Mud flat laser point cloud data classification method based on laser radar reflection intensity information

Abstract

The invention relates to the technical field of laser radar point cloud data classification, in particular to a mudflat laser point cloud data classification method based on laser radar reflection intensity information, and the method comprises the steps: 1, determining a mudflat collection range and collection time; 2, collecting laser point cloud data of a flight area to obtain initial laser radar point cloud data; 3, constructing a triangular initial sliding window for the initial laser radar point cloud data under the single unmanned aerial vehicle air strip; 4, dynamically shrinking the side length of the sliding window to adapt to the actual distribution characteristics of the sea surface and the beach area; step 5, optimizing classification processing of the point cloud data; step 6, obtaining accurate mudflat point cloud data; step 7, iteration of the sliding window is carried out; and 8, processing the air strips one by one according to the steps 3 to 7 to generate laser point cloud data of the mud flat area. And through intensity information difference analysis in the window, laser point cloud data noise brought by residual seawater is extracted, and pure beach laser radar data is obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of laser radar point cloud data classification, and in particular to a method for classifying tidal flat laser point cloud data based on laser radar reflection intensity information. Background Art

[0002] In the beach area, because the terrain of the beach area is changeable and the foundation of the muddy area is relatively soft, for large areas of beaches, traditional manual measurement methods rely on fixed-point measurement. For large areas of beaches, obtaining comprehensive and detailed data not only requires a lot of manpower, time and resources, but also makes it difficult to cover all the detailed features, and there are safety hazards for surveyors on the beach.

[0003] Drones equipped with laser radar can obtain high-precision beach elevation data when the seawater recedes. Compared with traditional measurement methods, drones equipped with laser radar to measure beaches have the characteristics of accurately obtaining complex terrain, quickly covering large areas, reducing the time and risk of manual measurement, and having low cost and high safety. However, the beach terrain is bumpy and uneven, and some seawater will remain after the tide recedes. The reflection intensity information contains the reflection intensity information of the residual seawater. The water quality of seawater in each period or each area is different, so the reflection intensity information of the seawater surface is also inconsistent. The existing technology lacks a simple method to distinguish the laser reflection data of the beach and the residual seawater surface, which affects the accuracy of the elevation data. Summary of the invention

[0004] In view of the problems existing in the prior art, in order to explore a laser point cloud data that can simply and quickly distinguish beach terrain and residual seawater, the purpose of the present invention is to propose a beach and seawater point cloud data classification method based on laser radar reflection intensity, which uses the difference in laser radar reflection intensity to classify and distinguish the laser point cloud data of the beach and seawater area, removes the residual seawater reflection points, and obtains more accurate tidal flat elevation data.

[0005] To achieve the above object, the technical solution adopted by the present invention is: a method for classifying tidal flat laser point cloud data based on laser radar reflection intensity information, comprising the following steps: Step 1: Based on the depth contours and coastline, obtain the theoretical maximum range of the beach exposed after the sea water recedes and the high tide. According to the predicted tide level at the lowest tide, obtain the operation time range and determine the tidal flat collection range and collection time. Step 2: Set the collection route. The UAV equipped with the laser radar flies over the mudflat collection area at low tide to collect laser point cloud data in the flight area. The initial laser radar point cloud data is obtained by solving the data solving software supporting the UAV and the laser radar equipment. Step 3: Construct a triangular initial sliding window for the initial lidar point cloud data under a single UAV flight path and set the initial side length; Step 4: According to the laser point cloud data and laser reflection intensity information, dynamically shrink the side length of the sliding window to adapt to the actual distribution characteristics of the sea surface and beach area; Step 5: When point cloud data are detected on both sides of the 90° angle of the sliding window, expand along the oblique direction to form a quadrilateral window with the same size as the adjusted sliding window to optimize the classification processing of the point cloud data; Step 6: In the quadrilateral window, the difference in reflection intensity between the tidal flat and the seawater is analyzed based on the laser radar reflection intensity information, the boundary between the tidal flat and the sea surface is calculated and extracted, the point cloud data is classified and noise is filtered out, and accurate tidal flat point cloud data is obtained; Step 7: After completing the classification of the point cloud data of the current quadrilateral window, the quadrilateral window moves to the next area along the flight strip according to the preset step length, repeating the boundary line detection and data classification process until the point cloud data of the entire flight strip is covered, and the sliding window is iterated; Step 8: Perform steps 3 to 7 for each flight strip. After classifying each flight strip and filtering out impurity points, integrate the point cloud data of all flight strips to generate a single-strip laser point cloud dataset of the tidal flat area without impurities. Integrate the single-strip laser point cloud dataset of the tidal flat area to generate the laser point cloud data of the tidal flat area.

[0006] In the above-mentioned tidal flat laser point cloud data classification method based on laser radar reflection intensity information, in step 1, the collection time is 80 minutes before to 10 minutes after the lowest tide, and the total time is 90 minutes.

[0007] In the above-mentioned tidal flat laser point cloud data classification method based on laser radar reflection intensity information, in the step 2, the drone collection route is arranged parallel to the isobath, the length of the collection route does not exceed the flight distance that the drone can fly in 90 minutes, and the collection route is set from close to the coastline to far away from the coastline, and then the collection route close to the isobath is collected at the lowest tide.

