Single-photon laser data on-orbit fine denoising and grading method

Abstract

The invention relates to an on-orbit fine denoising and grading method for single-photon laser data. The method comprises the following steps: constructing a surface type database; segmenting the laser point cloud data at equal intervals in the along-rail scanning direction; fitting the photon density distribution characteristics of the laser point cloud data segmented in the previous step; calculating a photon grading threshold according to a fitting result, and judging the confidence grade of photons; and performing photon fine denoising and grading on each section of laser point cloud data based on the photon grading threshold. According to the satellite-borne single-photon laser point cloud data on-orbit fine denoising and confidence grading method, on-orbit fine denoising and confidence grading of satellite-borne single-photon laser point cloud data are achieved, photon fine denoising point cloud products are rapidly provided for users, and data input is provided for satellite-borne subsequent point cloud advanced product generation.

Description

Technical Field

[0001] The present invention belongs to the technical field of on-orbit processing of laser satellite data, and relates to a method for precise on-orbit denoising and grading of single-photon laser data. Background Art

[0002] The laser altimeter using single-photon detection has significant advantages such as high repetition frequency, multi-beam, and high precision. However, the characteristics of single-photon detection also lead to large noise and signal probability distribution in its data. From the original detection data to the product, the single-photon data must go through a large number of denoising and filtering processes. To realize the generation of spaceborne single-photon data products, it is necessary to carry out real-time denoising and filtering of spaceborne single-photon data. Summary of the Invention

[0003] The technical problem solved by the present invention is: overcoming the deficiencies of the prior art, and proposing a method for precise on-orbit denoising and grading of single-photon laser data, realizing on-orbit precise denoising and confidence grading of spaceborne single-photon laser point cloud data, quickly providing users with photon-precise denoised point cloud products, and providing data input for the generation of subsequent high-level spaceborne point cloud products.

[0004] The technical solution adopted by the present invention is:

[0005] A method for precise on-orbit denoising and grading of single-photon laser data, comprising:

[0006] Constructing a surface type database;

[0007] Segmenting the laser point cloud data at equal intervals in the along-track scanning direction;

[0008] Fitting the photon density distribution characteristics of the laser point cloud data after the previous segmentation;

[0009] Calculating the photon grading threshold according to the fitting result, and judging the confidence grading of the photons;

[0010] Performing photon precise denoising on each segment of the laser point cloud data based on the confidence grading of the photons.

[0011] Preferably, the method for fitting the photon density distribution characteristics of the laser point cloud data is as follows:

[0012] 2.1 Constructing a fine histogram and a coarse histogram for the laser point cloud data before each segment of coarse denoising, and calculating the background noise of each segment of the laser point cloud data accordingly;

[0013] 2.2 Using the rotatable ellipse method to calculate the maximum photon density around each laser point in each segmented laser point cloud data, and denoting the maximum photon density around the i-th laser point point i as max_counts i , where the value range of i is 1, 2, 3,..., n;

[0014] 2.3 Calculate max_counts i The photon density histogram distribution hist_counts i ;

[0015] 2.4 Perform Gaussian mixture model fitting on the photon density histogram distribution to obtain the fitting results of the multi-Gaussian distribution.

[0016] Preferably, the background noise B of each segment of the laser point cloud data noise Satisfies:

[0017]

[0018] Where:

[0019] C hist Is the total photon time number of the fine histogram of this segment of the laser point cloud data;

[0020] C maxbin Is the maximum value in the coarse histogram of this segment of the laser point cloud data;

[0021] N hwbins Is the number of layers in the fine histogram of this segment of the laser point cloud data;

[0022] N hw2sw Is the ratio between the vertical unit of the coarse histogram and the vertical unit of the fine histogram of this segment of the laser point cloud data;

[0023] B noise Is the average noise of a single layer in the coarse histogram of this segment of the laser point cloud data;

[0024] SS is the vertical unit size of the coarse histogram of this segment of the laser point cloud data;

[0025] δ is the segmentation interval of the segment, and the laser point cloud data is equally spaced divided along the track scanning direction according to this segmentation interval.

