CN117274831B - Offshore turbid water body depth inversion method based on machine learning and hyperspectral satellite remote sensing image - Google Patents
Offshore turbid water body depth inversion method based on machine learning and hyperspectral satellite remote sensing image Download PDFInfo
- Publication number
- CN117274831B CN117274831B CN202311131798.9A CN202311131798A CN117274831B CN 117274831 B CN117274831 B CN 117274831B CN 202311131798 A CN202311131798 A CN 202311131798A CN 117274831 B CN117274831 B CN 117274831B
- Authority
- CN
- China
- Prior art keywords
- data
- remote sensing
- water depth
- satellite remote
- sensing image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 title claims abstract description 190
- 238000000034 method Methods 0.000 title claims abstract description 42
- 238000010801 machine learning Methods 0.000 title claims abstract description 40
- 238000012937 correction Methods 0.000 claims abstract description 29
- 238000012545 processing Methods 0.000 claims abstract description 17
- 238000007781 pre-processing Methods 0.000 claims abstract description 15
- 238000003384 imaging method Methods 0.000 claims abstract description 10
- 238000012549 training Methods 0.000 claims abstract description 10
- 238000004422 calculation algorithm Methods 0.000 claims description 18
- 238000004364 calculation method Methods 0.000 claims description 13
- 238000002310 reflectometry Methods 0.000 claims description 11
- 230000003595 spectral effect Effects 0.000 claims description 5
- 230000003044 adaptive effect Effects 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 4
- 230000002159 abnormal effect Effects 0.000 claims description 3
- 238000002790 cross-validation Methods 0.000 claims description 3
- 238000000926 separation method Methods 0.000 claims description 2
- 210000000476 body water Anatomy 0.000 abstract description 8
- 238000004088 simulation Methods 0.000 description 6
- 238000011160 research Methods 0.000 description 5
- 241000272525 Anas platyrhynchos Species 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000007613 environmental effect Effects 0.000 description 3
- 238000005259 measurement Methods 0.000 description 3
- 238000001228 spectrum Methods 0.000 description 3
- 230000005855 radiation Effects 0.000 description 2
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 description 1
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012876 topography Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/766—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using regression, e.g. by projecting features on hyperplanes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/30—Assessment of water resources
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Software Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Astronomy & Astrophysics (AREA)
- Remote Sensing (AREA)
- Image Processing (AREA)
Abstract
The invention provides a near-shore turbid water body water depth inversion method based on machine learning and hyperspectral satellite remote sensing images, which comprises the following steps: obtaining a Landsat-8 original hyperspectral satellite remote sensing image of a target area; preprocessing the acquired Landsat-8 original hyperspectral satellite remote sensing image; extracting a corresponding data set of the preprocessed hyperspectral satellite remote sensing image data points by using a remote sensing image processing platform; calculating corresponding instantaneous tide level data of water depth points in the actually measured water depth data set according to the imaging time of the satellite remote sensing image, carrying out tide correction, and carrying out data matching with the satellite remote sensing image data set; combining the extracted hyperspectral satellite remote sensing image characteristic dataset and the tidal corrected actual measured water depth dataset to construct a training sample dataset and train a machine learning model; and inverting the water depth data of the target area by using the trained machine learning model, and performing reverse tide correction on the inversion result. The applicability of the satellite remote sensing inversion water depth method in the near-shore turbid water body is improved.
Description
Technical Field
The invention relates to the technical field of hyperspectral satellite remote sensing image processing technology and offshore turbid water body water depth inversion, in particular to a method for offshore turbid water body water depth inversion based on machine learning and hyperspectral satellite remote sensing images.
Background
As an important topography element, accurate water depth data has important significance for research of offshore areas and construction and maintenance of ocean engineering facilities. With the continuous development and perfection of the traditional sounding technology, the traditional sounding tools such as the shipborne sonar, the airborne laser radar and the like are widely applied, but the application cost is high, the manual consumption is large, the measurement period is long, the measurement is easily limited by natural conditions such as environment, weather and the like, and the large-scale measurement work can not be carried out in certain areas, particularly in partial offshore shallow water areas. In recent years, satellite remote sensing technology is rapidly developed, and inversion of water depth data by using satellite remote sensing images has become one of research hotspots. Compared with the traditional water depth measuring means, the satellite remote sensing inversion water depth method is low in cost, can meet the dynamic and large-range water depth observation requirement of a certain water area in a short time, is not limited by natural conditions such as regional positions and environments, and can achieve higher space-time resolution in inversion water depth data.
