Coral sand reef sea area water depth detection method, device, equipment and storage medium
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- GUANGDONG LABORATORY OF SOUTHERN OCEAN SCIENCE AND ENGINEERING (GUANGZHOU)
- Filing Date
- 2026-03-26
- Publication Date
- 2026-06-23
Smart Images
Figure CN122265722A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of marine remote sensing monitoring technology, and in particular to a method, apparatus, equipment and storage medium for detecting water depth in coral sand island and reef waters. Background Technology
[0002] Water depth data from coral sand island and reef areas (such as the Spratly Islands) is crucial for marine resource development, island and reef maintenance, navigation safety, and ecological protection, serving as indispensable core foundational data for these fields. Currently, water depth detection in such areas primarily relies on field measurements and remote sensing inversion technology. Among these, remote sensing inversion technology, with its significant advantages of wide coverage and high detection efficiency, has become the mainstream technical approach for acquiring water depth data.
[0003] However, existing remote sensing depth inversion methods have many limitations. Most methods primarily rely on the band reflectance of optical remote sensing images to construct inversion models. However, coral sand reefs possess unique high-albedo seabed characteristics, which easily lead to image reflectance distortion, severely interfering with the accuracy of feature extraction. Furthermore, existing methods fail to fully consider the regional characteristics of coral sand reefs, resulting in final inversion accuracy that does not meet the needs of practical applications. This, to some extent, restricts the smooth progress and in-depth advancement of related work in coral sand reef areas. Summary of the Invention
[0004] Based on this, the purpose of this application is to provide a method, apparatus, equipment and storage medium for detecting water depth in coral sand island and reef waters, so as to solve the problems of insufficient inversion accuracy caused by the high albedo of the coral sand island and reef seabed and the lack of consideration of regional characteristics in the existing remote sensing water depth inversion methods for coral sand island and reef waters.
[0005] The water depth detection method for coral sand island and reef waters described in this application includes the following steps:
[0006] Acquire first optical remote sensing image data and measured water depth data of the coral sand island and reef sea area; the measured water depth data includes the actual water depth data of several locations within the coral sand island and reef sea area. Based on the first optical remote sensing image data, the visible light chromaticity fusion features corresponding to the several locations are obtained; A training dataset is constructed based on the visible light chromaticity fusion features of the aforementioned locations and the corresponding real water depth data; a preset model is trained based on the training dataset to obtain a water depth inversion model. Acquire second optical remote sensing image data of the coral sand island and reef sea area to be detected; based on the second optical remote sensing image data, acquire visible light chromaticity fusion features corresponding to each target location in the coral sand island and reef sea area to be detected; input the visible light chromaticity fusion features corresponding to each target location into the water depth inversion model to obtain water depth data of each target location in the coral sand island and reef sea area to be detected.
[0007] This application also provides a water depth detection device for coral sand island and reef waters, including: The data acquisition module is used to acquire first optical remote sensing image data and measured water depth data of the coral sand island and reef sea area; the measured water depth data includes the actual water depth data of several locations within the coral sand island and reef sea area. The feature processing module is used to obtain visible light chromaticity fusion features corresponding to the several locations based on the first optical remote sensing image data; The model training module is used to construct a training dataset based on the visible light chromaticity fusion features of the aforementioned locations and the corresponding real water depth data, and to train a preset model based on the training dataset to obtain a water depth inversion model. The water depth detection module is used to acquire second optical remote sensing image data of the coral sand island and reef sea area to be detected; based on the second optical remote sensing image data, it acquires visible light chromaticity fusion features corresponding to each target location in the coral sand island and reef sea area to be detected; and inputs the visible light chromaticity fusion features corresponding to each target location into the water depth inversion model to obtain water depth data of each target location in the coral sand island and reef sea area to be detected.
[0008] This application also provides a computer device, including a processor, a memory, and a computer-readable program stored in the memory, wherein the computer-readable program, when executed by the processor, implements the steps of the method described in any one of the embodiments of this application.
[0009] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed, controls the computer device to implement the method described in any one of the embodiments of this application.
[0010] This application embodiment simultaneously collects first optical remote sensing image data and measured water depth data, abandoning the traditional method of solely relying on the reflectance of optical remote sensing image bands. Instead, it obtains visible light chromaticity fusion features corresponding to several locations based on the first optical remote sensing image data. Because coral sand reefs have high albedo, traditional methods easily distort image reflectance and interfere with feature extraction accuracy. Visible light chromaticity fusion features effectively reduce this interference, more accurately reflecting the characteristics of the sea area and providing key feature support for building a high-precision model. In terms of model training, a training dataset is constructed using the extracted visible light chromaticity fusion features and corresponding actual measured water depth data. This dataset is then used to train a preset model to obtain a water depth inversion model. This training method, based on actual data and incorporating unique features, fully considers the regional characteristics of coral sand reefs, enabling the model to better adapt to this special sea environment and avoiding the problem of insufficient inversion accuracy caused by existing methods failing to fully integrate regional characteristics. During actual detection, second optical remote sensing image data of the sea area to be detected is acquired, the visible light chromaticity fusion features of the target location are extracted, and the data is input into the trained water depth inversion model to obtain accurate water depth data. The embodiments of this application effectively overcome many drawbacks of the prior art, significantly improve the accuracy and reliability of water depth detection in coral sand island and reef waters, and provide more accurate and effective water depth data support for practical applications such as marine resource development, island and reef maintenance, navigation safety and ecological protection, thus powerfully promoting the development and progress of related fields.
