A bridge point cloud multi-view projection view point screening method based on view angle consistency

Abstract

The application discloses a bridge point cloud multi-view projection view point screening method based on view angle consistency, and steps include: 1) using a fast ground detection algorithm to obtain a visible point internal component visible prompt information proportion; 2) based on the component visible prompt information proportion, calculating the component visibility rate; 3) selecting a view point with a component visibility rate greater than a preset threshold, and calculating the view angle consistency rate of the selected view point; 4) based on the view angle consistency rate, screening out the view point with the view angle consistency. The application evaluates the visible information of the visible point that can be obtained by the view point, and finally realizes the view point screening according to the view angle consistency between the view points, so that the view point capable of clearly presenting various components is obtained, and serves a downstream multi-view-based segmentation task.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of information processing, and particularly relates to a bridge point cloud multi-view projection view point screening method based on consistent view angles. BACKGROUND

[0002] With the rapid development of bridge engineering and the increase in the number of aging bridges, safety detection and structural health monitoring of bridges have become increasingly important. Point cloud data, as a high-precision three-dimensional measurement technology, has been widely used in bridge structure detection and component identification. However, the complexity and large amount of data of point cloud data make it challenging to extract effective bridge information directly from three-dimensional bridge point clouds.

[0003] Therefore, using multi-view projection to obtain point cloud multi-view, and using a two-dimensional multi-view segmentation network or a visual large model to realize rapid segmentation of bridge components has become a current research hotspot. However, existing point cloud multi-view projection methods only generate uniform ring or spherical view points centered on point clouds, without considering the screening of occluded view points. These methods are difficult to deal with complex bridge structures, and are prone to problems such as view angle occlusion and unclear component identification, which affect the extraction of component information and the accuracy of subsequent segmentation. Therefore, a bridge point cloud multi-view projection view point screening method that can ensure component visibility is needed. SUMMARY

[0004] The purpose of the present application is to provide a bridge point cloud multi-view projection view point screening method based on consistent view angles, comprising the following steps:

[0005] 1) generating a fixed number of candidate view points uniformly centered on the center point of the bridge point cloud;

[0006] 2) voxelizing the bridge point cloud and recording the index and occupancy state of each voxel;

[0007] 3) calculating the voxel center coordinates of the occupied voxels and recording the correspondence between the occupied voxels and the points contained therein;

[0008] 4) confirming the visible voxels and obtaining the corresponding visible points of each visible voxel based on the ray traversal voxel algorithm;

[0009] 5) obtaining the proportion of component visibility cues in the visible points using a fast ground detection algorithm;

[0010] 6) calculating the component visibility rate based on the proportion of component visibility cues;

[0011] 7) selecting view points with a component visibility rate greater than a preset threshold and calculating the view angle consistency rate of the selected view points;

[0012] 8) screening view points with consistent view angles based on the view angle consistency rate.

[0013] Further, in step 1), the step of uniformly generating a fixed number of candidate viewpoints includes:

[0014] 1.1) Standardize the bridge point cloud;

[0015] 1.2) Set the elevation angle to 0° and the horizontal angle to 360°×(i-1) / N at a distance of d units from the center point of the point cloud with the center point of the point cloud as the rotation center and the Z-axis of the bridge point cloud as the central axis, to obtain N candidate viewpoints of the uniform ring point cloud, and calculate the three-dimensional coordinates u of each viewpoint according to the elevation angle and the horizontal angle. i =(u ix ,u iy ,u iz ). i is the serial number of the candidate viewpoint.

[0016] Further, in step 2), the step of voxelizing the bridge point cloud and recording the index and occupancy state of each voxel includes:

[0017] 2.1) Voxelize the standardized bridge point cloud and divide the point cloud into a voxelization grid G with a voxel size of voxel_size;

[0018] 2.2) Record the index and occupancy state o i ∈{0,1} of each spatial voxel; if o i =1, it is considered that the voxel is an occupied voxel and there is a three-dimensional point in the voxel; otherwise, o i =0, it is considered that the voxel is an empty voxel and there is no three-dimensional point in the voxel.ix,iy,iz are the coordinates of the index in the x,y,z directions.

[0019] Further, in step 3), the voxel center coordinates of the occupied voxel are v j =(v jx ,v jy ,v jz ), and the corresponding relationship between the occupied voxel and the points contained therein is vp j ={p k |p k ∈P and p k in v j}. P is the bridge point cloud.

