Satellite-borne laser photon layering multi-scale denoising method and system based on minimum spanning tree
Through the minimum spanning tree and box graph anomaly point detection algorithm, the accuracy problem of satellite-borne laser photon denoising under density heterogeneity in different water depth areas is solved, and high-precision subsea photon signal extraction is achieved, simplifying the denoising process And improve the reliability of signal extraction.
Patent Information
- Application Number
- CN202411040431.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-31
- Publication Date
- 2024-11-19
AI Technical Summary
The existing satellite-borne laser photon denoising method is based on point density clustering algorithm, which is difficult to accurately denoise under the density heterogeneity of different water depth areas, which easily leads to high-density noise on the nearshore surface as signals and low-density signals on the seabed as noise. question.
Using a multi-scale denoising method of satellite-borne laser photons based on the minimum spanning tree, the photon separation between sea surface and water body, the photon distance is standardized, and the minimum spanning tree is constructed and segmented, and the box-shaped graph anomaly point detection algorithm is used to achieve different water depths. Accurate extraction of regional undersea photon signals.
It effectively overcomes the shortcomings of point density clustering algorithm and obtains high-precision and high-confidence subsea satellite-borne laser photon signal extraction results. It is simple and reliable, and is suitable for denoising satellite-borne laser photons in different water depth areas.
Smart Images

Figure CN118981003A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a minimum spanning tree-based satellite-borne laser photon hierarchical multi-scale denoising method and system, belonging to the technical field of satellite-borne laser ranging data processing and application. Background Art
[0002] As the areas with the most concentrated human activities and the most prosperous economy, the geospatial information data of shallow sea areas near the coast and islands and reefs is an important basic strategic resource, providing basic data and important basis for marine engineering, marine military and marine scientific research. However, the waters around the coast and islands and reefs are shallow, with large tidal ranges, many reefs and extremely complex marine environments. The underwater topography measurement has always been a major difficulty in the field of marine surveying and mapping. Remote sensing satellites have the characteristics of wide observation range, short revisit cycle and low economic cost. Satellite derived bathymetry (SDB) has become an important way to obtain shallow sea topography through remote sensing satellite images. SDB usually requires measured water depth data as control points to establish an inversion model. The measured water depth data traditionally relied on comes from shipborne single-beam, multi-beam or airborne laser depth sounding systems; however, shipborne single-beam, multi-beam and airborne laser radar depth sounding systems are limited by factors such as weather conditions, observation range, minimum water level, airspace control and economic benefits. It is difficult to achieve large-scale, high-frequency and low-cost near-coastal and island reef shallow sea topography data, especially for areas with rights disputes and border restrictions. In September 2018, the National Aeronautics and Space Administration (NASA) launched a new generation of photon counting laser radar satellite ICESat-2 (Ice, Cloud, and Land Elevation Satellite-2), which can effectively overcome the problem of obtaining measured water depth control point data.
[0003] ICESat-2 is the world's first and only satellite-borne photon counting lidar satellite. The Advanced Topographic Laser Altimeter System (ATLAS) carried by the platform can transmit and receive photons and record the round-trip time. The system uses single-photon counting to emit laser pulses. The error of each photon flight time recorded is within 800ps, and the distance resolution along the track can reach 0.7m. The ICESat-2 satellite orbit altitude is about 500km, with an inclination of 92°, and the observation range covers the area of 88°S~88°N on the surface latitude. Different from the previous generation of satellite full-waveform laser altimeter systems, ATLAS has the characteristics of micro-pulses, high repetition rate and high precision. The 532nm wavelength green light band electromagnetic waves it uses can penetrate water and obtain underwater photon signals. Studies have shown that ICESat-2 laser photons can penetrate to a depth of about 38m under the sea. Since ATLAS transmits and receives weak signals, affected by environmental factors such as the solar background, atmosphere, and the system's own noise, there is a lot of noise in the acquired satellite laser data, and the amount and distribution of noise vary greatly in different trajectory data. In order to obtain accurate satellite laser photon signals on the sea surface and the seabed, it is necessary to denoise the original satellite laser data. Existing satellite laser photon denoising methods are mostly based on point density clustering algorithms such as DBSCAN and OPTICS for photon denoising, which can easily cause the phenomenon of "high-density noise points near the sea surface are identified as signals, and low-density signal points on the seabed are identified as noise", making it difficult to obtain high-precision satellite laser photon denoising results.