[0011] To better understand and implement this application, the following detailed description is provided in conjunction with the accompanying drawings. Attached Figure Description
[0012] Figure 1 This is a flowchart illustrating the water depth detection method for coral sand island and reef areas according to an embodiment of this application. Figure 2 This is a schematic diagram of the structure of a computer device according to an embodiment of this application. Detailed Implementation
[0013] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings. Wherein, when the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements.
[0014] It should be understood that the embodiments described below do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this application.
[0015] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application are also intended to include the plural forms unless the context clearly indicates otherwise. Furthermore, in the description of this application, unless otherwise stated, “a plurality” means two or more. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more associated listed items, for example, A and / or B, which can represent: A alone, A and B together, and B alone; the character “ / ” generally indicates that the preceding and following objects are in an “or” relationship.
[0016] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, this information should not be limited to these terms, and these terms are only used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence, nor should they be construed as indicating or implying relative importance. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances. Depending on the context, the word "if" as used in this application can be interpreted as "when," "when," or "in response to determination."
[0017] Please refer to Figure 1 The water depth detection method for coral sand island and reef waters described in this application includes the following steps: S101: Acquire first optical remote sensing image data and measured water depth data of the coral sand island and reef sea area; the measured water depth data includes the actual water depth data of several locations within the coral sand island and reef sea area. S102: Based on the first optical remote sensing image data, obtain the visible light chromaticity fusion features corresponding to the several locations; S103: Construct a training dataset based on the visible light chromaticity fusion features of the aforementioned locations and the corresponding real water depth data, and train a preset model based on the training dataset to obtain a water depth inversion model; S104: Acquire second optical remote sensing image data of the coral sand island and reef sea area to be detected; based on the second optical remote sensing image data, acquire visible light chromaticity fusion features corresponding to each target location in the coral sand island and reef sea area to be detected; input the visible light chromaticity fusion features corresponding to each target location into the water depth inversion model to obtain water depth data of each target location in the coral sand island and reef sea area to be detected.
[0018] This application embodiment simultaneously collects first optical remote sensing image data and measured water depth data, abandoning the traditional method of solely relying on the reflectance of optical remote sensing image bands. Instead, it obtains visible light chromaticity fusion features corresponding to several locations based on the first optical remote sensing image data. Because coral sand reefs have high albedo, traditional methods easily distort image reflectance and interfere with feature extraction accuracy. Visible light chromaticity fusion features effectively reduce this interference, more accurately reflecting the characteristics of the sea area and providing key feature support for building a high-precision model. In terms of model training, a training dataset is constructed using the extracted visible light chromaticity fusion features and corresponding actual measured water depth data. This dataset is then used to train a preset model to obtain a water depth inversion model. This training method, based on actual data and incorporating unique features, fully considers the regional characteristics of coral sand reefs, enabling the model to better adapt to this special sea environment and avoiding the problem of insufficient inversion accuracy caused by existing methods failing to fully integrate regional characteristics. During actual detection, second optical remote sensing image data of the sea area to be detected is acquired, the visible light chromaticity fusion features of the target location are extracted, and the data is input into the trained water depth inversion model to obtain accurate water depth data. The embodiments of this application effectively overcome many drawbacks of the prior art, significantly improve the accuracy and reliability of water depth detection in coral sand island and reef waters, and provide more accurate and effective water depth data support for practical applications such as marine resource development, island and reef maintenance, navigation safety and ecological protection, thus powerfully promoting the development and progress of related fields.
[0019] The water depth detection method for coral sand islands and reefs described in this application uses computer equipment as the execution subject, and the following describes each step.
[0020] For step S101, first optical remote sensing image data and measured water depth data of the coral sand island and reef sea area are acquired; the measured water depth data includes the actual water depth data of several locations within the coral sand island and reef sea area.
[0021] Coral sand island and reef waters refer to the waters surrounding islands and reefs where coral sand is the main substrate type. The substrate is mainly composed of coral debris and sandy sediments, and the optical properties of the water are affected by the reflection of the substrate, water depth, and water composition.