[0008] In the above-mentioned tidal flat laser point cloud data classification method based on laser radar reflection intensity information, in step 3, the initial sliding window is an isosceles right triangle, one of the two right-angled sides is perpendicular to the direction of the flight strip, and the other is parallel to the direction of the flight strip. The initial side length is set to twice the flight altitude according to the width of the best point quality point cloud data obtained by the laser radar under the ideal field of view angle, L0=2h, where h is the flight altitude.

[0009] In the above-mentioned method for classifying tidal flat laser point cloud data based on laser radar reflection intensity information, step 4 comprises: Step 4-1: Place the initial sliding window at the end of the flight strip, with one 45° angle facing the land side of the flight strip, the other 45° angle perpendicular to the flight strip direction, and a 90° right angle facing the sea side. The initial right angle side L0=2h; Step 4-2: According to the width of the sea surface reflection point cloud, gradually shrink the side length of the sliding window, and use the formula for the side length after shrinkage: Calculate the length of the side after contraction, where is the shrinkage factor, is the width of the sea water point cloud; Step 4-3: When both right-angled sides of the sliding window detect the seawater point cloud data, stop shrinking and record the actual lengths of the two right-angled sides at this time. and , take the larger value and Compare and finalize .

[0010] In the above-mentioned tidal flat laser point cloud data classification method based on laser radar reflection intensity information, in step 4-2, it is defined that the reflection intensity in the vertical direction along the navigation strip is lower than the threshold value. The point cloud range, , The point cloud data of a single route currently being processed, It is the dividing threshold of the reflection intensity between sea water and mudflats.

[0011] In the above-mentioned method for classifying tidal flat laser point cloud data based on laser radar reflection intensity information, step 5 comprises: Step 5-1: Take the triangle window when the contraction stops as the reference, copy an identical triangle along the hypotenuse, make the hypotenuses of the two triangles overlap, and form a parallelogram window; Step 5-2: Calculate the total area of ​​the parallelogram window: , by using parallelogram windows to expand window coverage.

[0012] In the above-mentioned method for classifying tidal flat laser point cloud data based on laser radar reflection intensity information, step 6 comprises: Step 6-1: Reflection intensity difference analysis, LiDAR point cloud data ,in is the 3D coordinate of the point cloud, is the reflection intensity, N is the total number of point clouds, and the reflection intensity threshold is set based on the fact that the reflection intensity of the dry tidal flat area is higher and the reflection intensity of the tidal flat area is lower. As a basis for preliminary classification; Step 6-2: Boundary extraction, within the sliding window, according to the reflection intensity The point cloud data is divided into two categories: mudflats and seawater:

[0013] , dividing line Defined as and The spatial boundary of can be calculated by the gradient change of the reflection intensity in the window: ,in It represents the spatial gradient of reflection intensity, reflecting the sudden change in intensity at the interface between the mudflat and the sea water; Step 6-3: Filtering the point cloud on the seawater side: The point cloud data in the For subsequent processing; Step 6-4: Classification and filtering of residual noise in the tidal flat area. In the process, we further classify and remove the residual seawater points, and set the secondary filtering conditions based on the elevation and reflection intensity: ,in, For the window The average elevation of the tidal flats is used to distinguish low-lying tidal pools from dry mudflats; Secondary reflection intensity threshold; Step 6-5: Elimination Then we get the accurate local point cloud .

[0014] In the above-mentioned method for classifying tidal flat laser point cloud data based on laser radar reflection intensity information, step 7 comprises: Step 7-1: Step size definition, the parallelogram sliding window moves along the extension direction of the flight strip, and the moving step size for: ,in, is the step size scaling factor; Step 7-2: Movement process: The window moves along the direction of the flight path Distance, the new window center position is updated as: ,in, The center coordinates of the current window; The unit vector of the flight path direction; The center coordinates of the next window; Step 7-3: Repeat step 6 at the new window position to continuously generate new local fine point clouds; Step 7-4: Move the parallelogram window to the end of the flight strip to obtain point cloud data covering the entire flight strip .

[0015] The above-mentioned tidal flat laser point cloud data classification method based on laser radar reflection intensity information includes, in step 8: in the process of integrating the single-line laser point cloud data set of the tidal flat area, it is necessary to keep the point cloud plane coordinate system consistent, the elevation benchmark consistent, the overlapping area data between the flight strips, and choose to smooth or interpolate the overlapping area to maintain the continuity of the data set. ,in, The point cloud of the overlapping area, the point cloud data after optimization and smoothing is the final point cloud data ,in .

[0016] The beneficial effect of the present invention's method for classifying tidal flat laser point cloud data based on laser radar reflection intensity information is that the present invention method accurately extracts the laser point cloud data noise caused by residual seawater through intensity information difference analysis within the window, ensuring that pure beach laser radar data is obtained. The method can adaptively adjust the classification parameters according to the changes in seawater reflection intensity of beaches in different regions and different water quality conditions, thereby effectively adapting to complex tidal flat environments. Compared with traditional methods, the present invention significantly improves the accuracy of coastal tidal flat DEM data, and provides accurate elevation basic data for subsequent terrain analysis and coastal engineering.

