CN109785261A - A kind of airborne LIDAR three-dimensional filtering method based on gray scale volume element model - Google Patents
A kind of airborne LIDAR three-dimensional filtering method based on gray scale volume element model Download PDFInfo
- Publication number
- CN109785261A CN109785261A CN201910027077.0A CN201910027077A CN109785261A CN 109785261 A CN109785261 A CN 109785261A CN 201910027077 A CN201910027077 A CN 201910027077A CN 109785261 A CN109785261 A CN 109785261A
- Authority
- CN
- China
- Prior art keywords
- gray scale
- ground
- element model
- volume elements
- airborne lidar
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Image Processing (AREA)
Abstract
The present invention provides a kind of airborne LIDAR three-dimensional filtering method based on gray scale volume element model, is related to Remote Sensing Data Processing technical field.This method reads original airborne LIDAR point cloud data first, forms original airborne LIDAR point cloud data collection；Airborne LIDAR point cloud data rule is turned into gray scale volume element model, wherein gray scale is that the discretization of the mean intensity of laser point in volume elements indicates；The minimum characteristic of local elevation based on ground point chooses ground seed voxel, and then marks with its three-dimensional communication and the close non-zero value volume elements of voxel values, terrain slope is ground volume elements.Airborne LIDAR three-dimensional filtering method provided by the invention based on gray scale volume element model, based on three-dimensional communication region building theory, point cloud data filtering is used for using intensity and terrain slope information as auxiliary information, more effective information can be provided for the differentiation of ground and non-ground target, to improve filtering accuracy, and extends the three-dimensional filtering method based on volume element model and be suitable for more complicated scene.
Description
Technical field
The present invention relates to Remote Sensing Data Processing technical fields more particularly to a kind of based on the airborne of gray scale volume element model
LIDAR three-dimensional filtering method.
Background technique
Airborne laser radar (Light Detection and Ranging, i.e. LIDAR) data filtering is current point cloud number
Digital elevation model (Digital Elevation is directly affected according to the precision of the key technology in process field, filtering
Model, DEM) etc. the quality of products and subsequent such as building, vegetation target classification precision.Therefore, for LIDAR point
The research of cloud data filtering techniques is of great significance.
Domestic and foreign scholars have made intensive studies around point cloud data filtering, propose very more filtering algorithms, classical
Such as constructed based on interpolation, based on the gradient, based on mathematical morphology, based on three-dimensional communication set and filtering based on cluster segmentation
Method etc..Data structure used by these filtering algorithms includes grid grid, irregular triangle network (TIN), volume elements and discrete
Point cloud etc..Grid grid and TIN are 2.5 dimensional data structures, and expressing airborne LIDAR point cloud data with it will lead to information loss simultaneously
Further influence the integrality of the filter result based on such data structure.Space structure and topology information inside discrete point cloud
It is difficult to be utilized, leads to filtering algorithm difficult design, inefficiency based on such data structure；The ruler of each node of Octree
Very little different, the syntople between each node is difficult to set up, and this also increases the filtering algorithm designs based on such data structure
Difficulty.And the filtering algorithm based on voxel data structure can be very good to avoid the defect of the above method, this is because: (1)
Volume elements structure is true three-dimensional data structure；(2) adjacent and topology information is implied between its internal volume elements.Thus it is based on volume elements knot
The filtering algorithm design of structure is simpler.But the existing filtering algorithm based on voxel data structure is all made of two-value body at present
Meta structure (i.e. volume elements assignment is whether containing laser point to distinguish assignment 1 and 0 according in volume elements, wherein 1 value volume elements corresponds to target, 0
Value volume elements corresponds to background) (Liying Wang, Yan Xu, and Yu Li.Aerial LIDAR point cloud
voxelization with its 3D ground filtering application[J].Photogrammetric
Engineering and Remote Sensing, 2017,83 (2): 95-107).This filtering algorithm thinks ground target meeting
Three-dimensional communication set is formed, but when other targets (such as short vegetation, building facade) and ground target are connected to form three-dimensional connect
When logical set, which can not be effectively distinguish the two.In addition, the algorithm does not consider terrain slope information, thus lead
Cause the filtering accuracy of algorithm lower.
Summary of the invention
It is a kind of based on gray scale volume elements mould the technical problem to be solved by the present invention is in view of the above shortcomings of the prior art, provide
The airborne LIDAR three-dimensional filtering method of type solves the problems, such as effective differentiation on connected ground and non-ground target.