[0026] Preferably, the implementation method of the step 2.2 is as follows:

[0027] 2.2.1 Take the ellipse as the statistical region, and record the ratio of the major axis to the minor axis of the ellipse as radio;

[0028] 2.2.2 Set the major axis and minor axis parameters of the ellipse according to the surface type and the noise rate;

[0029] Consult the surface type database according to the position where this segment of the laser point cloud is located to obtain the surface type where this point cloud segment is located;

[0030] Select different photon number thresholds noise_points_num for different surface types;

[0031] According to the noise_points_num and the background noise rate B of this laser point cloud segment noise Calculate the major axis and minor axis of the ellipse;

[0032] The ellipse can rotate, and the major axis direction traverses the angular range of 0 - 180°, that is, the major axis starts from the horizontal scanning direction and rotates counterclockwise. Take a value at every interval of angle theta, and calculate the i-th laser point point in this laser point cloud segment i The number of photons in the ellipse at each of the above interval angles, where the value range of i is 1, 2, 3, …, n;

[0033] Find the maximum value of the number of photons of the i-th laser point at each interval angle, and denote it as the maximum photon density max_counts around the i-th laser point point i i

[0034] Preferably, the major axis E of the ellipse x and the minor axis E y Satisfy:

[0035] E x = radio·E y , and it is required that radio ≥ 1.

[0036] Preferably, in step 3.4, the fitting results Gaussian of the two Gaussian distributions total Satisfy

[0037] Gaussian total = Gaussian1 + Gaussian2;

[0038] Gaussian1 is the Gaussian distribution of the clustering of noise photons, and Gaussian2 is the Gaussian distribution of the clustering of signal photons;

[0039]

[0040] Among them, Gaussian1 and Gaussian2 are two Gaussian curves, where a1, b1, c1, a2, b2, c2 are the parameters of the Gaussian distribution curve.

[0041] Preferably, when there is an intersection point between the two Gaussian distribution curves, the method for calculating the photon classification threshold according to the fitting result is as follows:

[0042] Calculate the position index_equal of the intersection point between the two Gaussian distribution curves;

[0043] Calculate ​​

[0044] Calculate the photon classification threshold according to the following formula:

[0045]

[0046] signal_low = min(noise_high, index_equal)

[0047] where noise_high is the noise threshold, signal_high is the high signal threshold, signal_medium is the medium signal threshold, signal_low is the low signal threshold, and a1, b1, c1, a2, b2, c2 are the parameters of the Gaussian distribution curve.

[0048] Preferably, when there is no intersection between the two Gaussian distribution curves, the method for calculating the photon classification threshold according to the fitting result is as follows:

[0049] Calculate

[0050] Calculate the photon classification threshold according to the following formula:

[0051]

[0052] signal_low = noise_high

[0053] where noise_high is the noise threshold, signal_high is the high signal threshold, signal_medium is the medium signal threshold, signal_low is the low signal threshold, and a1, b1, c1, a2, b2, c2 are the parameters of the Gaussian distribution curve.

[0054] Preferably, the method for judging the confidence level classification of photons is as follows:

[0055] In each segment of laser point cloud data, photons with a maximum photon density greater than signal_high are defined as high-confidence photons, photons with a maximum photon density less than or equal to signal_high and greater than signal_medium are defined as medium-confidence photons, photons with a maximum photon density less than or equal to signal_medium and greater than signal_low are defined as low-confidence photons, and photons with a maximum photon density less than or equal to signal_low are defined as noise photons.

[0056] A terminal device, comprising:

[0057] A memory for storing instructions executed by at least one processor;

[0058] A processor for executing instructions stored in a memory to implement the above method.

