A method and system for extracting shoreline wave height based on single-point lidar data
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SOUTH CHINA SEA INST OF OCEANOLOGY CHINESE ACAD OF SCI
- Filing Date
- 2025-09-10
- Publication Date
- 2026-07-17
Smart Images
Figure CN121383971B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of marine observation data processing technology, and in particular to a method and system for extracting shoreline wave heights based on single-point lidar data. Background Technology
[0002] As the most extensive ecosystem on Earth, the ocean plays a vital role in both climate change and human activities. Among these, ocean waves, being the most common ocean undulation phenomenon, are beneficial for real-time monitoring in marine engineering safety, early warning of shoreline erosion, and the development of marine energy.
[0003] However, existing wave height (wave height refers to the height of waves) extraction techniques for shoreline survey lines are limited by equipment and measurement methods, resulting in insufficient accuracy and reliability of extracted wave height parameters under complex sea conditions. Specifically, the methods for obtaining shoreline topography and wave height parameters are relatively simple and difficult to observe over long periods. Existing laser measurement methods are typically mounted on aircraft or ships, which rarely consider the continuity of data collection and are greatly affected by weather and topography. The deployment and maintenance costs of these devices are high, and wave height calculations usually require combining information from the survey platform, lacking data cleaning and extraction. This leads to insufficient accuracy and precision of the final wave height parameters, making reliable long-term wave monitoring impossible. Summary of the Invention
[0004] This application provides a method and system for extracting wave height from shoreline survey lines based on single-point lidar data. By combining four technologies—precise trajectory segmentation, coordinate transformation optimization, intelligent data cleaning, and multi-algorithm comprehensive calculation—it achieves efficient and high-precision wave height extraction. This enables radar acquisition to be limited to specific locations, climates, and times, achieving reliable long-term wave monitoring and solving the problem of insufficient precision and accuracy in obtaining wave height parameters in existing technologies.
[0005] Firstly, this application provides a method for extracting the wave height of beach survey lines based on single-point lidar data, including:
[0006] Acquire periodically collected single-point lidar data, wherein the single-point lidar data is elliptical trajectory point cluster data obtained by periodically monitoring the beach by a single-point lidar at a preset frequency;
[0007] The lidar data is preprocessed by format conversion and visualized. The spatiotemporal distribution data of the left and right sections of the elliptical trajectory are segmented from the visualized data corresponding to the lidar data to obtain the three-dimensional data to be converted.
[0008] Based on the three-dimensional data to be converted, coordinate system transformation is performed through fitting optimization to convert the coordinate system of the three-dimensional data to be converted into the geodetic coordinate system, thus obtaining the data to be cleaned;
[0009] Based on the spatial density and data characteristics of the data to be cleaned, the data to be cleaned is subjected to data cleaning and gridding processing to obtain test data that retains local features;
[0010] A dual-mode processing architecture is constructed based on the fusion of the time-domain zero-crossing method and the frequency-domain wave spectrum analysis method. The wave height is extracted from the test data through the dual-mode processing architecture to obtain the final wave height parameters.
[0011] Optionally, the lidar data undergoes format conversion preprocessing and visualization analysis. The spatiotemporal distribution data of the left and right sections of the elliptical trajectory are segmented from the visualization data corresponding to the lidar data to obtain the three-dimensional data to be converted, including:
[0012] The lidar data is processed by data format conversion, and key data is extracted, including terrain identifiers, time, center point distance, motor encoder readings, and three-dimensional coordinates.
[0013] Based on the key data, a visualization is constructed. On the visualization, invalid data located on the shore is filtered out by visually calibrating the left and right sections of the elliptical trajectory to obtain spatiotemporal distribution data. The elliptical trajectory is the visualized laser acquisition data trajectory.
[0014] Based on the characteristics of the point cloud trajectory analysis of the spatiotemporal distribution data, noise is filtered out from the dense data at the front end and the divergent data at the back end of the spatiotemporal distribution data to obtain the three-dimensional data to be converted.
[0015] Optionally, based on the three-dimensional data to be converted, coordinate system transformation is performed through fitting optimization to convert the coordinate system of the three-dimensional data to be converted into a geodetic coordinate system, resulting in data to be cleaned, including:
[0016] The least squares method is used to perform planar fitting on the three-dimensional data to be converted according to the time series, and the fitted three-dimensional data is obtained.
[0017] Using the obtained three-dimensional coordinate points as a reference, matrix operations are performed on the data coordinates and rotation matrix in the fitted three-dimensional data to obtain rotated three-dimensional data, wherein the data coordinates in the rotated three-dimensional data are represented by data three-dimensional coordinates.
[0018] Extract the initial rotation angle from the fitted 3D data and use it as the observation point;
[0019] Based on the observation points, the cross-sections in the rotated three-dimensional data are dynamically and iteratively adjusted until the instrument installation error and data fitting error are eliminated, thus obtaining the data to be cleaned.
[0020] The initial rotation angle is the rotation angle used when processing the fitted three-dimensional data.
[0021] Optionally, the least squares method is used to perform planar fitting on the three-dimensional data to be transformed according to the time series, to obtain fitted three-dimensional data, including:
[0022] Input data is selected from the three-dimensional data to be converted according to the time series.
[0023] The least squares method is used to perform plane fitting on the input data, based on... Solve for the plane parameters;
[0024] Using plane parameters as input, according to Extract the plane normal vector;
[0025] Obtain the target normal vector ;
[0026] Introducing Rodriguez rotation, according to Calculate the axis of rotation, and according to Calculate the rotation angle;
[0027] Using the rotation axis and rotation angle as references, the plane normal vector and the target normal vector are compared. When the comparison result indicates that the plane normal vector and the target normal vector do not coincide, according to... Construct a rotation matrix to obtain the fitted 3D data;
[0028] in, It is the identity matrix. Let be the cross product matrix of the rotation axes.
[0029] Optionally, based on the observation points, the cross-sections in the rotated 3D data are dynamically and iteratively adjusted until instrument installation errors and data fitting errors are eliminated, resulting in data to be cleaned, including:
[0030] Obtain a preset first adjustment value and a second adjustment value, both of which are used to adjust the rotation angle;
[0031] Based on the first adjustment value, the observation point is dynamically and iteratively adjusted using an angle iteration algorithm. In each adjustment, the rotational three-dimensional data is processed, and the first rotation angle is determined based on the output first cross-sectional image.
[0032] Based on the second adjustment value, the first rotation angle is dynamically iteratively adjusted, and in each adjustment, the rotation three-dimensional data is processed. Based on the output second cross-sectional image, the target rotation angle is determined, and the target rotation angle is the rotation angle corresponding to the instrument setup angle.