At present, a satellite remote sensing inversion water depth method is mainly based on an optical radiation transmission theory, and the water depth value is inverted by utilizing the water-leaving reflectivity data. However, the water-leaving reflectivity data is easily influenced by the turbidity of the water body, the submarine geology and other regional environmental characteristics, so that the uncertainty of the water depth inversion result is increased, and the applicability of the conventional satellite remote sensing water depth inversion method in a turbid water area is influenced. In recent years, a machine learning algorithm is rapidly developed, mathematical relations between data features and corresponding targets can be well learned, a data set with relatively complex and stronger uncertainty is processed, predictions are rapidly made, influences of various noises and interferences on prediction results are greatly reduced, good application results are provided in the field of water depth inversion, but most of the machine learning algorithm only selects wave band information of satellite remote sensing images as input features to establish a water depth inversion model, the input features are limited, and relatively reliable high-resolution water depth inversion data cannot be obtained.
Disclosure of Invention
According to the shortcomings of the existing water depth inversion method, the method for inverting the water depth of the near-shore turbid water body based on machine learning and hyperspectral satellite remote sensing images is provided, and the spectral characteristics and the spatial characteristics of Landsat 8 hyperspectral satellite remote sensing images are utilized to be combined with normalized differential water body indexes to be used as input characteristics of a machine learning model; and a gradient lifting decision regression tree is selected as a base learner, and is fused with a self-adaptive lifting algorithm, so that a water depth inversion model based on a machine learning algorithm is established, and the applicability of the satellite remote sensing inversion water depth method in a near-shore turbid water body is improved.
The invention adopts the following technical means:
A method for inverting the water depth of a near-shore turbid water body based on machine learning and hyperspectral satellite remote sensing images comprises the following steps:
S1, acquiring Landsat-8 original hyperspectral satellite remote sensing images of a target area;
s2, preprocessing the acquired Landsat-8 original hyperspectral satellite remote sensing image by using a remote sensing image processing platform;
S3, extracting a corresponding data set of image data points by using a remote sensing image processing platform based on the preprocessed hyperspectral satellite remote sensing image;
s4, calculating corresponding instantaneous tide level data of water depth points in the actually measured water depth data set according to imaging time of the satellite remote sensing image, correcting tide, and performing data matching with the satellite remote sensing image data set;
S5, combining the extracted hyperspectral satellite remote sensing image characteristic dataset and the tidal corrected actual measured water depth dataset to construct a training sample dataset and train a machine learning model;
S6, inverting the water depth data of the target area by using the trained machine learning model, and performing reverse tide correction on the inversion result.
Further, the Landsat-8 original hyperspectral satellite remote sensing image obtained in the step S1 has cloud cover rate meeting the standard without cloud cover.
Further, the step S2 specifically includes:
S21, performing geometric correction, radiometric calibration and atmospheric correction preprocessing operations on the acquired Landsat-8 original hyperspectral satellite remote sensing image by using a relevant tool of a remote sensing image processing platform, and splicing satellite remote sensing images after mosaic processing;
S22, extracting a water body by a threshold method according to the normalized differential water body index, and carrying out amphibious separation, wherein the calculation formula of the normalized differential water body index is as follows:
Wherein NDWI represents a normalized differential water index; r rs (green) and R rs (NIR) respectively represent the water-leaving reflectivity data of a green wave band and a near infrared wave band in the Landsat-8 satellite remote sensing image.
Further, in the step S3, the corresponding data set of the extracted image data points includes: and the blue wave band of the Landsat-8 hyperspectral satellite remote sensing image, the green wave band of the Landsat-8 hyperspectral satellite remote sensing image and the red wave band of the Landsat-8 hyperspectral satellite remote sensing image are respectively corresponding to the image data points, and the water-leaving reflectivity data and the geographic coordinate data of the Landsat-8 hyperspectral satellite remote sensing image are obtained.
Further, the step S4 specifically includes:
S41, calculating instantaneous tide level data of a water depth point in the measured water depth data set corresponding to the imaging time of the satellite remote sensing image by using a relevant tide level calculation toolbox, and performing tide correction work, wherein a correction formula is as follows:
dt＝dm+e
Wherein d t represents instantaneous water depth data; d m represents measured water depth data; e represents instantaneous water depth data, and if the water depth data is smaller than 0m after tidal correction, rejecting;
s42: and selecting a pixel point closest to the actually measured water depth data as a matching data point of the water depth point, and eliminating the actually measured water depth data point if the matched reflectivity data is abnormal.