[0022] The first optical remote sensing image data is a visible light band remote sensing image covering the coral sand island and reef waters. This data is used to extract the visible light band reflectance of sample locations, construct visible light colorimetric fusion features, and form a training dataset together with measured water depth data. This dataset is used to train the water depth inversion model, providing an input data source for model training. In this embodiment, the first optical remote sensing image data can be acquired by a satellite or airborne platform equipped with an optical sensor.
[0023] Measured water depth data refers to the actual water depth values at several specific locations within the coral sand island and reef waters, obtained through on-site measurement methods. This data serves as supervisory labels for model training, ensuring that the model learns the relationship between actual water depth and features. On-site measurement methods can include sonar detection, depth sounder measurements, etc.
[0024] This step aims to collect foundational data. First-stage optical remote sensing imagery of the coral sand island reef area is acquired using optical remote sensing equipment, covering optical information over a wide area. Simultaneously, actual water depth data at several locations within the area is obtained through field measurements. This measured water depth data provides training labels for subsequent model training and is crucial for ensuring the accuracy of water depth inversion.
[0025] For step S102, based on the first optical remote sensing image data, the visible light chromaticity fusion features corresponding to the several locations are obtained.
[0026] Visible light chromaticity fusion features are comprehensive feature representations obtained by processing various information contained in the visible light band of optical remote sensing images. They can more comprehensively and accurately reflect the color, texture, and other characteristics of features in coral sand island and reef waters. Compared to single-information features, they can effectively reduce interference caused by factors such as special seabed conditions, improving the accuracy and reliability of feature extraction. The various information contained in the visible light band can specifically include light or chromaticity information of different dimensions, such as visible light band reflectance, tristimulus values, and / or uniform chromaticity space characteristics. In this embodiment, the visible light chromaticity fusion feature can be a fusion of visible light band reflectance and uniform chromaticity space characteristics; in other embodiments, tristimulus values can be further fused.
[0027] This step, based on the acquired first optical remote sensing image data, constructs visible light chromaticity fusion features for several locations determined in S101.
[0028] In one embodiment, the visible light chromaticity fusion feature includes visible light band reflectance and uniform chromaticity space features.
[0029] Step S102, based on the first optical remote sensing image data, is a step of obtaining visible light chromaticity fusion features corresponding to the plurality of locations, including: Step S1021: Extract the visible light band reflectance corresponding to the several locations from the first optical remote sensing image data.
[0030] Visible light reflectance refers to an object's ability to reflect light in the visible light band (typically light with wavelengths between 380 and 780 nanometers). It is expressed as the ratio of reflected light intensity to incident light intensity and reflects the optical properties of an object's surface within the visible light range. Different ground features exhibit varying visible light reflectance, making it one of the important characteristics for distinguishing ground features. In this embodiment, the visible light reflectance includes red light reflectance, green light reflectance, and blue light reflectance.
[0031] This step is the initial stage of feature extraction. Using specific algorithms and image processing techniques, the reflectance values in the visible light band for each selected location are accurately obtained from remote sensing imagery. This reflectance data reflects the visible light reflection characteristics of these locations and forms the basis for subsequent calculations of other color-related features, providing crucial information for a comprehensive description of the optical characteristics of the coral sand island and reef waters.
[0032] Step S1022: Based on the visible light band reflectance of the several locations, the tristimulus values corresponding to the several locations are obtained by linear transformation using a standard chromaticity system; based on the tristimulus values of the several locations, the uniform chromaticity space characteristics corresponding to the several locations are obtained by nonlinear transformation using a uniform chromaticity space.
[0033] The standard colorimetric system is a scientific system for quantitatively describing and measuring color. It specifies the methods of color representation and calculation rules. Through linear transformation, the spectral information of color can be converted into tristimulus values, thereby providing a precise quantitative description of color.
[0034] In the standard colorimetric system, the tristimulus values are three basic parameters describing color perception, corresponding to the human eye's sensitivity to the three primary colors of red, green, and blue, respectively. They are represented by X, Y, and Z, and can quantitatively represent the amount of color stimulation, serving as the basis for calculating other color-related parameters.
[0035] The uniform color space (CIE-Lab) is a color space established to more uniformly represent color differences. In this space, the distance between two points can more accurately reflect the degree of color difference perceived by the human eye. The uniform color space has perceptual uniformity and can better stretch spectral differences caused by changes in water depth.
[0036] Uniform chromaticity space characteristics describe color properties in a uniform chromaticity space, typically represented by specific coordinate values. They more accurately reflect subtle differences between colors, providing more reasonable parameters for a comprehensive description of object colors. In this embodiment, uniform chromaticity space characteristics include lightness characteristics, red-green hue characteristics, and yellow-blue hue characteristics.
[0037] After obtaining the visible light reflectance, this step converts the reflectance data into tristimulus values using the linear transformation formula specified by the standard colorimetric system. Specifically, assuming the reflectances of the red, green, and blue bands of the remote sensing image are R, G, and B, respectively, they first need to be normalized and Gamma corrected (depending on the sensor characteristics, linear RGB can usually be directly converted). The core formula is as follows, and the transformation matrix adopts the CIE 1931 standard (D65 white point):
[0038] Among them, the Y value directly corresponds to the brightness information perceived by the human eye. In coral reef waters, this value decreases exponentially with water depth and is the dominant factor in the inversion.