[0033] At the target rotation angle, the visualization point cloud distribution data corresponding to the rotation three-dimensional data is used for verification to obtain the data to be cleaned after eliminating instrument installation errors and data fitting errors.
[0034] Optionally, based on the spatial density and data characteristics of the data to be cleaned, the data to be cleaned is subjected to data cleaning and gridding processing to obtain test data that retains local features, including:
[0035] Time series data, straight-line distance, and elevation data are selected from the data to be cleaned as the data to be analyzed.
[0036] The spatial density and data characteristics of the data to be analyzed are analyzed. The data to be analyzed is constrained according to the spatial density, and the data to be analyzed is cleaned and corrected based on the data characteristics to obtain the first cleaned data.
[0037] When the first cleaned data conforms to the preset distribution rules, Chebyshev's inequality is used to process outliers in the first cleaned data to obtain the second cleaned data.
[0038] The second cleaning data is mapped to grid points of a standardized grid to obtain the data to be tested.
[0039] Optionally, the spatial density and data characteristics of the data to be analyzed are analyzed, the data to be analyzed is constrained according to the spatial density, and the data to be analyzed is cleaned and corrected based on the data characteristics to obtain first cleaned data, including:
[0040] Descriptive statistical and visual analyses are performed on the data to be analyzed to determine the distribution status and numerical characteristics;
[0041] Based on the distribution status and digital characteristics, the dispersion characteristics of the data are analyzed to determine the spatial density and data properties;
[0042] Based on the spatial density, the data interval of the data to be analyzed is calibrated to constrain the spatial range;
[0043] The data to be analyzed within the specified spatial range is corrected for missing and negative values to obtain cleaned data.
[0044] Optionally, Chebyshev's inequality is used to process outliers in the first cleaned data to obtain the second cleaned data, including:
[0045] Using Chebyshev's inequality, according to The first cleaned data is subjected to outlier removal within a threshold range, and the outlier is replaced with the nearest valid value to update the first cleaned data;
[0046] The time-series data of the first cleaned data is subjected to fragmentation mitigation processing, and the timestamps are adaptively adjusted to update the first cleaned data.
[0047] The spatial data density of the first cleaned data is detected according to a preset threshold to determine the range of empty data;
[0048] The first cleaned data is emptied according to the emptied data range, and the first cleaned data is updated to obtain the second cleaned data;
[0049] in, This indicates the input sample data. This represents the mean of the dataset. The standard deviation of the dataset is represented by the standard deviation of the dataset. The threshold parameter is set.
[0050] Optionally, a dual-mode processing architecture is constructed based on the fusion of the time-domain zero-crossing method and the frequency-domain wave spectrum analysis method. The wave height is extracted from the measured data using this dual-mode processing architecture to obtain the final wave height parameters, including:
[0051] Based on the data to be tested, the wave cycle is segmented by zero intersection detection, and the average wave height and maximum wave height of the large wave are calculated.
[0052] Based on the data to be measured, a Hamming window is used to suppress spectral leakage by FFT transform to generate a one-sided power spectral density, and the one-sided power spectral density is optimized by filtering and binning smoothing. The effective wave height is then inverted by zero-order moment integration.
[0053] The average wave height, maximum wave height, and significant wave height of the large wave are used as the final wave height parameters.
[0054] Secondly, this application provides a shoreline survey wave height extraction system based on single-point lidar data, comprising:
[0055] The data preprocessing module is used to acquire periodically collected single-point lidar data, which is elliptical trajectory point cluster data obtained by single-point lidar periodically monitoring the beach at a preset frequency; the lidar data is preprocessed by format conversion and visualization analysis, and the spatiotemporal distribution data of the left and right sections of the elliptical trajectory are segmented from the visualization data corresponding to the lidar data to obtain the three-dimensional data to be converted.
[0056] The coordinate transformation module is used to perform coordinate system transformation processing on the three-dimensional data to be transformed by fitting and optimization, so as to convert the coordinate system of the three-dimensional data to be transformed into the geodetic coordinate system and obtain the data to be cleaned.
[0057] The intelligent cleaning and anomaly detection module is used to perform data cleaning and gridding processing on the data to be cleaned based on the spatial density and data characteristics of the data to be cleaned, so as to obtain test data that retains local features;
[0058] The wave height calculation module is used to construct a dual-mode processing architecture based on the fusion of the time-domain zero-crossing method and the frequency-domain wave spectrum analysis method. The wave height is extracted from the test data through the dual-mode processing architecture to obtain the final wave height parameters.
[0059] In summary, this application first periodically converts and visualizes the complex elliptical trajectory data collected by single-point lidar, reconstructing it into intuitive spatiotemporal distribution data along the cross-section. Then, the coordinate system of the data is converted to a geodetic coordinate system through fitting optimization. Intelligent data cleaning, based on spatial density and data characteristics, cleans the converted data and performs gridding to diagnose data anomalies, remove data errors, and preserve local features. Finally, a dual-modal processing architecture is constructed by combining the zero-point method and spectral analysis to extract wave height from the data and output the final wave height parameters. Therefore, this application achieves efficient and high-precision wave height extraction, enabling reliable and stable long-term observation. This allows radar-acquired data to be used for wave height analysis not limited to specific locations, climates, and times, and can be widely applied to monitoring shoreline topography and wave parameters, solving the problem of insufficient precision and accuracy in wave height parameter acquisition using existing technologies. Attached Figure Description
[0060] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.
[0061] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0062] Figure 1 A flowchart illustrating a method for extracting shoreline wave height based on single-point lidar data, provided in an embodiment of this application;
[0063] Figure 2 This is a flowchart illustrating the steps of a method for extracting shoreline wave height based on single-point lidar data, provided in an optional embodiment of this application.
[0064] Figure 3 This application provides an example of a laser acquisition data trajectory diagram;
[0065] Figure 4 This is a schematic diagram of noise data provided as an example in this application;
[0066] Figure 5 This is a schematic diagram of data rotation provided as an example in this application;
[0067] Figure 6 This is a schematic diagram of synchronous rotation of a lidar provided as an example in this application;
[0068] Figure 7 This application provides an example of a cross-sectional view corresponding to 107.1 degrees;
[0069] Figure 8 This application provides an example of a comparison image of elevation data before and after cleaning;
[0070] Figure 9 This is an example of a comparison diagram before and after time-coherent processing provided in this application;
[0071] Figure 10 This is a schematic diagram illustrating the wave height extracted from single-point lidar data, provided as an example in this application.
[0072] Figure 11 This is a structural block diagram of a beach survey wave height extraction system based on single-point lidar data, provided in an embodiment of this application. Detailed Implementation
[0073] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0074] To facilitate understanding of the embodiments of this application, further explanations and descriptions will be provided below in conjunction with the accompanying drawings and specific embodiments. These embodiments do not constitute a limitation on the embodiments of this application.