[0020] Further, in step 4), based on the ray traversal voxel algorithm, the step of confirming the visible voxel and obtaining the corresponding visible point of each visible voxel includes:

[0021] 4.1) Construct a voxel as the starting point of the light v j =(v jx ,v jy ,v jz), to v j -u i For light ray advancing direction, the light ray r ji is initialized as

[0022] 4.2) Calculate the time interval Δt X required for the light ray to advance one voxel along three axis directions jx = voxel_size / (v ix -u Y ), Δt jy = voxel_size / (v iy -u Z ), Δt jz = voxel_size / (v iz -u X ), and initialize the time when the light ray enters the next voxel along three axis directions, i.e., let T X ← Δt Y , T Y ← Δt Z , T Z ← Δt jx ;

[0023] 4.3) According to the light ray direction, obtain the index increments of the voxel along three axes Δx = sign(v ix -u jy ), Δy = sign(v iy -u jz ), Δz = sign(v iz -u ji );

[0024] If the index of the voxel where the current light ray is located is within the voxel grid G, then go to step 4.4), otherwise, obtain all voxels r X through which the light ray passes and go to step 4.5);

[0025] 4.4) If min{T Y , T Z , T X} = T X , then update the X-axis index of the current accessed voxel ix = ix + Δx, update the next voxel access time of the X-axis T X = T X + Δt ji , and add the information of the current voxel to the set of voxels r X through which the light ray passes;

[0026] If min{T Y , T Z , T YThen the Y-axis index of the current voxel is updated iy=iy+Ay, and the next voxel access time in the Y-axis is updated T Y =T Y +At Y , and the current voxel information is added to the voxel set r ji ;

[0027] If min{T X , T Y , T Z} = T Z , then the Z-axis index of the current voxel is updated iz=iz+Az, and the next voxel access time in the Z-axis is updated T Z =T Z +At Z , and the current voxel information is added to the voxel set r ji , and step 4.4) is repeated;

[0028] 4.5) The occupancy state of all voxels r ji through which the light ray passes is counted. If , the voxel is a visible voxel; if , the voxel is a non-visible voxel;

[0029] 4.6) The points corresponding to all visible voxels are extracted as visible points P k .

[0030] Further, in step 5), the step of obtaining the proportion of visible cues of the internal structure in the visible points using the fast ground detection algorithm includes:

[0031] 5.1) The visible points are divided into different sectors S s in the XY plane, i.e.:

[0032]

[0033] wherein n c =(x c -x k , y c -y k , z c -z k ), n i =(x i -x k , y i -y k , z i -z k ); θ(n c , n i ) is the angle between two vectors, and the value range is [0, 2π); Δα is a preset sector angle; (x c, y c , z c ) is the center point coordinate of the bridge point cloud; (x k , y k , z k ) is the viewpoint coordinate; (x i , y i , z i ) is the coordinate of any point in the bridge point cloud;

[0034] The points in one sector are as follows:

[0035] P ks = {p i ∈ P k | s(p i ) = s} (2)

[0036] 5.2) The point cloud in the same sector is divided into different blocks according to the distance between the point and the viewpoint , if , then

[0037] 5.3) The two-dimensional representation of the lowest point of each block in the same sector is linearly fitted using a two-dimensional linear fitting method, and a straight line line s is obtained.

[0038] The two-dimensional representation of the lowest point of each block is as follows:

[0039]

[0040] 5.4) The distance of each point in each sector to the straight line line s is calculated, and the points with a distance less than a threshold value are taken as the component visual prompt information P kg .

[0041] Further, in step 6), the calculation formula of the component visibility is as follows:

[0042]

[0043] In the formula, r is the component visibility, P kg is the number of points of the component visual prompt information, and P k is the number of visible points.

[0044] Further, in step 7), the calculation formula of the viewpoint angle consistency rate is as follows:

[0045]

[0046] In the formula, r threshold is a preset threshold value. r i , ri-1 is the visibility rate of the component for the view point i, i-1; c is the view angle consistency rate.

[0047] Further, in step 8), the step of screening the view points with view angle consistency is: based on the component consistency rate of each view point, screening the view points with the component consistency rate in the interval [c low ,c high ] as the final screening points U selected ={u j ,j=1,...,m}.

[0048] Further, the screening points are used for multi-view projection of the bridge point cloud; the multi-view projection of the bridge point cloud is used for semantic segmentation of the bridge point cloud, component identification and / or reconstruction of the bridge completion model.

[0049] The technical effect of the present application is self-evident. The present application evaluates the visible information of the view points that can be obtained by the view points, and finally realizes view point screening according to the view angle consistency between the view points, so as to obtain view points that can clearly present various components, serving downstream segmentation tasks based on multi-view.

[0050] The present application reduces the calculation cost of view point corresponding visible point extraction by voxelizing the point cloud; realizes visible point extraction by extracting visible voxels, effectively solves the extraction of view point visible points under the premise of no semantic information; obtains the proportion of component visible prompt information by rapid ground detection algorithm detection, which can facilitate the calculation of view point component visibility, and calculate the view angle consistency of the view point based on the change of the component visible prompt information of the adjacent view angle. By screening the view points with certain component visibility according to the view angle consistency, the view points that do not have irrelevant information (such as retaining wall, etc.) shielding and clearly provide component information can be screened out, which is beneficial to the improvement of subsequent segmentation accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0051] Figure 1 A fixed number of candidate view points provided by the embodiment of the present application are shown in the schematic diagram;

[0052] Figure 2 A bridge point cloud voxelization schematic diagram provided by the embodiment of the present application is shown in the schematic diagram;

[0053] Figure 3 A point cloud voxel occupancy state and visible state schematic diagram provided by the embodiment of the present application is shown in the schematic diagram;

[0054] Figure 4 A view point and some voxel center light ray schematic diagram provided by the embodiment of the present application is shown in the schematic diagram;

[0055] Figure 5 A view point corresponding visible point schematic diagram provided by the embodiment of the present application is shown in the schematic diagram;

[0056] Figure 6 A schematic diagram of a detection result of visible prompt information of a component in a certain view point provided for an embodiment of the present application is shown in FIG. 6.

