on remote sensing image bottom identification according to the embodiment of the present invention. Figure Ⅰ ; Figure 5 This is an example of the result after preliminary water depth inversion in the multi-temporal fusion uncontrolled water depth inversion method based on remote sensing image bottom identification according to the embodiment of the present invention. Figure II ; Figure 6 This is an example of the result of multi-temporal uncontrolled water depth inversion fusion in the multi-temporal fusion uncontrolled water depth inversion method based on remote sensing image bottom soil recognition according to the embodiment of the present invention. Figure Ⅰ ; Figure 7 This is an example of the result of multi-temporal uncontrolled water depth inversion fusion in the multi-temporal fusion uncontrolled water depth inversion method based on remote sensing image bottom soil recognition according to the embodiment of the present invention. Figure II . DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0018] Example 1 This embodiment provides a multi-temporal fusion uncontrolled water depth inversion method based on remote sensing image bottom identification, which is applied to the water depth inversion of shallow sea areas without water depth control points. Figure 1 , the method comprises the following steps: Step S1: Collect and process multi-temporal remote sensing image data and measured data. Specifically: In this step, the data of multi-temporal remote sensing images are preprocessed to obtain accurate remote sensing reflectance, which is then used for subsequent water depth inversion. After collecting multispectral remote sensing images of different time series, the preprocessing to be carried out specifically includes geometric correction, radiation calibration, atmospheric correction, flare correction and water-land separation, among which: geometric correction can eliminate errors, ensure the correct correspondence between the pixel points and the actual geographical location in the data of multi-temporal remote sensing images, and thus ensure the consistency of the data of multi-temporal remote sensing images during fusion. Radiation calibration can convert the DN value of the original image into a radiance value, and different sensors have different radiation calibration methods. Atmospheric correction can eliminate the interference of factors such as atmospheric scattering and aerosols, and then obtain a surface reflectance close to the real one; after comparing atmospheric correction methods such as FLAASH, dark pixels and 6S, the FLAASH atmospheric correction method with the best effect is selected in this embodiment. Flare correction can remove the interference of solar flares on the sea surface, and the specific method of median filtering is adopted in this embodiment. Water-land separation can extract water body data, and the normalized difference water index (NDWI) is used in this embodiment to achieve it.

[0019] In this step, the measured data is standardized for accuracy verification. According to the acquisition time of the collected multi-temporal remote sensing images, the measured data is tidal corrected for the corresponding images.

[0020] Step S2: Determine the inversion factor and construct the MM-P-DLA model, i.e., a physically based dual-band log-linear model with multi-temporal fusion. Specifically: In this step, the correlation analysis between the reflectivity of each band and the water depth in the multi-temporal remote sensing images is carried out to obtain the correlation coefficient between the reflectivity of each band and the water depth. The blue and green bands in the correlation coefficient are used as inversion factors to construct the MM-P-DLA model.

[0021] Step S3: Extract characteristic pixels based on the bottom soil identification of multi-temporal remote sensing images. Specifically: In this step, different bottom materials are distinguished according to the brightness of multi-temporal remote sensing images in shallow water areas, and different types of feature pixels are extracted from the pre-processed multi-temporal remote sensing images. Please refer to Figure 2 and Figure 3 , specifically including adjacent pixel pairs of the same depth but different substrates, pixels of the same substrate but different depths, pixels in deep water areas, and pixels near the waterline.

[0022] Step S4: Perform preliminary water depth inversion based on the MM-P-DLA model to obtain the water depth inversion results of each single time phase, which can be referred to Figure 4 and Figure 5 . Specifically: In this step, the water depth inversion results of each single time phase are obtained according to the preliminary inversion algorithm in the MM-P-DLA model. The expression of the preliminary inversion algorithm is as follows: , In the formula, for water depth; is the sum of the diffuse attenuation coefficients of the green band, is the ratio of the attenuation coefficients of the blue and green bands; , , is the remote sensing reflectivity of the secondary water surface, is the remote sensing reflectance of the sub-water surface in the optical deep water area; 1 and 2 represent the blue band and the green band respectively. , are the weight feature vectors of the blue and green bands respectively, is the bottom parameter. Among them, for the expression of the preliminary inversion algorithm, it is necessary to solve the optimal band rotation coefficient unit vector , bottom parameters , blue-green band attenuation coefficient ratio and the sum of the diffuse attenuation coefficients of the green band , as follows: Ⅰ. Optimal band rotation coefficient unit vector