[0039] After obtaining the tristimulus values, a nonlinear transformation method in a uniform chromaticity space is used to convert these tristimulus values into features in the uniform chromaticity space. Specifically, the core formula is as follows:
[0040]
[0041]
[0042] in, , , (Reference white point under D65 standard light source). This transformation introduces nonlinear characteristics through the function f(t). The nonlinear transformation acts as a feature enhancement and can be used as a strong feature for depth inversion. The function f(t) is defined as: when t> When, f(t) = When t At that time, f(t) = 1 / 3 In the coral sand reefs (Brightness) independently characterizes the light transmittance of a water body; (Redness / greenness) changes sensitively in shallow coral areas and can effectively distinguish coral reefs (leaning towards green) from deep water areas (leaning towards blue / black). (Yellowish-blue tint) reacts strongly to sandy (yellowish) substrates.
[0043] Step S1023: Based on the visible light band reflectance and uniform chromaticity space characteristics corresponding to the several positions, obtain the visible light chromaticity fusion characteristics corresponding to the several positions.
[0044] This embodiment extracts visible light reflectance from the first optical remote sensing image data, directly obtaining basic optical information of various locations in the sea area within the visible light range, providing the original basis for subsequent feature calculations. Next, the visible light reflectance is converted into tristimulus values using a linear transformation of the standard chromaticity system. Then, the tristimulus values are converted into uniform chromaticity space features through a nonlinear transformation of the uniform chromaticity space. This fully considers the visual characteristics of the human eye, making the representation of color features more consistent with actual perception and more accurately describing subtle differences between colors. It comprehensively characterizes the optical properties of the coral sand reef sea area from different levels, and the constructed visible light chromaticity fusion feature can more realistically and accurately reflect the actual situation of the sea area. Compared with single features, the visible light chromaticity fusion feature can overcome the image reflectance distortion problem caused by the high albedo of the coral sand reef seabed, providing richer and more accurate information. This lays a solid foundation for the subsequent training of a high-precision water depth inversion model, thereby effectively improving the accuracy of water depth detection.
[0045] In one embodiment, the visible light chromaticity fusion feature further includes tristimulus values; Step S1023, which involves obtaining the visible light chromaticity fusion features corresponding to the aforementioned locations based on the visible light band reflectance and uniform chromaticity space characteristics, further includes: Step S10231: Based on the visible light band reflectance, tristimulus values and uniform chromaticity space characteristics corresponding to the several positions, obtain the visible light chromaticity fusion characteristics corresponding to the several positions.
[0046] This embodiment integrates three features—visible light reflectance, tristimulus values, and uniform chromaticity space characteristics—to form a visible light chromaticity fusion feature. This feature combines descriptions of land cover characteristics from different dimensions. The tristimulus values can quantitatively represent the relative amounts of the three primary colors of light required to produce a certain color. It serves as an important bridge connecting the optical properties of objects with human color perception. This feature effectively overcomes the interference of special factors such as the high albedo of coral sand islands and reefs on traditional single features. It comprehensively and accurately reflects the characteristics of marine land cover from multiple perspectives, thereby improving the feature's ability to represent the characteristics of land cover.
[0047] In one embodiment, before step S1021, which involves extracting the visible light band reflectance corresponding to the plurality of locations from the first optical remote sensing image data, the following steps are included: Step S1020: Radiometric calibration, atmospheric correction and water surface flare removal are performed sequentially on the first optical remote sensing image data to obtain the processed first optical remote sensing image data.
[0048] First, radiometric calibration is performed on the raw first-order optical remote sensing image data. By establishing a quantitative relationship between the sensor's output signal and the incident radiation, the digital quantization values of the image are converted into radiance values. This eliminates the influence of differences in sensor performance on the data, enabling the image data to more accurately reflect the radiation status of ground objects and providing reliable basic data for subsequent processing. Next, atmospheric correction is performed. Since various components in the atmosphere absorb and scatter solar radiation, the radiation information received by the remote sensor contains atmospheric interference. Atmospheric correction uses specific algorithms and models to remove these atmospheric effects, obtaining the true reflectance information of ground objects, further improving the quality and accuracy of the image data, making the image closer to the actual state of the ground objects. Finally, water flare removal processing is performed, targeting strong flare areas in the image caused by specular reflection of sunlight on the water surface. These flares can obscure the true information of the water surface, affecting the accurate extraction of water features. By employing appropriate methods and techniques, such as image feature analysis and physical models, water flares are removed, restoring the true reflectance of the water surface, resulting in processed first-order optical remote sensing image data, providing high-quality imagery for subsequent operations such as extracting reflectance in the visible light band.