[0057] Figure 7 A schematic diagram of a selection result of a bridge point cloud projection view angle provided for an embodiment of the present application is shown in FIG. 7. DETAILED DESCRIPTION

[0058] The present application is further described below in conjunction with embodiments, but should not be understood as limiting the above-mentioned subject matter of the present application to the following embodiments. Various substitutions and modifications can be made according to ordinary technical knowledge and conventional means in the art without departing from the above-mentioned technical idea of the present application, and all of them should be included in the protection scope of the present application.

[0059] Embodiment 1

[0060] Referring to Figures 1 to 7 A bridge point cloud multi-view projection view point screening method based on view angle consistency includes the following steps:

[0061] 1) A fixed number of candidate view points are uniformly generated with the center point of the bridge point cloud as the center;

[0062] 2) The bridge point cloud is voxelized, and the index and occupancy state of each voxel are recorded;

[0063] 3) The voxel center coordinates of the occupied voxels are calculated, and the correspondence between the occupied voxels and the points contained therein is recorded;

[0064] 4) Based on the ray traversal voxel algorithm, the visible voxels are confirmed, and the corresponding visible points of each visible voxel are obtained;

[0065] 5) The visible prompt information proportion of the components in the visible points is obtained by using a fast ground detection algorithm;

[0066] 6) The component visibility rate is calculated based on the component visible prompt information proportion;

[0067] 7) The view points with a component visibility rate greater than a preset threshold are selected, and the view angle consistency rate of the selected view points is calculated;

[0068] 8) The view points with view angle consistency are screened based on the view angle consistency rate.

[0069] In step 1), the step of uniformly generating a fixed number of candidate view points includes:

[0070] 1.1) Standardize the bridge point cloud;

[0071] 1.2) Set the elevation angle to 0° and the horizontal angle to 360°x(i-1) / N at a distance of d units from the center point of the point cloud, taking the center point of the point cloud as the center of rotation and the Z-axis of the bridge point cloud as the central axis, to obtain N candidate viewpoints u of the uniform ring point cloud i = (u ix , u iy , u iz ). i is the serial number of the candidate viewpoint. u ix , u iy , u iz are the coordinates in the x, y, and z directions;

[0072] In step 2), the bridge point cloud is voxelized, and the index and occupancy state of each voxel are recorded, including:

[0073] 2.1) Voxelize the standardized bridge point cloud, and divide the point cloud into a voxelized grid G with a voxel size of voxel_size;

[0074] 2.2) Record the index and occupancy state o i ∈{0,1} of each spatial voxel; if o i =1, it is considered that the voxel is an occupied voxel, and there is a three-dimensional point in the voxel; otherwise, o i =0, it is considered that the voxel is an empty voxel, and there is no three-dimensional point in the voxel. ix, iy, iz are the coordinates of the index in the x, y, and z directions.

[0075] In step 3), the voxel center coordinates of the occupied voxel are v j = (v jx , v jy , v jz ), and the correspondence between the occupied voxel and the points contained therein is vp j = {p k |p k ∈P and p k in v j}. P is the bridge point cloud.

[0076] In step 4), the visible voxels are confirmed, and the corresponding visible points of each visible voxel are obtained, including:

[0077] 4.1) Construct a light ray r j = (v jx , v jy , v jz ) with the voxel as the starting point v j -u i as the direction of the light ray, and initialize the light ray r ji currently passing through the voxel to

[0078] 4.2) Calculate the time interval Δt required for the light ray to advance one voxel along the three axes X = voxel_size / (v jx -u ix ), Δt Y = voxel_size / (v jy -u iy ), Δt Z = voxel_size / (v jz -u iz ), and initialize the time for the light ray to enter the next voxel along the three axes, i.e., let T X ← Δt X , T Y ← Δt Y , T Z ← Δt Z ;

[0079] 4.3) According to the direction of the light ray, obtain the index increments of the voxel along the three axes Δx = sign(v jx -u ix ), Δy = sign(v jy -u iy ), Δz = sign(v jz -u iz );

[0080] If the index of the voxel where the current light ray is located is within the voxel grid G, then go to step 4.4), otherwise, obtain all voxels r ji through which the light ray passes and go to step 4.5);

[0081] 4.4) If min{T X , T Y , T Z} = T X , then update the X-axis index of the currently accessed voxel ix = ix + Δx, update the next voxel access time along the X-axis T X = T X + Δt X , and add the information of the current voxel to the light ray r ji ;

[0082] If min{T X , T Y , T Z} = T Y , then update the Y-axis index of the currently accessed voxel iy = iy + Δy, update the next voxel access time along the Y-axis T Y = T Y + Δt Y , and add the information of the current voxel to the light ray r ji ;

[0083] If min{T X ,T Y ,T Z} = T Z , the Z-axis index of the current access voxel is updated as iz = iz + Δz, the next voxel access time of the Z-axis is updated as T Z = T Z + Δt Z , and the current voxel information is added to the ray r ji , and step 4.2) is returned.