[0017] The geometric characteristics of the triangular window facilitate sliding along the flight strip, and the directionality of the right-angle edge adapts to the spatial distribution of the flight strip data; the initial size ensures sufficient data coverage and provides an initial reference for subsequent window contraction based on reflection intensity. The window size is adaptively adjusted through reflection intensity to ensure that the window is neither too large nor too small, avoiding the inclusion of too much irrelevant data while avoiding missing key features; The formula balances the constraints of the initial settings and actual data and improves the classification accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A tidal timetable for querying a tidal station provided in an embodiment of the present invention; Figure 2 A schematic diagram of the route layout direction for obtaining drone data provided in an embodiment of the present invention; Figure 3 A schematic diagram of the initial sliding window size for classifying beach and sea surface laser radar data intensity information provided in an embodiment of the present invention; Figure 4 A schematic diagram of a sliding window contraction factor for beach and sea surface laser radar data intensity information classification provided in an embodiment of the present invention; Figure 5 A schematic diagram of a sliding window determined when classifying beach and sea surface laser radar data intensity information provided in an embodiment of the present invention; Figure 6A schematic diagram of laser radar intensity information of a beach and sea surface provided in an embodiment of the present invention; Figure 7 This is a schematic diagram of intensity information of beach lidar data after classification provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to enable those skilled in the art to better understand the technical solution of the present invention, the technical solution of the present invention is described below in conjunction with specific implementation methods and drawings.

[0020] Example 1 like Figure 1-Figure 7 As shown, a method for classifying tidal flat laser point cloud data based on laser radar reflection intensity information includes the following steps.

[0021] Step 1: Based on the depth contours and coastline, obtain the theoretical maximum exposed beach area after the sea water recedes and the high tide. According to the predicted tidal height at the lowest tide, obtain the operation time range and determine the tidal flat collection range and collection time. The collection time is 80 minutes before to 10 minutes after the lowest tide, and the total time is 90 minutes.

[0022] Step 2: Set the collection route. The drone equipped with a lidar flies over the mudflat collection area at low tide to collect laser point cloud data in the flight area. The initial lidar point cloud data is obtained by solving the data solving software supporting the drone and lidar equipment.

[0023] The drone collection route is arranged parallel to the isobath. The length of the collection route does not exceed the flight distance that the drone can fly in 90 minutes. The collection route is set from close to the coastline to far away from the coastline, and then the collection route close to the isobath is collected at the lowest tide.

[0024] Step 3: Construct a triangular initial sliding window for the initial lidar point cloud data under a single drone flight path, and set the initial side length. The initial sliding window is an isosceles right triangle, with one right-angle side perpendicular to the flight path and the other parallel to the flight path. The initial side length is set to twice the flight height according to the width of the lidar point cloud data with the best point quality under the ideal field of view, L0=2h, where h is the flight height.

[0025] Step 4: According to the laser point cloud data and laser reflection intensity information, the side length of the sliding window is shrunk. The initial sliding window is placed at the end of the flight path. According to the width of the sea surface reflection point cloud, the side length of the sliding window is gradually shrunk until the sea water point cloud data is detected on both sides of the 90° angle of the sliding window. The formula for the side length after shrinkage is L=max(aL0, Tsea water), where a is the shrinkage factor, Tsea water is the sea water point cloud reflection intensity, and L0 is the UAV altitude.

[0026] Step 5: When point cloud data are detected on both sides of the 90° angle of the sliding window, the sliding window is copied along the oblique direction to form a quadrilateral window with the same size as the adjusted sliding window.

[0027] Step 6: In the quadrilateral window, analyze the difference in reflection intensity between the tidal flat and the seawater based on the LiDAR reflection intensity information, calculate and extract the boundary between the tidal flat and the sea surface, filter out the point cloud data on the seawater side, and further classify the residual seawater and tidal flat point cloud data in the tidal flat area, eliminate the seawater noise points in the tidal flat area, and obtain accurate tidal flat point cloud data.

[0028] Step 7: After completing the classification of the point cloud data of the current quadrilateral window, the quadrilateral window moves to the next area along the flight strip according to the preset step size, repeating the boundary line detection and data classification process until the point cloud data of the entire flight strip is covered, and the sliding window is iterated.

[0029] Step 8: Perform steps 3 to 7 for each flight strip. After classifying each flight strip and filtering out impurity points, integrate the point cloud data of all flight strips to generate a laser point cloud dataset of the tidal flat area without impurities. Integrate the laser point cloud dataset of the tidal flat area to generate the laser point cloud data of the tidal flat area.

[0030] Example 2 This embodiment proposes a beach and sea surface laser point cloud data classification method based on laser radar reflection intensity information, which can automatically adjust the size of the sliding window to improve the accuracy of the elevation data of the tidal flat area. In the specific method, the side length of the initial sliding window is 2 times the flight height. By shrinking the sliding window and adjusting it according to the laser reflection intensity information, the optimal size of the sliding window is finally obtained to adapt to the characteristics of the actual data.

[0031] like Figure 1-Figure 7 As shown, specifically including: (1) Determination of beach collection scope.

[0032] According to the topographic map of the area where the beach is located, the positions of the isobaths and coastlines in the topographic map are obtained, and the maximum range of the beach that is theoretically exposed after the sea water recedes at high tide is determined based on the isobaths and coastlines.