Further, the step S5 specifically includes:
s51, selecting the blue wave band of the Landsat-8 hyperspectral satellite remote sensing image, the green wave band of the Landsat-8 hyperspectral satellite remote sensing image and the red wave band of the Landsat-8 hyperspectral satellite remote sensing image corresponding to the image data points as spectral features in the input data features;
S52, adopting a universal transverse ink card support grid system, and selecting longitude values and latitude values corresponding to the wave bands in the step S51 in the satellite remote sensing image as space features in the input data features;
s53, selecting a normalized differential water index as the last input data characteristic, wherein a data sample set is expressed as follows:
D＝{(x1,y1),(x2,y2),...,(xi,yi)}，i＝1，2，...，N
Wherein ,xi＝{Rrs(blue)i,Rrs(green)i,Rrs(red)i,Longitudie,Latituid,e NDWiW denotes the ith input data feature; y i represents a target value corresponding to the data characteristic, namely measured water depth data after tidal correction;
S54, performing Z-Score standardized preprocessing on all data in a data sample set, wherein the preprocessing formula is as follows:
In the method, in the process of the invention, Representing the i-th normalized data in the dataset; d i denotes the i-th data in the original data; μ represents the mean of the raw data; sigma represents standard deviation of the original data;
S55, selecting a gradient lifting decision regression tree as a base learner, fusing the gradient lifting decision regression tree with an adaptive lifting algorithm, and establishing a water depth inversion model based on a machine learning algorithm;
S56, inputting a sample data set into the established water depth inversion model, training the water depth inversion model, selecting a Bayesian optimization algorithm to find the optimal super-parameter combination of the model during training, and applying a k-fold cross validation method to prevent overfitting.
Further, the step S6 specifically includes:
s61, inverting the water depth data of the target area by using the trained water depth inversion model to obtain instantaneous water depth inversion data;
S62, calculating instantaneous tide level data corresponding to the imaging time of the satellite remote sensing image of the target area by using a relevant tide level calculation toolbox, and carrying out reverse tide correction on instantaneous water depth inversion data to obtain steady-state water depth data.
Compared with the prior art, the invention has the following advantages:
1. The near-shore turbid water body water depth inversion method based on the machine learning and hyperspectral satellite remote sensing image provided by the invention uses Landsat-8 satellite remote sensing images, is convenient to acquire, has rich spectral band information, has high resolution, and can meet the water depth data set requirements in most application scenes.
2. According to the offshore turbid water body water depth inversion method based on the machine learning and hyperspectral satellite remote sensing image, on the basis of a conventional common method, the longitude value (Longitude), the Latitude value (Latitude) and the normalized differential water body index (NDWI) of each data point are additionally selected as input features, so that the water body spectrum and the spatial features can be fully utilized, the interference of complex environmental information of the offshore turbid water body is effectively reduced, the learning capacity of a model is further enhanced, the water depth inversion accuracy of the model is improved, and meanwhile, good robustness is ensured;
3. compared with other common water depth data sets, the water depth data set utilized by the inversion method provided by the invention has good space coverage rate and resolution, and compared with other satellite remote sensing inversion water depth results, the inversion method based on machine learning and hyperspectral satellite remote sensing images is more reliable, and has considerable application prospects in offshore sea area research and ocean engineering projects.
Based on the reasons, the method can be widely popularized in the fields of hyperspectral satellite remote sensing image processing technology, near-shore turbid water body water depth inversion and the like.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings may be obtained according to the drawings without inventive effort to a person skilled in the art.
Fig. 1 is a position diagram of a duck green river mouth area provided by an embodiment of the invention.
FIG. 2 is a block diagram of a water depth inversion model based on an AdaBoost-GBDT machine learning algorithm provided by an embodiment of the invention.
Fig. 3 is a spatial distribution diagram of 4 water depth data products in a duck green river mouth area according to an embodiment of the present invention.
Detailed Description
In order that those skilled in the art will better understand the present invention, a technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in which it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, shall fall within the scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and the claims of the present invention and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the invention described herein may be implemented in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
The invention provides a near-shore turbid water body water depth inversion method based on machine learning and hyperspectral satellite remote sensing images, which comprises the following steps:
S1, acquiring Landsat-8 original hyperspectral satellite remote sensing images of a target area;
s2, preprocessing the acquired Landsat-8 original hyperspectral satellite remote sensing image by using a remote sensing image processing platform;
S3, extracting a corresponding data set of image data points by using a remote sensing image processing platform based on the preprocessed hyperspectral satellite remote sensing image;
s4, calculating corresponding instantaneous tide level data of water depth points in the actually measured water depth data set according to imaging time of the satellite remote sensing image, correcting tide, and performing data matching with the satellite remote sensing image data set;
S5, combining the extracted hyperspectral satellite remote sensing image characteristic dataset and the tidal corrected actual measured water depth dataset to construct a training sample dataset and train a machine learning model;
S6, inverting the water depth data of the target area by using the trained machine learning model, and performing reverse tide correction on the inversion result.