In order to solve the above technical problems, the technical solution used in the present invention is: a kind of machine based on gray scale volume element model
Carry LIDAR three-dimensional filtering method, comprising the following steps:
Step 1: reading original airborne LIDAR point cloud data, form original airborne LIDAR point cloud data collection；
Step 2: original airborne LIDAR point cloud data collection rule being turned into gray scale volume element model, wherein gray scale is in volume elements
The discretization expression of the mean intensity of laser point, method particularly includes:
Step 2.1: concentrating rejecting abnormalities data from original airborne LIDAR point cloud data, obtain rejecting abnormalities data set；
Step 2.1.1: the frequency that original airborne LIDAR point cloud data concentrates each laser point height value is counted, and with straight
The form visualization display statistical result of square figure；
Step 2.1.2: highest elevation threshold value T corresponding with real terrain and atural object is determinedheWith lowest elevation threshold value Tle；
Step 2.1.3: concentrating each laser point for original airborne LIDAR point cloud data, if its height value is higher than highest
Elevation threshold value TheOr it is lower than lowest elevation threshold value Tle, then the laser point is height anomaly data, is rejected, and otherwise retaining should
Laser point obtains and rejects height anomaly data set；
Step 2.1.4: statistics rejects the frequency of the intensity value of each laser point in height anomaly data set, and with histogram
Form visualization display statistical result；
Step 2.1.5: maximum intensity threshold value T corresponding with real terrain and atural object is determinedhiWith minimum intensity threshold value Tli；
Step 2.1.6: for each laser point in height anomaly data set is rejected, if its intensity value is higher than maximum intensity threshold
Value ThiOr it is lower than minimum intensity threshold value Thi, then the laser point is intensity abnormal data, it is rejected, otherwise retains the laser point,
Final obtain rejects elevation and intensity abnormal data set；
Step 2.2: rejecting abnormalities data set rule is turned into gray scale volume element model；
Step 2.2.1: the spatial dimension of data set is indicated with the axial parallel bounding box of rejecting abnormalities data set；
Step 2.2.2: calculated body element resolution ratio, that is, voxel size, the resolution ratio on x, y, z direction is according to rejecting abnormalities number
It is determined according to the equalization point spacing of concentration laser point；
Step 2.2.3: it is carried out according to axial parallel bounding box of the resolution ratio on x, y, z direction to rejecting abnormalities data set
Division obtains three-dimensional grid, each three-dimensional grid unit is known as volume elements；
Step 2.2.4: laser point each in rejecting abnormalities data set is mapped to three-dimensional grid, and then is wrapped according in volume elements
The Intensity attribute of the laser point contained is each volume elements assignment, and each voxel values discretization is finally obtained gray scale body to { 0 ..., 255 }
Meta-model；
Step 3: it is theoretical based on the building of three-dimensional communication region, the ground volume elements in gray scale volume element model is detected, is had
Body method are as follows:
Step 3.1: the minimum characteristic of local elevation based on ground point, the minimum non-zero value of elevation from gray scale volume element model
Volume elements is as ground seed voxel set；
Step 3.1.1: in the horizontal direction, piecemeal is carried out to gray scale volume element model using the Grid size of setting, and take
The minimum non-zero value volume elements of elevation is seed voxel in each piece；
Step 3.1.2: it is searched in each piece and is less than difference in height threshold value T with the difference in height of ground seedeNon-zero value volume elements be
Ground seed voxel obtains encrypted ground seed voxel set Vs, wherein s=1,2 ...；
Step 3.2: label ground seed voxel and with its three-dimensional communication and the close volume elements of the gradient, gray scale is constituted three
Dimension connected region is ground voxel data collection, completes the airborne LIDAR three-dimensional filtering based on gray scale volume element model；
Step 3.2.1: the frequency histogram of the gray value of non-zero value volume elements in gray scale volume element model is calculated, Gaussian Mixture is used
Models fitting gray scale frequency histogram determines grey value profile range corresponding with ground target；
Step 3.2.2: to any ground seed voxel Vs, traverse gray scale volume element model in Current terrestrial seed voxel Vs
Three-dimensional communication, gray value are located in the corresponding tonal range of ground target and local terrain slope is less than gradient threshold value TsIt is all
Unmarked volume elements, and it is labeled as Lg, until having marked all ground seed voxel VsThree-dimensional communication region, i.e. ground volume elements collection
It closes；Wherein, the determination of gradient threshold value is adaptive, method particularly includes: to current seed voxel Vs, detect in its spatial neighborhood
Whether marked ground volume elements is contained, if so, then the ruling grade value between existing ground volume elements is determined as in the neighborhood
Terrain slope threshold value Ts；Otherwise, 90 ° are set by terrain slope threshold value.