[0059] The beneficial effects of the present invention compared with the prior art are as follows:

[0060] (1) In the present invention, when calculating the photon density, a rotatable ellipse is used for the calculation area, and the density value of each photon takes the maximum value at various rotation angles. This density calculation method can avoid the influence of slope on photon density calculation and ensure the accuracy of photon density calculation under various slopes of the terrain.

[0061] (2) The present invention dynamically sets the axis length parameters of the photon density statistical ellipse area in segments according to the prior knowledge of the surface type and the photon noise situation, ensuring the universality of the method under various surface types and environmental conditions.

[0062] (3) The intensity and signal-to-noise ratio of the laser point cloud signal in the present invention are greatly related to the surface reflectivity, surface slope, and atmospheric environment, so they are constantly changing in orbit. The present invention statistically analyzes the photon distribution characteristics in segments, dynamically calculates the classification threshold according to the distribution characteristics, and classifies the photons based on the threshold to achieve precise denoising and photon classification under various environmental conditions. Description of the Drawings

[0063] Figure 1 is a surface type database, where 09 represents the ocean, 10 represents land (non-forest land), 11 represents land (forest land), and 13 represents ice and snow.

[0064] Figure 2 is a schematic diagram of the result after real-time rough denoising of single-photon laser data in orbit (in the figure, the blue is the original photon point cloud data, and the red is the photon point cloud data after rough denoising. The input point cloud data of the present invention is the blue and red parts in the figure).

[0065] Figure 3 is a schematic diagram of the rotatable ellipse for photon density calculation.

[0066] Figure 4 is a schematic diagram of the photon density histogram distribution and Gaussian mixture model fitting.

[0067] Figure 5 is a schematic diagram of signal photon classification, where (a) is the case where there is an intersection point between two Gaussian distribution curves; (b) is the case where there is no intersection point between two Gaussian distribution curves.

[0068] Figure 6 is the area of the real-time compression embodiment of single-photon laser data.

[0069] Figure 7 is obtained by using the method of the present invention Figure 2The overall results of precise denoising and grading of data. In the figure, red represents high-confidence photons; blue represents medium-confidence photons; cyan represents low-confidence photons; and green represents noise photons.

[0070] Figure 8 For Figure 7 A section of the data, the photon data from the 73.3 to 73.4 section;

[0071] Figure 9 For Figure 8 The photon density histogram distribution and Gaussian mixture model fitting results of the photon point cloud data in

[0072] Figure 10 For Figure 8 The precise denoising and photon grading results of the photon point cloud data in , where red represents high-confidence photons; blue represents medium-confidence photons; cyan represents low-confidence photons; and green represents noise photons. Specific implementation manners

[0073] The present invention will be further described below with reference to the accompanying drawings.

[0074] The present invention provides a method for in-orbit precise denoising and grading of spaceborne laser point clouds. First, a prior database of surface types is constructed, then the point cloud data is segmented and fitted, the photon grading threshold is calculated based on the fitting results, and then the precise denoising and grading results of the photons are obtained based on the grading threshold.

[0075] The specific technical solutions include:

[0076] I. Construction of the surface type database

[0077] Construct a surface type database, including four surface types: ocean, land (non-forest land), land (forest land), and ice and snow. The grid scale of the database is 0.25°×0.25°. As Figure 1 shown.

[0078] II. Segmentation of the laser point cloud data

[0079] The input laser point cloud data of the present invention is the photon point cloud data before and after spaceborne real-time rough denoising. As Figure 2 shown, it is a schematic diagram of the result of in-orbit real-time rough denoising of single-photon laser data. In the figure, blue represents the original photon point cloud data, and red represents the photon point cloud data after rough denoising. The input point cloud data of the present invention is the blue and red parts in the figure. Based on this point cloud data, precise denoising and photon grading are performed.