In a specific implementation, as shown in fig. 1, in the step S1, the target area is a duck-green river mouth area, 2 original hyperspectral satellite remote sensing images of Landsat-8 are obtained in total, and the cloud coverage of the images is 0.02, so that the standard without cloud coverage is satisfied.
In specific implementation, as a preferred embodiment of the present invention, the step S2 specifically includes:
S21, performing geometric correction, radiometric calibration and atmospheric correction preprocessing operations on the acquired Landsat-8 original hyperspectral satellite remote sensing image by using a relevant tool of a remote sensing image processing platform, and splicing satellite remote sensing images after mosaic processing; in this embodiment, since the used Landsat-8Collection 2Level-1 level high-spectrum satellite remote sensing image has been subjected to geometric correction preprocessing operation, in this embodiment, only the Flash atmospheric correction module in ENVI 5.6 is used to perform radiation calibration and atmospheric correction preprocessing operations on the acquired Landsat-8 high-spectrum satellite remote sensing image, and the satellite remote sensing image after mosaic processing is spliced;
S22, extracting a water body by a threshold method according to a Normalized differential water body Index (Normalized DIFFERENCE WATER Index, NDWI), and separating the water and the land, wherein the calculation formula of the Normalized differential water body Index is as follows:
Wherein NDWI represents a normalized differential water index; r rs (green) and R rs (NIR) respectively represent the reflectance data of the green wave band (green) and the near infrared wave band (NEAR INFRARED, NIR) in the Landsat-8 satellite remote sensing image.
In a specific implementation, as a preferred embodiment of the present invention, in the step S3, the corresponding data set of the extracted image data points includes: and the blue wave band (blue) of the Landsat-8 hyperspectral satellite remote sensing image, the green wave band (green) of the Landsat-8 hyperspectral satellite remote sensing image and the red wave band (red) of the Landsat-8 hyperspectral satellite remote sensing image which correspond to the image data points respectively, and geographic coordinate data. In this embodiment, the ENVI 5.6 toolbox of the remote sensing image processing platform is utilized to extract the corresponding dataset of image data points. The geographic coordinates are selected from universal transverse ink card support grid system (UTM).
In specific implementation, as a preferred embodiment of the present invention, the step S4 specifically includes:
S41, calculating instantaneous tide level data of a water depth point in the measured water depth data set corresponding to the imaging time of the satellite remote sensing image by using an MATLAB TMD tool box, and performing tide correction, wherein the correction formula is as follows:
dt＝dm+e
Wherein d t represents instantaneous water depth data; d m represents measured water depth data; e represents instantaneous water depth data, and if the water depth data is smaller than 0m after tidal correction, rejecting;
s42: and selecting a pixel point closest to the actually measured water depth data as a matching data point of the water depth point, and eliminating the actually measured water depth data point if the matched reflectivity data is abnormal.
In specific implementation, as a preferred embodiment of the present invention, the step S5 specifically includes:
s51, selecting the blue wave band of the Landsat-8 hyperspectral satellite remote sensing image, the green wave band of the Landsat-8 hyperspectral satellite remote sensing image and the red wave band of the Landsat-8 hyperspectral satellite remote sensing image corresponding to the image data points as spectral features in the input data features;
S52, adopting a universal transverse ink card support grid system (UTM), and selecting a longitude value (Longitude) and a Latitude value (Latitude) corresponding to the wave band in the step S51 in the satellite remote sensing image as space characteristics in the input data characteristics;
S53, selecting a Normalized Differential Water Index (NDWI) as the last input data characteristic, wherein the data sample set is expressed as follows:
D＝{(x1,y1),(x2,y2),...,(xi,yi)}，i＝1，2，...，N
Wherein ,xi＝{Rrs(blue)i,Rrs(green)i,Rrs(red)i,Longitudi eLatituide NDWiW denotes the ith input data feature; y i represents a target value corresponding to the data characteristic, namely measured water depth data after tidal correction;
S54, in order to ensure the effectiveness of data and eliminate the influence caused by different data sizes, performing Z-Score standardized preprocessing on all data in a data sample set, wherein the preprocessing formula is as follows:
In the method, in the process of the invention, Representing the i-th normalized data in the dataset; d i denotes the i-th data in the original data; μ represents the mean of the raw data; sigma represents standard deviation of the original data;
S55, selecting a gradient lifting decision regression tree (Gradient Boosting Decision Tree Regressor, GBDTR) as a base learner, fusing the gradient lifting decision regression tree with an adaptive lifting algorithm (Adaptive Boosting, adaBoost), and establishing a water depth inversion model AdaBoost-GBDT based on a machine learning algorithm, as shown in FIG. 2.