[0033] (2) Determination of beach collection time.

[0034] According to the location of the beach, check the tide schedule of the nearby tide gauge station to determine the lowest tide time in a month. According to the predicted tide level at the lowest tide, determine the operation time range.

[0035] The operation time is determined based on the lowest tide time in the tidal timetable. The operation time is from 80 minutes before to 10 minutes after the lowest tide time, and the total time is 90 minutes.

[0036] (3) Beach lidar point cloud data collection.

[0037] The drone is equipped with a laser radar and flies over the beach area at low tide to collect laser point cloud data of the flight area, including reflection data of the dry beach, tide pool and sea surface.

[0038] The layout of the drone routes is parallel to the isobaths, and the route length does not exceed the range that the drone can fly in 90 minutes. Priority is given to collecting routes close to the coastline, and then collecting routes close to the isobaths at the lowest tide.

[0039] (4) LiDAR point cloud data solution.

[0040] According to the data solving software supporting the UAV and LiDAR equipment, the initial LiDAR point cloud data is solved, including the point cloud data of the dry beach, tide pool and part of the sea surface.

[0041] (5) Construction of a single flight strip sliding window.

[0042] An initial sliding window is constructed for the point cloud data under a single flight strip. The window is an equilateral triangle, in which the two long sides are perpendicular to the flight strip direction and parallel to the flight strip direction, with an angle of 90°, and the initial side length is set to twice the flight altitude L0=2h.

[0043] The initial sliding window is an equilateral triangle, in which the two long sides are perpendicular to the direction of the flight strip and along the direction of the flight strip, with an angle of 90°, and the initial side length is twice the flight altitude L0=2h flight altitude, which is based on the width assumption that the laser radar obtains the best point quality point cloud data under the ideal field of view angle (-45° to +45°).

[0044] (6) Adjustment and contraction of the sliding window.

[0045] According to the laser point cloud data and laser reflection intensity information, the side length of the sliding window is shrunk. The initial equilateral triangle sliding window is placed at the end of the flight strip, with one 45° angle facing the land end of the flight strip and the other 45° angle perpendicular to the flight strip direction; according to the width of the sea surface reflection point cloud, the side length of the window is gradually shrunk until the sea water point cloud data is detected on both sides of the 90° angle of the window; the shrunk side length is adjusted by the formula L = max(aL0, Tseawater), where a is the shrinkage factor and Tseawater is the reflection intensity of the sea water point cloud.

[0046] The sliding window shrinks, and the lengths of the two right-angled sides are obtained after the point cloud data are detected. The adjusted length Ladjustment = max(aL0, Tseawater), where a is the shrinkage factor.

[0047] (7) The sliding window is transformed into a quadrilateral window.

[0048] When point cloud data are detected on both sides of the 90° angle of the sliding window, the sliding window is copied along the oblique direction to form a quadrilateral window. The size of the quadrilateral window is consistent with the size of the adjusted sliding window.

[0049] The sliding window is transformed into a quadrilateral window, and after the shrinking triangular window, the quadrilateral window is formed by copying and expanding along the hypotenuse of the original regular triangle sliding window, which is used for subsequent point cloud data classification operations, and its side length is: Lquadrilateral=Ladjustment.

[0050] (8) Classification of point cloud data of a single flight path.

[0051] In the sliding window, the difference in reflection intensity between the tidal flat and the seawater is analyzed based on the laser radar reflection intensity information, and the boundary between the tidal flat and the seawater surface is calculated and extracted. The point cloud data on the seawater side is filtered out; in the tidal flat area, the residual seawater and tidal flat point cloud data are further classified, and the seawater noise points in the tidal flat area are eliminated to obtain accurate tidal flat point cloud data.

[0052] (9) Sliding window iteration.

[0053] After completing the classification of the point cloud data of the current sliding window, the sliding window moves to the next area along the flight strip according to the preset step size, and repeats the boundary line detection and data classification process until the point cloud data of the entire flight strip is covered.

[0054] (10) Classification of point cloud data by flight.

[0055] Steps (5) to (9) are processed for each flight strip. After each flight strip is classified and impurity points are filtered out, the point cloud data of all flight strips are integrated to generate a laser point cloud dataset of the tidal flat area without impurities.

[0056] (11)Data integration.

[0057] The tidal flat point cloud data obtained after all flight strips are processed are integrated to generate laser point cloud data of the tidal flat area.

[0058] Example 3 like Figure 1-Figure 7 As shown, this embodiment proposes a beach and sea surface point cloud data classification method based on laser radar reflection intensity information, and improves the accuracy of the elevation data of the tidal flat area by dynamically adjusting the sliding window size. The initial sliding window side length is twice the flight height, and then the window size is adaptively adjusted according to the reflection intensity information to adapt to the actual terrain and data characteristics.

[0059] (1) Determination of beach collection scope.

[0060] Based on the national elevation standard, the contour lines of the beach are obtained according to the topographic map of the beach area. Combined with the tide table published by the tide gauge station near the beach during low tide, the lowest point elevation that can be exposed on the beach under the national elevation datum is calculated based on the relationship between the tide height, depth datum and national elevation datum during low tide: ,in, It is the lowest point elevation that can be exposed under the national elevation benchmark at low tide; The height of the tide during low tide; The difference between the tidal datum and the national elevation datum (such as Figure 1 middle, ).