[0080] Segment the laser point cloud data in the along-track scanning direction. Due to the changes in surface characteristics and lighting conditions, the noise rate and signal-to-noise ratio characteristics of the single-photon point cloud data are in a changing process in the along-track scanning direction. Segment the laser point cloud data at equal intervals in the along-track scanning direction, and the segmentation interval is selected according to the characteristics of the point cloud. For example, δ = 140m. All subsequent operation steps of the present invention are carried out with the segmentation interval as the basic unit.

[0081] III. Piecewise fitting of the photon density distribution characteristics of the laser point cloud

[0082] Statistically analyze the photon density distribution characteristics of the laser point cloud data after the previous segmentation. Under certain lighting conditions, the photon density distribution of the single-photon laser point cloud has relatively fixed distribution characteristics for specific ground objects and can be decomposed into the superposition of several fixed distributions.

[0083] 1. Calculation of the background noise of the segmented laser point cloud

[0084] 1) Generation of a fine histogram

[0085] Taking the segmentation interval δ = 140m as the unit and the vertical height direction with HS meters (such as 3m) as the unit, construct a fine histogram of the laser point cloud data (the original photon point cloud data, Figure 2 blue part);

[0086] 2) Generation of a coarse histogram

[0087] Based on the above fine histogram, construct a coarse histogram. The vertical unit size SS of the coarse histogram is, for example, 48m;

[0088] 3) Calculation of the background noise

[0089] Calculate the background noise rate B noise for each segment of the photon point cloud data:

[0090]

[0091] Where:

[0092] C hist is the total photon time number of the fine histogram of the photon point cloud data of this segment; C maxbin is the maximum value in the coarse histogram of the photon point cloud data of this segment;

[0093] N hwbins is the number of layers in the fine histogram of the photon point cloud data of this segment;

[0094] N hw2swis the ratio between the vertical unit of the coarse histogram of this segment of photon point cloud data and the vertical unit of the fine histogram;

[0095] B noise is the average noise of a single layer in the coarse histogram of this segment of photon point cloud data;

[0096] SS is the vertical unit size of the coarse histogram of this segment of photon point cloud data;

[0097] δ is the segmentation interval of the segment, and the laser point cloud data is equally spaced divided along the track scanning direction according to this segmentation interval.

[0098] 2. Calculation of the photon density of the segmented laser point cloud

[0099] Calculate the maximum photon density around each laser point point1, point2,..., point in the laser point cloud data of each segment n ;

[0100] 1) The statistical area is set as an ellipse, and the ratio of the major axis to the minor axis of the ellipse is radio, for example

[0101] radio = 5, and it is required that radio ≥ 1;

[0102] 2) Adaptively set the ellipse axis length parameters according to the surface type and noise rate;

[0103] Consult the surface type database according to the position of this laser point cloud segment to obtain the surface type of this point cloud segment, including ocean, land (non-forest land), land (forest land) and ice and snow.

[0104] Select different photon number thresholds noise_points_num for different surface types.

[0105] For example: for land (non-forest land and forest land), select noise_points_num = 6; for ocean, select noise_points_num = 2; for ice and snow, select noise_points_num = 5.

[0106] According to noise_points_num and the background noise rate B of this segment noise calculate the major axis E x and the minor axis E y respectively as:

[0107]

[0108] The ellipse is rotatable, and the direction of its major axis traverses the angular range of 0 - 180°. A value is taken every theta angle, for example, 5°. That is, starting from the horizontal scanning direction, the major axis rotates counterclockwise, and the included angle between it and the horizontal scanning direction is theta = [0°, 5°, 10°, …, 175°], as Figure 3 shown.

[0109] Calculate the number of photons counts i at each laser point point i in the laser point cloud data at all rotation intervals (theta), where the value range of i is 1, 2, 3, …, n;

[0110] 3) Find the maximum value max_counts i of each photon in the counts i (theta) obtained in the previous step, max_counts i = max(counts i (theta)), and store the corresponding ellipse angle value theta_max

[0111] 3. Calculation of the photon density histogram for segmented laser point clouds

[0112] Calculate the histogram distribution hist_counts i of the max_counts i obtained in the previous step, as Figure 4 shown.