S56, inputting a sample data set into the established water depth inversion model, training the water depth inversion model, selecting a Bayesian optimization (Bayesian Optimization) algorithm to find the optimal super-parameter combination of the model during training, and applying a k-fold cross validation method to prevent overfitting.
In specific implementation, as a preferred embodiment of the present invention, the step S6 specifically includes:
s61, inverting the water depth data of a target area (a duck green river mouth area) by using the trained water depth inversion model to obtain instantaneous water depth inversion data;
S62, calculating instantaneous tide level data corresponding to imaging time of satellite remote sensing images of a target area (a duck green river mouth area) by using an MATLAB TMD tool box, and carrying out reverse tide correction on instantaneous water depth inversion data to obtain steady-state water depth data.
Example 1
In order to evaluate the accuracy of the water depth inversion data set in the embodiment, the water depth inversion results of the data set and a water depth logarithmic band ratio formula (table 1) of several common satellite remote sensing inversion water depths are compared; the water depth predictive performance of several common machine learning models was then compared to that of the AdaBoost-GBDT machine learning model, and the correlation coefficients (Correlation Coefficient, R), mean absolute error (Mean Absolute Error, MAE), root mean square error (Root Mean Square Error, RMSE), and mean absolute percent error (Mean Absolute Percentage Error, MAPE) were compared for four indices, with the comparison results shown in tables 2 and 3;
Table 16 common satellite remote sensing inversion water depth logarithmic band ratio calculation
TABLE 2 statistical results of the log band ratio calculation of the water depth inversion performance
As can be seen from tables 2 and 3, compared with the 6 log band ratio formulas, the water depth inversion model of the machine learning model has obvious advantages, and the water depth inversion accuracy of the machine learning model is better than the log band ratio formulas, both in the whole and in each water depth interval; compared with other machine learning models, the Adaboost-GBDT machine learning model has the best performance, and in addition, the water depth inversion performance of each machine learning model is obviously improved by introducing an AdaBoost algorithm. Experimental results show that the water depth inversion model based on the AdaBoost-GBDT machine learning algorithm has higher accuracy and reliability in the water depth inversion work of the near-shore turbid water body, and has good application prospect.
Table 3 statistical results of machine learning model water depth inversion performance
Example 2
In addition to the data accuracy aspect, the present example water depth inversion dataset was also compared to 2 other common water depth datasets: ETOPO Water depth data set, actually measured Water depth data set and logarithmic band ratio equation (f equation in Table 1, hereinafter this equation is referred to as logarithmic band ratio equation) mainly includes: firstly, the space coverage rate and resolution ratio of water depth data, and secondly, the application result of the water depth data on a FVCOM marine environment model is as follows:
as shown in fig. 3, the spatial coverage rate and resolution ratio of the 4 water depth data sets can be seen, the distribution condition of ETOPO water depth data sets (fig. 3 a) commonly used in offshore research is relatively uniform, all areas including offshore areas are covered by data, but the spatial resolution ratio is still very low, and the actual requirements of offshore environment research and coastal engineering are relatively difficult to meet; the measured water depth data set (fig. 3 b) has good reliability in terms of data accuracy, but due to the deficiency of the conventional water depth measuring tool, the coverage of the data is limited, the water depth data of a large amount of areas is sparse or even has no data, and the resolution of the water depth data of only a small amount of areas can meet the requirement. Compared with the two water depth data sets, the water depth data set based on the logarithmic band ratio formula has better resolution and wider data coverage (figure 3 c); likewise, the water depth dataset based on the AdaBoost-GBDT water depth inversion model also has high resolution and good spatial coverage.
Comparison of the results of the application of the 4 water depth datasets on FVCOM marine environmental models (table 4). On the tide level calculation results, the simulation results of the 4 water depth data sets all show higher R values, wherein the RMSE of the ETOPO water depth data simulation results is slightly higher than that of other 3 water depth data, the accuracy of the tide level simulation results of all water depth products is relatively similar, and the R values are all about 0.99; in terms of flow direction calculation, the result of ETOPO water depth data is poor, the simulation result of a water depth data set of a water depth inversion model based on an AdaBoost-GBDT machine learning algorithm in other 3 water depth data is the best in overall performance, the measured water depth data set is inferior, and the simulation result of an improved logarithmic band ratio calculation water depth data set is larger in error between a small tide period of a Y02 station and a large tide period of the Y03 station; in terms of flow rate calculation, the simulation result of the ETOPO water depth data set is poor, particularly in the period of a large tide, the result is far lower than the normal level, the actual measured water depth data set is poor in the period of the large tide at the Y03 station, and the water depth data set based on the AdaBoost-GBDT water depth inversion model is better in performance.