[0061] according to The theoretical exposed beach area after low tide can be determined based on the contour lines and coastline positions in the topographic map.

[0062] Example: From Figure 1 In the tide timetable, we can see that at 21:46, the lowest tide is 85cm, and the tide height reference plane is 254cm below the average sea level. =85cm+(-254cm)=-169cm, that is, the lowest point of the beach that can be exposed at 21:46 is -1.69 meters under the elevation benchmark. Check the contour lines in the topographic map and calculate = -1.69 m (a continuous line). The high and low tide times of the sampling area are shown in Table 1, and the high and low tide times are shown in Table 2. Figure 1 The tide table shown in represents astronomical tide data and does not include the impact of meteorological reasons. The tide height datum is 254cm below the mean sea level. The tide table datum uses the theoretical depth datum.

[0063] Table 1: Figure 1 The tide gauge station high and low tide timetable shown , Table 2: Figure 1 The high and low tide timetables at the tide gauge stations shown .

[0064] (2) Determination of beach collection time.

[0065] Depending on the location of the beach, check the tide schedule at the nearest tide gauge to determine the lowest tide time within a month. , determine the operation time range based on the tide level predicted at the lowest tide.

[0066] The operation time is determined based on the lowest tide time in the tidal timetable. The operation time is from 80 minutes before to 10 minutes after the lowest tide time, and the total time is 90 minutes.

[0067] (3) Beach lidar point cloud data collection.

[0068] Drone equipped with laser radar at low tide Fly over the beach area and collect laser point cloud data of the flight area, including the dry beach, tide pool and reflection data of the sea surface.

[0069] The route layout of the drone is parallel to the depth contour, and the route length does not exceed the range that the drone can fly in 90 minutes. Priority is given to collecting routes close to the coastline, and then collecting routes close to the depth contour at the lowest tide. ,in, is the flight speed of the UAV (unit: m / min); 90 is the operation time (unit: minute).

[0070] (4) LiDAR point cloud data solution.

[0071] The raw lidar data collected by the drone The data solving software supporting the LiDAR equipment is used to solve the initial LiDAR point cloud data, providing a global initial data set for subsequent classification and processing. . ,in Laser emission time, Laser beam angle, Distance measurement value (unit: meter), Reflection intensity, N is the total number of points. It is a three-dimensional point cloud format obtained directly from the drone data in step (3) without being processed by the supporting data solving software. ,in, is the 3D coordinate of the point cloud, is the reflection intensity, and N is the total number of point clouds.

[0072] (5) Construction of a single flight strip sliding window.

[0073] from Select the point cloud data of a flight path , For the single route point cloud data currently being processed, .

[0074] in It's a flight strip The two-dimensional projection area is determined by the flight zone boundary coordinates and Sure.

[0075] In the process of processing point cloud data under a single flight strip, an initial sliding window needs to be constructed first. The initial sliding window is designed as an isosceles right triangle.

[0076] The initial window definition and side length are as follows: Window geometry definition: One right-angled side is defined as perpendicular to the belt direction, that is, it forms a 90° angle with the extension direction of the navigation belt; the other right-angled side is parallel to the direction of the navigation belt and is along the extension direction of the navigation belt; the two right-angled sides are equal, forming an isosceles right triangle.

[0077] Initial side length calculation: The initial right-angle side length L0 is twice the flight altitude, that is, L0 = 2h, where h represents the flight altitude. The side length L0 = 2h is the point cloud coverage width calculated based on the ideal field of view of the LiDAR (-45° to +45°). Within this field of view, the LiDAR can obtain the best point cloud quality and density. Assuming the field of view is =90° (i.e. ±45°), point cloud coverage width Therefore, the initial side length L0 = 2h can cover the optimal point cloud data range within the flight strip, effectively capture local terrain features (changes at the junction of mudflats and sea water), and avoid reducing the calculation efficiency due to excessive window range.

[0078] The geometric characteristics of the triangular window facilitate sliding along the direction of the flight strip, while adapting to the spatial distribution of the flight strip data through the directionality of the right-angled sides; the initial size ensures sufficient data coverage and provides an initial reference for the subsequent window contraction based on reflection intensity.

[0079] (6) Adjustment and contraction of the sliding window.

[0080] According to the laser point cloud data and laser reflection intensity information, the side length of the sliding window is dynamically shrunk to adapt to the actual distribution characteristics of the sea surface and beach area.

[0081] The adjustment process is as follows: Initial window placement: The initial sliding window is an isosceles right triangle, placed at the end of the flight strip, with a 45° angle facing the land side of the flight strip, another 45° angle perpendicular to the flight strip direction, and a 90° right angle facing the sea side. The initial right-angle side L0=2h (h is the flight altitude, unit: meter).

[0082] Shrinkage basis: The window side length is gradually shrunk according to the width of the sea surface point cloud. The sea point cloud is identified by the laser reflection intensity I, and the sea reflection intensity is usually lower than that of the mudflat; the shrinking target is that the sea point cloud data is detected on both right-angle sides at the 90° angle of the window, that is, the window boundary just covers the boundary area between the sea and the mudflat.