The beneficial effects of adopting the technical scheme are that provided by the invention a kind of based on gray scale volume element model
Airborne LIDAR three-dimensional filtering method, by the building of three-dimensional communication region it is theoretical based on, using intensity and terrain slope information as
Auxiliary information is filtered for point cloud data, more effective information can be provided for the differentiation of ground and non-ground target, to improve
Filtering accuracy, and extend the three-dimensional filtering method based on volume element model and be suitable for more complicated scene.
Detailed description of the invention
Fig. 1 is a kind of airborne LIDAR three-dimensional filtering method based on gray scale volume element model provided in an embodiment of the present invention
Flow chart；
Fig. 2 is original airborne LIDAR point cloud data schematic diagram provided in an embodiment of the present invention, wherein (a) is Csite1 point
Cloud data (b) are Csite2 point cloud data, (c) are Csite3 point cloud data, are (d) Csite4 point cloud data, (e) are
Fsite5 point cloud data (f) is Fsite6 point cloud data, (g) is Fsite7 point cloud data；
Original airborne LIDAR point cloud data collection rule is turned to gray scale volume element model to be provided in an embodiment of the present invention by Fig. 3
Specific flow chart；
Fig. 4 is the schematic diagram that laser point provided in an embodiment of the present invention projection calculates convex hull and area；
Fig. 5 is the flow chart provided in an embodiment of the present invention that volume elements detection in ground is carried out to gray scale said three-dimensional body metadata set；
Fig. 6 is the ash of the non-zero value volume elements in the corresponding said three-dimensional body variable matrix V of sample samp41 provided in an embodiment of the present invention
Spend frequency histogram；
Fig. 7 is any seed voxel V in depth-first traversal gray scale volume element model provided in an embodiment of the present inventionsThree-dimensional
The flow chart of connected region.
Specific embodiment
With reference to the accompanying drawings and examples, specific embodiments of the present invention will be described in further detail.Implement below
Example is not intended to limit the scope of the invention for illustrating the present invention.
The present embodiment is by taking certain organizes remotely-sensed data as an example, using of the invention a kind of based on the airborne of gray scale volume element model
LIDAR three-dimensional filtering method is filtered analysis to this group of data.
A kind of airborne LIDAR three-dimensional filtering method based on gray scale volume element model, as shown in Figure 1, comprising the following steps:
Step 1: reading original airborne LIDAR point cloud data, form original airborne LIDAR point cloud data collection.
In the present embodiment, using International Photography measurement and remote sensing association (International Society for
Photogrammetry and Remote Sensing, ISPRS) third working group provide dedicated for filtering algorithm test
15 sample datas as experimental data, with the validity and feasibility of the method for inspection.As shown in Fig. 2, these sample datas
Shearing is likely encountered difficult situation, such as rough error which includes the scene of differing complexity and filtering from 7 test datas
The influence of point, complicated atural object, atural object be connected with ground, on slope or short vegetation, landform discontinuously etc., table 1 is listed
The characteristic and basic parameter of each sample data in the present embodiment.The first, last time echo and intensity letter of sample record point cloud
Breath (in sample data and do not include point cloud data strength information.Strength information herein is from corresponding 7 test datas
Extract gained).
The characteristic and basic parameter of 1 sample data of table
(15 sample datas are accurately classified as ground to the reference data that design uses ISPRS to provide in the present embodiment
Point and non-ground points) as normal data evaluation arithmetic accuracy.
In the present embodiment, original airborne LIDAR point cloud data P={ p is definedi(xi, yi, zi), i=1 ..., n }, wherein i
It is the index of original airborne LIDAR point cloud data, n is the number of original airborne LIDAR point cloud data, piIt is i-th of original machine
LIDAR point cloud data is carried, coordinate is (xi, yi, zi)。
Step 2: original airborne LIDAR point cloud data collection rule being turned into gray scale volume element model, as shown in figure 3, specific side
Method are as follows:
Step 2.1: concentrating rejecting abnormalities data from original airborne LIDAR point cloud data, obtain rejecting abnormalities data set.