[0113] 4. Gaussian mixture model fitting for the photon density histogram

[0114] Due to the distribution characteristics of the photon point cloud, its histogram shows certain clustering distribution characteristics. The noise photon density is low, the signal photon density is high, and it shows a normal distribution. The photon density histogram can be approximately fitted by multiple single Gaussian distributions, as Figure 4 shown. The expectation maximization algorithm is used for GMM fitting.

[0115] The GMM fitting result is the fitting of multiple Gaussian distributions, for example, K. The number of Gaussian distributions can be specifically selected according to the surface type and the signal-to-noise ratio of the point cloud data. Here, it is assumed that K = 2, as Figure 4 shown. Gaussian1 is the Gaussian distribution with a smaller number of photon densities, that is, the clustering of noise photons, and Gaussian2 is the Gaussian distribution with a larger number of photon densities, that is, the clustering of signal photons.

[0116]

[0117] Gaussian1 and Gaussian2 are two Gaussian curves, where a1, b1, c1, a2, b2, c2 are the parameters of the Gaussian distribution curves.

[0118] The fitting result of the multi-Gaussian distribution can be expressed as:

[0119] Gaussian total = Gaussian1 + Gaussian2

[0120] IV. Photon Classification Threshold Calculation

[0121] According to the fitting result of the Gaussian mixture model of the laser photon density histogram in the previous step, calculate the classification threshold of the signal photons. The signal photons are divided into four levels: high-confidence photons, medium-confidence photons, low-confidence photons, and noise photons. Due to the distribution of the actual photon data, it is possible that the intervals between the Gaussian distribution curves are relatively close and there are intersection points. The threshold calculation method is given in two cases: the existence and non-existence of intersection points between two Gaussian distribution curves.

[0122] 1. Threshold calculation when there are intersection points between two Gaussian distribution curves

[0123] If there are intersection points between two Gaussian distribution curves, as shown in (a) of Figure 5 shown.

[0124] 1) Calculate the position index_equal of the intersection point between the two Gaussian distribution curves;

[0125] 2) Define the noise threshold noise_high as a threshold of the photon density,

[0126] where b1, c1 are the parameters of the above Gaussian distribution;

[0127] 3) Perform other threshold calculations:

[0128] High signal threshold:

[0129] Medium signal threshold:

[0130] Low signal threshold: signal_low = min(noise_high, index_equal)

[0131] where b2, c2 are the parameters of the above Gaussian distribution; the above thresholds split the single-photon data into noise, high-confidence, medium-confidence, and low-confidence photons according to the thresholds.

[0132] 2. Threshold calculation when there are no intersection points between two Gaussian distribution curves

[0133] If there is no intersection between two Gaussian distribution curves, as shown in Figure 5 (b) below.

[0134] 1) Define the noise threshold noise_high as a threshold of photon density and calculate

[0135]

[0136] 2) Conduct other threshold calculations:

[0137] High signal threshold:

[0138] Medium signal threshold:

[0139] Low signal threshold: signal_low = noise_high

[0140] The above thresholds split the single-photon data into noise, high-confidence, medium-confidence, and low-confidence photons according to the thresholds.

[0141] 3. Determine the confidence level classification of photons based on the threshold range calculated above

[0142] According to the max_counts of the photons i Determine the confidence level classification of photons based on the above threshold range. Figure 5 The example diagram in shows the threshold ranges where photons with different confidence levels are located.

[0143] max_counts i Photons in the range of max_counts > signal_high are defined as high-confidence photons and are represented by

[0144] max_counts i <= signal_high and max_counts > signal_medium are defined as medium-confidence photons and are represented by in the figure;

[0145] max_counts i <= signal_medium and max_counts > signal_low are defined as low-confidence photons and are represented by in the figure;

[0146] max_counts iPhotons with <=noise_high are defined as noise photons and are represented in the figure by

[0147] ;

[0148] By removing the noise photons, the precise denoising of the photon data is completed.