[0083] Calculation of adjusted side length: right-angle side after contraction Determined by the following formula ,in, Shrinkage factor, 0< <1, adjusted according to the actual point cloud density, the recommended value range is 0.5~0.9, adjusted according to the point cloud density and terrain complexity; is the width of the seawater point cloud (unit: meter). It is defined as the vertical reflection intensity along the flight strip below the threshold The point cloud range, , The point cloud data of a single route currently being processed, The threshold of the reflection intensity between seawater and mudflats, Through experimental calibration, for example, taking the statistical upper limit of the seawater reflection intensity.

[0084] The shrinking process starts from the initial side length L0, and gradually reduces the length of the right-angle side. When two of the sides detect the seawater point cloud, there is a point P that satisfies When , stop shrinking and record the actual length of the two right-angle sides at this time and , take the larger value and Compare and finalize .

[0085] Adaptively adjust the window size by reflection intensity to ensure that the window is neither too large to include too much irrelevant data nor too small to miss key features. The formula balances the constraints of the initial settings and actual data and improves the classification accuracy.

[0086] (7) The sliding window is transformed into a quadrilateral window.

[0087] After the sliding window is adjusted, when the two right-angled sides at the 90° angle detect seawater point cloud data, the triangular window is expanded along the hypotenuse direction and transformed into a quadrilateral window to further optimize the classification processing of the point cloud data.

[0088] The specific process is as follows: Transformation conditions: The initial sliding window is an isosceles right triangle. After shrinkage, the length of the right angle side is , when two of the edges detect the seawater point cloud, that is, there is a point satisfy , it indicates that the window has adapted to the boundary area between seawater and mudflats, and the transformation is triggered at this time.

[0089] Quadrilateral window construction: Take the adjusted triangular window as the base, copy and expand along its hypotenuse to form a quadrilateral window. The collective characteristic of the quadrilateral window is that the length of the two parallel sides is equal to the length of the right-angle side of the triangle. , the length of the other two parallel sides (in the direction of the original hypotenuse) is the length of the hypotenuse of the triangle, that is, , based on the properties of an isosceles right triangle, the length of the hypotenuse is equal to the length of the right angle side times.

[0090] Definition of the side length of the quadrilateral window: The size of the quadrilateral window is consistent with the adjusted triangular window, and the length of its right angle side is .

[0091] The transformation process is: Keep the 90° position of the adjusted triangle window unchanged, copy an identical triangle along the hypotenuse, and make the hypotenuses of the two triangles overlap to form a parallelogram. The total area of ​​the quadrilateral window is: .

[0092] The purpose of the transformation is: the quadrilateral window enhances the ability to capture point cloud data in the interface area between seawater and mudflats by expanding the coverage; its regular geometric shape facilitates subsequent classification operations based on reflection intensity and improves the stability and accuracy of the algorithm.

[0093] (8) Classification of point cloud data of a single flight path.

[0094] In the sliding window, the difference in reflection intensity between the tidal flat and the seawater is analyzed based on the LiDAR reflection intensity information, and the boundary between the tidal flat and the sea surface is calculated and extracted. The point cloud data is then classified and noise filtered to obtain accurate tidal flat point cloud data. The specific steps are as follows.

[0095] Reflection intensity difference analysis: Analysis within a sliding window Reflection intensity difference, LiDAR point cloud data , is the 3D coordinate of the point cloud, is the reflection intensity, and N is the total number of point clouds; tidal flat areas, such as dry beaches, usually have higher reflection intensity, while seawater areas, including tide pools, have lower reflection intensity. Based on this feature, the reflection intensity threshold is set. As the basis for preliminary classification.

[0096] Boundary extraction: within the sliding window, according to the reflection intensity Will Point cloud data is divided into two categories: mudflats and seawater:

[0097] , dividing line Defined as and The spatial boundary of can be calculated by the gradient change of the reflection intensity in the window: ,in It represents the spatial gradient of reflection intensity, reflecting the sudden change in intensity at the interface between the mudflat and the sea water.

[0098] Seawater side point cloud filtering: The point cloud data in the For subsequent processing.

[0099] Classification and filtering of residual noise in the tidal flat area: There may be residual seawater points, such as tide pools or wetlands, which need to be further classified and removed. The secondary filtering conditions are set in combination with the elevation z and the reflection intensity I: ,in, For the window The average elevation of the tidal flats is used to distinguish low-lying tidal pools from dry mudflats; Secondary reflection intensity threshold , used to identify the low-reflection characteristics of residual seawater points.

[0100] Culling Finally, we can get accurate beach and tidal flat point cloud data. .

[0101] The boundary line is accurately located through reflection intensity threshold and gradient analysis to improve classification accuracy; secondary filtering combines elevation and intensity characteristics to effectively eliminate seawater noise in the tidal flat area and ensure high accuracy of point cloud data.

[0102] (9) Sliding window iteration.

[0103] After completing the classification of the point cloud data in the current sliding window, the sliding window moves to the next area along the flight strip according to the preset step size, and repeats the boundary line detection and data classification process until the point cloud data of the entire flight strip is covered.