Comparison of application results of 44 water depth data products in FVCOM ocean numerical model
In summary, in the above embodiment of the present application, by providing a method for inverting the depth of a near-shore turbid water body based on machine learning and hyperspectral satellite remote sensing images, the method mainly includes: by utilizing the data characteristics of Landsat 8 hyperspectral satellite images and combining with an AdaBoost-GBDT machine learning algorithm, the built near-shore turbid water body water depth inversion model can effectively reduce the interference of the turbid water body complex environment information, the inversion can obtain a more accurate water depth data set, and the water depth data set has good application prospect.
The foregoing embodiment numbers of the present invention are merely for the purpose of description, and do not represent the advantages or disadvantages of the embodiments.
In the foregoing embodiments of the present invention, the descriptions of the embodiments are emphasized, and for a portion of this disclosure that is not described in detail in this embodiment, reference is made to the related descriptions of other embodiments.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some or all of the technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit of the invention.
Claims (2)
1. A method for inverting the water depth of a near-shore turbid water body based on machine learning and hyperspectral satellite remote sensing images is characterized by comprising the following steps:
S1, acquiring Landsat-8 original hyperspectral satellite remote sensing images of a target area;
the Landsat-8 original hyperspectral satellite remote sensing image obtained in the step S1 has cloud cover rate meeting the standard of no cloud cover;
S2, preprocessing the acquired Landsat-8 original hyperspectral satellite remote sensing image by using a remote sensing image processing platform, wherein the method specifically comprises the following steps of:
S21, performing geometric correction, radiometric calibration and atmospheric correction preprocessing operations on the acquired Landsat-8 original hyperspectral satellite remote sensing image by using a relevant tool of a remote sensing image processing platform, and splicing satellite remote sensing images after mosaic processing;
S22, extracting a water body by a threshold method according to the normalized differential water body index, and carrying out amphibious separation, wherein the calculation formula of the normalized differential water body index is as follows:
Wherein NDWI represents a normalized differential water index; r rs (green) and R rs (NIR) respectively represent the water-leaving reflectivity data of a green wave band and a near infrared wave band in the Landsat-8 satellite remote sensing image;
S3, extracting a corresponding data set of image data points by using a remote sensing image processing platform based on the preprocessed hyperspectral satellite remote sensing image;
In step S3, the corresponding data set of the extracted image data points includes: the method comprises the steps of obtaining water-leaving reflectivity data and geographic coordinate data of a blue wave band of a Landsat-8 hyperspectral satellite remote sensing image, a green wave band of the Landsat-8 hyperspectral satellite remote sensing image and a red wave band of the Landsat-8 hyperspectral satellite remote sensing image, which correspond to image data points respectively;
s4, calculating corresponding instantaneous tide level data of water depth points in the actually measured water depth data set according to the imaging time of the satellite remote sensing image, correcting tide, and performing data matching with the satellite remote sensing image data set, wherein the method specifically comprises the following steps of:
S41, calculating instantaneous tide level data of a water depth point in the measured water depth data set corresponding to the imaging time of the satellite remote sensing image by using a relevant tide level calculation toolbox, and performing tide correction work, wherein a correction formula is as follows:
dt＝dm+e
Wherein d t represents instantaneous water depth data; d m represents measured water depth data; e represents instantaneous water depth data, and if the water depth data is smaller than 0m after tidal correction, rejecting;
s42: selecting a pixel point closest to the actually measured water depth data as a matching data point of the water depth point, and eliminating the actually measured water depth data point if the matched reflectivity data is abnormal;
S5, combining the extracted hyperspectral satellite remote sensing image characteristic dataset and the tidal corrected actual measured water depth dataset to construct a training sample dataset and train a machine learning model;
The step S5 specifically includes:
s51, selecting the blue wave band of the Landsat-8 hyperspectral satellite remote sensing image, the green wave band of the Landsat-8 hyperspectral satellite remote sensing image and the red wave band of the Landsat-8 hyperspectral satellite remote sensing image corresponding to the image data points as spectral features in the input data features;
S52, adopting a universal transverse ink card support grid system, and selecting longitude values and latitude values corresponding to the wave bands in the step S51 in the satellite remote sensing image as space features in the input data features;
s53, selecting a normalized differential water index as the last input data characteristic, wherein a data sample set is expressed as follows:
D＝{(x1,y1),(x2,y2),...,(xi,yi)}，i＝1，2，...，N
Wherein ,xi＝{Rrs(blue)i,Rrs(green)i,Rrs(red)i,Longitudei,Latitudei,NDWIi} denotes the ith input data feature; y i represents a target value corresponding to the data characteristic, namely measured water depth data after tidal correction; r rs (blue) represents the water-leaving reflectivity data of a blue wave band in the Landsat-8 satellite remote sensing image; r rs (red) represents the water-leaving reflectivity data of the red wave band in the Landsat-8 satellite remote sensing image; longitude i represents a longitude value corresponding to a band; latitude i represents the Latitude value corresponding to the band;
S54, performing Z-Score standardized preprocessing on all data in a data sample set, wherein the preprocessing formula is as follows:
In the method, in the process of the invention, Representing the i-th normalized data in the dataset; d i denotes the i-th data in the original data; μ represents the mean of the raw data; sigma represents standard deviation of the original data;
S55, selecting a gradient lifting decision regression tree as a base learner, fusing the gradient lifting decision regression tree with an adaptive lifting algorithm, and establishing a water depth inversion model based on a machine learning algorithm;
S56, inputting a sample data set into the established water depth inversion model, training the water depth inversion model, selecting a Bayesian optimization algorithm to find the optimal super-parameter combination of the model during training, and applying a k-fold cross validation method to prevent overfitting;
S6, inverting the water depth data of the target area by using the trained machine learning model, and performing reverse tide correction on the inversion result.