Step 2.1.1: the frequency that original airborne LIDAR point cloud data concentrates each laser point height value is counted, and with straight
The form visualization display statistical result of square figure；
Step 2.1.2: highest elevation threshold value T corresponding with real terrain and atural object is determinedheWith lowest elevation threshold value Tle；
Step 2.1.3: concentrating each laser point for original airborne LIDAR point cloud data, if its height value is higher than highest
Elevation threshold value TheOr it is lower than lowest elevation threshold value Tle, then the laser point is height anomaly data, is rejected, and otherwise retaining should
Laser point obtains and rejects height anomaly data set；
Step 2.1.4: statistics rejects the frequency of the intensity value of each laser point in height anomaly data set, and with histogram
Form visualization display statistical result；
Step 2.1.5: maximum intensity threshold value T corresponding with real terrain and atural object is determinedhiWith minimum intensity threshold value Tli；
Step 2.1.6: for each laser point in height anomaly data set is rejected, if its intensity value is higher than maximum intensity threshold
Value ThiOr it is lower than minimum intensity threshold value Thi, then the laser point is intensity abnormal data, it is rejected, otherwise retains the laser point,
Final obtain rejects elevation and intensity abnormal data set.
In the present embodiment, highest elevation threshold value The, lowest elevation threshold value Tle, maximum intensity threshold value ThiWith minimum intensity threshold
Value TliIt is constant, value need to be determined according to the space distribution situation of original airborne LIDAR point cloud data.
In the present embodiment, rejecting abnormalities data set is denoted as Q={ qi′(xi′, yi′, zi′), i '=1 ..., t }, wherein i ' is
The index of laser point in rejecting abnormalities data set, t are the number of laser point in rejecting abnormalities data set, qi′It is rejecting abnormalities data
The i-th ' a laser point is concentrated, coordinate is (xi′, yi′, zi′)。
Step 2.2: rejecting abnormalities data set rule is turned into gray scale volume element model.
Step 2.2.1: the spatial dimension of data set is indicated with the axial parallel bounding box of rejecting abnormalities data set.
In present embodiment, the spatial dimension of rejecting abnormalities data set Q can be by its axial parallel bounding box (Axis-
Aligned Bounding Box, AABB) it determines, AABB=(x, y, z) | xmin≤x≤xmax, ymin≤y≤ymax, zmin≤z
≤zmax, in formula, (xmax, ymax, zmax) and (xmin, ymin, zmin) respectively represent x, y and z coordinate in rejecting abnormalities data set
Maximum and minimum value.
Step 2.2.2: calculated body element resolution ratio, that is, voxel size, the resolution ratio on x, y, z direction is according to rejecting abnormalities number
It is determined according to the equalization point spacing of concentration laser point.
In present embodiment, calculation formula such as formula (1) institute of volume elements resolution ax x in the x, y, z-directions, Δ y, Δ z
Show:
Wherein, Sxy={ (xi′, yi′), i '=1 ..., t } Sxz={ (xi′, zi′), i '=1 ..., t }, Syz={ (yi′,
zi′), i '=1 ..., t } it is respectively rejecting abnormalities data set Q two dimension point set, C obtained by the projection in XOY, XOZ and YOZ plane
() is the convex hull of corresponding point set, and A () is the area of corresponding convex hull, as shown in Figure 4.
Step 2.2.3: axial parallel bounding box is divided to obtain three-dimensional lattice according to the resolution ratio on x, y, z direction
Net, each three-dimensional grid unit are known as volume elements.
In the present embodiment, based on voxel resolution, (axial parallel bounding box can be divided into three by Δ x, Δ y, Δ z)
Grid is tieed up, can be indicated with said three-dimensional body variable matrix.If V is the volume elements set in said three-dimensional body variable matrix, as shown in formula (2):
V={ vj(rj, cj, lj), j=1 ..., m } (2)
Wherein, j is volume elements index；M is volume elements number；vjIt is the voxel values of j-th of volume elements；(rj, cj, lj) it is j-th of volume elements
Coordinate in volume elements array, rj, cj, ljThe respectively row, column and level number of coordinate.