[0075] Figure 1 A flowchart illustrating a method for extracting shoreline wave height based on single-point lidar data, provided in this application embodiment, is shown below. The method may specifically include the following steps:
[0076] Step 110: Acquire periodically collected single-point lidar data.
[0077] The single-point lidar data refers to the elliptical trajectory point cluster data obtained by periodically monitoring the beach at a preset frequency using a single-point lidar.
[0078] In this embodiment, a single-point lidar device is used to monitor the beach around the clock at a certain frequency, periodically collecting complex elliptical trajectory point cluster data as single-point lidar data. This elliptical trajectory can be understood as a laser trajectory. Preferably, the collection frequency can be 10 minutes per hour. After each data collection, the data is stored in a file according to a certain format, such as generating a bin file every hour. This embodiment does not impose any restrictions on this.
[0079] The single-point lidar data mainly includes seven types of data, which are used as key data for extracting wave height. For example, single-point lidar data includes, but is not limited to: terrain identifiers, time, distance from the lidar center point, motor encoder readings, and coordinate data (the coordinate axes are usually xyz) collected in a coordinate system with the lidar center as the origin.
[0080] Step 120: Perform format conversion preprocessing and visualization analysis on the lidar data, and segment the spatiotemporal distribution data of the left and right cross sections of the elliptical trajectory from the visualization data corresponding to the lidar data to obtain the three-dimensional data to be converted.
[0081] In this embodiment, the acquired LiDAR data is stored in a file according to a certain format. To facilitate subsequent processing, the data can be segmented and reconstructed. Specifically, the file is first converted to a common format through format conversion, making it easier to acquire and analyze the laser trajectory. Then, the converted data is processed using visualization technology to construct a visualization interface including visualization graphics. In the visualization interface, the data acquired by the single-point LiDAR is displayed in the form of elliptical trajectory point cloud data. By performing visualization analysis on the data, the cross-sections of the elliptical trajectory are calibrated. These cross-sections typically include a left cross-section and a right cross-section, referred to as the left and right cross-sections. In this embodiment, the calibration of the left and right cross-sections is used to filter out invalid data. For example, cross-section calibration can clearly identify the valid and invalid areas, allowing the data in the invalid areas to be filtered out.
[0082] In particular, when calibrating the left and right cross sections to filter out invalid data, overly scattered data (usually regarded as noise data or invalid data) can be filtered out, reducing the impact of noise data and invalid data on subsequent processing.
[0083] Finally, the segmented left and right cross-sectional data are reconstructed according to time series and spatial location to form spatiotemporal distribution data. In this data, the data can exist in both temporal and spatial distribution forms. Spatial distribution mainly reflects the changes in data along the cross-section within a certain area; temporal distribution can be understood as the time stamp sequence of data collection, which reflects the changes in waves at different times. Therefore, through the segmentation and reconstruction of the data, the elliptical trajectory is divided into left and right cross-sectional spatiotemporal distribution data, allowing for a direct representation of the dynamic changes of waves on the cross-section over time.
[0084] In this embodiment, the spatiotemporal distribution data is used as the three-dimensional data to be converted for subsequent coordinate system transformation processing. The three-dimensional data to be converted includes the pre-processed single-point lidar xyz three-dimensional data.
[0085] Step 130: Based on the three-dimensional data to be converted, coordinate system transformation is performed through fitting optimization to convert the coordinate system of the three-dimensional data to be converted into the geodetic coordinate system, thereby obtaining the data to be cleaned.
[0086] In the specific implementation, for the single-point lidar xyz 3D data obtained after preliminary processing, the coordinate system of the data is the instrument coordinate system. How to convert the data stored in the instrument coordinate system to the geodetic coordinate system is one of the key technical problems solved in this embodiment, that is, how to convert the coordinate system of the data from the instrument coordinate system to the geodetic coordinate system. This embodiment can process the 3D data through fitting optimization methods such as plane fitting, rotation matrix construction and transformation dynamic optimization. The main task is to convert the instrument coordinate system in the 3D data to the geodetic coordinate system. Through coordinate system transformation processing, instrument installation deviation and data fitting error can also be eliminated. The obtained data is used as the data to be cleaned for the data cleaning process.
[0087] Step 140: Based on the spatial density and data characteristics of the data to be cleaned, perform data cleaning and gridding processing on the data to be cleaned to obtain test data that retains local features.
[0088] In this embodiment, when processing the data to be cleaned, the data required for the final analysis is selected, including but not limited to three types of data: time series, straight-line distance, and elevation data. Then, a complete data cleaning and gridding process is performed to address issues such as missing values, outliers, and temporal anomalies in the dataset. Standardization is then used to provide high-quality input for subsequent wave height extraction, resulting in high-quality standardized test data.
[0089] For data cleaning, the main approach is to analyze the characteristics of the three types of data mentioned above, including spatial density and data properties, to determine the scope of data cleaning, and to perform unified cleaning on all types of data. During data cleaning, missing values, negative values, outliers, etc. are handled uniformly.
[0090] For gridding processing, this embodiment outputs a standardized grid after cleaning the data and retains local features to facilitate subsequent wave height extraction.
[0091] Step 150: Construct a dual-mode processing architecture based on the fusion of the time-domain zero-crossing method and the frequency-domain wave spectrum analysis method. Use the dual-mode processing architecture to extract wave height from the data to be measured to obtain the final wave height parameters.
[0092] In its implementation, this embodiment introduces the time-domain zero-crossing method and the frequency-domain wave spectrum analysis method, fusing the two algorithms to construct a dual-modal processing architecture (or dual-modal computational architecture), and innovatively constructing a cross-method result complementarity mechanism. Specifically, the time-domain zero-crossing method is used to process the measured data to extract wave height parameters, while the frequency-domain wave spectrum analysis method is used simultaneously to process the measured data to invert the significant wave height. Combining the wave height parameters and the significant wave height, the final wave height parameters are obtained.
[0093] As can be seen, this embodiment first periodically converts and visualizes the complex elliptical trajectory data collected by single-point lidar, reconstructing it into intuitive spatiotemporal distribution data along the cross-section. Then, the coordinate system of the data is converted to a geodetic coordinate system through fitting optimization. Intelligent data cleaning, based on spatial density and data characteristics, cleans the converted data and performs gridding, thereby diagnosing data anomalies, removing data errors, and preserving local features. Finally, a dual-modal processing architecture is constructed by combining the zero-point crossing method and spectral analysis to extract wave height from the data and output the final wave height parameters. Therefore, this embodiment achieves efficient and high-precision wave height extraction, enabling reliable and stable long-term observation. This allows radar-acquired data to be analyzed for wave height regardless of specific location, climate, or time, and can be widely applied to monitoring shoreline topography and wave parameters, solving the problem of insufficient precision and accuracy in wave height parameter acquisition using existing technologies.