[0049] This embodiment eliminates the impact of variations in remote sensor performance on the data through radiometric calibration, ensuring that the image data accurately reflects the radiometric characteristics of ground objects and providing a reliable foundation for subsequent processing. Atmospheric correction removes atmospheric interference from the remote sensing image, obtaining true reflectance information of ground objects and further improving the quality of the image data, making the image closer to the actual state of the ground objects. Water flare removal eliminates specific interference phenomena in water surface areas, restoring the true reflectance of the water surface and preventing flares from masking the extraction of water body features. The high-quality image data obtained after this series of processing ensures accurate extraction of visible light reflectance.
[0050] In one embodiment, before step S1020, which sequentially performs radiometric calibration, atmospheric correction, and water flare removal on the first optical remote sensing image data, the following steps are included: Step S10201: Obtain the albedo characteristics of the coral sand substrate in the coral sand island and reef sea area, and determine the substrate correction coefficient corresponding to the albedo characteristics.
[0051] Albedo refers to the ratio of the radiant energy reflected by an object to the incident radiant energy. The albedo characteristics of coral sand substrate refer to the reflectance / albedo properties of coral sand sediments at the bottom of coral sand islands and reefs in the visible light band. It reflects the ability of coral sand sediments to reflect light of different wavelengths and is an important parameter for describing the optical properties of coral sand.
[0052] The substrate correction factor is a coefficient determined based on the albedo characteristics of coral sand substrates and used to correct the reflectance of remote sensing images. Because the high albedo of coral sand can cause image reflectance distortion, the substrate correction factor is used to eliminate this effect, making the image reflectance closer to reality.
[0053] The albedo characteristics of coral sand substrate in coral sand island and reef waters are obtained through methods such as field measurements, literature reviews, and existing databases. Field measurements utilize specialized optical instruments to measure the reflectance of coral sand under different lighting conditions to obtain accurate albedo data. Literature reviews and databases provide existing research results and data references. Then, based on the obtained albedo characteristics and combined with the imaging principles and optical characteristics of remote sensing images, the corresponding substrate correction coefficients are determined. The determination of substrate correction coefficients needs to consider the relationship between coral sand albedo and image reflectance, as well as the influence of different wavelengths, and usually requires extensive experiments and data analysis to establish a suitable model or algorithm to accurately calculate the substrate correction coefficients.
[0054] Step S10202: Multiply the image reflectance parameter of the first optical remote sensing image data with the substrate correction coefficient to obtain the first optical remote sensing image data after substrate reflectance correction.
[0055] Image reflectance parameter refers to the quantitative parameter in the first optical remote sensing image data that characterizes the reflectivity of the water and seabed of coral sand islands and reefs to visible light. It is a core optical parameter inherent in the image. After seabed correction, this parameter can eliminate the interference of coral sand seabed reflection, improve the accuracy of the parameter, and provide a basis for subsequent extraction of visible light band reflectance at several locations, construction of tristimulus values, and uniform color space characteristics.
[0056] After determining the substrate correction coefficient, the image reflectance parameter in the first optical remote sensing image data was corrected. Specifically, the image reflectance parameter of each pixel was multiplied by the corresponding substrate correction coefficient. This multiplication operation eliminated the influence of the high albedo of coral sand on the image reflectance, resulting in the first optical remote sensing image data with substrate reflectance correction. The corrected image data more accurately reflects the true reflectance of ground features, providing a more reliable basis for subsequent radiometric calibration, atmospheric correction, and water surface flare removal.
[0057] This embodiment significantly improves the accuracy and reliability of subsequent water depth detection by correcting the seabed reflectance of the first optical remote sensing image data. Coral sand has high albedo, which can cause distortion of the reflectance of remote sensing images and affect the accurate extraction of water features. By acquiring the albedo characteristics of the coral sand seabed and determining the seabed correction coefficient, the image reflectance parameters are corrected, effectively eliminating the influence of the seabed on the image reflectance and making the image data closer to the true reflectance of ground features.
[0058] For step S103, a training dataset is constructed based on the visible light chromaticity fusion features of the aforementioned locations and the corresponding real water depth data; a preset model is trained based on the training dataset to obtain a water depth inversion model.
[0059] The training dataset includes visible light chromaticity fusion features extracted from optical remote sensing images at various locations, along with corresponding actual water depth measurements. The visible light chromaticity fusion features serve as training samples, while the corresponding real water depth data serve as training labels.
[0060] The preset model refers to a pre-defined machine learning model suitable for regression prediction, specifically a gradient boosting regression model. In this embodiment, the preset model is used to learn the mapping relationship between visible light chromaticity fusion features and actual water depth. Using visible light chromaticity fusion features as input and actual water depth as the supervision label, training is completed through iterative optimization of loss values or selection of candidate parameters to obtain the water depth inversion model.