Step 2.2.4: laser point each in rejecting abnormalities data set is mapped to three-dimensional grid, and then is wrapped according in volume elements
The Intensity attribute of the laser point contained is each volume elements assignment, and each voxel values discretization is finally obtained gray scale body to { 0 ..., 255 }
Meta-model；
In the present embodiment, each laser point in rejecting abnormalities data set Q is mapped to 3D grid, and then wrap according in volume elements
The Intensity attribute of the laser point contained is each volume elements assignment.Wherein, the volume elements containing laser point is assigned a value of laser point intensity value (if containing
Have multiple laser points, then the intensity value of the minimum laser point of assignment elevation), the volume elements without containing laser point be assigned a value of 0, further will
Each volume elements assignment discretization obtains each voxel values to { 0 ..., 255 }.Gray scale volume element model is obtained as a result, is completed different to rejecting
The regularization of regular data collection.
Step 3: it is theoretical based on the building of three-dimensional communication region, the ground volume elements in gray scale volume element model is detected, is had
Body process is as shown in Figure 5.
Step 3.1: the minimum characteristic of local elevation based on ground point, the minimum non-zero value of elevation from gray scale volume element model
Volume elements is as seed voxel set Vs, wherein s=1,2 ....
Step 3.1.1: in the horizontal direction, piecemeal is carried out to gray scale volume element model using the Grid size of setting, and take
The minimum non-zero value volume elements of elevation is seed voxel in each piece.
In the present embodiment, Grid size need to determine that, if landform is more complicated, Grid size takes according to the complexity of landform
Value is smaller, conversely, can suitably increase Grid size.But minimum Grid size cannot be less than original airborne LIDAR point cloud data
The size of maximum target structure (such as building) is concentrated, otherwise, the point inside max architecture will be misjudged as ground seed voxel.
Step 3.1.2: it is searched in each piece and is less than difference in height threshold value T with the height difference of ground seedeNon-zero value volume elements be ground
Face seed obtains encrypted ground seed set Vs, wherein s=1,2 ...；
In present embodiment, height difference threshold value TeFor constant, value need to be according to the space of original airborne LIDAR point cloud data
Distribution situation determines.
Step 3.2: label ground seed voxel and with its three-dimensional communication and the close volume elements of the gradient, gray scale is constituted three
Dimension connected region is ground voxel data collection, completes the airborne LIDAR three-dimensional filtering based on gray scale volume element model；
Step 3.2.1: the frequency histogram of the gray value of non-zero value volume elements in gray scale volume element model is calculated, Gaussian Mixture is used
Model (Gaussian Mixture Model, GMM) is fitted gray scale frequency histogram, determines gray value corresponding with ground target
Distribution.
In the present embodiment, by taking sample samp41 as an example, the frequency of the gray value of the non-zero value volume elements in its corresponding V is counted,
And shown with represented as histograms, as shown in Figure 6.It will be appreciated from fig. 6 that multimodality is presented in intensity profile therein.If assuming multimodal point
Cloth is normal state multi-modal, then can be used gauss hybrid models to be fitted the histogram in Fig. 1, obtain each Gaussian Profile
Distribution characteristics parameter (mean value, standard deviation).It is that filtering obtains ground volume elements for application purpose of the invention, so being faced with ground
The distribution characteristics parameter μ for the Gaussian Profile answered, σ and distribution [0,20] are used for determining the tonal range of ground target, in detail
Thin scheme is as follows: if enabling μ-ml× σ=0, μ+mr× σ=20, then can determine mlAnd mr, thus can determine the gray scale of ground target
Range, i.e. [μ-m σ, μ+m σ], wherein multiplier m=max { ml, mr, m takes mlAnd mrMaximum value be pair based on Gaussian Profile
What title property determined.In addition, it is necessary to be noted that in tonal range that there are negatives, then 0 is set by minimum value but do not included
0。
Step 3.2.2: to any ground seed voxel Vs, traverse gray scale volume element model in Current terrestrial seed voxel Vs
Three-dimensional communication, gray value are located in the corresponding tonal range of ground target and local terrain slope is less than gradient threshold value TsIt is all
Unmarked volume elements, and it is labeled as Lg, until having marked all ground seed voxel VsThree-dimensional communication region, i.e. ground volume elements collection
It closes；Wherein, the determination of gradient threshold value is adaptive, method particularly includes: to current seed voxel Vs, detect in its spatial neighborhood
Whether marked ground volume elements is contained, if so, then the ruling grade value between existing ground volume elements is determined as in the neighborhood
Terrain slope threshold value Ts；Otherwise, 90 ° are set by terrain slope threshold value.
In the present embodiment, any seed voxel V in depth-first traversal gray scale volume element modelsThree-dimensional communication region tool
Body method is as shown in Figure 7.