[0094] Reference Figure 2 The diagram illustrates a step-by-step flowchart of a method for extracting shoreline wave height based on single-point lidar data according to an optional embodiment of this application. The method may specifically include the following steps:
[0095] Step 210: Acquire periodically collected single-point lidar data.
[0096] The single-point lidar data refers to the elliptical trajectory point cluster data obtained by periodically monitoring the beach at a preset frequency using a single-point lidar.
[0097] Step 220: Perform data format conversion processing on the lidar data and extract key data.
[0098] The key data includes terrain identifiers, time, center point distance, motor encoder readings, and three-dimensional coordinates.
[0099] Preferably, when the storage file for the collected single-point LiDAR data can be in bin format, the bin file can be converted to a txt file, with each txt file containing seven key data categories. The txt files are then input separately, and files with incorrect formats are automatically identified and processed, with the filename recorded in the log.
[0100] Step 230: Construct a visualization based on the key data. On the visualization, filter out invalid data located on the shore by visually calibrating the left and right cross-sections of the elliptical trajectory to obtain spatiotemporal distribution data.
[0101] The elliptical trajectory is a visualized laser acquisition data trajectory.
[0102] Step 240: Based on the characteristics of the point cloud trajectory of the spatiotemporal distribution data analysis, noise is filtered out from the dense data at the front end and the divergent data at the back end of the spatiotemporal distribution data to obtain the three-dimensional data to be converted.
[0103] Steps 230-240 are described uniformly as follows:
[0104] In this embodiment, because the instrument (which can be understood as a single-point lidar device) is positioned so that the trajectory's leading edge is located on a calm beach, the collected data includes some invalid points located on the shore. Therefore, for the acquired key data, the corresponding data file can be used as input to identify the laser trajectory and segment it. (Refer to...) Figure 3 As shown, during segmentation, data can be visualized and analyzed using software such as CloudCompare. By visually calibrating the left and right sections of the elliptical trajectory, most invalid points located on the shore can be filtered out.
[0105] Reference Figure 4 As shown, after initially filtering out invalid points, based on the characteristics of the data point cloud trajectory, data that is too dense at the front end and too scattered at the back end can be directly filtered out, that is, the dense data at the front end and the scattered data at the back end are filtered out, thereby reducing the impact of noisy and invalid data on subsequent processing.
[0106] Therefore, this embodiment improves the accuracy and precision of the extracted wave height parameters by accurately segmenting and filtering out some invalid data through trajectory segmentation, thus avoiding the influence of invalid data on the subsequent extraction of wave height parameters.
[0107] Step 250: Based on the three-dimensional data to be converted, coordinate system transformation is performed through fitting optimization to convert the coordinate system of the three-dimensional data to be converted into the geodetic coordinate system, thereby obtaining the data to be cleaned.
[0108] Optionally, this embodiment performs coordinate system transformation processing based on the three-dimensional data to be transformed, through fitting optimization, to convert the coordinate system of the three-dimensional data to be transformed into a geodetic coordinate system, thereby obtaining the data to be cleaned. This may include the following sub-steps:
[0109] Sub-step 2501: Using the least squares method, the three-dimensional data to be converted is fitted to a plane according to the time series to obtain the fitted three-dimensional data.
[0110] In related technologies, high-precision coordinate system transformation is mainly limited by two technical challenges: how to convert data stored in the instrument coordinate system to data stored in the geodetic coordinate system, and how to make adaptive and rapid adjustments according to different instrument angles.
[0111] To address the technical challenge of converting data stored in the instrument coordinate system to the geodetic coordinate system, this embodiment introduces plane fitting, rotation matrix construction, and dynamic optimization of the transformation, ultimately achieving high-precision conversion from the instrument coordinate system to the geodetic coordinate system.
[0112] Specifically, a number of data points (including at least several days' worth of data) are extracted from the 3D data to be converted according to a time series (e.g., at least several days). The least squares method is used to perform plane fitting on the data, solving for the plane parameters and then the relevant normal vectors to construct a rotation matrix. By processing the rotation matrix, the rotated 3D coordinates of the data are obtained. The coordinates of the 3D data to be converted are then updated based on these coordinates to obtain the fitted 3D data.
[0113] In an optional embodiment, the above-mentioned least squares method is used to perform planar fitting on the three-dimensional data to be converted according to the time series to obtain fitted three-dimensional data. Specifically, this may include: selecting input data from the three-dimensional data to be converted according to the time series; and performing planar fitting on the input data using the least squares method, based on... Solve for the plane parameters; using the plane parameters as input, according to Extract plane normal vector; obtain target normal vector Introducing Rodriguez rotation, according to Calculate the axis of rotation, and according to Calculate the rotation angle; using the rotation axis and rotation angle as a reference, compare the plane normal vector and the target normal vector, and when the comparison result indicates that the plane normal vector and the target normal vector do not coincide, according to... Construct a rotation matrix to obtain the fitted 3D data; where, It is the identity matrix. Let be the cross product matrix of the rotation axes.
[0114] The relevant parameters of this embodiment are described in detail below:
[0115] parameter These are all coefficients of the plane equation, used to collectively form the normal vector of the plane; All of these are three-dimensional coordinate values in the instrument coordinate system, i.e., the original data coordinates collected by the single-point lidar; is the constant term in the plane equation.
[0116] Plane normal vector Extracted by least-squares plane fitting of raw data from at least several days, reflecting the normal direction of the plane containing the data point cloud in the instrument coordinate system; target normal vector. The normal vector representing the geodetic coordinate system (usually understood as perpendicular to the horizontal plane of the earth and pointing upwards) is the target direction for coordinate system transformation.
[0117] The axes of rotation are calculated using Rodriguez's rotation formula. and rotation angle The rotation axis is calculated by the cross product of the original plane normal vector and the target normal vector, and can be understood as the perpendicular direction of the plane containing the two normal vectors, i.e., the axis around which the coordinate system rotates; the rotation angle is calculated by the dot product of the two normal vectors and the inverse cosine function, and can be understood as the angle required to rotate the original plane normal vector to the target normal vector, reflecting the angle between the two coordinate systems.
[0118] Using the above parameters as input, a rotation matrix can be constructed. It is used to convert three-dimensional coordinates in the instrument coordinate system to coordinates in the geodetic coordinate system.
[0119] Before constructing the rotation matrix in this embodiment, the plane normal vector and the target normal vector are compared based on the rotation axis and rotation angle. If the two do not coincide, the rotation matrix is constructed according to the rotation matrix formula.
[0120] Sub-step 2502 involves using the obtained three-dimensional coordinate points as a reference to perform matrix operations on the data coordinates and rotation matrix in the fitted three-dimensional data to obtain rotated three-dimensional data.