[0149] The present invention also provides a terminal device, including: a memory for storing instructions executed by at least one processor; a processor for executing the instructions stored in the memory to implement the above method.

[0150] Embodiment

[0151] Select the single-photon laser simulation data from the rising area of the Sichuan Basin to the Qinghai-Tibet Plateau for precise denoising and confidence level grading. This area transitions from the Sichuan Basin to the Qinghai-Tibet Plateau (longitude 103.5°, latitude 31°), with the elevation rising from hundreds of meters to the order of four or five thousand meters, and the height difference changing by more than four thousand meters, showing significant changes, such as Figure 6 shown.

[0152] Figure 2 The input data is given, which is Figure 6 a section of the photon point cloud data of the trajectory, where the blue is the original photon point cloud data and the red is the photon point cloud data after rough denoising. The present invention performs precise denoising and confidence level grading on the basis of the Figure 2 data.

[0153] Figure 7 The overall results of precise denoising and grading of the Figure 2 data obtained by using the method of the present invention are given. Figure 8 The Figure 7 data is given, which is the photon data from section 73.3 to 73.4. Figure 9 The Figure 8 photon density histogram distribution and Gaussian mixture model fitting results of the photon point cloud data in are given. Figure 10 The Figure 8 precise denoising and photon grading results of the photon point cloud data in are given. This embodiment performs precise denoising and photon grading on a section of photon data with a large height difference change. It can be seen from the figure that the precise denoising and confidence level grading of photons can be accurately achieved.

[0154] The present invention realizes precise denoising and grading on the basis of rough denoising of single-photon point cloud data, can quickly provide users with single-photon data after precise denoising, and also provides data input for the generation of single-photon high-level products.

[0155] The parts not detailed in the present invention belong to the common general knowledge of those skilled in the art.

Claims

1. A method for on-orbit precise denoising and grading of single-photon laser data, characterized in that, Including: Construct a surface type database; Segment the lidar point cloud data at equal intervals in the along-track scanning direction; Fit the photon density distribution characteristics of the lidar point cloud data after the previous segmentation; Calculate the photon classification threshold according to the fitting result and judge the confidence classification of photons; Perform photon fine denoising on each segment of lidar point cloud data based on the confidence classification of photons.

2. A method for on-orbit precise denoising and grading of single-photon laser data according to claim 1, characterized in that, The method for fitting the photon density distribution characteristics of the lidar point cloud data is as follows: 2.1 Construct a fine histogram and a coarse histogram for each segment of lidar point cloud data before rough denoising, and calculate the background noise of each segment of lidar point cloud data accordingly; 2.2 Calculate the maximum photon density around each laser point in each segmented laser point cloud data using the rotatable ellipse method. The maximum photon density around the \(i\)th laser point point i is denoted as max_counts i , where the value range of \(i\) is 1, 2, 3, …, \(n\); 2.3 Calculate max_counts i Photon density histogram distribution hist_counts of i ; 2.4 Fit the photon density histogram distribution with a Gaussian mixture model to obtain the fitting result of a multi-Gaussian distribution.

3. A method for on-orbit precise denoising and grading of single-photon laser data according to claim 2, characterized in that, The background noise B of each segment of laser point cloud data noise Satisfies: Where: C hist is the total number of photon times for the fine histogram of this segment of laser point cloud data; C maxbin is the maximum value in the rough histogram of this segment of laser point cloud data; N hwbins is the number of layers in the fine histogram of this segment of laser point cloud data; N hw2sw is the ratio between the vertical unit of the coarse histogram of the laser point cloud data segment and the vertical unit of the fine histogram; B noise is the average noise of a single layer in the rough histogram of the laser point cloud data for this segment; SS is the vertical unit size of the coarse histogram of this segment of lidar point cloud data; δ is the segmentation interval, and the lidar point cloud data is equally spaced in the along-track scanning direction according to this segmentation interval.