[0023] The dataset of adjacent pixel pairs with the same depth but different substrates is applied to the optimal band rotation coefficient unit vector To solve the problem, adjacent pixels can be recorded as and , using the minimum optimization function, solve according to the following solution formula: , In the formula, To minimize the function, Represents a pair of adjacent pixels of the same depth but different substrates. and They correspond to two different types of substrate in a pair of adjacent pixels. is the number of adjacent pixel pairs.

[0024] II. Sediment parameters

[0025] The strong absorption of water in the near-infrared band can be used to make the water body appear black in the near-infrared band image to identify the waterline, and then extract the pixels near the waterline. When the water depth of the waterline is 0, It can be approximately expressed as ; Therefore, solve the bottom parameters When , pixels of different bottom types at the waterline can be extracted, and then the reflectivity can be calculated. And take the average value, where The substrate type of the pixel.

[0026] III. Blue-green band attenuation coefficient ratio and the sum of the diffuse attenuation coefficients of the green band

[0027] The expressions of the preliminary inversion algorithm are applied to the blue and green bands respectively, and then the variables are eliminated. It can be seen that the ratio of the blue-green band attenuation coefficient is It can be regarded as a collection based on the same substrate but different depth standards. The regression slope of the data set.

[0028] The sum of the diffuse attenuation coefficients in the green band It can be solved by the QAA algorithm or based on the Hydrolight simulation. The QAA algorithm can use remote sensing reflectivity to solve the inherent optical properties of water bodies such as water absorption coefficient and water backscattering coefficient, and use the inherent optical properties to invert water quality parameters, including the diffuse attenuation coefficient of the green band. In the Hydrolight simulation, the inherent optical properties of the water body, bottom reflection characteristics and other parameters are input to simulate radiation and other output values.

[0029] Step S5: Based on the single-phase water depth inversion results, multi-phase uncontrolled water depth inversion fusion is performed based on the MM-P-DLA model to obtain the final inversion result, which can be referred to Figure 6 and Figure 7 . Specifically: In this step, multi-phase uncontrolled water depth inversion fusion is a method that uses remote sensing image data from multiple time points, combines it with a water depth inversion model for comprehensive processing, and then improves the accuracy and reliability of water depth inversion information through fusion. Specifically, on the basis of obtaining the water depth inversion results of each single-phase, the data at different time points are analyzed in time series and averaged, and then the water depth inversion results of multiple time points can be fused. This fusion process can effectively offset the errors that may exist in a single data source, and optimize and improve the accuracy and reliability of the final water depth inversion results. Among them, the expression of the fusion processing algorithm is as follows: , , In the formula, To minimize the function, is the value of the final inversion result.

[0030] To sum up, in the multi-temporal fusion uncontrolled water depth inversion method based on remote sensing image bottom identification based on high-resolution satellite remote sensing image bottom identification provided in this embodiment, by executing steps S1 to S5, the efficient inversion of shallow sea uncontrolled water depth can be completed with high inversion accuracy; this multi-temporal fusion uncontrolled water depth inversion method based on remote sensing image bottom identification is suitable for different satellites and different sea areas, and the water depth information obtained based on this method can be directly used in shallow sea navigation, island and reef development, ecological restoration and other projects.

[0031] It should be noted that the parts that are not described in detail or in detail in the above scheme, such as the specific implementation process of each preprocessing in step S1, the specific principles and processes of correlation analysis in step S2, etc., are all prior art, and do not belong to the improvements made by the present invention on the prior art, nor do they belong to the scope of protection of the technical scheme of the present invention. Therefore, they will not be repeated in this article.

[0032] Of course, the above contents are only preferred embodiments of the present invention and cannot be considered to limit the scope of the embodiments of the present invention. The present invention is not limited to the above examples, and equal changes and improvements made by ordinary technicians in the technical field within the essential scope of the present invention should all fall within the scope of the patent coverage of the present invention.