[0004] For example, Chinese patent publication number CN114494287A discloses a method for processing long-distance laser radar point cloud data. The method uses an optical scanning system to scan the target environment for data collection, collects the flight time of photons at each pixel point, and derives and extracts the collected data; applies the point cloud data segmentation method to calculate the filter parameters to complete parameter adaptation; applies the principal component analysis technology to reduce the dimension of the data; applies the density-based spatial clustering denoising (DBSCAN) algorithm to cluster the point cloud, deletes outliers, i.e., noise points, and finally upgrades the obtained two-dimensional point cloud to three-dimensional point cloud data to complete the denoising filtering of the original point cloud data. The DBSCAN algorithm based on the idea of point density clustering is used for photon denoising. When applied to ocean satellite laser photons, it is also easy to cause the phenomenon of "high-density noise points near the sea surface are identified as signals, and low-density signal points on the seabed are identified as noise", making it difficult to obtain high-precision satellite laser photon denoising results.
[0005] In view of the above problems, the present invention proposes a spaceborne laser photon hierarchical multi-scale denoising method based on minimum spanning tree. Summary of the invention
[0006] The purpose of the present invention is to propose a hierarchical multi-scale denoising method and system for satellite-borne laser photons based on a minimum spanning tree, so as to solve the problem that it is difficult to accurately denoise satellite-borne laser photons with density heterogeneity in different water depth areas based on a point density clustering algorithm.
[0007] The present invention discloses a method for layered multi-scale denoising of satellite-borne laser photons based on a minimum spanning tree, comprising separation of satellite-borne laser photons on the sea surface and in the water body, standardization of the distance of satellite-borne laser photons, layered multi-scale denoising of satellite-borne laser photons and refinement of the denoising results of satellite-borne laser photons. The layered multi-scale denoising of satellite-borne laser photons realizes denoising of satellite-borne laser photons on the seabed in different water depth areas by constructing and dividing a minimum spanning tree.
[0008] Firstly, the separation of satellite laser photons from the sea surface and water bodies is realized based on the kernel density evaluation idea; then, the distance and elevation of photons along the track are standardized based on the Z-score model to eliminate the magnitude difference between the distance and elevation of photons along the track; secondly, a hierarchical multi-scale denoising method for water body satellite laser photons based on the minimum spanning tree is proposed to achieve accurate extraction of seabed satellite laser photons; finally, the denoised seabed satellite laser photons are refined based on the box plot outlier detection algorithm to eliminate the influence of outlier photon points on the extraction of seabed satellite laser photons. It can achieve accurate denoising of satellite laser photons with density heterogeneity in different water depths and obtain accurate extraction results of seabed satellite laser photons; it is simple, reliable, accurate and easy to implement.
[0009] The specific steps include:
[0010] Step S1, separation of sea surface and water body satellite-borne laser photons: based on the kernel density evaluation algorithm, the distribution of shallow sea satellite-borne laser photons is evaluated to achieve the separation of sea surface satellite-borne laser photons and water body satellite-borne laser photons, and eliminate the influence of sea surface satellite-borne laser photons on the extraction of seabed satellite-borne laser photons;
[0011] Step S2, spaceborne laser photon distance standardization: based on the Z-score model, the shallow sea spaceborne laser photon along-track distance and elevation are standardized;
[0012] Step S3, hierarchical multi-scale denoising of satellite-borne laser photons: denoising of laser photons in different water depth areas is achieved by constructing and segmenting a minimum spanning tree;
[0013] Step S4, refinement of satellite-borne laser photon denoising results: refine the denoised seabed satellite-borne laser photons based on the box plot to eliminate the influence of abnormal photon points on the extraction of seabed satellite-borne laser photons.
[0014] Preferably, in step S1, a Gaussian kernel density function is used to calculate the Gaussian kernel density value of satellite-borne laser photons, and the elevation corresponding to the maximum Gaussian kernel density value is obtained. Satellite-borne laser photons corresponding to the Gaussian kernel density value within the elevation range value are selected to obtain sea surface satellite-borne laser photon signals. The values greater than the elevation range value are sea surface noise, and the values less than the elevation range value are seabed satellite-borne laser photon signals, thereby achieving separation of sea surface satellite-borne laser photons and water body satellite-borne laser photons (i.e., seabed satellite-borne laser photons).