[0121] In the rotated three-dimensional data, the data coordinates are represented by data three-dimensional coordinates.
[0122] In this embodiment, by inputting three-dimensional coordinate points, the data coordinates and rotation matrix are correctly operated to obtain the rotated three-dimensional coordinates, thus converting the three-dimensional coordinates in the instrument coordinate system to the geodetic coordinate system and achieving precise alignment between different coordinate systems.
[0123] like Figure 5 As shown, Figure 5 The blue plane represents the plane fitted to the unrotated data, while the yellow plane is the target plane after rotation. The coordinate system transformation involves rotating the plane from the downward direction indicated by the blue arrow to the direction of the red arrow. At this point, on the new data plane (with the normal vector perpendicular to the ground and upward), it's equivalent to the original instrument coordinate system also undergoing a corresponding angle rotation. Figure 6 As shown.
[0124] Sub-step 2503: Extract the initial rotation angle from the fitted 3D data and use it as the observation point.
[0125] Sub-step 2504: Based on the observation points, the cross-sections in the rotated three-dimensional data are dynamically and iteratively adjusted until the instrument installation error and data fitting error are eliminated, and the data to be cleaned is obtained.
[0126] The initial rotation angle is the rotation angle used when processing the fitted three-dimensional data.
[0127] A unified description is provided for sub-steps 2501-2504:
[0128] In its implementation, this embodiment introduces a dynamic iterative adjustment algorithm for the rotation angle to enable adaptive and rapid adjustments based on different instrument angles. This algorithm iteratively tests and selects the instrument's angle for adaptive adjustment. Specifically, based on a plane fitting process, the rotation angle is obtained and used as the initial rotation angle, with this angle as the observation point. The rotation angle is adjusted iteratively, and an image is output at each adjustment. Based on the output image, it is determined whether the current rotation angle corresponds to the instrument's mounting angle, thus ultimately selecting the rotation angle corresponding to the instrument's mounting angle. Therefore, combined with verification using visualized point cloud distribution, this embodiment eliminates instrument installation deviations and data fitting errors.
[0129] In one optional embodiment, this embodiment, based on the observation point, dynamically and iteratively adjusts the cross-sections in the rotated 3D data until instrument installation errors and data fitting errors are eliminated, obtaining the data to be cleaned. Specifically, this may include: obtaining preset first adjustment values and second adjustment values, both of which are used to adjust the rotation angle; dynamically and iteratively adjusting the observation point using an angle iteration algorithm based on the first adjustment value, and processing the rotated 3D data in each adjustment, determining a first rotation angle based on the output first cross-sectional image; dynamically and iteratively adjusting the first rotation angle based on the second adjustment value, and processing the rotated 3D data in each adjustment, determining a target rotation angle based on the output second cross-sectional image, the target rotation angle being the rotation angle corresponding to the instrument setup angle; and verifying the data based on the visualized point cloud distribution data corresponding to the rotated 3D data at the target rotation angle to obtain the data to be cleaned after eliminating instrument installation errors and data fitting errors.
[0130] Reference Figure 7 As shown, based on the formula in the aforementioned embodiment, the initial rotation angle is... An angle iteration program is constructed using an angle iteration algorithm. Two rotation angle adjustment values are set: a first adjustment value and a second adjustment value. Since this embodiment first uses the first adjustment value for iterative adjustment, the first adjustment value can be greater than the second adjustment value. This allows for quick locking of the range of rotation angles corresponding to the instrument setup angle using the first adjustment value, followed by precise adjustment using the second adjustment value.
[0131] In this embodiment, the dynamic iterative adjustment of the rotation angle can be considered as a two-round iterative process. In the first round of iterative adjustment, the three-dimensional rotation data is input into the angle iteration program, where an arbitrary rotation angle slightly smaller than the initial value (i.e., the first adjustment value) is input. Preferably, the first adjustment value is 1 degree. At this time, one cross-sectional view corresponding to the three-dimensional rotation data is output every 1 degree, i.e., the first cross-sectional view. Assuming 10 iterations cover the range of rotation angles corresponding to the instrument setup angle, a total of 10 cross-sectional views can be obtained. Based on the analysis of the cross-sectional views output in the first round of iterations, the elimination of instrument installation errors and data fitting errors is assessed, and the rotation angle that eliminates the most errors is selected as the first rotation angle.
[0132] Then, based on the first rotation angle, a second round of iterative adjustment is performed using a second adjustment value, preferably 0.1 degrees. At this point, an iteration is performed for each 0.1-degree angle. The elimination of instrument installation errors and data fitting errors is analyzed based on the output second cross-sectional diagram. The rotation angle that eliminates both instrument installation and data fitting errors is taken as the target rotation angle, i.e., the rotation angle corresponding to the final instrument setup angle. Through automatic iteration based on the input angle, combined with verification using visualized point cloud distribution, instrument installation deviations and data fitting errors can be eliminated.
[0133] Therefore, this embodiment constructs a data rotation matrix by fitting a plane and matching the target normal vector, and then uses visualized data cutting and angle iterative optimization to achieve the optimal transformation from instrument coordinates to geodetic coordinates.
[0134] Step 260: Select time series data, straight-line distance and elevation data from the data to be cleaned as the data to be analyzed.
[0135] Step 270: Analyze the spatial density and data characteristics of the data to be analyzed, constrain the data to be analyzed according to the spatial density, and perform data cleaning and data correction on the data to be analyzed based on the data characteristics to obtain the first cleaned data.
[0136] Steps 260-270 are described uniformly as follows:
[0137] The final required data is selected from the data processed in the above steps, including but not limited to three types of data: time series, straight-line distance, and elevation data, as the data to be analyzed. Then, a complete data cleaning and gridding process is performed on it to solve problems such as missing values, outliers, and temporal anomalies in the dataset, and to provide high-quality input for subsequent wave height extraction through standardization.
[0138] Specifically, the rotated data can now intuitively reflect the straight-line distance and elevation of the measurement point from the radar installation location. At this point, statistical analysis is performed on the three selected data types to determine spatial density and data characteristics. Based on spatial density, a constraint range is selected for the data, thereby identifying valid data and reducing scattered, erroneous, and noisy data. Then, data cleaning and correction are performed on the data to be analyzed based on its characteristics. Data characteristics can be understood as the inherent properties of different data types, such as data structure, type, and whether the data can be negative (e.g., rotated data should not have negative values). During data cleaning and correction, a unified missing value and negative value detection process is performed on all types of data (rotated data should not have negative values). If data requiring correction (i.e., missing values) is detected, it is replaced with the nearest valid value.