[0104] The specific steps are as follows: Iteration trigger: Current sliding window (quadrilateral window, side length ) After completing the classification of the mudflat and seawater point clouds, the local fine point cloud is obtained .

[0105] Step size definition: The sliding window moves along the extension direction of the flight strip, the extension direction of the flight strip is parallel to a right angle edge, and the moving step size for ,in, is the step size factor ( ), controls the relationship between the moving distance and the window size, and takes a value between 0.3 and 0.5 to balance coverage and computational efficiency; The length of the right-angle side of the window after adjustment, in meters.

[0106] Movement process: The window moves along the direction of the flight path Distance, the new window center position is updated as: ,in, The center coordinates of the current window; The unit vector representing the direction of the flight path, Indicates the center coordinates of the next window.

[0107] Iterative operation: Repeat the above steps at the new window position to detect the boundary line based on the reflection intensity I, and then classify the tidal flat point cloud and seawater point cloud , filter out The seawater noise in .

[0108] Termination condition: The window moves to the end of the flight strip, covering the point cloud data of the single flight strip currently being processed Represents the point cloud data in iterative processing, through the calculation formula Determine the point cloud data location at the end of the flight path, where: is the total length of the flight strip (unit: meter); Ceiling function.

[0109] Step Length Adaptive design based on Ensure that window coverage is consistent with terrain changes; the iterative process efficiently covers the flight strip to avoid data omissions or duplicate processing.

[0110] (10) Classification of point cloud data for each flight path.

[0111] right Each flight path in the process is processed from step (5) to step (9) one by one to complete the classification and impurity filtering of point cloud data. The specific steps are as follows: Single strip processing: Strip aggregation Each flight path The following processing is performed respectively, where M is the total number of flight strips.

[0112] Execute step (5) to construct the initial sliding window (isosceles right triangle, ); Execute step (6) to adjust the window side length to ; Execute step (7) to transform into a quadrilateral window ( ); Execute step (8) to classify the tidal flat point cloud based on reflection intensity And filter out seawater noise; Execute step (9) according to the step size Iterate the sliding window until it covers All point cloud data .

[0113] Single strip results: For each strip After processing, a subset of impurity-free mudflat point clouds is obtained: ,in, For flight strip The accurate tidal flat point cloud after classification and filtering.

[0114] (11)Data integration.

[0115] The tidal flat point cloud data obtained after processing each flight strip are integrated to obtain a laser point cloud data set of the tidal flat area without impurities.

[0116] Merge the tidal flat point cloud subsets of all flight strips to obtain a complete tidal flat point cloud dataset for the beach area: .

[0117] During the integration process, ensure that the point cloud plane coordinate system is consistent, the elevation datum is consistent, select the 1985 national elevation datum, and the overlapping area data between flight strips can be smoothed or interpolated to optimize the continuity of the data set.

[0118] ,in, is the point cloud of the overlapping area.

[0119] The point cloud data after optimization and smoothing is the final point cloud data ,in .

[0120] The above embodiments are only for illustrating the inventive concept and features of the present invention, and their purpose is to enable those skilled in the art to understand the content of the present invention and implement it accordingly, and they cannot be used to limit the protection scope of the present invention. Any equivalent changes or modifications made based on the essence of the content of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for classifying tidal flat laser point cloud data based on laser radar reflection intensity information, characterized in that: The following steps are involved: Step 1: Based on the depth contours and coastline, obtain the theoretical maximum range of the beach exposed after the sea water recedes and the high tide. According to the predicted tide level at the lowest tide, obtain the operation time range and determine the tidal flat collection range and collection time. Step 2: Set the collection route. The UAV equipped with the laser radar flies over the mudflat collection area at low tide to collect laser point cloud data in the flight area. The initial laser radar point cloud data is obtained by solving the data solving software supporting the UAV and the laser radar equipment. Step 3: Construct a triangular initial sliding window for the initial lidar point cloud data under a single UAV flight path and set the initial side length; Step 4: According to the laser point cloud data and laser reflection intensity information, dynamically shrink the side length of the sliding window to adapt to the actual distribution characteristics of the sea surface and beach area; Step 5: When point cloud data are detected on both sides of the 90° angle of the sliding window, expand along the oblique direction to form a quadrilateral window with the same size as the adjusted sliding window to optimize the classification processing of the point cloud data; Step 6: In the quadrilateral window, the difference in reflection intensity between the tidal flat and the seawater is analyzed based on the laser radar reflection intensity information, the boundary between the tidal flat and the sea surface is calculated and extracted, the point cloud data is classified and noise is filtered out, and accurate tidal flat point cloud data is obtained; Step 7: After completing the classification of the point cloud data of the current quadrilateral window, the quadrilateral window moves to the next area along the flight strip according to the preset step length, repeating the boundary line detection and data classification process until the point cloud data of the entire flight strip is covered, and the sliding window is iterated; Step 8: Perform steps 3 to 7 for each flight strip. After classifying each flight strip and filtering out impurity points, integrate the point cloud data of all flight strips to generate a single-strip laser point cloud dataset of the tidal flat area without impurities. Integrate the single-strip laser point cloud dataset of the tidal flat area to generate the laser point cloud data of the tidal flat area.

2. The method for classifying tidal flat laser point cloud data based on laser radar reflection intensity information according to claim 1 is characterized in that: In step 1, the collection time is from 80 minutes before to 10 minutes after the lowest tide, and the total time is 90 minutes.

3. The method for classifying tidal flat laser point cloud data based on laser radar reflection intensity information according to claim 1 is characterized in that: In step 2, the drone collection route is arranged parallel to the isobath, the length of the collection route does not exceed the flight distance that the drone can fly in 90 minutes, and the collection route is set from close to the coastline to far away from the coastline, and then the collection route close to the isobath is collected at the lowest tide.

4. The method for classifying tidal flat laser point cloud data based on laser radar reflection intensity information according to claim 1 is characterized in that: In step 3, the initial sliding window is an isosceles right triangle, one of the two right-angled sides is perpendicular to the direction of the flight strip, and the other is parallel to the direction of the flight strip. The initial side length is set to twice the flight altitude according to the width of the point cloud data with the best point quality obtained by the laser radar under the ideal field of view angle, L0=2h, where h is the flight altitude.

5. The method for classifying tidal flat laser point cloud data based on laser radar reflection intensity information according to claim 1 is characterized in that: The step 4 comprises: Step 4-1: Place the initial sliding window at the end of the flight strip, with one 45° angle facing the land side of the flight strip, the other 45° angle perpendicular to the flight strip direction, and a 90° right angle facing the sea side. The initial right angle side L0=2h; Step 4-2: According to the width of the sea surface reflection point cloud, gradually shrink the side length of the sliding window, and use the formula for the side length after shrinkage: Calculate the length of the side after contraction, where is the shrinkage factor, is the width of the sea water point cloud; Step 4-3: When both right-angled sides of the sliding window detect the seawater point cloud data, stop shrinking and record the actual lengths of the two right-angled sides at this time. and , take the larger value and Compare and finalize .

6. The method for classifying tidal flat laser point cloud data based on laser radar reflection intensity information according to claim 5 is characterized in that: In step 4-2, it is defined that the reflection intensity in the vertical direction along the flight strip is lower than the threshold The point cloud range, , The point cloud data of a single route currently being processed, It is the dividing threshold of the reflection intensity between sea water and mudflats.

7. The method for classifying tidal flat laser point cloud data based on laser radar reflection intensity information according to claim 5 is characterized in that: The step 5 comprises: Step 5-1: Take the triangle window when the contraction stops as the reference, copy an identical triangle along the hypotenuse, make the hypotenuses of the two triangles overlap, and form a parallelogram window; Step 5-2: Calculate the total area of ​​the parallelogram window: , by using parallelogram windows to expand window coverage.

8. The method for classifying tidal flat laser point cloud data based on laser radar reflection intensity information according to claim 5 is characterized in that: The step 6 comprises: Step 6-1: Reflection intensity difference analysis, LiDAR point cloud data ,in is the 3D coordinate of the point cloud, is the reflection intensity, N is the total number of point clouds, and the reflection intensity threshold is set based on the fact that the reflection intensity of the dry tidal flat area is higher and the reflection intensity of the tidal flat area is lower. As a basis for preliminary classification; Step 6-2: Boundary extraction, within the sliding window, according to the reflection intensity The point cloud data is divided into two categories: mudflats and seawater: , , dividing line Defined as and The spatial boundary of can be calculated by the gradient change of the reflection intensity in the window: ,in It represents the spatial gradient of reflection intensity, reflecting the sudden change in intensity at the interface between the mudflat and the sea water; Step 6-3: Filtering the point cloud on the seawater side: The point cloud data in the For subsequent processing; Step 6-4: Classification and filtering of residual noise in the tidal flat area. In the process, we further classify and remove the residual seawater points, and set the secondary filtering conditions based on the elevation and reflection intensity: ,in, For the window The average elevation of the tidal flats is used to distinguish low-lying tidal pools from dry mudflats; Secondary reflection intensity threshold; Step 6-5: Elimination Then we get an accurate local point cloud .

9. The method for classifying tidal flat laser point cloud data based on laser radar reflection intensity information according to claim 8 is characterized in that: The step 7 comprises: Step 7-1: Step size definition, the parallelogram sliding window moves along the extension direction of the flight strip, and the moving step size for: ,in, is the step size scaling factor; Step 7-2: Movement process: The window moves along the direction of the flight path Distance, the new window center position is updated as: ,in, The center coordinates of the current window; The unit vector of the flight path direction; The center coordinates of the next window; Step 7-3: Repeat step 6 at the new window position to continuously generate new local fine point clouds; Step 7-4: Move the parallelogram window to the end of the flight strip to obtain point cloud data covering the entire flight strip .

10. The method for classifying tidal flat laser point cloud data based on laser radar reflection intensity information according to claim 9, characterized in that: Step 8 includes: In the process of integrating the laser point cloud data set of a single strip in the tidal flat area, it is necessary to keep the point cloud plane coordinate system consistent, the elevation benchmark consistent, the overlapping area data between the strips, and choose to smooth or interpolate the overlapping area to maintain the continuity of the data set. ,in, The point cloud of the overlapping area, the point cloud data after optimization and smoothing is the final point cloud data ,in .

Images