[0015] Preferably, the elevation corresponding to the maximum Gaussian kernel density value is h, that is, the center position of the sea surface at this time; the elevation range value is [h-β, h+β], the calculation formula of Gaussian kernel density is shown in formula (1), and the judgment criteria of sea surface, sea surface noise and seabed satellite-borne laser photons are shown in formula (2):
[0016]
[0017] Where: n is the number of satellite-borne laser photons; W is the kernel density function adjustment constant; x is the satellite-borne laser photon elevation; β is the sea surface fluctuation constant; cls noise ,cls sea-surface and cls seaf l oor They are sea surface noise, sea surface satellite-borne laser photon signals and seabed satellite-borne laser photon category labels respectively.
[0018] Specifically, in step 1, it is proposed to use a kernel density evaluation algorithm to achieve the separation of sea surface and water body photons, where the sea surface fluctuation constant can be set to 0.5.
[0019] Preferably, the calculation formula for normalizing the along-track distance and elevation of shallow sea star-borne laser photons based on the Z-score model in step S2 is as follows:
[0020]
[0021]
[0022] in: and are the original distance and standardized distance of the satellite-borne laser photon labeled i; μ is the mean distance of the satellite-borne laser photons; δ is the standard deviation of the distance of the satellite-borne laser photons; and n is the number of satellite-borne laser photons.
[0023] The Z-score model is used to standardize the along-track distance and elevation of shallow sea star-borne laser photons to eliminate the magnitude difference between the along-track distance and elevation of photons.
[0024] Preferably, the step S3 specifically includes the following steps:
[0025] Step S31, minimum spanning tree construction: clustering the photons based on the point space distance, adding them to the minimum spanning tree in turn according to the distance between the satellite-borne laser photons, and constructing the minimum spanning tree;
[0026] Step S32, determining the optimal segmentation threshold: fitting a straight line equation based on the first and last endpoints of the branch length trend graph of the minimum spanning tree, calculating the distance from the point on the trend graph to the fitted straight line, and the point with the longest distance is the optimal segmentation threshold for the minimum spanning tree;
[0027] Step S33, minimum spanning tree segmentation: in the minimum spanning tree, branches whose length is greater than the optimal segmentation threshold are broken, and the minimum spanning tree is segmented into multiple subtrees; the number of satellite-borne laser photons contained in the subtrees is counted, and the subtrees whose satellite-borne laser photons are less than the set minimum photon number threshold are marked as noise, otherwise they are marked as signals, thereby realizing denoising of satellite-borne laser photons.
[0028] Preferably, the calculation formula for the distance between satellite-borne laser photons in step S31 is as follows:
[0029]
[0030] in: is the distance between the spaceborne laser photons of tags i and j; SLP x i* and SLP h i* are the along-track distance and elevation of the satellite-borne laser photon with label i, respectively; and are the along-track distance and elevation of the satellite-borne laser photon of label j, respectively;
[0031] The optimal segmentation threshold calculation formula in step S32 is as follows:
[0032]
[0033] Where: (x i ,y i ) is the branch number and length in the trend graph; k and b are the slope and intercept of the fitting line respectively; m is the number of branches in the minimum spanning tree; L max is the distance corresponding to the optimal segmentation threshold.
[0034] Preferably, the minimum photon number threshold set in step S33 is used to judge the subtree photon number based on multi-scale threshold denoising, and the water depth is divided into four areas, namely: The minimum photon number thresholds corresponding to the four regions are 12, 9, 6, and 3, respectively; that is, the water area is divided into four layers of regions to determine the number of subtree nodes;
[0035] Where: D low and Dhigh are the elevations of the shallowest and deepest seafloor satellite-borne laser photons, respectively; T is the elevation difference between the shallowest and deepest seafloor satellite-borne laser photons, that is, T = |D low -D high |.
[0036] Preferably, in step S4, the calculation of the refined processing of the denoised seafloor satellite laser photons based on the box plot is as follows:
[0037]
[0038] in: and are the category label and elevation of the spaceborne laser photon of label i; cls seafloor and cls noise They respectively indicate that the laser photons are seafloor satellite-borne laser photon signals or noise; Q1 and Q3 are the The number of laser photons and The laser photon elevation corresponding to the number of laser photons; IQR is The number of laser photons and The laser photon elevation difference corresponding to the number of laser photons.