In the present embodiment, it is based on VsThe formula for calculating gradient E with ground volume elements marked in its spatial neighborhood is shown in formula
(3):
Wherein, (rp, cp, lp) it is VsCoordinate；(re, ce, le) it is marked ground voxel coordinates, t is in spatial neighborhood
Marked ground volume elements number.
In the present embodiment, different filter results can be obtained using different neighborhood sizes in above-mentioned labeling process.Most
Good neighborhood scale will determine in an experiment.
The resulting ground data of filtering method proposed by the present invention is expressed with volume elements, and is then discrete in reference data
LIDAR laser point expression-form.Method precision proposed by the present invention, this implementation are compared and then evaluated to do with reference data
In example, then laser point first in statistics this method ground volume elements detected is compared with reference data, and then with I class
Error (ground point mistake is divided into non-ground points ratio), II class error (non-ground points mistake is divided into ground point ratio), overall error
The index quantifications such as (ratio of the laser point of mistake point) and Kappa coefficient evaluate filtering algorithm precision.
In the present embodiment, when neighborhood scale is respectively 6,18,26,51,56 and 64,15 are tested using the method for the present invention
Data are filtered, and the Kappa coefficient of corresponding filter result is as shown in table 2.Data in the table are intended to examine or check different field
Influence of the size to filter result precision, and thereby determine that optimal neighborhood size.
The Kappa coefficient of the filtering method of the different neighborhood sizes of table 2
As shown in Table 2, the average Kappa coefficient of 6,18,26,51,56 and 64 neighborhoods be respectively 71.64%, 78.73%,
80.81%, 84.34%, 83.29% and 83.28%.This shows: (1) the maximum Kappa coefficient of 51 neighbor assignments, therefore, from
From the point of view of Kappa coefficient index, 51 neighborhoods are best neighborhood sizes.(2) increase of neighborhood size is not meant to filtering accuracy
Necessarily improve.This is because: the information of ground target can pass through connectivity, gray scale defined in gray scale said three-dimensional body metadata set
It is propagated with terrain slope intensity similarity.By taking 6 neighborhoods as an example, the information of ground target can only to its upper and lower, left and right, it is preceding,
6 volume elements are propagated afterwards, thus cause seed voxel that can only propagate to level terrain around it.If can increase neighborhood size, such as
26 neighborhoods, the comparison increased direction of propagation of 6 neighborhoods may be included in more ground volume elements, to improve the accurate of filtering
Property.This can explain why 26 neighborhoods compare 6 neighborhoods filtering algorithm Kappa coefficient it is higher.But if neighborhood scale
It is excessive, then some non-ground volume elements mistakes may be divided into ground volume elements, so as to cause the reduction of filtering accuracy.This can be explained
Why the precision of 56 neighborhoods, 51 neighborhoods of comparison is declined instead.
It using reference data is standard under 51 neighborhood sizes of 15 sample datas using the method for the present invention in the present embodiment
The quantitative assessment that carries out of filtering accuracy, the precision of filter result is as shown in table 3.
The precision of 3 filter result of table
| Sample | I class error (%) | II class error (%) | Overall error (%) | Kappa coefficient (%) |
| samp11 | 13.49 | 19.80 | 16.18 | 66.86 |
| samp12 | 3.72 | 3.54 | 3.63 | 92.74 |
| samp21 | 1.12 | 3.47 | 1.64 | 95.25 |
| samp22 | 3.32 | 5.53 | 4.01 | 90.71 |
| samp23 | 2.78 | 5.37 | 4.00 | 91.96 |
| samp24 | 2.32 | 9.38 | 4.26 | 89.21 |
| samp31 | 0.94 | 1.94 | 1.40 | 97.18 |
| samp41 | 2.11 | 3.99 | 3.05 | 93.89 |
| samp42 | 1.24 | 5.99 | 0.79 | 98.10 |
| samp51 | 3.51 | 15.85 | 6.20 | 81.62 |
| samp52 | 1.03 | 31.12 | 4.19 | 75.28 |
| samp53 | 1.58 | 48.23 | 3.46 | 52.93 |
| samp54 | 1.91 | 6.88 | 4.58 | 90.83 |
| samp61 | 0.33 | 31.62 | 1.40 | 76.30 |
| samp71 | 2.36 | 20.28 | 4.39 | 77.95 |
As shown in Table 3: the error distribution of I class error, II class error and overall error be 0~14%, 1~49%, 0~
17%；The average Kappa coefficient of filtering is up to 84.34%.To demonstrate the validity of method proposed by the present invention.