[0061] This step utilizes the visible light chromaticity fusion features constructed in S102 at several locations, along with the corresponding real water depth data, to build a training dataset. This dataset contains the correspondence between visible light chromaticity fusion features and real water depth, serving as the foundational sample for model training. Based on this training dataset, the preset model is trained. During training, the model continuously adjusts its parameters, learning the intrinsic relationship between visible light chromaticity fusion features and real water depth. After multiple iterations and optimizations, a water depth inversion model capable of accurately predicting water depth based on visible light chromaticity fusion features is obtained.
[0062] In one embodiment, the preset model may employ a gradient boosting regression model. Gradient boosting regression is a regression algorithm based on ensemble learning. It iteratively constructs multiple weak learners (typically decision trees) and weights and combines the predictions of these weak learners to obtain a strong learner. In each iteration, the model constructs new weak learners based on the residuals of the previous iteration (i.e., the difference between the predicted and actual values), aiming to minimize the residuals and gradually improve the model's prediction accuracy.
[0063] In one embodiment, step S103, which involves training a preset model based on the training dataset to obtain a water depth inversion model, includes: Step S1031: Obtain several candidate parameter combinations of the preset model; sequentially select each candidate parameter combination as the model parameter of the preset model; input the visible light chromaticity fusion features of each location in the training dataset into the preset model to obtain the predicted water depth data for each location; calculate the corresponding loss value based on the predicted water depth data and the corresponding real water depth data for each location; determine the preset model with the smallest loss value as the trained water depth inversion model.
[0064] This embodiment focuses on training a gradient boosting regression model. Based on the characteristics and experience of gradient boosting regression models, several different parameter value combinations are generated beforehand, covering a range of parameters where the model may have good performance. One set of candidate parameter combinations is selected sequentially and used as the current model parameters for the preset model. Then, the visible light chromaticity fusion features at each location in the training dataset are input into the preset model with the set of parameters. Based on the input visible light chromaticity fusion features and the current parameter combination, the preset model outputs predicted water depth data for each location. Next, based on these predicted water depth data and the corresponding real water depth data in the training dataset, the loss value of the model under this parameter set is calculated to evaluate the predictive performance of the model under this parameter combination. After completing the above operations for all candidate parameter combinations, the magnitude of the loss values corresponding to each set of parameters is compared. The preset model corresponding to the parameter set with the smallest loss value is selected as the trained water depth inversion model, which has the best water depth prediction capability under the current conditions.
[0065] This embodiment fully leverages the parameter sensitivity of the gradient boosting regression model. By comprehensively searching different parameter combinations, it can find model parameters that are more suitable for the relationship between optical remote sensing image features and water depth in coral sand island and reef waters. Compared to randomly setting parameters or adjusting parameters solely based on experience, this approach more systematically and comprehensively optimizes model performance, effectively improving the model's prediction accuracy and stability.
[0066] In this embodiment, the preferred preset model is the XGBoost (Extreme Gradient Boosting Tree) model. The core parameters of the XGBoost model are designed specifically for the characteristics of coral sand island and reef data as follows: 1. Feature Engineering: Constructing the input feature matrix ,in The number of samples, i.e. Each pixel contains three visible light band reflectance values (R, G, B) and three uniform color space features. , , That is, 6 feature dimensions: [R,G,B, , , By introducing , , As input features, this allows the model to learn not only the variation of "light intensity" with depth (through R, G, B, ... It can also learn the shift in "water color" with depth (through...). , ).
[0067] 2. Model training strategy: The objective function is to minimize the mean square error between the predicted and measured water depths.
[0068] in This is a regularization term used to control the complexity of the tree and prevent overfitting in the coral reef edge region.
[0069] In this embodiment, the learning rate of the XGBoost model is set to 0.01 to 0.1; the maximum tree depth is set to 3 to 8; the subsample ratio is set to 0.6 to 0.9; the column sample ratio is set to 0.6 to 0.9; and the mean squared error (MSE) is used as the objective function.
[0070] In one specific embodiment, the core parameters are set as follows: Learning Rate: 0.05 (A smaller learning rate to prevent overfitting and adapt to sparse water depths).
[0071] max_depth:6 (capable of capturing complex terrain abrupt changes at the edge of the coral reef).
[0072] colsample_bytree=0.8 (randomly samples 80% of the features when building each tree, forcing the model to use CIE visible chromaticity fusion features for splitting, rather than relying too much on a single band).
[0073] Feature splitting and weight analysis: During model training, the XGBoost model automatically calculates the contribution of each feature to the gain.
[0074] Experimental results show that, compared to existing technologies that use a single visible light band feature, this embodiment, by introducing uniform color space features, Features helped the model improve its classification of shallow and deep water; The features contribute significantly in the 5-15 meter transition zone, helping the model distinguish between coral reefs and the deep sea; The feature makes a significant contribution in extremely shallow water areas of 0-5 meters, effectively correcting the problem of depth underestimation caused by coral sand bleaching.
[0075] For step S104, acquire the second optical remote sensing image data of the coral sand island and reef sea area to be detected; based on the second optical remote sensing image data, acquire the visible light chromaticity fusion features corresponding to each target location in the coral sand island and reef sea area to be detected; input the visible light chromaticity fusion features corresponding to each target location into the water depth inversion model to obtain the water depth data of each target location in the coral sand island and reef sea area to be detected.
[0076] The second optical remote sensing image data is similar to the first optical remote sensing image data. It was also acquired through optical remote sensing technology, but it is for the coral sand island and reef sea area to be detected and is used for water depth detection in practical applications.
[0077] In practical applications, firstly, second optical remote sensing image data of the coral sand reef area to be detected is acquired. Then, based on this image data, visible light chromaticity fusion features corresponding to each target location in the area to be detected are constructed using the same method as in S102. Finally, these visible light chromaticity fusion features of the target locations are input into the water depth inversion model trained in S103. The model outputs water depth data for each target location, thus completing the water depth detection of the coral sand reef area to be detected.
[0078] This application also provides a water depth detection device for coral sand island and reef waters, including: The data acquisition module is used to acquire first optical remote sensing image data and measured water depth data of the coral sand island and reef sea area; the measured water depth data includes the actual water depth data of several locations within the coral sand island and reef sea area. The feature processing module is used to obtain visible light chromaticity fusion features corresponding to the several locations based on the first optical remote sensing image data; the visible light chromaticity fusion features include visible light band reflectance and uniform chromaticity space features. The model training module is used to construct a training dataset based on the visible light chromaticity fusion features of the aforementioned locations and the corresponding real water depth data, and to train a preset model based on the training dataset to obtain a water depth inversion model. The water depth detection module is used to acquire second optical remote sensing image data of the coral sand island and reef sea area to be detected; based on the second optical remote sensing image data, it acquires visible light chromaticity fusion features corresponding to each target location in the coral sand island and reef sea area to be detected; and inputs the visible light chromaticity fusion features corresponding to each target location into the water depth inversion model to obtain water depth data of each target location in the coral sand island and reef sea area to be detected.
[0079] It should be noted that the water depth detection device for coral sand islands and reefs provided in the above embodiments is only illustrated by the division of the above functional modules when performing the water depth detection method for coral sand islands and reefs. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the water depth detection device for coral sand islands and reefs provided in the above embodiments and the water depth detection method for coral sand islands and reefs belong to the same concept, and its implementation process is detailed in the above embodiments of the water depth detection method for coral sand islands and reefs, which will not be repeated here.
[0080] Please refer to Figure 2 This application also provides a computer device 301, including: a processor 302, a memory 303, and a computer program 304 stored in the memory 303 and executable on the processor 302. When the processor 302 executes the computer program 304, it implements the steps of the method as described in any one of the embodiments of this application.
[0081] The processor 302 may include one or more processing cores. The processor 302 connects to various parts within the computer device 301 via various interfaces and lines. It executes various functions of the computer device 301 and processes data by running or executing instructions, programs, code sets, or instruction sets stored in the memory 303, and by accessing data stored in the memory 303. Optionally, the processor 302 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 302 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required to be displayed on the touch screen; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 302 and may be implemented as a separate chip.
[0082] The memory 303 may include random access memory (RAM) or read-only memory. Optionally, the memory 303 may include a non-transitory computer-readable storage medium. The memory 303 may be used to store instructions, programs, code, code sets, or instruction sets. The memory 303 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch instructions), instructions for implementing the various method embodiments described above, etc.; the data storage area may store data involved in the various method embodiments described above, etc. Optionally, the memory 303 may also be at least one storage device located remotely from the aforementioned processor 302.
[0083] This application also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the steps of the method described in any one of the embodiments of this application. That is, those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. This program is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The computer program may include computer program code, which may be in the form of source code, object code, executable file, or some intermediate form. The aforementioned storage medium includes: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content contained in computer-readable media may be appropriately added to or subtracted from the requirements of legislation and patent practice in a jurisdiction. For example, in some jurisdictions, computer-readable media may not include electrical carrier signals and telecommunication signals, in accordance with legislation and patent practice.
[0084] The above embodiments are merely illustrative of several implementation methods of this application, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of this application, and this application also intends to include these modifications and variations.
Claims
1. A method for measuring water depth in coral sand island and reef waters, characterized in that, Includes the following steps: Acquire first optical remote sensing image data and measured water depth data of the coral sand island and reef sea area; the measured water depth data includes the actual water depth data of several locations within the coral sand island and reef sea area. Based on the first optical remote sensing image data, the visible light chromaticity fusion features corresponding to the several locations are obtained; A training dataset is constructed based on the visible light chromaticity fusion features of the aforementioned locations and the corresponding real water depth data; A preset model is trained based on the training dataset to obtain a water depth inversion model; Acquire second optical remote sensing image data of the coral sand island and reef sea area to be detected; based on the second optical remote sensing image data, acquire visible light chromaticity fusion features corresponding to each target location in the coral sand island and reef sea area to be detected; input the visible light chromaticity fusion features corresponding to each target location into the water depth inversion model to obtain water depth data of each target location in the coral sand island and reef sea area to be detected.