[0139] Optionally, this embodiment analyzes the spatial density and data characteristics of the data to be analyzed, constrains the data to be analyzed according to the spatial density, and performs data cleaning and correction on the data to be analyzed based on the data characteristics to obtain first cleaned data. This may include: performing descriptive statistical analysis and visualization analysis on the data to be analyzed to determine the distribution state and numerical characteristics; analyzing the dispersion characteristics of the data according to the distribution state and numerical characteristics to determine the spatial density and data characteristics; using the spatial density as a benchmark, calibrating the data to be analyzed to constrain the spatial range; and performing missing value and negative value correction processing on the data to be analyzed within the spatial range to obtain cleaned data.
[0140] In the specific implementation, descriptive statistical analysis and visualization analysis are performed on the three selected data types to observe their distribution and numerical characteristics. It can be found that the data exhibits near-end dispersion and far-end dispersion. These intervals are mostly comprised of beach noise points at the front and error points caused by laser divergence at the back. Therefore, further calibration of the data intervals is necessary. At this point, interval selection (i.e., constraining the spatial range) can be made based on the statistical results. For example, when performing statistical analysis on a dataset, it is found that there are very few data points after 100m. In this case, the farthest distance can be calibrated as 100m. Based on the cross-sectional diagram and on-site survey, the first 50m is basically beach, so the closest distance can be calibrated as 50m. After constraining the spatial range, data cleaning and correction are performed.
[0141] Step 280: When the first cleaned data conforms to the preset distribution rules, the Chebyshev inequality is used to process outliers in the first cleaned data to obtain the second cleaned data.
[0142] Reference Figure 8 The elevation data cleaning and comparison chart is shown. In the specific implementation, after the above data is uniformly cleaned, outlier handling can be performed according to the characteristics of each data point. First, the distribution rules of the data are analyzed. If the distribution rules conform to preset distribution rules, the outlier handling process is performed. For example, when the three types of data after cleaning have inconsistent distribution patterns and are not all normally distributed, it is determined that they conform to preset distribution rules. Subsequently, Chebyshev's inequality is used and different thresholds are set to remove outliers, resulting in the second cleaned data. Among these, outliers are replaced by the nearest valid value.
[0143] In an optional embodiment, this embodiment uses Chebyshev's inequality to process outliers in the first cleaned data to obtain the second cleaned data. Specifically, this may include: using Chebyshev's inequality, according to... Outliers within a threshold range are removed from the first cleaned data, and the nearest valid value is used to replace the outliers, thus updating the first cleaned data. Time-series data of the first cleaned data undergoes breakage mitigation processing, with timestamps adaptively adjusted, and the first cleaned data is updated. Spatial data density of the first cleaned data is detected according to a preset threshold to determine the range of nullable data. The first cleaned data is then nulled according to the defined nullable data range, and the first cleaned data is updated to obtain the second cleaned data. This indicates the input sample data. This represents the mean of the dataset. The standard deviation of the dataset is represented by the standard deviation of the dataset. The threshold parameter is set.
[0144] In this embodiment, Chebyshev's inequality is used to handle outliers and update the cleaned data. Specifically, in the formula of this embodiment: It represents a single sample of data in a dataset (such as a specific value in time series, straight-line distance, or elevation data). It represents the mean of the dataset, reflecting the central tendency of the data. The standard deviation of a dataset reflects the degree of dispersion of the data. : This is a set threshold parameter used to define the range of anomalies. In this embodiment, different thresholds can be set according to the characteristics of different data. value.
[0145] Reference Figure 9 As shown, it should be noted that, in addition to outliers, time-series data may experience discontinuities or sudden reversals in the 10-minute timeframe due to sudden equipment malfunctions or filtering during previous processing. These anomalies will introduce noise into subsequent gridding interpolation, thus requiring specific additional processing of the time series. This solution detects time-series data; if the jump between two consecutive times exceeds a certain value, it adaptively adjusts subsequent timestamps to ensure data continuity and mitigate discontinuities.
[0146] After performing the above cleaning process, the data is further processed. Spatial data density is detected using a preset threshold, such as 0.1 meters per 50 points, and abnormal equipment states are identified simultaneously and recorded in the output log. Data with densities that do not meet the requirements is cleared to prevent significant errors in subsequent wave height calculations and ensure accuracy. This embodiment uses a 0.1-meter-per-50-point threshold to detect spatial data density and performs data cleaning simultaneously to prevent abnormal values from appearing in the extracted wave height parameters.
[0147] Step 290: Map the second cleaning data to grid points of a standardized grid to obtain the data to be tested.
[0148] Finally, this embodiment outputs the processed data as a standardized grid, mapping various data types to grid points, preserving local features, and facilitating subsequent wave height extraction.
[0149] Step 300: Construct a dual-mode processing architecture based on the fusion of the time-domain zero-crossing method and the frequency-domain wave spectrum analysis method. Use the dual-mode processing architecture to extract wave height from the data to be measured to obtain the final wave height parameters.
[0150] In an optional embodiment, the above-mentioned dual-modal processing architecture, based on the fusion of the time-domain zero-crossing method and the frequency-domain wave spectrum analysis method, is used to extract wave height from the test data to obtain the final wave height parameters. Specifically, this may include: based on the test data, using zero-intersection detection to segment the wave period and calculating the average wave height and maximum wave height of the large wave; based on the test data, using Hamming window-based FFT transform to suppress spectral leakage to generate a one-sided power spectral density, and performing dual optimization of filtering and binning smoothing on the one-sided power spectral density, and inverting the effective wave height through zero-moment integration; using the average wave height, maximum wave height, and effective wave height of the large wave as the final wave height parameters.
[0151] Reference Figure 10 As shown, in the specific implementation, a dual-modal computational architecture integrating the time-domain zero-crossing method and the frequency-domain wave spectrum analysis method is used to achieve high-precision extraction of coastal wave parameters. Specifically, firstly, zero-intersection detection is used to segment the wave period and calculate the average wave height (Hs) and maximum wave height (Hmax) of the first 1 / 3 of the large wave. Simultaneously, a one-sided power spectral density is generated based on the FFT transform to suppress spectral leakage using the Hamming window. After dual optimization of filtering and binning smoothing, the effective wave height is inverted through zero-moment integration.
[0152] Therefore, this embodiment innovatively constructs a cross-method result complementarity mechanism, automatically replacing the NaN value of the zero-point method with the spectral analysis result, overcoming the failure problem in low signal-to-noise ratio scenarios. The entire process adopts a batch automatic processing engine, dynamically identifying cross-sectional data types and realizing spatial location mapping, and outputting a wave height matrix by combining spatiotemporal correlation markers (extracting timestamps from files), supporting long-term shoreline wave height monitoring. At the engineering level, vectorization operations and pre-sorting accelerate calculations, and robust designs such as null value safety handling and matrix dimension verification significantly improve the monitoring reliability of lidar in complex sea conditions.