In the present embodiment, using (Liying Wang, Yan Xu, the and Yu Li.Aerial such as the method for the present invention and Wang
LIDAR point cloud voxelization with its 3D ground filtering application[J]
.Photogrammetric Engineering and Remote Sensing, 2017,83 (2): 95-107) propose based on
The overall error comparing result of the three-dimensional filtering method of two-value volume element model is as shown in table 4.
4 the method for the present invention of table is compared with the overall error of the three-dimensional filtering method based on two-value volume element model
| Sample | The overall error (%) of the methods of Wang | The overall error (%) of the method for the present invention |
| samp11 | 19.49 | 16.18 |
| samp12 | 4.02 | 3.63 |
| samp21 | 2.05 | 1.64 |
| samp22 | 4.97 | 4.01 |
| samp23 | 5.91 | 4.00 |
| samp24 | 6.34 | 4.26 |
| samp31 | 1.58 | 1.40 |
| samp41 | 2.17 | 3.05 |
| samp42 | 1.07 | 0.79 |
| samp51 | 8.09 | 6.20 |
| samp52 | 4.90 | 4.19 |
| samp53 | 3.46 | 3.46 |
| samp54 | 5.62 | 4.58 |
| samp61 | 1.93 | 1.40 |
| samp71 | 5.42 | 4.39 |
| It is average | 5.13 | 4.21 |
As shown in Table 4: the overall error of the method for the present invention is significantly lower than the existing three-dimensional filtering based on two-value volume element model
Algorithm, i.e. the method for the present invention precision are higher.
Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although
Present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: it still may be used
To modify to technical solution documented by previous embodiment, or some or all of the technical features are equal
Replacement；And these are modified or replaceed, model defined by the claims in the present invention that it does not separate the essence of the corresponding technical solution
It encloses.
Claims (6)
1. a kind of airborne LIDAR three-dimensional filtering method based on gray scale volume element model, it is characterised in that: the following steps are included:
Step 1: reading original airborne LIDAR point cloud data, form original airborne LIDAR point cloud data collection；
Step 2: original airborne LIDAR point cloud data collection rule being turned into gray scale volume element model, wherein gray scale is laser in volume elements
The discretization expression of the mean intensity of point, method particularly includes:
Step 2.1: concentrating rejecting abnormalities data from original airborne LIDAR point cloud data, obtain rejecting abnormalities data set；
Step 2.2: rejecting abnormalities data set rule is turned into gray scale volume element model；
Step 3: it is theoretical based on the building of three-dimensional communication region, the ground volume elements of gray scale volume element model is detected, specific method
Are as follows:
Step 3.1: the minimum characteristic of local elevation based on ground point, the minimum non-zero value volume elements of elevation from gray scale volume element model
As ground seed voxel set；
Step 3.2: label ground seed voxel and with its three-dimensional communication and the close volume elements of the gradient, gray scale is constituted it is three-dimensional even
Logical region is ground voxel data collection, completes the airborne LIDAR three-dimensional filtering based on gray scale volume element model.
2. a kind of airborne LIDAR three-dimensional filtering method based on gray scale volume element model according to claim 1, feature exist
In: the step 2.1 method particularly includes:
Step 2.1.1: the frequency that original airborne LIDAR point cloud data concentrates each laser point height value is counted, and with histogram
Form visualization display statistical result；
Step 2.1.2: highest elevation threshold value T corresponding with real terrain and atural object is determinedheWith lowest elevation threshold value Tle；
Step 2.1.3: concentrating each laser point for original airborne LIDAR point cloud data, if its height value is higher than highest elevation
Threshold value TheOr it is lower than lowest elevation threshold value Tle, then the laser point is height anomaly data, is rejected, otherwise retains the laser
Point obtains and rejects height anomaly data set；
Step 2.1.4: statistics rejects the frequency of the intensity value of each laser point in height anomaly data set, and in the form of histogram
Visualization display statistical result；
Step 2.1.5: maximum intensity threshold value T corresponding with real terrain and atural object is determinedhiWith minimum intensity threshold value Tli；
Step 2.1.6: for each laser point in height anomaly data set is rejected, if its intensity value is higher than maximum intensity threshold value Thi
Or it is lower than minimum intensity threshold value Thi, then the laser point is intensity abnormal data, is rejected, otherwise retains the laser point, finally
It obtains and rejects elevation and intensity abnormal data set.