2. The method for inverting the water depth of a near-shore turbid water body based on machine learning and hyperspectral satellite remote sensing images according to claim 1, wherein the step S6 specifically comprises:
s61, inverting the water depth data of the target area by using the trained water depth inversion model to obtain instantaneous water depth inversion data;
S62, calculating instantaneous tide level data corresponding to the imaging time of the satellite remote sensing image of the target area by using a relevant tide level calculation toolbox, and carrying out reverse tide correction on instantaneous water depth inversion data to obtain steady-state water depth data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311131798.9A CN117274831B (en) | 2023-09-04 | 2023-09-04 | Offshore turbid water body depth inversion method based on machine learning and hyperspectral satellite remote sensing image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311131798.9A CN117274831B (en) | 2023-09-04 | 2023-09-04 | Offshore turbid water body depth inversion method based on machine learning and hyperspectral satellite remote sensing image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117274831A CN117274831A (en) | 2023-12-22 |
| CN117274831B true CN117274831B (en) | 2024-06-14 |
Family
ID=89201859
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311131798.9A Active CN117274831B (en) | 2023-09-04 | 2023-09-04 | Offshore turbid water body depth inversion method based on machine learning and hyperspectral satellite remote sensing image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117274831B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117688847B (en) * | 2024-02-02 | 2024-05-31 | 自然资源部第二海洋研究所 | Sea gas oxygen flux remote sensing inversion method based on control mechanism and machine learning |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109657392A (en) * | 2018-12-28 | 2019-04-19 | 北京航空航天大学 | A kind of high-spectrum remote-sensing inversion method based on deep learning |
Family Cites Families (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9046363B2 (en) * | 2012-04-27 | 2015-06-02 | SATOP GmbH | Using multispectral satellite data to determine littoral water depths despite varying water turbidity |
| CN110009213B (en) * | 2019-03-28 | 2021-10-22 | 交通运输部水运科学研究所 | Method for tracking, monitoring and evaluating ecological influence of channel engineering based on long-time sequence satellite remote sensing data |
| JP7491779B2 (en) * | 2020-08-31 | 2024-05-28 | 株式会社日立製作所 | Flood depth estimation device and method |
| CN113639716A (en) * | 2021-07-29 | 2021-11-12 | 北京航空航天大学 | Depth residual shrinkage network-based water depth remote sensing inversion method |
| CN113793374B (en) * | 2021-09-01 | 2023-12-22 | 自然资源部第二海洋研究所 | Method for inverting water depth based on water quality inversion result by improved four-band remote sensing image QAA algorithm |
| CN114117886A (en) * | 2021-10-28 | 2022-03-01 | 南京信息工程大学 | Water depth inversion method for multispectral remote sensing |
| CN114201732A (en) * | 2021-11-24 | 2022-03-18 | 中国人民解放军92859部队 | Sentinel-2A image-based shallow sea water depth inversion method |
| CN114724045A (en) * | 2022-03-14 | 2022-07-08 | 陕西九州遥感信息技术有限公司 | Method for inverting submarine topography of shallow sea on land |
| CN114993268A (en) * | 2022-04-13 | 2022-09-02 | 南京信息工程大学 | Water depth inversion method and device combined with Catboost and storage medium |
| CN115376002A (en) * | 2022-08-29 | 2022-11-22 | 上海海洋大学 | Multispectral satellite remote sensing depth measurement method based on stacking integrated model |
| CN115690385A (en) * | 2022-09-07 | 2023-02-03 | 北京海舶无人船科技有限公司 | Water quality prediction method, system, equipment and medium based on multispectral image |
| CN115856925A (en) * | 2022-10-26 | 2023-03-28 | 中国科学院西安光学精密机械研究所 | Multispectral remote sensing image water depth inversion method, medium and equipment based on chart data |
| CN115761463A (en) * | 2022-11-02 | 2023-03-07 | 中国地质调查局军民融合地质调查中心 | Shallow sea water depth inversion method, system, equipment and medium |