[0153] In summary, the solution provided in this application embodiment can extract the wave height of the beach survey line by combining four technologies: precise trajectory segmentation, coordinate transformation optimization, intelligent data cleaning, and multi-algorithm comprehensive calculation. This achieves efficient and high-precision wave height extraction, making radar data acquisition not limited to specific locations, climates, and times. This embodiment is based on a fully automated processing architecture. After the first round of parameter calibration, it can automatically extract wave heights from the same batch (location) of data without manual intervention, supporting batch data processing. Simultaneously, it uses visual calibration to assist in data segmentation, reducing data errors and removing useless data, thus improving processing efficiency. In coordinate system transformation, it overcomes the shortcomings of traditional manual parameter fixing and angle fine-tuning, which accumulates large errors and is time-consuming. It innovatively integrates a normal vector fitting + angle iterative optimization method. A complete and meticulous data cleaning scheme is adopted to further improve the accuracy of the final results. Data density—equipment linkage detection—is introduced, monitoring data characteristics while processing data to provide feedback on abnormal equipment operating states and reduce the impact of abnormal data. Furthermore, this algorithm overcomes terrain and climate adaptability limitations, enabling long-term wave monitoring in different terrain scenarios, and rich visualization options simplify data research and quality assessment.
[0154] It should be noted that, for the sake of simplicity, the method embodiments are described as a series of actions. However, those skilled in the art should know that the embodiments of this application are not limited to the described order of actions, because according to the embodiments of this application, some steps may be performed in other orders or simultaneously.
[0155] like Figure 11 As shown in the figure, this application embodiment also provides a shoreline survey wave height extraction system 1100 based on single-point lidar data, including:
[0156] The data preprocessing module 1110 is used to acquire periodically collected single-point lidar data, which is elliptical trajectory point cluster data obtained by the single-point lidar periodically monitoring the beach at a preset frequency; the lidar data is preprocessed by format conversion and visualization analysis, and the spatiotemporal distribution data of the left and right sections of the elliptical trajectory are segmented from the visualization data corresponding to the lidar data to obtain the three-dimensional data to be converted.
[0157] The coordinate transformation module 1120 is used to perform coordinate system transformation processing on the three-dimensional data to be transformed by fitting optimization, and to transform the coordinate system of the three-dimensional data to be transformed into the geodetic coordinate system to obtain the data to be cleaned.
[0158] The intelligent cleaning and anomaly detection module 1130 is used to perform data cleaning and gridding processing on the data to be cleaned based on the spatial density and data characteristics of the data to be cleaned, so as to obtain test data that retains local features.
[0159] The wave height calculation module 1140 is used to construct a dual-mode processing architecture based on the fusion of the time-domain zero-crossing method and the frequency-domain wave spectrum analysis method. The wave height is extracted from the data to be measured through the dual-mode processing architecture to obtain the final wave height parameters.
[0160] It should be noted that the beach survey line wave height extraction system based on single-point lidar data provided in this application embodiment can execute the beach survey line wave height extraction method based on single-point lidar data provided in any embodiment of this application, and has the corresponding functions and beneficial effects of the execution method.
[0161] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0162] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.
Claims
1. A method for extracting wave height of beach survey lines based on single-point lidar data, characterized in that, include: Acquire periodically collected single-point lidar data, wherein the single-point lidar data is lidar data obtained by periodically monitoring the beach at a preset frequency using a single-point lidar. The lidar data is preprocessed by format conversion and visualized. The spatiotemporal distribution data of the left and right sections of the elliptical trajectory are segmented from the visualized data corresponding to the lidar data to obtain the three-dimensional data to be converted. Based on the three-dimensional data to be converted, coordinate system transformation is performed through fitting optimization to convert the coordinate system of the three-dimensional data to be converted into the geodetic coordinate system, thus obtaining the data to be cleaned; Based on the spatial density and data characteristics of the data to be cleaned, the data to be cleaned is subjected to data cleaning and gridding processing to obtain test data that retains local features; A dual-mode processing architecture is constructed based on the fusion of the time-domain zero-crossing method and the frequency-domain wave spectrum analysis method. The wave height is extracted from the test data through the dual-mode processing architecture to obtain the final wave height parameters. The process involves preprocessing the lidar data through format conversion and visualization analysis. The spatiotemporal distribution data of the left and right sections of the elliptical trajectory is segmented from the visualization data corresponding to the lidar data to obtain the 3D data to be converted. This includes: performing data format conversion on the lidar data to extract key data, including terrain identifiers, time, center point distance, motor encoder readings, and 3D coordinates; constructing a visualization based on the key data; visually calibrating the left and right sections of the elliptical trajectory on the visualization, filtering out invalid data located on the shoreline to obtain spatiotemporal distribution data, where the elliptical trajectory is the visualized laser acquisition data trajectory; and analyzing the characteristics of the point cloud trajectory based on the spatiotemporal distribution data to filter out noise from the dense data at the front end and the divergent data at the back end of the spatiotemporal distribution data to obtain the 3D data to be converted. Based on the three-dimensional data to be converted, coordinate system transformation is performed through fitting optimization to convert the coordinate system of the three-dimensional data to be converted into a geodetic coordinate system, resulting in data to be cleaned. This includes: using the least squares method to perform planar fitting on the three-dimensional data to be converted according to a time series, obtaining fitted three-dimensional data; using the obtained three-dimensional coordinate points as a reference, performing matrix operations on the data coordinates and rotation matrix in the fitted three-dimensional data, obtaining rotated three-dimensional data, where the data coordinates are represented by data three-dimensional coordinates; extracting an initial rotation angle from the fitted three-dimensional data as an observation point; and based on the observation point, dynamically iteratively adjusting the cross-sections in the rotated three-dimensional data until instrument installation errors and data fitting errors are eliminated, resulting in data to be cleaned; wherein, the initial rotation angle is the rotation angle used when rotating the fitted three-dimensional data.
2. The method according to claim 1, characterized in that, The least squares method is used to perform planar fitting on the three-dimensional data to be transformed according to the time series, resulting in fitted three-dimensional data, including: Input data is selected from the three-dimensional data to be converted according to the time series. The least squares method is used to perform plane fitting on the input data, based on... Solve for the plane parameters; Using plane parameters as input, according to Extract the plane normal vector; Obtain the target normal vector ; Introducing Rodriguez rotation, according to Calculate the axis of rotation, and according to Calculate the rotation angle; Using the rotation axis and rotation angle as references, the plane normal vector and the target normal vector are compared. When the comparison result indicates that the plane normal vector and the target normal vector do not coincide, according to... Construct a rotation matrix to obtain the fitted 3D data; in, It is the identity matrix. Let be the cross product matrix of the rotation axes.