[0084] 4.5) Count the occupancy state of all voxels r ji through which the ray passes, if , the voxel is a visible voxel; if , the voxel is an invisible voxel.

[0085] 4.6) Extract the points corresponding to all visible voxels as visible points P k .

[0086] In step 5), the steps for obtaining the proportion of visible cues in the visible points include:

[0087] 5.1) Divide the visible points into different sectors S s in the XY plane, that is:

[0088]

[0089] where n c = (x c -x k , y c -y k , z c -z k ), n i = (x i -x k , y i -y k , z i -z k ); θ(n c , n i ) is the angle between two vectors, and its value range is [0, 2π); Δα is a preset sector angle; (x c , y c , z c ) is the center point coordinate of the bridge point cloud; (x k , y k , z k ) is the viewpoint coordinate; (x i , y i , z i() represents the coordinates of any point in the bridge point cloud;

[0090] The points in a sector are shown below:

[0091] P ks ={p i ∈P k |s(p i )=s} (2)

[0092] 5.2) Divide the point cloud within the same sector into different blocks based on the distance between the points and the viewpoint. In the middle, when dividing, if but

[0093] 5.3) Using a two-dimensional line fitting method, the lowest point of each block within the same sector is fitted with a straight line to obtain the line. s ;

[0094] The lowest point of each block is shown below:

[0095]

[0096] In the formula, take Calculate the lowest point of the z-coordinate p′ i .

[0097] 5.4) Calculate the distance from each point in each sector to the line. s The distance is used to select points whose distance is less than a threshold as the component's visual prompt information P. kg .

[0098] In step 6), the component visibility is as follows:

[0099]

[0100] In the formula, r represents the visibility of the component.

[0101] In step 7), the viewpoint consistency rate is as follows:

[0102]

[0103] In the formula, r threshold This is a preset threshold. i r i-1 denoted as , where is the component visibility rate at viewpoints i and i-1; c is the viewpoint consistency rate.

[0104] In step 8), the step of selecting viewpoints with visual consistency is as follows: based on the component consistency rate of each viewpoint, select viewpoints whose component consistency rate falls within the interval [c]. low ,c highThe view point inside the bridge point cloud as the final screening point U selected = {u j , j = 1, …, m}.

[0105] The screening point is used for bridge point cloud multi-view projection; the bridge point cloud multi-view projection is used for bridge point cloud semantic segmentation, bridge component identification and / or bridge completion model reconstruction.

[0106] Embodiment 2:

[0107] A bridge point cloud multi-view projection view point screening method based on view angle consistency, comprising the following steps:

[0108] 1) Uniformly generate a fixed number of candidate view points around the bridge point cloud;

[0109] 2) Voxelize the bridge point cloud, record the index and occupation state of each voxel;

[0110] 3) Calculate the voxel center coordinates of the occupied voxels, and record the correspondence between the occupied voxels and the points contained therein;

[0111] 4) Based on the ray traversal voxel algorithm, confirm the visible voxels, and obtain the corresponding visible points of each visible voxel;

[0112] 5) Use the fast ground detection algorithm to obtain the component visible prompt information proportion in the visible point;

[0113] 6) Based on the component visible prompt information proportion, calculate the component visibility rate;

[0114] 7) Select the view points with a component visibility rate greater than a preset threshold, and calculate the view angle consistency rate of the selected view points;

[0115] 8) Based on the view angle consistency rate, screen out the view points with view angle consistency.

[0116] Embodiment 3:

[0117] A bridge point cloud multi-view projection view point screening method based on view angle consistency, the technical content is the same as that of embodiment 2, further, in step 1), the step of uniformly generating a fixed number of candidate view points comprises:

[0118] 1.1) Standardize the bridge point cloud;

[0119] 1.2) Take the point cloud center point as the rotation center, take the Z axis of the bridge point cloud as the central axis, set the elevation angle to 0° and the horizontal angle to 360°×(i-1) / N at a distance of d units from the point cloud, to obtain N candidate view points u i = (u ix , u iy , u iz ).

[0120] Embodiment 4:

[0121] A bridge point cloud multi-view projection view point screening method based on consistent view angles, the technical content is the same as any one of embodiments 2-3, further, in step 2), the bridge point cloud is voxelized, and the step of recording the index and occupancy state of each voxel includes:

[0122] 2.1) voxelizing the standardized bridge point cloud, and dividing the point cloud into a voxelized grid G with a voxel size of voxel_size;

[0123] 2.2) recording the index of each spatial voxel and the occupancy state o i ∈{0,1}; if o i =1, it is considered that the voxel is an occupied voxel, and there is a three-dimensional point in the voxel; otherwise, o i =0, it is considered that the voxel is an empty voxel, and there is no three-dimensional point in the voxel.