| CN116187168A (en) * | 2022-12-30 | 2023-05-30 | 中国航天科技创新研究院 | Method for improving water depth inversion accuracy based on neural network-gravity information wavelet decomposition |
| CN116011342B (en) * | 2023-02-08 | 2023-07-21 | 中国自然资源航空物探遥感中心 | All-weather reconstruction method for high-resolution thermal infrared surface temperature |
| CN116543312B (en) * | 2023-05-08 | 2023-10-20 | 中国水利水电科学研究院 | Construction and updating method of remote sensing inversion sample library of water quality parameters of large-scale river and lake |
-
2023
- 2023-09-04 CN CN202311131798.9A patent/CN117274831B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109657392A (en) * | 2018-12-28 | 2019-04-19 | 北京航空航天大学 | A kind of high-spectrum remote-sensing inversion method based on deep learning |
Non-Patent Citations (1)
| Title |
|---|
| 基于遥感影像的水深反演方法研究;孟然;《中国优秀硕士学位论文全文数据库 基础科学辑》;20230215;第2023年卷(第2期);第18、21页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117274831A (en) | 2023-12-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11521379B1 (en) | Method for flood disaster monitoring and disaster analysis based on vision transformer | |
| CN112070234B (en) | Water chlorophyll and phycocyanin land-based remote sensing machine learning algorithm under complex scene | |
| CN113156439B (en) | SAR wind field and sea wave joint inversion method and system based on data driving | |
| CN117274831B (en) | Offshore turbid water body depth inversion method based on machine learning and hyperspectral satellite remote sensing image | |
| Chybicki | Mapping south baltic near-shore bathymetry using Sentinel-2 observations | |
| CN102176001A (en) | Permeable band ratio factor-based water depth inversion method | |
| CN111008664B (en) | Hyperspectral sea ice detection method based on space-spectrum combined characteristics | |
| CN112013822A (en) | Multispectral remote sensing water depth inversion method based on improved GWR model | |
| CN114297938B (en) | Inversion method of optical shallow water depth based on neural network | |
| CN111738347B (en) | Sea wave direction spectrum correction method and device, storage medium and electronic equipment | |
| Sowers et al. | Standardized geomorphic classification of seafloor within the United States Atlantic canyons and continental margin | |
| CN114676773A (en) | Arctic sea ice classification method based on SAR data | |
| CN115097451A (en) | Sea wave multi-parameter inversion method and system based on SAR satellite data | |
| CN116721243B (en) | Deep learning atmosphere correction method and system based on spatial spectrum feature constraint | |
| CN117664874A (en) | Estuary area salt tide tracing monitoring method and system based on space-earth combination | |
| Abessolo et al. | Error proxies in video-based depth inversion: Temporal celerity estimation | |
| CN115235431B (en) | Shallow sea water depth inversion method and system based on spectrum layering | |
| Liu et al. | Evaluation of the effectiveness of multiple machine learning methods in remote sensing quantitative retrieval of suspended matter concentrations: A case study of Nansi Lake in North China | |
| CN112782690B (en) | Offshore non-ocean waveform detection classification method and system for spaceborne radar altimeter | |
| Torres et al. | Preliminary investigation on chlorophyll-a and total suspended matter concentration in Manila Bay using Himawari-8 AHI and Sentinel-3 OLCI C2RCC | |
| CN112434590A (en) | SAR image wind stripe identification method based on wavelet transformation | |
| Gabr et al. | Assessment of genetic algorthim in developing bathymetry using multispectral landsat images | |
| Kurniawan et al. | Challenges of Acquisition Bathymetry Information on PlanetScope Data and Nautical Chart: Experiment Based on IHO S-44 Total Vertical Uncertainty in Multi-Method Satellite-Derived Bathymetry | |
| Xie et al. | New Reference Bathymetric Point Cloud Datasets Derived from ICESat-2 Observations: A Case in the Caribbean Sea | |
| CN117523321B (en) | Optical shallow water classification method based on passive remote sensing spectral image application neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |