A method for measuring the volume of unordered point clouds based on voxelization and filling algorithms

Abstract

The present invention discloses a method for measuring the volume of unordered point clouds based on voxelization and filling algorithms, belonging to the technical field of point cloud mapping, including: determining algorithm parameters; inputting point cloud data and performing direct filtering slicing according to the determined slicing operation parameters; only considering the non-slicing coordinate direction and performing two-dimensional voxelization according to the voxelization resolution; constructing a two-dimensional voxel matrix; using the X-scan line filling algorithm to fill the two-dimensional voxel matrix; traversing the two-dimensional voxel matrix and counting the slice unit quantity value of the point cloud; obtaining the area of the point cloud slice and the volume of the sliced point cloud data, and accumulating them into the point cloud volume; adjusting the starting point of the slice, iterating until the sum of the starting point of the slice and the slice step is greater than the highest point of the point cloud data, and obtaining the volume of the entire point cloud data. The present invention achieves a balance between efficiency and accuracy by setting different parameters, is faster than the slicing method, reduces errors, improves the point cloud slicing method, and enhances the practicality and engineering level.

Description

Technical Field

[0001] The present invention relates to the technical field of point cloud mapping, and in particular to a method for measuring the volume of unordered point clouds based on voxelization and filling algorithms. Background Art

[0002] The methods for calculating the volume of point clouds are generally divided into four categories: convex hull algorithm, model reconstruction method, projection method, and slicing method.

[0003] The convex hull algorithm uses a convex hull model to approximately represent an irregular object, and then calculates the sum by slicing and dividing the convex hull model, or decomposes the convex hull model into two upper and lower triangular mesh surfaces, and uses the orthographic projection method to obtain the projection volumes of the two. It is only applicable to convex models, and the error of non-convex models is relatively large.

[0004] The model reconstruction method obtains the volume by using triangular patches to construct a physical model. This algorithm is greatly affected by the point cloud density, the number of generated triangular meshes, and the point accuracy, and is prone to generate holes.

[0005] The projection method first triangulates the projection of the point cloud, then constructs a pentahedron by the projection points and their original corresponding points, and obtains the total volume by accumulating the volumes of the pentahedrons. This algorithm is also prone to generate holes. For the above algorithms, whether constructing a physical model from the three-dimensional point cloud first and then calculating the volume, or directly calculating the volume based on the three-dimensional point cloud by geometric methods, when there are problems such as uneven density of the three-dimensional point cloud collected by lidar, or transition zones or transition lines in spatial objects, the error of reconstructing the three-dimensional model is relatively large, and the volume calculation accuracy is not high.

[0006] The slicing method slices the point cloud along a certain coordinate axis direction, then calculates the areas of the upper and lower surfaces of the slice, and obtains the total volume by accumulating the slice volumes. This method is affected by the slice thickness. The smaller the slice thickness, the higher the calculation accuracy but the lower the calculation efficiency.

[0007] In order to obtain more scene information and more comprehensive parameter information of the point cloud data. Today, with the increasing popularity of the point cloud data acquisition method, it is urgent to develop a new method for measuring the volume of point clouds. Summary of the Invention

[0008] The technical problem to be solved by the present invention is to provide a method for measuring the volume of unordered point clouds based on voxelization and filling algorithms. For point cloud objects with different irregular shapes, a balance between efficiency and accuracy is achieved by setting different parameters. It is faster than the slicing method, reduces errors, improves the point cloud slicing method, and improves the practicality and engineering level.

[0009] To solve the above technical problem, the technical solution adopted by the present invention is:

[0010] A method for measuring the volume of unordered point clouds based on voxelization and filling algorithms, comprising the following steps:

[0011] Step 1: Determine the algorithm parameters, including the point cloud slicing operation parameters and the voxelization resolution;

[0012] Step 2: Input the point cloud data, and perform a pass-through filtering slice on the point cloud data along a specific direction according to the determined slicing operation parameters;

[0013] Step 3: Only consider the non-slicing coordinate direction of the sliced point cloud data, perform two-dimensional voxelization according to the voxelization resolution, construct a voxel grid, calculate the voxel number of each voxel in the sliced data, and randomly sample the points with the same voxel number;

[0014] Step 4: Construct a two-dimensional voxel matrix according to the voxelization resolution and the boundary information of the slice;

[0015] Step 5: Use the X scan line filling algorithm to fill the two-dimensional voxel matrix processed in Step 4, and set the matrix elements within the boundary to 1;

[0016] Step 6: Traverse the two-dimensional voxel matrix, count the number of matrix elements with a value of 1, and use it as the value of the slice unit quantity of the point cloud;

[0017] Step 7: Take the product of the voxelization resolution and the voxelization resolution as the basic area unit, obtain the area of the point cloud slice and the volume of the sliced point cloud data, and accumulate them into the point cloud volume;

[0018] Step 8: Adjust the starting point of the slice, and iterate Steps 1 to 6 until the sum of the starting point of the slice and the slice step size is greater than the highest point of the point cloud data, and obtain the volume of the entire point cloud data.

[0019] A further improvement of the technical solution of the present invention lies in that: Step 1 specifically includes the following steps:

[0020] Step 1.1: Set the point cloud slicing operation parameters, including the slicing height h and the number of slices l;

[0021] Step 1.2: Set the voxelization resolution r of the point cloud.

[0022] A further improvement of the technical solution of the present invention lies in that: when the slicing axis is the z-axis, Step 2 specifically includes the following steps:

[0023] Step 2.1: Set the slice step size of the point cloud slice:

[0024]

[0025] where w is the slice step size, Z max 、Z min are the maximum and minimum values of the z-axis of the point cloud slice;

[0026] Step 2.2: Set the starting point s and the ending point e of the slicing operation:

[0027] s = Z min + a (2)

[0028] e = s + h (3)

[0029] Where s is the starting point of the Z-axis of the point cloud slice for the slicing operation, e is the ending point of the Z-axis of the point cloud slice for the slicing operation, and a is the slicing increment;

[0030] Step 2.3: Take [s, e] as the range of action of the pass-through filter, and take the point cloud data points within this range as the slice data for the j-th slice area calculation. The points outside this range are discarded.

[0031] A further improvement of the technical solution of the present invention is that when the slicing axis is the z-axis, step 3 specifically includes the following steps:

[0032] Step 3.1: Obtain the maximum and minimum values of the non-slicing axes X and Y of the sliced point cloud;

[0033] Step 3.2: Calculate and obtain the unit D x , D y :

[0034]

[0035] Where D x is the basic unit of the X-axis after voxelization, D y is the basic unit of the Y-axis after voxelization, X min is the minimum value of the X-axis of the sliced point cloud, X max is the maximum value of the X-axis of the sliced point cloud, Y min is the minimum value of the Y-axis of the sliced point cloud, Y max is the maximum value of the Y-axis of the sliced point cloud;

[0036] Step 3.3: For the point P i (x i , y i , z i ), calculate and obtain the voxel number v i :

[0037] v i = (x i - X min ) + (y i - Y min ) * D x (5)

[0038] Step 3.4: Add the voxel number information to P i :

[0039] P i =(x i ,y i ,z i ,v i )(6)

[0040] Step 3.5: Randomly sample data points with the same voxel number:

[0041] P i = random(v i )(7).

[0042] A further improvement of the technical solution of the present invention is that when the slicing axis is the z-axis, step 4 specifically includes the following steps:

[0043] Step 4.1: Construct a two-dimensional voxel matrix M with dimensions D x *D y , and initialize all element values in the two-dimensional voxel matrix M to 0;

[0044] Step 4.2: Fill the two-dimensional voxel matrix elements according to the voxel numbers of the data points, and set the matrix element corresponding to the point cloud voxel number to 1. For the point P i

[0045]

[0046] where x m is the x coordinate of the voxel matrix corresponding to the voxel number of the point P i , y m is the y coordinate of the voxel matrix corresponding to the voxel number of the point P i , is the element corresponding to the x m th row and the y m th column of the two-dimensional voxel matrix M;

[0047] Step 4.3: Iterate steps 4.1 to 4.2 until all data points are traversed, and map the data points into the two-dimensional voxel matrix M.

[0048] A further improvement of the technical solution of the present invention is that step 5 specifically includes the following steps:

[0049] Step 5.1: Traverse the row vectors of the entire matrix and merge the matrix boundary points. For the left boundary, set the boundary point as the rightmost point of the left boundary point set, that is, the point with the largest x value; for the right boundary, set the boundary point as the leftmost point of the right boundary point set, that is, the point with the smallest x value, so that there are 2 boundary points in each row;

[0050] Step 5.2: Match the boundary points of each row of the matrix. Two boundary points in one row are used as a group, and the boundary merging is as followsFigure 5 as shown

[0051] Step 5.3: Set all matrix elements between each group of boundary points to 1.

[0052] A further improvement of the technical solution of the present invention lies in that: Step 7 specifically includes the following steps:

[0053] Step 7.1: Use the product of the voxelization resolution r and the voxelization resolution r as the basic voxel unit area after slice voxelization;

[0054] Step 7.2: Obtain the area S of the slice point cloud data j :

[0055] S j = n * r * r (10)

[0056] Step 7.3: Obtain the volume V of the slice point cloud data j :

[0057] V j = S j * w (11)

[0058] Step 7.4: Accumulate V j into the point cloud volume V.

[0059] A further improvement of the technical solution of the present invention lies in that: Step 8 specifically includes the following steps:

[0060] Step 8.1: Increase the slice increment value:

[0061] a = w * j (12)

[0062] where j is the number of slices;

[0063] Step 8.2: Judge the increment condition. If at this time Z min + α > Z max , then the algorithm process is completed, and the point cloud volume V is returned; otherwise, go to Step 8.3;

[0064] Step 8.3: Iterate Steps 2 to 8 until the entire point cloud data is sliced.

[0065] Due to the adoption of the above technical solution, the technical progress achieved by the present invention is:

[0066] 1. By combining the graphics filling method and the traditional slicing method to obtain the volume of point cloud data, the present invention realizes the rapid and high-precision measurement of the volume parameters of a large amount of unordered and irregular point cloud data.

[0067] 2. The present invention obtains the slice area by using the X-scan line algorithm to fill the slice data. Compared with the traditional method of fitting the slice boundary and then integrating, the speed is less affected by the boundary complexity and data complexity. The time complexity of boundary filling depends on the data scale, which improves the real-time performance and robustness of the algorithm. Description of the Drawings

[0068] Figure 1 is the overall system architecture diagram of the present invention;

[0069] Figure 2 is the schematic diagram of the slicing operation;

[0070] Figure 3 is the schematic diagram of the slicing result;

[0071] Figure 4 is the schematic diagram of the voxel matrix construction and boundary filling result;

[0072] Figure 5 is the schematic diagram of boundary point merging;

[0073] Figure 6 is the schematic diagram of the X-scan line filling algorithm;

[0074] Figure 7(a) is the schematic diagram of the slice point cloud data before filling, and Figure 7(b) is the schematic diagram of the voxel filling operation of the filled slice point cloud data;

[0075] Figure 8 is the schematic diagram of the voxel filling result of the voxel matrix. Detailed Embodiment

[0076] The present invention will be further described in detail below with reference to the drawings and embodiments:

[0077] As Figure 1 shown, a method for measuring the volume of unordered point clouds based on voxelization and filling algorithms includes the following steps:

[0078] Step 1: Determine the algorithm parameters, including the point cloud slicing operation parameters and the voxelization resolution;

[0079] Specifically, it includes the following steps:

[0080] Step 1.1: Set the point cloud slicing operation parameters, including the slicing height h and the number of slices l;

[0081] Step 1.2: Set the point cloud voxelization resolution r;

[0082] Step 2: Input the point cloud data, and perform straight-through filtering slicing on the point cloud data along a specific direction according to the determined slicing operation parameters;

[0083] When the slicing axis is the z-axis, it specifically includes the following steps:

[0084] Step 2.1: Set the slicing step of the point cloud slice:

[0085]

[0086] where w is the slicing step, and Z max and Z min are the maximum and minimum values of the Z-axis of the point cloud slice;

[0087] Step 2.2: Set the starting point s and ending point e of the slicing operation:

[0088] s = Z min + a (2)

[0089] e = s + h (3)

[0090] where s is the starting point of the Z-axis of the point cloud slice for the slicing operation, e is the ending point of the Z-axis of the point cloud slice for the slicing operation, and a is the slicing increment;

[0091] Step 2.3: Take [s, e] as the range of the pass-through filter, and take the point cloud data points within this range as the slice data for the j-th slice area calculation, and the points outside this range are discarded;

[0092] The specific slicing operation is as Figure 2 shown, and the slicing operation result is as Figure 3 shown.

[0093] Step 3: Only consider the non-slicing coordinate directions of the sliced point cloud data, perform two-dimensional voxelization according to the voxelization resolution, construct a voxel grid, calculate the voxel number of each slice data, and randomly sample the points with the same voxel number;

[0094] When the slicing axis is the z-axis, it specifically includes the following steps:

[0095] Step 3.1: Obtain the maximum and minimum values of the non-slicing axes X and Y of the sliced point cloud;

[0096] Step 3.2: Calculate and obtain the unit D x , D y :

[0097]

[0098] where D x is the basic unit of the X-axis after voxelization, D y is the basic unit of the Y-axis after voxelization,, X min is the minimum value of the X-axis of the sliced point cloud, X max is the maximum value of the X-axis of the sliced point cloud, Y min is the minimum value of the Y-axis of the sliced point cloud, Ymax is the maximum value of the Y-axis of the sliced point cloud;

[0099] Step 3.3: For point P i (x i ,y i ,z i ), calculate and obtain the voxel number v i ,:

[0100] v i = (x i - X min) + (y i - Y min ) * D x (5)

[0101] Step 3.4: Add the voxel number information to P i :

[0102] P i = (x i ,y i ,z i ,v i ) (6)

[0103] Step 3.5: Randomly sample data points with the same voxel number:

[0104] P i = random(v i ) (7).

[0105] Step 4: Construct a two-dimensional voxel matrix according to the voxelization resolution and the boundary information of the slice;

[0106] When the slice axis is the z-axis, it specifically includes the following steps:

[0107] Step 4.1: Construct a two-dimensional voxel matrix M with dimensions D x * D y , and initialize all element values in the two-dimensional voxel matrix M to 0;

[0108] Step 4.2: Fill the two-dimensional voxel matrix elements according to the voxel numbers of the data points, and set the matrix element corresponding to the point cloud voxel number to 1. For point P i

[0109]

[0110] where x m is the x coordinate of the voxel matrix corresponding to the voxel number of point P i , y m is the y coordinate of the voxel matrix corresponding to the voxel number of point P i , It is the element corresponding to the x m -th row and y m -th column of the two-dimensional voxel matrix M;

[0111] Step 4.3: Iterate steps 4.1 to 4.2 until all data points are traversed, and map the data points into the two-dimensional voxel matrix M;

[0112] The constructed two-dimensional voxel matrix and the boundary filling effect are as Figure 4 shown.

[0113] Step 5: Use the X-scan line filling algorithm to fill the two-dimensional voxel matrix processed in step 4, and set the matrix elements inside the boundary to 1;

[0114] Specifically, it includes the following steps:

[0115] Step 5.1: Traverse the row vectors of the entire matrix, merge the matrix boundary points. For the left boundary, set the boundary point as the rightmost point of the left boundary point set, that is, the point with the largest x value; for the right boundary, set the boundary point as the leftmost point of the right boundary point set, that is, the point with the smallest x value, so that there are 2 boundary points in each row;

[0116] Step 5.2: Match the boundary points in each row of the matrix. Two boundary points in one row are taken as a group, and the boundary merging is as Figure 5 shown;

[0117] Step 5.3: Set all matrix elements between each group of boundary points to 1;

[0118] The X-scan line filling algorithm is as Figure 6 shown, the filling operation is shown in Figure 7, and the two-dimensional voxel matrix after filling is shown as follows.

[0119] Step 6: Traverse the two-dimensional voxel matrix, count the number of matrix elements with a value of 1, and use it as the number of slice units of the point cloud;

[0120] Step 7: Take the product of the voxelization resolution and the voxelization resolution as the basic area unit, obtain the area of the point cloud slice and the volume of the sliced point cloud data, and accumulate them into the point cloud volume;

[0121] Specifically, it includes the following steps:

[0122] Step 7.1: Take the product of the voxelization resolution r and the voxelization resolution r as the basic voxel area after slice voxelization;

[0123] Step 7.2: Obtain the area S j :

[0124] S j = n * r * r (10)

[0125] Step 7.3: Obtain the volume V of the sliced point cloud data j :

[0126] V j = S j * w (11)

[0127] Step 7.4: Accumulate V j into the point cloud volume V.

[0128] Step 8: Adjust the starting point of the slice, iterate Steps 1 - 6 until the sum of the starting point of the slice and the slice step size is greater than the highest point of the point cloud data, and obtain the volume of the entire point cloud data;

[0129] Specifically, it includes the following steps:

[0130] Step 8.1: Increase the slice increment value:

[0131] a = w * j (12)

[0132] where j is the number of times of slicing;

[0133] Step 8.2: Judge the increment condition. If at this time Z min + α > Z max , the algorithm process is completed, return the point cloud volume V, otherwise enter Step 8.3;

[0134] Step 8.3: Iterate Steps 2 - 8 until the entire point cloud data is sliced.

[0135] In summary, the present invention combines the point cloud slicing method in the traditional measurement method with the efficient filling algorithm X - scan line algorithm in computer graphics, constructs the point cloud topological structure based on the voxelization method, reduces the amount of point cloud data to be processed through voxel downsampling, and then discretely approximates the point cloud slice area by filling the slice boundary through the computer graphics filling algorithm, thus avoiding solving the plane boundary expression and greatly reducing the volume calculation time. For different irregularly shaped point cloud objects, different parameters can be set to achieve a balance between efficiency and accuracy. This fast calculation method is faster than the slicing method and has a smaller error. This fast method improves the point cloud slicing method and is a more practical and engineering - oriented method.

Claims

1. A method for measuring the volume of unordered point clouds based on voxelization and filling algorithms, characterized in that: It includes the following steps: Step 1: Determine the algorithm parameters, including the point cloud slicing operation parameters and the voxelization resolution; Step 2: Input the point cloud data, and perform a pass-through filtering slice on the point cloud data along a specific direction according to the determined slicing operation parameters; Step 3: Only consider the non-slicing coordinate direction of the sliced point cloud data, perform two-dimensional voxelization according to the voxelization resolution, construct a voxel grid, calculate the voxel number of each voxel of the sliced data, and randomly sample the points with the same voxel number; Step 4: Construct a two-dimensional voxel matrix according to the voxelization resolution and the boundary information of the slice; Step 5: Use the X scan line filling algorithm to fill the two-dimensional voxel matrix processed in Step 4, and set the matrix elements within the boundary to 1; Step 6: Traverse the two-dimensional voxel matrix, count the number of matrix elements with a value of 1, and use it as the value of the slice unit quantity of the point cloud; Step 7: Take the product of the voxelization resolution and the voxelization resolution as the basic unit of area, obtain the area of the point cloud slice and the volume of the sliced point cloud data, and accumulate them into the point cloud volume; Step 8: Adjust the starting point of the slice, and iterate Steps 1 to 6 until the sum of the starting point of the slice and the slice step size is greater than the highest point of the point cloud data, and obtain the volume of the entire point cloud data.

2. A method for measuring the volume of unordered point clouds based on voxelization and filling algorithms according to claim 1, characterized in that: Step 1 specifically includes the following steps: Step 1.1: Set the point cloud slicing operation parameters, including the slice height h and the number of slices l; Step 1.2: Set the voxelization resolution r of the point cloud.