4. A method for on-orbit precise denoising and grading of single-photon laser data according to claim 2, characterized in that The implementation method of step 2.2 is as follows: 2.2.1 Use an ellipse as the statistical region, and record the ratio of the major axis to the minor axis of the ellipse as radio; 2.2.2 Set the major axis and minor axis parameters of the ellipse according to the surface type and noise rate; Consult the surface type database according to the position of this lidar point cloud segment to obtain the surface type of this point cloud segment; Select different photon number thresholds noise_points_num for different surface types; According to noise_points_num and the background noise rate B of this laser point cloud segment noise Calculate the major axis and minor axis of the ellipse; The ellipse can rotate, and the direction of its major axis traverses the angular range of 0 - 180°. That is, the major axis starts from the horizontal scanning direction and rotates counterclockwise. A value is taken at every interval of angle theta, and the i-th laser point point in the laser point cloud segment is calculated. i The number of photons in the ellipse at each of the above interval angles, where the value range of i is 1, 2, 3, …, n; Find the maximum number of photons of the \(i\)-th laser point at each interval angle, denoted as the \(i\)-th laser point point i The maximum photon density max_counts around it i .

5. A method for on-orbit precise denoising and grading of single-photon laser data according to claim 4, characterized in that Major axis E of the ellipse x and minor axis E y Satisfy: E x = radio·E y , where radio ≥ 1 is required.

6. A method for on-orbit precise denoising and grading of single-photon laser data according to claim 2, characterized in that, In step 2.4, the fitting results Gaussian of the two Gaussian distributions total satisfy Gaussian total = Gaussian1 + Gaussian2; Gaussian1 is the Gaussian distribution of the clustering of noise photons, and Gaussian2 is the Gaussian distribution of the clustering of signal photons; Among them, Gaussian1 and Gaussian2 are two Gaussian curves, where a1, b1, c1, a2, b2, c2 are the parameters of the Gaussian distribution curve.

7. A method for on-orbit precise denoising and grading of single-photon laser data according to claim 2, characterized in that, When there is an intersection between the two Gaussian distribution curves, the method for calculating the photon classification threshold according to the fitting result is as follows: Calculate the position index_equal of the intersection between the two Gaussian distribution curves; Calculation Calculate the photon classification threshold according to the following formula: signal_low = min(noise_high, index_equal) Where noise_high is the noise threshold, signal_high is the high signal threshold, signal_medium is the medium signal threshold, signal_low is the low signal threshold, and a1, b1, c1, a2, b2, c2 are the parameters of the Gaussian distribution curve.

8. A method for on-orbit precise denoising and grading of single-photon laser data according to claim 2, characterized in that When there is no intersection between the two Gaussian distribution curves, the method for calculating the photon classification threshold according to the fitting result is as follows: Calculation Calculate the photon classification threshold according to the following formula: signal_low = noise_high Where noise_high is the noise threshold, signal_high is the high signal threshold, signal_medium is the medium signal threshold, signal_low is the low signal threshold, and a1, b1, c1, a2, b2, c2 are the parameters of the Gaussian distribution curve.

9. A method for on-orbit precise denoising and grading of single-photon laser data according to claim 1, characterized in that Judge the confidence classification of photons, and the method is as follows: In each segment of the laser point cloud data, photons with a maximum photon density greater than signal_high are defined as high-confidence photons, photons with a maximum photon density less than or equal to signal_high and greater than signal_medium are defined as medium-confidence photons, photons with a maximum photon density less than or equal to signal_medium and greater than signal_low are defined as low-confidence photons, and photons with a maximum photon density less than or equal to signal_low are defined as noise photons.

10. A terminal device, characterized in that, Including: A memory for storing instructions executed by at least one processor; A processor for executing the instructions stored in the memory to implement the method according to any one of claims 1-9.

Images