[0124] Embodiment 5:

[0125] A bridge point cloud multi-view projection view point screening method based on consistent view angles, the technical content is the same as any one of embodiments 2-4, further, in step 3), the voxel center coordinates of the occupied voxel are v j =(v jx ,v jy ,v jz ), and the corresponding relationship between the occupied voxel and the points contained therein is vp j ={p k |p k ∈P and p k in v j}.

[0126] Embodiment 6:

[0127] A bridge point cloud multi-view projection view point screening method based on consistent view angles, the technical content is the same as any one of embodiments 2-5, further, in step 4), the step of confirming the visible voxel and obtaining the corresponding visible point of each visible voxel includes:

[0128] 4.1) constructing a light ray r ji with the voxel as the starting point v j =(v jx ,v jy ,v jz ) and the direction v j -u i , and initializing the light ray r ji currently passing through the voxel as

[0129] 4.2) Calculate the time interval Δt required for the ray to advance one voxel along the three axes X = voxel_size / (v jx -u ix ), Δt Y = voxel_size / (v jy -u iy ), Δt Z = voxel_size / (v jz -u iz ), and initialize the time for the ray to enter the next voxel along the three axes, i.e., let T X ← Δt X , T Y ← Δt Y , T Z ← Δt Z ;

[0130] 4.3) According to the direction of the ray, obtain the index increments of the voxel along the three axes Δx = sign(v jx -u ix ), Δy = sign(v jy -u iy ), Δz = sign(v jz -u iz );

[0131] If the index of the voxel where the current ray is located is within the voxel grid G, then go to step 4.4), otherwise, obtain all voxels r ji through which the ray passes and go to step 4.5);

[0132] 4.4) If min{T X , T Y , T Z} = T X , then update the X-axis index of the currently visited voxel ix = ix + Δx, update the next voxel access time along the X-axis T X = T X + Δt X , and add the information of the current voxel to the ray r ji ;

[0133] If min{T X , T Y , T Z} = T Y , then update the Y-axis index of the currently visited voxel iy = iy + Δy, update the next voxel access time along the Y-axis T Y = T Y + Δt Y , and add the information of the current voxel to the ray r ji ;

[0134] If min{T X ,T Y ,T Z}=T Z , then the Z-axis index of the current access voxel is updated as iz=iz+Δz, the Z-axis next voxel access time is updated as T Z =T Z +Δt Z , and the current voxel information is added to the ray r ji , and step 4.2) is returned.

[0135] 4.5) Count the occupancy state of all voxels r ji through which the ray passes, if , the voxel is a visible voxel; if , the voxel is an invisible voxel.

[0136] 4.6) Extract the points corresponding to all visible voxels as visible points P k .

[0137] Embodiment 7:

[0138] A bridge point cloud multi-view projection view point screening method based on consistent view angles, the technical content is the same as any one of embodiments 2-6, further, in step 5), the step of obtaining the proportion of visible point internal member visible prompt information includes:

[0139] 5.1) Divide the visible points into different sectors S s in the XY plane, that is:

[0140]

[0141] Wherein, n c =(x c -x k ,y c -y k ,z c -z k ), n i =(x i -x k ,y i -y k ,z i -z k ); θ(n c ,n i ) is the angle between two vectors, and the value range is [0, 2π);

[0142] The points in a sector are as follows:

[0143] P ks ={p i ∈Pk |s(p i )=s} (2)

[0144] 5.2)Divide the point cloud in the same sector into different blocks according to the distance between the point and the viewpoint , when dividing, if , then

[0145] 5.3) Use a two-dimensional linear fitting method to perform linear fitting on the lowest point of each block in the same sector to obtain a straight line line s .

[0146] Wherein, the lowest point of each block is as follows:

[0147]

[0148] 5.4) Calculate the distance from each point in each sector to the straight line line s , and take the points with a distance less than a threshold value as the component visible prompt information P kg .

[0149] Example 8:

[0150] A bridge point cloud multi-view projection viewpoint screening method based on consistent view angles, the technical content of any one of embodiments 2-7, further, in step 6), the component visibility is as follows:

[0151]

[0152] In the formula, r is the component visibility.

[0153] Example 9:

[0154] A bridge point cloud multi-view projection viewpoint screening method based on consistent view angles, the technical content of any one of embodiments 2-8, further, in step 7), the view angle consistency rate of the viewpoint is as follows:

[0155]

[0156] In the formula, r threshold is a preset threshold value.

[0157] Example 10:

[0158] A bridge point cloud multi-view projection viewpoint screening method based on consistent view angles, the technical content of any one of embodiments 2-9, further, in step 8), the step of screening out the viewpoint with consistent view angles is: based on the component consistency rate of each viewpoint, screening out the component consistency rate in the interval [c low ,c highThe view point inside the bridge point cloud as the final screening point U selected = {u j j = 1,..., m}.

[0159] Embodiment 11:

[0160] A bridge point cloud multi-view projection view point screening method based on view angle consistency, the technical content is the same as any one of embodiments 2-10, further, the screening point is used for bridge point cloud multi-view projection, serving bridge point cloud semantic segmentation, component identification and bridge completion model reconstruction and other applications.