Claims

1. A multi-temporal fusion uncontrolled water depth inversion method based on remote sensing image bottom identification, characterized in that: The steps include: Step S1: Collect and process multi-temporal remote sensing image data and measured data; Step S2: determine the inversion factor and construct the MM-P-DLA model, i.e., a physically based dual-band log-linear model with multi-temporal fusion; Step S3: extracting characteristic pixels based on bottom soil identification of multi-temporal remote sensing images; Step S4: Perform preliminary water depth inversion based on the MM-P-DLA model to obtain water depth inversion results for each single time phase; Step S5: Based on the single-phase water depth inversion results, multi-phase uncontrolled water depth inversion fusion is performed based on the MM-P-DLA model to obtain the final inversion result.

2. The multi-temporal fusion uncontrolled water depth inversion method based on remote sensing image bottom identification according to claim 1 is characterized in that: In the step S1: pre-processing the data of multi-temporal remote sensing images for water depth inversion, and standardizing the measured data for accuracy verification.

3. The multi-temporal fusion uncontrolled water depth inversion method based on remote sensing image bottom identification according to claim 1 is characterized in that: In the step S2: correlation analysis is performed on the reflectivity of each band and the water depth in the multi-temporal remote sensing image to obtain the correlation coefficient between the reflectivity of each band and the water depth, and the blue and green bands in the correlation coefficient are used as inversion factors to construct the MM-P-DLA model.

4. The multi-temporal fusion uncontrolled water depth inversion method based on remote sensing image bottom identification according to claim 1 is characterized in that: In step S3: different bottom materials are distinguished according to the brightness of the multi-temporal remote sensing images in shallow water areas, and adjacent pixel pairs of the same depth but different bottom materials, pixels of the same bottom material but different depths, deep water area pixels and waterline pixels are extracted from the processed multi-temporal remote sensing images.

5. The multi-temporal fusion uncontrolled water depth inversion method based on remote sensing image bottom identification according to claim 1 is characterized in that: In step S4: the water depth inversion results of each single time phase are obtained according to the preliminary inversion algorithm in the MM-P-DLA model. The expression of the preliminary inversion algorithm is as follows: ， In the formula, for water depth; is the sum of the diffuse attenuation coefficients of the green band, is the ratio of the attenuation coefficients of the blue and green bands; , , is the remote sensing reflectivity of the secondary water surface, is the remote sensing reflectance of the sub-water surface in the optical deep water area; 1 and 2 represent the blue band and the green band respectively. , are the weight feature vectors of the blue and green bands respectively, It is the bottom parameter.

6. The multi-temporal fusion uncontrolled water depth inversion method based on remote sensing image bottom identification according to claim 5 is characterized in that: In step S5: multi-temporal uncontrolled water depth inversion fusion is performed according to the fusion processing algorithm in the MM-P-DLA model. The expression of the fusion processing algorithm is as follows: ， , In the formula, To minimize the function, is the value of the final inversion result.

7. The multi-temporal fusion uncontrolled water depth inversion method based on remote sensing image bottom identification according to claim 2 is characterized by: The preprocessing includes geometric correction, radiation calibration, atmospheric correction, flare correction and water-land separation.

8. The multi-temporal fusion uncontrolled water depth inversion method based on remote sensing image bottom identification according to claim 5 is characterized by: For the expression of the preliminary inversion algorithm, it is necessary to solve the optimal band rotation coefficient unit vector , the solution formula is as follows: ， In the formula, To minimize the function, Represents a pair of adjacent pixels of the same depth but different substrates. and They correspond to two different types of substrate in a pair of adjacent pixels. is the number of adjacent pixel pairs.

9. The multi-temporal fusion uncontrolled water depth inversion method based on remote sensing image bottom identification according to claim 5 is characterized by: For the expression of the preliminary inversion algorithm, it is necessary to solve the bottom parameters , when solving, extract the pixels of different bottom types at the waterline and calculate according to the reflectivity And take the average value, where The substrate type of the pixel.

Images