2. The method for detecting water depth in coral sand island and reef waters according to claim 1, characterized in that, The visible light chromaticity fusion features include visible light band reflectance and uniform chromaticity space features; The step of obtaining visible light chromaticity fusion features corresponding to the several locations based on the first optical remote sensing image data includes: Extract the visible light reflectance corresponding to the several locations from the first optical remote sensing image data; Based on the visible light reflectance at the aforementioned locations, the tristimulus values corresponding to the aforementioned locations are obtained through a linear transformation using a standard chromaticity system; based on the tristimulus values at the aforementioned locations, the uniform chromaticity space characteristics corresponding to the aforementioned locations are obtained through a nonlinear transformation using a uniform chromaticity space. Based on the visible light band reflectance and uniform chromaticity space characteristics corresponding to the aforementioned locations, the visible light chromaticity fusion characteristics corresponding to the aforementioned locations are obtained.
3. The method for detecting water depth in coral sand island and reef waters according to claim 2, characterized in that, The visible light chromaticity fusion feature also includes tristimulus values; The step of obtaining the visible light chromaticity fusion features corresponding to the aforementioned locations based on the visible light band reflectance and uniform chromaticity space characteristics further includes: Based on the visible light band reflectance, tristimulus values, and uniform chromaticity space characteristics corresponding to the aforementioned locations, the visible light chromaticity fusion characteristics corresponding to the aforementioned locations are obtained.
4. The method for detecting water depth in coral sand island and reef waters according to claim 2, characterized in that, Before the step of extracting the visible light band reflectance corresponding to the several locations from the first optical remote sensing image data, the following steps are included: The first optical remote sensing image data is subjected to radiometric calibration, atmospheric correction, and water surface flare removal processing in sequence to obtain the processed first optical remote sensing image data.
5. The method for detecting water depth in coral sand island and reef waters according to claim 4, characterized in that, Before performing radiometric calibration, atmospheric correction, and water surface flare removal on the first optical remote sensing image data, the following steps are included: Obtain the albedo characteristics of the coral sand substrate in the coral sand island and reef area, and determine the substrate correction coefficient corresponding to the albedo characteristics; The image reflectance parameter of the first optical remote sensing image data is multiplied by the substrate correction coefficient to obtain the first optical remote sensing image data after substrate reflectance correction.
6. The method for detecting water depth in coral sand island and reef waters according to claim 2, characterized in that, The visible light band reflectance includes red light band reflectance, green light band reflectance and blue light band reflectance; the uniform color space characteristics include lightness characteristics, red-green hue characteristics and yellow-blue hue characteristics.
7. The method for measuring water depth in coral sand island and reef waters according to any one of claims 1 to 6, characterized in that, The steps for training a preset model based on the training dataset to obtain a water depth inversion model include: Obtain several candidate parameter combinations of the preset model; sequentially select each candidate parameter combination as the model parameter of the preset model, input the visible light chromaticity fusion features of each location in the training dataset into the preset model to obtain the predicted water depth data for each location; calculate the corresponding loss value based on the predicted water depth data and the corresponding real water depth data for each location; determine the preset model with the smallest loss value as the trained water depth inversion model.
8. A water depth detection device for coral sand island and reef sea areas, characterized in that, include: The data acquisition module is used to acquire first optical remote sensing image data and measured water depth data of the coral sand island and reef waters. The measured water depth data includes the actual water depth data at several locations within the coral sand island and reef area; The feature processing module is used to obtain visible light chromaticity fusion features corresponding to the several locations based on the first optical remote sensing image data; The model training module is used to construct a training dataset based on the visible light chromaticity fusion features of the aforementioned locations and the corresponding real water depth data, and to train a preset model based on the training dataset to obtain a water depth inversion model. The water depth detection module is used to acquire second optical remote sensing image data of the coral sand island and reef sea area to be detected; based on the second optical remote sensing image data, it acquires visible light chromaticity fusion features corresponding to each target location in the coral sand island and reef sea area to be detected; and inputs the visible light chromaticity fusion features corresponding to each target location into the water depth inversion model to obtain water depth data of each target location in the coral sand island and reef sea area to be detected.
9. A computer device, characterized in that, It includes a processor, a memory, and a computer-readable program stored in the memory, which, when executed by the processor, implements the steps of the method as described in any one of claims 1-7.
10. A computer-readable storage medium, characterized in that, It stores a computer program thereon, which, when executed, controls the computer device to implement the method as described in any one of claims 1-7.