[0161] Embodiment 12:

[0162] A bridge point cloud multi-view projection view point screening method based on view angle consistency, the steps are as follows:

[0163] S101, taking the bridge point cloud as the center, uniformly generating a fixed number of candidate view points;

[0164] S102, voxelizing the bridge point cloud, recording the index and occupation state of each voxel;

[0165] S103, calculating the voxel center coordinates of the occupied voxels, and recording the information of all points in the voxels;

[0166] S104, based on the ray traversal voxel algorithm, confirming the visible voxels, and obtaining the visible points corresponding to each view point;

[0167] S105, using a fast ground detection algorithm to obtain the proportion of visible component prompt information in the visible points;

[0168] S106, based on the detected proportion of visible component prompt information, calculating the visible rate of the component;

[0169] S107, selecting the view points with a visible rate of the component greater than a certain threshold, and calculating the view angle consistency rate of each view point;

[0170] S108, according to the view angle consistency rate obtained by S107, screening out the view points with view angle consistency.

[0171] In the present application, the cost of extracting the visible point corresponding to the viewpoint is reduced by voxelizing the point cloud; the extraction of the visible point is realized by extracting the visible voxel, effectively solving the extraction of the visible point of the viewpoint under the premise of no semantic information; the proportion of the visible prompt information of the component is obtained by the rapid ground detection algorithm, which can facilitate the calculation of the visibility of the viewpoint, and the viewpoint consistency is calculated based on the change of the visible prompt information of the adjacent view angle. By selecting the view angle consistency of the viewpoint with certain component visibility, the viewpoint that provides clear component information without irrelevant information (such as retaining wall) shielding can be screened out, which is beneficial to the improvement of subsequent segmentation accuracy.

[0172] S101, generating a fixed number of candidate viewpoints uniformly around the bridge point cloud as the center;

[0173] In specific implementation, first, the bridge point cloud is standardized, and the center point of the point cloud is taken as the rotation center, the Z-axis of the bridge point cloud is taken as the central axis, the elevation angle is set to 0° at a distance of d=2 units from the point cloud, and the horizontal angle is set to 360°×(i-1) / 20 in turn, to obtain 20 candidate viewpoints u i ix iy iz around the point cloud, as shown in Figure 1 .

[0174] S102, voxelizing the bridge point cloud and recording the index and occupancy state of each voxel;

[0175] In specific implementation, the standardized bridge point cloud is voxelized, and the point cloud is divided into a voxel grid G with a voxel size of voxel_size=0.01. The index and occupancy state o i ∈{0,1} of each spatial voxel are recorded; if o i =1, it is considered that the voxel is an occupied voxel, and the voxel contains three-dimensional points; otherwise, o i =0, it is considered that the voxel is an empty voxel, and there are no three-dimensional points in it. The occupied voxel of the bridge point cloud in the implementation example is shown in Figure 2 .

[0176] S103, calculating the voxel center coordinates of the occupied voxel and recording the information of all points in the voxel;

[0177] In specific implementation, for each occupied voxel, the voxel center coordinates v j = (v jx , v jy , v jz ) are calculated, and the corresponding relationship vp j = {p k ​​​|p k ∈P and p k in v j}

[0178] S104. Based on the ray traversal voxel algorithm, identify the visible voxels and obtain the viewpoint corresponding to each viewpoint;

[0179] In practice, for each viewpoint u i =(u ix ,u iy ,u iz The corresponding visible voxels are extracted using the ray traversal voxel algorithm, thereby realizing the extraction of visible points.

[0180] Visual voxel extraction: Assuming there is a viewpoint u i =(u ix ,u iy ,u iz ), construct its relationship with the occupied voxel v j =(v jx ,v jy ,v jz The light between ( ).

[0181] Starting with voxels j =(v jx ,v jy ,v jz ), with v j -u i Let r be the direction of the light ray. ji The current voxel is initialized to Calculate the time interval Δt required for a ray to travel one voxel along each of the three axes. X =0.01 / (v jx -u ix ), Δt Y =0.01 / (v jy -u iy ), Δt Y =0.01 / (v jz -u iz And initialize T with the time for entering the next voxel in the current three directions. X ←Δt X T Y ←Δt Y T Z ←Δt Z Simultaneously, based on the direction of the light, the index increment Δx = sign(v) of the voxel along the three axes is obtained. jx -u ix ), Δy = sign(v jy -u iy), Δz = sign(v jz . -u iz .

[0182] If the current ray is still in the voxel grid G, compare T X , T Y , T Z , the next access voxel coordinate is obtained, specifically as follows:

[0183] If min{T X , T Y , T Z} = T X , the X-axis index of the current access voxel is updated ix = ix + Δx, the next voxel access time on the X-axis is updated T X = T X + Δt X , and the current voxel information is added to the ray r ji ;

[0184] If min{T X , T Y , T Z} = T Y , the Y-axis index of the current access voxel is updated iy = iy + Δy, the next voxel access time on the Y-axis is updated T Y = T Y + Δt Y , and the current voxel information is added to the ray r ji ;

[0185] If min{T X , T Y , T Z} = T Z , the Z-axis index of the current access voxel is updated iz = iz + Δz, the next voxel access time on the Z-axis is updated T Z = T Z + Δt Z , and the current voxel information is added to the ray r ji

[0186] Repeat the above ray access voxel obtaining process until the current ray is outside the voxel grid G, and obtain all voxels r ji through which the ray passes.

[0187] Statistical occupancy state of all voxels r ji through which the ray passes, if , the voxel is a visible voxel; if , the voxel is an invisible voxel.

[0188] Visible point extraction: for each viewpoint u i = (uix ,u iy ,u iz ), the visible point P k .

[0189] S105, using a fast ground detection algorithm, obtaining the visible point inside the member visible prompt information proportion;

[0190] In specific implementation, for each view point P k , using a fast ground detection method, obtaining the visible point inside the bridge member visible prompt information proportion.

[0191] First, according to formula (1) the visible point is divided into different sectors S s of XY plane:

[0192]

[0193] Wherein, n c =(x c -x k ,y c -y k ,z c -z k ), n i =(x i -x k ,y i -y k ,z i -z k ); θ(n c ,n i ) is the angle between two vectors, the value range is [0, 2π). Therefore, the point in a sector can be recorded as:

[0194] P ks ={p i ∈P k |s(p i )=s} (2)

[0195] In the same sector, the point cloud is divided into different blocks according to the distance between the point and the view point, that is Then p i ∈P ks .

[0196] At the same time, the lowest point of each block is selected and represented as (3) formula form:

[0197]

[0198] The lowest point of each block in the same sector is fitted with a straight line using a two-dimensional straight line fitting method to obtain a straight line line s The distance of each point in each sector to the straight line line s is calculated, and the points with a distance less than a threshold value 0.01 are considered as the component visual prompt information P kg .

[0199] S106, based on the detected component visual prompt information proportion, calculating the component visual rate;

[0200] In specific implementation, for each visual point, based on the component visual prompt information obtained by S105, the component visual rate is calculated using formula (4):

[0201]

[0202] S107, selecting the visual point with a component visual rate greater than a certain threshold value, and calculating the visual angle consistency rate of each visual point;

[0203] In specific implementation, the visual point with a component visual rate greater than r threshold =0.2 is selected, and the visual angle consistency rate of the visual point is calculated using formula (5):

[0204]

[0205] S108, according to the visual angle consistency rate obtained by S107, screening out the visual point with visual angle consistency.

[0206] In specific implementation, according to the component consistency rate of each visual point calculated by S107, the visual point with a component consistency rate in the interval [0.2, 5] is screened out as the final screening point U selected ={u j , j=1,..., m}.

Claims

1. A method for view point screening of bridge point cloud multi-view projection based on view consistency, characterized in that, The method comprises the following steps: 1) generating a fixed number of candidate viewpoints uniformly around the center point of the bridge point cloud; 2) voxelizing the bridge point cloud and recording the index and occupancy state of each voxel; 3) calculating the voxel center coordinates of the occupied voxels and recording the correspondence between the occupied voxels and the points contained therein; 4) confirming the visible voxels and obtaining the corresponding visible points of each visible voxel based on the ray traversal voxel algorithm; 5) obtaining the visible component prompt information in the visible point by using a fast ground detection algorithm; 6) calculating the component visibility rate based on the component visible prompt information; 7) selecting the viewpoints with a component visibility rate greater than a preset threshold and calculating the view angle consistency rate of the selected viewpoints; 8) screening the viewpoints with view angle consistency based on the view angle consistency rate; In step 5), the step of obtaining the visible component prompt information in the visible point by using a fast ground detection algorithm comprises: 5.1) dividing the visible points into different sectors S of the XY plane s i.e.: wherein n c =(x c -x k ,y c -y k ,z c -z k ), n i =(x i -x k ,y i -y k ,z i -z k ); θ(n c ,n i ) is the angle between two vectors, and the value range is [0, 2π); s(p i ) is the sector to which the visible point p i belongs; Δα is a preset sector angle; (x c , y c , z c ) is the center point coordinate of the bridge point cloud; (x k , y k , z k ) is the view point coordinate; (x i , y i , z i ) is the coordinate of any point in the bridge point cloud. The points in a sector are as follows: P ks = {p i ∈ P k | s(p i ) = s} (2) wherein P ks is a point set in a sector; 5.2) Divide the point cloud in the same sector into different blocks according to the distance between the point and the viewpoint In the division, if then 5.3) Using a two-dimensional linear fitting method to fit a line to the two-dimensional representation of the lowest points of each block in the same sector, obtaining the line line s ; The two-dimensional representation of the lowest point of each block is as follows: 5.4) Calculate the distance of each point in each sector to the line line s and use the points with a distance less than a threshold as visual cues for the component. In step 6), the calculation formula of the component visibility rate is as follows: where r is the component visibility, P kg is the number of points of the component visible cue information, P k is the number of points of the visible point; In step 7), the calculation formula of the view angle consistency rate of the viewpoint is as follows: In the formula, r threshold is a preset threshold; r i , r i-1 is the component visibility rate of the viewpoint i, i-1; and c is the angle consistency rate.