3. A method for measuring the volume of unordered point clouds based on voxelization and filling algorithm according to claim 2, characterized in that: When the slicing axis is the z-axis, Step 2 specifically includes the following steps: Step 2.1: Set the slice step size of the point cloud slice: where w is the slicing step, and Z max and Z min are the maximum and minimum values of the Z-axis of the point cloud slice; Step 2.2: Set the starting point s and the ending point e of the slicing operation: s = Z min + a(2) e = s + h (3) where s is the starting point of the z-axis of the point cloud slice for the slicing operation, e is the ending point of the z-axis of the point cloud slice for the slicing operation, and a is the slice increment; Step 2.3: Take [s, e] as the pass-through filtering action interval, take out the point cloud data points within this interval as the sliced data for the j-th slice area calculation, and discard the points outside this interval.

4. A method for measuring the volume of unordered point clouds based on voxelization and filling algorithms according to claim 1, characterized in that: When the slicing axis is the z-axis, Step 3 specifically includes the following steps: Step 3.1: Obtain the maximum and minimum values of the non-slicing axes X and Y of the sliced point cloud; Step 3.2: Calculate and obtain unit D x , D y : Among them, r is the voxelization resolution, D x is the basic unit of the X-axis after voxelization, D y is the basic unit of the Y-axis after voxelization, X min is the minimum value of the X-axis of the sliced point cloud, X max is the maximum value of the X-axis of the sliced point cloud, Y min is the minimum value of the Y-axis of the sliced point cloud, Y max is the maximum value of the Y-axis of the sliced point cloud; Step 3.3: For point P i (x i , y i , z i ), calculate and obtain the voxel number v i : v i = (x i - X min ) + (y i - Y min ) * D x (5) Step 3.4: Add the voxel number information to P i in: P i = (x i , y i , z i , v i ) (6) Step 3.5: Randomly sample the data points with the same voxel number: P i = random(v i ) (7).

5. A method for measuring the volume of unordered point clouds based on voxelization and filling algorithms according to claim 4, characterized in that: When the slicing axis is the z-axis, Step 4 specifically includes the following steps: Step 4.1: Construct a two-dimensional voxel matrix M with dimension D x *D y and initialize all element values in the two-dimensional voxel matrix M to 0; Step 4.2: Fill the two-dimensional voxel matrix elements according to the voxel numbers of the data points, set the matrix elements corresponding to the voxel numbers of the point cloud to 1, for point P i where x m is the x - coordinate of the voxel matrix corresponding to the voxel number of point P i , y m is the y - coordinate of the voxel matrix corresponding to the voxel number of point P i , and is the element corresponding to the x m -th row and the y m -th column of the two - dimensional voxel matrix M; Step 4.3: Iterate Steps 4.1 to 4.2 until all data points are traversed, and map the data points to the two-dimensional voxel matrix M.

6. A method for measuring the volume of unordered point clouds based on voxelization and filling algorithms according to claim 1, characterized in that: Step 5 specifically includes the following steps: Step 5.1: Traverse the row vectors of the entire matrix, merge the matrix boundary points. For the left boundary, set the boundary point as the rightmost point of the left boundary point set, that is, the point with the largest x value; for the right boundary, set the boundary point as the leftmost point of the right boundary point set, that is, the point with the smallest x value, so that there are 2 boundary points in each row; Step 5.2: Match the boundary points of each row of the matrix. Take two boundary points in a row as a group and perform boundary merging; Step 5.3: Set all matrix elements between each group of boundary points to 1.

7. A method for measuring the volume of unordered point clouds based on voxelization and filling algorithms according to claim 3, characterized in that: Step 7 specifically includes the following steps: Step 7.1: Take the product of the voxelization resolution r and the voxelization resolution r as the basic unit area of the voxel after slice voxelization; Step 7.2: Obtain the area S of the sliced point cloud data j : S j = n * r * r (10) Step 7.3: Obtain the volume V of the sliced point cloud data j : V j = S j * w (11) Step 7.4: Add V j to the point cloud volume V.

8. A method for measuring the volume of unordered point clouds based on voxelization and filling algorithms according to claim 7, characterized in that: Step 8 specifically includes the following steps: Step 8.1: Increase the slice increment value: a = w * j (12) where j is the number of slices; Step 8.2: Determine the incremental condition. If at this time Z min +a > Z max , then the algorithm process is completed, and the point cloud volume V is returned; otherwise, go to Step 8.3; Step 8.3: Iterate steps 2 to 8 until the entire point cloud data is sliced.

Images