[0039] The denoised seafloor satellite laser photons are refined based on the box plot outlier detection algorithm. The outlier detection is performed using segmented distance along the track, and the segmented distance can be set to 20m.
[0040] A system for realizing the above-mentioned minimum spanning tree-based on-board laser photon hierarchical multi-scale denoising method of the present invention comprises:
[0041] Sea surface and water body satellite laser photon separation module: used to evaluate the distribution of shallow sea satellite laser photons based on the kernel density evaluation algorithm, realize the separation of sea surface satellite laser photons and water body satellite laser photons, and eliminate the influence of sea surface satellite laser photons on the extraction of seabed satellite laser photons;
[0042] Satellite-borne laser photon distance standardization module: used to standardize the along-track distance and elevation of shallow-water satellite-borne laser photons based on the Z-score model;
[0043] Spaceborne laser photon hierarchical multi-scale denoising module: used to denoise laser photons in different water depths by constructing and splitting the minimum spanning tree;
[0044] Satellite laser photon denoising result refinement module: used to refine the denoised seabed satellite laser photons based on the box plot to eliminate the influence of abnormal photon points on the extraction of seabed satellite laser photons.
[0045] Compared with the existing technology, the minimum spanning tree based on-board laser photon hierarchical multi-scale denoising method and system of the present invention have the following advantages:
[0046] The present invention discloses a method and system for hierarchical multi-scale denoising of satellite-borne laser photons based on a minimum spanning tree. By constructing and splitting a minimum spanning tree and based on the idea of hierarchical multi-scale subtree node judgment, the present invention can well couple the inconsistent density distribution of shallow-sea satellite-borne laser photons in different water depths, overcome the phenomenon that satellite-borne laser photon denoising algorithms based on the idea of point density clustering are prone to "identifying high-density noise points near the sea surface as signals and low-density signal points on the seabed as noise", and obtain accurate and high-confidence seabed satellite-borne laser photon signal extraction results. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] Figure 1 This is a schematic diagram of the distribution of shallow sea star-borne laser photons in Example 1 of the present invention;
[0048] Figure 2 This is a kernel density evaluation diagram of the shallow sea star-borne laser photon distribution in Example 1 of the present invention;
[0049] Figure 3 This is a schematic diagram of the normalized along-track distance and elevation of satellite-borne laser photons in Example 1 of the present invention;
[0050] Figure 4 This is a schematic diagram of the process of constructing a minimum spanning tree of satellite-borne laser photons in Example 1 of the present invention;
[0051] Figure 5 This is a schematic diagram of calculating the optimal segmentation threshold of the minimum spanning tree in Example 1 of the present invention;
[0052] Figure 6 This is a diagram of the minimum spanning tree segmentation result in Example 1 of the present invention. DETAILED DESCRIPTION
[0053] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.
[0054] Example 1
[0055] This embodiment discloses a method for layered multi-scale denoising of satellite-borne laser photons based on a minimum spanning tree. The method can be supported by computer software to automatically run the process, including:
[0056] (1) Separation of satellite-borne laser photons on the sea surface and in the water: The distribution of satellite-borne laser photons in shallow water is evaluated based on the kernel density evaluation algorithm to achieve separation of satellite-borne laser photons on the sea surface and photons in the water, thereby eliminating the influence of satellite-borne laser photons on the seabed on the extraction of satellite-borne laser photons.
[0057] Among the satellite-borne laser photons near the coast and island reefs, the satellite-borne laser photons on the sea surface usually have more obvious distribution characteristics, such as Figure 1 As shown; accurately extracting sea surface satellite laser photons and removing photon noise above the sea surface can eliminate the influence of sea surface satellite laser photons and photon noise above the sea surface on the extraction of seabed satellite laser photon signals. The Gaussian kernel density function can effectively count the density distribution of spatial points. The denser the spatial points, the larger the Gaussian kernel density value of the points, and vice versa. Satellite-borne laser photons are usually most densely distributed at the sea surface. In this embodiment, the Gaussian kernel function is used to calculate the Gaussian kernel density value of satellite-borne laser photons to obtain the elevation h corresponding to the maximum Gaussian kernel density value, which is the center position of the sea surface at this time; the satellite-borne laser photons corresponding to the Gaussian kernel density within [h-β, h+β] are considered to be sea surface satellite laser photons, thereby obtaining sea surface satellite laser photon signals; those greater than h+β are considered to be sea surface noise, and those less than h-β are seabed satellite laser photons, such as Figure 2 The calculation formula of Gaussian kernel density is shown in formula (1), and the judgment criteria of sea surface, sea surface noise and seabed satellite-borne laser photons are shown in formula (2).
[0058]
[0059]
[0060] Where: n is the number of satellite laser photons; W is the kernel density function adjustment constant; x is the satellite laser photon elevation; β is the sea surface fluctuation constant, usually set to 0.5; cls noise ,cls sea-surface and cls seafloor They are sea surface noise, sea surface satellite-borne laser photon signals and seabed satellite-borne laser photon category labels respectively.
[0061] (2) Standardization of spaceborne laser photon distance: Based on the Z-score model, the along-track distance and elevation of shallow-water spaceborne laser photons are standardized to eliminate the magnitude difference between the along-track distance and elevation of photons.
[0062] exist Figure 1In the figure, the along-track distance is in the thousands, and the elevation is in the tens. The inconsistent magnitudes will affect the distance measurement between satellite-borne laser photons, and further affect the denoising effect of satellite-borne laser photons. In order to eliminate the impact of the inconsistent magnitudes of along-track distance and elevation on the distance measurement between satellite-borne laser photons, it is necessary to standardize the along-track distance and elevation of laser photons so that the two types of distance measurements are within the same magnitude range. Z-Score standardization is the most commonly used data standardization processing algorithm in data analysis and mining systems. It can truly reflect the relative standard distance of the average of a number of distances, convert data of different magnitudes into the same magnitude, and uniformly measure them with the calculated Z-Score value to ensure comparability between data. Therefore, in this embodiment, Z-Score standardization is used to standardize the along-track distance and elevation of satellite-borne laser photons to eliminate the impact of the inconsistent magnitudes of along-track distance and elevation of laser photons on the denoising effect of satellite-borne laser photons. The along-track distance and elevation of photons after Z-Score standardization are as follows. Figure 3 The Z-Score standardization calculation is as follows:
[0063]
[0064] in: and are the original distance and the standardized distance of the satellite-borne laser photon labeled i; μ is the mean distance of the satellite-borne laser photons; δ is the standard deviation of the distance of the satellite-borne laser photons; and n is the number of satellite-borne laser photons that need to be standardized.
[0065] (3) Layered multi-scale denoising of satellite-borne laser photons: By constructing and splitting the minimum spanning tree, laser photons in different water depth areas can be denoised.
[0066] The distribution of satellite-borne laser photons usually becomes sparser with increasing seawater depth, and the distribution density of photons varies at different water depths. The existing photon denoising process based on the idea of point density clustering can easily cause the phenomenon that high-density noise points on the sea surface are identified as signals and low-density signal points on the seabed are identified as noise. The spatial point data clustering process based on the minimum spanning tree can effectively retain spatial point data with different density distributions and eliminate the influence of spatial outliers on the clustering of spatial points with different densities. The spatial point clustering algorithm based on the minimum spanning tree has been widely used in image segmentation, spatial data mining and other fields, and has achieved good clustering effects. Based on the spatial distribution characteristics of satellite-borne laser photons, this embodiment proposes a hierarchical multi-scale denoising method for seabed satellite-borne laser photons based on the minimum spanning tree. The method is mainly divided into three parts: ① minimum spanning tree construction; ② optimal segmentation threshold determination; ③ minimum spanning tree segmentation.
[0067] ① Minimum spanning tree construction
[0068] In the minimum spanning tree construction process, photons are clustered based on the point space distance, and added to the minimum spanning tree in sequence according to the distance between photons. The schematic diagram of the minimum spanning tree construction process is shown in the figure. Figure 4 The distance calculation formula between satellite-borne laser photons is as follows:
[0069]
[0070] in: is the distance between the satellite-borne laser photons of tags i and j; and are the along-track distance and elevation of the satellite-borne laser photon with label i, respectively; and are the along-track distance and elevation of the satellite-borne laser photon of label j, respectively.
[0071] ②Determination of the optimal segmentation threshold
[0072] The minimum spanning tree can include photons in the tree in order according to the density of photon distribution. By breaking the branches between the minimum spanning trees, the tree is divided into multiple small branches. The generated multiple small branches form a cluster, which is the cluster segmentation result of the satellite-borne laser photons. How to break the branches to generate the best cluster segmentation result requires obtaining the best branch segmentation threshold. In this embodiment, the lengths of the branches in the minimum spanning tree are first arranged from small to large, such as Figure 5 As shown. Figure 5 In the image, the length of the branch increases with the continuous addition of photons. When the length of the branch increases dramatically at a certain node, this node is the optimal segmentation threshold of the branch. Figure 5 The length of the branches of the initial minimum spanning tree is almost unchanged, and the corresponding area is the shallow water photon signal; the length of the subsequent branches begins to increase slowly, and the corresponding area is the deep water photon signal; the area corresponding to the drastic change node is the photon noise; based on the minimum spanning tree, the satellite-borne laser photon signals with different density distributions can be effectively retained. Figure 5 According to the fitting straight line equation of the first and last endpoints of the branch length trend chart, the distance from the point on the trend chart to the fitting straight line is calculated. The point with the longest distance is considered to be the minimum generated optimal segmentation threshold. The optimal segmentation threshold calculation formula is as follows:
[0073]
[0074] Where: (x i ,y i ) is the branch number and length in the trend graph; k and b are the slope and intercept of the fitting line respectively; m is the number of branches in the minimum spanning tree; L max is the distance corresponding to the optimal segmentation threshold.
[0075] ③ Minimum spanning tree partitioning
[0076] In the minimum spanning tree, the branches whose length is greater than the optimal split threshold are broken, and the minimum spanning tree is split into multiple subtrees, such as Figure 6 As shown. The number of photons contained in the subtree is counted, and the subtree with a photon number less than the set minimum photon number threshold is marked as noise, otherwise it is marked as a signal, thereby realizing the denoising of satellite-borne laser photons. As the water depth increases, the distribution of satellite-borne laser photons becomes sparse, and the photon signals in shallow water are denser than those in deep water. In order to accurately retain the photon signals at different water depths, the number of subtree photons is judged based on the multi-scale threshold denoising idea; this embodiment divides the water depth into four regions, namely The photon number thresholds corresponding to the four regions are 12, 9, 6, and 3 respectively; multi-scale photon number threshold denoising can eliminate the densely distributed noise in the shallow water area and retain the sparsely distributed photon signals in the deep water area. low and D high are the elevations of the shallowest and deepest seafloor satellite-borne laser photons, respectively; T is the elevation difference between the shallowest and deepest seafloor satellite-borne laser photons, that is, T = |D low -D high |.
[0077] (4) Refinement of satellite-borne laser photon denoising results: The denoised seabed satellite-borne laser photons are refined based on the box plot to eliminate the influence of abnormal photon points on the extraction of seabed satellite-borne laser photons.
[0078] After the minimum spanning tree-based hierarchical multi-scale seabed satellite laser photon denoising, some discrete and densely distributed laser photon noise may also be marked as signals. Based on the characteristics of the continuous spatial distribution of seabed satellite laser photons, photons that do not extend continuously in the terrain are marked as noise, thereby achieving the removal of discrete and densely distributed laser photon noise. The outlier detection algorithm based on the box plot has been widely used in fields such as spatial data mining, and can effectively remove dense noise points with abnormal distribution outside signal points. In this project, the initial denoising results are optimized based on the box plot to eliminate the discrete and densely distributed laser photon anomalies and obtain accurate seabed satellite laser photon signal extraction results. The optimization calculation based on the box plot denoising results is as follows:
[0079]
[0080] in: and are the category label and elevation of the spaceborne laser photon of label i; cls seafloor and cls noise They respectively indicate that the laser photons are seafloor satellite-borne laser photon signals or noise; Q1 and Q3 are the The number of laser photons and The laser photon elevation corresponding to the number of laser photons; IQR is The number of laser photons and The laser photon elevation difference corresponding to the number of laser photons.
[0081] Example 2
[0082] This embodiment discloses a system for implementing a minimum spanning tree based on-board laser photon hierarchical multi-scale denoising method described in Embodiment 1, comprising:
[0083] Sea surface and water body satellite laser photon separation module: used to evaluate the distribution of shallow sea satellite laser photons based on the kernel density evaluation algorithm, realize the separation of sea surface satellite laser photons and water body satellite laser photons, and eliminate the influence of sea surface satellite laser photons on the extraction of seabed satellite laser photons;
[0084] Satellite-borne laser photon distance standardization module: used to standardize the along-track distance and elevation of shallow-water satellite-borne laser photons based on the Z-score model;
[0085] Spaceborne laser photon hierarchical multi-scale denoising module: used to denoise laser photons in different water depths by constructing and splitting the minimum spanning tree;
[0086] Satellite laser photon denoising result refinement module: used to refine the denoised seabed satellite laser photons based on the box plot to eliminate the influence of abnormal photon points on the extraction of seabed satellite laser photons.
[0087] The above description is only a preferred specific implementation manner of this embodiment, but the protection scope of this embodiment is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of this embodiment within the technical scope disclosed in this embodiment, and they should be covered by the protection scope of this embodiment.
Claims
1. A hierarchical multi-scale denoising method for satellite-borne laser photons based on minimum spanning tree, characterized in that: The invention comprises separation of satellite-borne laser photons on the sea surface and in the water body, standardization of distance of satellite-borne laser photons, layered multi-scale denoising of satellite-borne laser photons and refinement of denoising results of satellite-borne laser photons. The layered multi-scale denoising of satellite-borne laser photons realizes denoising of satellite-borne laser photons on the seabed in different water depth areas by constructing and dividing a minimum spanning tree.
2. According to the minimum spanning tree based on-board laser photon hierarchical multi-scale denoising method of claim 1, it is characterized in that: The specific steps include: Step S1, separation of sea surface and water body satellite-borne laser photons: based on the kernel density evaluation algorithm, the distribution of shallow sea satellite-borne laser photons is evaluated to achieve the separation of sea surface satellite-borne laser photons and water body satellite-borne laser photons, and eliminate the influence of sea surface satellite-borne laser photons on the extraction of seabed satellite-borne laser photons; Step S2, spaceborne laser photon distance standardization: based on the Z-score model, the shallow sea spaceborne laser photon along-track distance and elevation are standardized; Step S3, hierarchical multi-scale denoising of satellite-borne laser photons: denoising of laser photons in different water depth areas is achieved by constructing and segmenting a minimum spanning tree; Step S4, refinement of satellite-borne laser photon denoising results: refine the denoised seabed satellite-borne laser photons based on the box plot to eliminate the influence of abnormal photon points on the extraction of seabed satellite-borne laser photons.
3. The method for hierarchical multi-scale denoising of satellite-borne laser photons based on minimum spanning tree according to claim 2 is characterized in that: In step S1, a Gaussian kernel density function is used to calculate the Gaussian kernel density value of satellite-borne laser photons, and the elevation corresponding to the maximum Gaussian kernel density value is obtained. Satellite-borne laser photons corresponding to the Gaussian kernel density value within the elevation range are selected to obtain sea surface satellite-borne laser photon signals. The values greater than the elevation range value are sea surface noise, and the values less than the elevation range value are seabed satellite-borne laser photon signals, thereby achieving separation of sea surface satellite-borne laser photons and water body satellite-borne laser photons.
4. The method for hierarchical multi-scale denoising of satellite-borne laser photons based on minimum spanning tree according to claim 3 is characterized in that: The elevation corresponding to the maximum Gaussian kernel density value is h, which is the center position of the sea surface at this time; the elevation range value is [h-β,h+β]. The calculation formula of Gaussian kernel density is shown in formula (1), and the judgment criteria of sea surface, sea surface noise and seabed satellite-borne laser photons are shown in formula (2): Where: n is the number of satellite-borne laser photons; W is the kernel density function adjustment constant; x is the satellite-borne laser photon elevation; β is the sea surface fluctuation constant; cls noise ,cls sea-surface and cls seafloor They are sea surface noise, sea surface satellite-borne laser photon signals and seabed satellite-borne laser photon category labels respectively.
5. The method for hierarchical multi-scale denoising of satellite-borne laser photons based on minimum spanning tree according to claim 2 is characterized in that: The calculation formula for normalizing the along-track distance and elevation of shallow sea star-borne laser photons based on the Z-score model in step S2 is as follows: in: and are the original distance and standardized distance of the satellite-borne laser photon labeled i; μ is the mean distance of the satellite-borne laser photons; δ is the standard deviation of the distance of the satellite-borne laser photons; and n is the number of satellite-borne laser photons.
6. The method for hierarchical multi-scale denoising of satellite-borne laser photons based on minimum spanning tree according to claim 2 is characterized in that: The step S3 specifically includes the following steps: Step S31, minimum spanning tree construction: clustering the photons based on the point space distance, adding them to the minimum spanning tree in turn according to the distance between the satellite-borne laser photons, and constructing the minimum spanning tree; Step S32, determining the optimal segmentation threshold: fitting a straight line equation based on the first and last endpoints of the branch length trend graph of the minimum spanning tree, calculating the distance from the point on the trend graph to the fitted straight line, and the point with the longest distance is the optimal segmentation threshold for the minimum spanning tree; Step S33, minimum spanning tree segmentation: in the minimum spanning tree, branches whose length is greater than the optimal segmentation threshold are broken, and the minimum spanning tree is segmented into multiple subtrees; the number of satellite-borne laser photons contained in the subtrees is counted, and the subtrees whose satellite-borne laser photons are less than the set minimum photon number threshold are marked as noise, otherwise they are marked as signals, thereby realizing denoising of satellite-borne laser photons.
7. The method for hierarchical multi-scale denoising of satellite-borne laser photons based on minimum spanning tree according to claim 6, characterized in that: The calculation formula of the distance between satellite-borne laser photons in step S31 is as follows: in: is the distance between the satellite-borne laser photons of tags i and j; and are the along-track distance and elevation of the satellite-borne laser photon with label i, respectively; and are the along-track distance and elevation of the satellite-borne laser photon of label j, respectively; The optimal segmentation threshold calculation formula in step S32 is as follows: Where: (x i ,y i ) is the branch number and length in the trend graph; k and b are the slope and intercept of the fitting line respectively; m is the number of branches in the minimum spanning tree; L max is the distance corresponding to the optimal segmentation threshold.
8. The method for hierarchical multi-scale denoising of satellite-borne laser photons based on minimum spanning tree according to claim 6, characterized in that: The minimum photon number threshold set in step S33 is used to judge the subtree photon number based on multi-scale threshold denoising, and the water depth is divided into four areas, namely The minimum photon number thresholds corresponding to the four regions are 12, 9, 6, and 3, respectively; Where: D low and D high are the elevations of the shallowest and deepest seafloor satellite-borne laser photons, respectively; T is the elevation difference between the shallowest and deepest seafloor satellite-borne laser photons, that is, T = |D low -D high |.
9. The method for hierarchical multi-scale denoising of satellite-borne laser photons based on minimum spanning tree according to claim 2, characterized in that: In step S4, the calculation of the refined processing of the denoised seafloor satellite laser photons based on the box plot is as follows: in: and are the category label and elevation of the spaceborne laser photon of label i; cls seafloor and cls noise They respectively indicate that the laser photons are seafloor satellite-borne laser photon signals or noise; Q1 and Q3 are the The number of laser photons and The laser photon elevation corresponding to the number of laser photons; IQR is The number of laser photons and The laser photon elevation difference corresponding to the number of laser photons.
10. A system for implementing the minimum spanning tree based on-board laser photon hierarchical multi-scale denoising method according to any one of claims 1 to 9, characterized in that: include: Sea surface and water body satellite laser photon separation module: used to evaluate the distribution of shallow sea satellite laser photons based on the kernel density evaluation algorithm, realize the separation of sea surface satellite laser photons and water body satellite laser photons, and eliminate the influence of sea surface satellite laser photons on the extraction of seabed satellite laser photons; Satellite-borne laser photon distance standardization module: used to standardize the along-track distance and elevation of shallow-water satellite-borne laser photons based on the Z-score model; Spaceborne laser photon hierarchical multi-scale denoising module: used to denoise laser photons in different water depths by constructing and splitting the minimum spanning tree; Satellite laser photon denoising result refinement module: used to refine the denoised seabed satellite laser photons based on the box plot to eliminate the influence of abnormal photon points on the extraction of seabed satellite laser photons.