2. The view point screening method for bridge point cloud multi-view projection based on consistent view angles according to claim 1, characterized in that, In step 1), the step of uniformly generating a fixed number of candidate viewpoints comprises: 1.1) standardizing the bridge point cloud; 1.2) Set the elevation angle to 0° and the horizontal angle to 360°x(i-1) / N at a distance of d units from the center point of the point cloud, with the Z-axis of the bridge point cloud as the central axis, to obtain N candidate viewpoints of the uniform ring point cloud, and calculate the three-dimensional coordinates of each viewpoint u i ix iy iz ; i is the serial number of the candidate viewpoint; u ix iy iz is the coordinate in the x, y, and z directions.​​​​​ 3. The view point screening method for bridge point cloud multi-view projection based on consistent view angles according to claim 1, characterized in that, In step 2), the step of voxelizing the bridge point cloud and recording the index and occupancy state of each voxel comprises: 2.1) voxelizing the standardized bridge point cloud and dividing the point cloud into a voxelized grid G with a voxel size of voxel_size; 2.2) Record the index of each spatial voxel and the occupancy state o i ∈ {0, 1}; if o i = 1, it is considered that the voxel is an occupied voxel, and there is a three-dimensional point in the voxel; otherwise o i = 0, it is considered that the voxel is an empty voxel, and there is no three-dimensional point in the voxel; ix, iy, iz are the coordinates of the index in the x, y, z directions.

4. The view point screening method for bridge point cloud multi-view projection based on consistent view angles according to claim 1, characterized in that, In step 3), the voxel center coordinates of the occupied voxel are v j =(v jx , v jy , v jz ), the correspondence between the occupied voxel and the points contained in the interior is vp j ={p k |p k ∈P and p k in v j}; P is a bridge point cloud.

5. The view point screening method for bridge point cloud multi-view projection based on consistent view angles according to claim 1, characterized in that, In step 4), the step of confirming the visible voxels and obtaining the corresponding visible points of each visible voxel based on the ray traversal voxel algorithm comprises: 4.1) Construct a ray with origin at the voxel center v j = (v jx , v jy , v jz ) and direction v j - u i , and initialize the ray r ji to be currently passing through the voxel v = (v i , v ix , v iy ) and u iz = (u i , u ix , u iy ). 4.2) Calculate the time interval Δt required for the ray to advance along the three axes by one voxel X = voxel_size / (v jx - u ix ), Δt Y = voxel_size / (v jy - u iy ), Δt Z = voxel_size / (v jz - u iz ), and initialize the time to enter the next voxel along the three axes, i.e. set T X ← Δt X , T Y ← Δt Y , T Z ← Δt Z ; 4.3) depending on the direction of light ray advancement, get the index increments of the voxel along the three axes Δx = sign(v jx -u ix ), Δy = sign(v jy -u iy ), Δz = sign(v jz -u iz ); u ix , u iy , u iz are the coordinates of the candidate point in the x, y, z directions; If the current ray is in the voxel grid G, then go to step 4.4), otherwise, get all the voxels r that the ray passes through ji and go to step 4.5); 4.4) If min{T X ,T Y ,T Z} = T X , then the X-axis index of the current access voxel is updated ix = ix + Δx, the X-axis next voxel access time is updated T X = T X + Δt X , and the current voxel information is added to the voxel set r ji through which the ray passes. If min{T X ,T Y ,T Z} = T Y , then the Y-axis index of the current access voxel is updated iy = iy + Δy, the Y-axis next voxel access time is updated T Y = T Y + Δt Y , and the current voxel information is added to the voxel set r ji through which the light ray passes. If min{T X ,T Y ,T Z} = T Z , then the Z-axis index of the current access voxel is updated as iz = iz + Δz, the next voxel access time in the Z-axis is updated as T Z = T Z + Δt Z , the current voxel information is added to the voxel set r ji that the ray passes through, and step 4.4) is repeated. 4.5) Count all voxels r through which the ray passes ji if the occupancy state of r is then r is a visible voxel; if the occupancy state of r is then r is a non-visible voxel; 4.6) Extract all the points corresponding to the visible voxels as visible points P k .

6. The view point screening method for bridge point cloud multi-view projection based on consistent view angles according to claim 1, characterized in that, In step 8), the step of screening the viewpoints with consistent view angles is: based on the component consistency rate of each viewpoint, screening the viewpoints with the component consistency rate in the interval [c low ,c high ] as the final screening points U selected ={u j ,j=1,...,m}.

7. The view point screening method for bridge point cloud multi-view projection based on consistent view angles according to claim 1, characterized in that: The screening points are used for bridge point cloud multi-view projection; the bridge point cloud multi-view projection is used for bridge point cloud semantic segmentation, component recognition and / or bridge completion model reconstruction.

Images