3. The method according to claim 2, characterized in that, Based on the observation points, the cross-sections in the rotated 3D data are dynamically and iteratively adjusted until instrument installation errors and data fitting errors are eliminated, resulting in data to be cleaned, including: Obtain a preset first adjustment value and a second adjustment value, both of which are used to adjust the rotation angle; Based on the first adjustment value, the observation point is dynamically and iteratively adjusted using an angle iteration algorithm. In each adjustment, the rotational three-dimensional data is processed, and the first rotation angle is determined based on the output first cross-sectional image. Based on the second adjustment value, the first rotation angle is dynamically iteratively adjusted, and in each adjustment, the rotation three-dimensional data is processed. Based on the output second cross-sectional image, the target rotation angle is determined, and the target rotation angle is the rotation angle corresponding to the instrument setup angle. At the target rotation angle, the visualization point cloud distribution data corresponding to the rotation three-dimensional data is used for verification to obtain the data to be cleaned after eliminating instrument installation errors and data fitting errors.
4. The method according to claim 1, characterized in that, Based on the spatial density and data characteristics of the data to be cleaned, the data to be cleaned is subjected to data cleaning and gridding processing to obtain test data that retains local features, including: Time series data, straight-line distance, and elevation data are selected from the data to be cleaned as the data to be analyzed. The spatial density and data characteristics of the data to be analyzed are analyzed. The data to be analyzed is constrained according to the spatial density, and the data to be analyzed is cleaned and corrected based on the data characteristics to obtain the first cleaned data. When the first cleaned data conforms to the preset distribution rules, Chebyshev's inequality is used to process outliers in the first cleaned data to obtain the second cleaned data. The second cleaning data is mapped to grid points of a standardized grid to obtain the data to be tested.
5. The method according to claim 4, characterized in that, Analyze the spatial density and data characteristics of the data to be analyzed, constrain the data to be analyzed according to the spatial density, and perform data cleaning and correction on the data to be analyzed based on the data characteristics to obtain first cleaned data, including: Descriptive statistical and visual analyses are performed on the data to be analyzed to determine the distribution status and numerical characteristics; Based on the distribution status and digital characteristics, the dispersion characteristics of the data are analyzed to determine the spatial density and data properties; Based on the spatial density, the data interval of the data to be analyzed is calibrated to constrain the spatial range; The data to be analyzed within the specified spatial range is corrected for missing and negative values to obtain cleaned data.
6. The method according to claim 4, characterized in that, Outlier processing of the first cleaned data is performed using Chebyshev's inequality to obtain the second cleaned data, which includes: Using Chebyshev's inequality, according to The first cleaned data is subjected to outlier removal within a threshold range, and the outlier is replaced with the nearest valid value to update the first cleaned data; The time-series data of the first cleaned data is subjected to fragmentation mitigation processing, and the timestamps are adaptively adjusted to update the first cleaned data. The spatial data density of the first cleaned data is detected according to a preset threshold to determine the range of empty data; The first cleaned data is emptied according to the emptied data range, and the first cleaned data is updated to obtain the second cleaned data; in, This indicates the input sample data. This represents the mean of the dataset. The standard deviation of the dataset is represented by the standard deviation of the dataset. The threshold parameter is set.
7. The method according to claim 1, characterized in that, A dual-mode processing architecture is constructed based on the fusion of the time-domain zero-crossing method and the frequency-domain wave spectrum analysis method. This architecture is used to extract wave height parameters from the measured data, yielding the final wave height parameters, including: Based on the data to be tested, the wave cycle is segmented by zero intersection detection, and the average wave height and maximum wave height of the large wave are calculated. Based on the data to be measured, a Hamming window is used to suppress spectral leakage by FFT transform to generate a one-sided power spectral density, and the one-sided power spectral density is optimized by filtering and binning smoothing. The effective wave height is then inverted by zero-order moment integration. The average wave height, maximum wave height, and significant wave height of the large wave are used as the final wave height parameters.
8. A system for extracting shoreline wave height based on single-point lidar data, characterized in that, include: The data preprocessing module is used to acquire periodically collected single-point lidar data, which is elliptical trajectory point cluster data obtained by single-point lidar periodically monitoring the beach at a preset frequency; the lidar data is preprocessed by format conversion and visualization analysis, and the spatiotemporal distribution data of the left and right sections of the elliptical trajectory are segmented from the visualization data corresponding to the lidar data to obtain the three-dimensional data to be converted. The coordinate transformation module is used to perform coordinate system transformation processing on the three-dimensional data to be transformed by fitting and optimization, so as to convert the coordinate system of the three-dimensional data to be transformed into the geodetic coordinate system and obtain the data to be cleaned. The intelligent cleaning and anomaly detection module is used to perform data cleaning and gridding processing on the data to be cleaned based on the spatial density and data characteristics of the data to be cleaned, so as to obtain test data that retains local features; The wave height calculation module is used to construct a dual-mode processing architecture based on the fusion of the time-domain zero-crossing method and the frequency-domain wave spectrum analysis method. The wave height is extracted from the data to be measured through the dual-mode processing architecture to obtain the final wave height parameters. The process involves preprocessing the lidar data through format conversion and visualization analysis. The spatiotemporal distribution data of the left and right sections of the elliptical trajectory is segmented from the visualization data corresponding to the lidar data to obtain the 3D data to be converted. This includes: performing data format conversion on the lidar data to extract key data, including terrain identifiers, time, center point distance, motor encoder readings, and 3D coordinates; constructing a visualization based on the key data; visually calibrating the left and right sections of the elliptical trajectory on the visualization, filtering out invalid data located on the shoreline to obtain spatiotemporal distribution data, where the elliptical trajectory is the visualized laser acquisition data trajectory; and analyzing the characteristics of the point cloud trajectory based on the spatiotemporal distribution data to filter out noise from the dense data at the front end and the divergent data at the back end of the spatiotemporal distribution data to obtain the 3D data to be converted. Based on the three-dimensional data to be converted, coordinate system transformation is performed through fitting optimization to convert the coordinate system of the three-dimensional data to be converted into a geodetic coordinate system, resulting in data to be cleaned. This includes: using the least squares method to perform planar fitting on the three-dimensional data to be converted according to a time series, obtaining fitted three-dimensional data; using the obtained three-dimensional coordinate points as a reference, performing matrix operations on the data coordinates and rotation matrix in the fitted three-dimensional data, obtaining rotated three-dimensional data, where the data coordinates are represented by data three-dimensional coordinates; extracting an initial rotation angle from the fitted three-dimensional data as an observation point; and based on the observation point, dynamically iteratively adjusting the cross-sections in the rotated three-dimensional data until instrument installation errors and data fitting errors are eliminated, resulting in data to be cleaned; wherein, the initial rotation angle is the rotation angle used when rotating the fitted three-dimensional data.