3. a kind of airborne LIDAR three-dimensional filtering method based on gray scale volume element model according to claim 1, feature exist
In: the step 2.2 method particularly includes:
Step 2.2.1: the spatial dimension of data set is indicated with the axial parallel bounding box of rejecting abnormalities data set；
Step 2.2.2: calculated body element resolution ratio, that is, voxel size, the resolution ratio on x, y, z direction is according to rejecting abnormalities data set
The equalization point spacing of middle laser point determines；
Step 2.2.3: it is divided according to axial parallel bounding box of the resolution ratio on x, y, z direction to rejecting abnormalities data set
Three-dimensional grid is obtained, each three-dimensional grid unit is known as volume elements；
Step 2.2.4: laser point each in rejecting abnormalities data set is mapped to three-dimensional grid, and then includes according in volume elements
The Intensity attribute of laser point obtains gray scale volume elements mould finally by each voxel values discretization to { 0 ..., 255 } for each volume elements assignment
Type.
4. a kind of airborne LIDAR three-dimensional filtering method based on gray scale volume element model according to claim 1, feature exist
In: the step 3.1 method particularly includes:
Step 3.1.1: in the horizontal direction, piecemeal is carried out to gray scale volume element model using the Grid size of setting, and take each piece
The minimum non-zero value volume elements of interior elevation is seed voxel；
Step 3.1.2: it is searched in each piece and is less than difference in height threshold value T with the difference in height of ground seedeNon-zero value volume elements be ground
Seed voxel obtains encrypted ground seed voxel set Vs, wherein s=1,2 ....
5. a kind of airborne LIDAR three-dimensional filtering method based on gray scale volume element model according to claim 4, feature exist
In: the step 3.2 method particularly includes:
Step 3.2.1: the frequency histogram of the gray value of non-zero value volume elements in gray scale volume element model is calculated, gauss hybrid models are used
It is fitted gray scale frequency histogram, determines grey value profile range corresponding with ground target；
Step 3.2.2: to any ground seed voxel Vs, traverse gray scale volume element model in Current terrestrial seed voxel VsIt is three-dimensional
Connection, gray value are located in the corresponding tonal range of ground target and local terrain slope is less than gradient threshold value TsAll do not mark
Remember volume elements, and is labeled as Lg, until having marked all ground seed voxel VsThree-dimensional communication region, i.e. ground volume elements set.
6. a kind of airborne LIDAR three-dimensional filtering method based on gray scale volume element model according to claim 5, feature exist
In: the determination of gradient threshold value described in step 3.2.2 is adaptive, method particularly includes: to current seed voxel Vs, detect it
Whether contain marked ground volume elements in spatial neighborhood, if so, then determining the ruling grade value between existing ground volume elements
For the terrain slope threshold value T in the neighborhoods；Otherwise, 90 ° are set by terrain slope threshold value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910027077.0A CN109785261B (en) | 2019-01-11 | 2019-01-11 | Airborne LIDAR three-dimensional filtering method based on gray voxel model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910027077.0A CN109785261B (en) | 2019-01-11 | 2019-01-11 | Airborne LIDAR three-dimensional filtering method based on gray voxel model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109785261A true CN109785261A (en) | 2019-05-21 |
| CN109785261B CN109785261B (en) | 2023-03-17 |
Family
ID=66500206
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910027077.0A Active CN109785261B (en) | 2019-01-11 | 2019-01-11 | Airborne LIDAR three-dimensional filtering method based on gray voxel model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109785261B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110517220A (en) * | 2019-06-10 | 2019-11-29 | 长安大学 | A kind of surface of aggregate quantity detection method based on laser three-D data |
| CN112099046A (en) * | 2020-09-16 | 2020-12-18 | 辽宁工程技术大学 | Airborne LIDAR three-dimensional plane detection method based on multi-valued voxel model |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130101175A1 (en) * | 2011-10-21 | 2013-04-25 | James D. Lynch | Reimaging Based on Depthmap Information |
| CN103679655A (en) * | 2013-12-02 | 2014-03-26 | 河海大学 | LiDAR point cloud filter method based on gradient and area growth |
| CN105844602A (en) * | 2016-04-01 | 2016-08-10 | 辽宁工程技术大学 | Airborne LIDAR point cloud 3D filtering method based on volume elements |
| CN106022259A (en) * | 2016-05-20 | 2016-10-12 | 江苏得得空间信息科技有限公司 | Laser-point cloud based method for extracting mountainous road by use of three-dimensional characteristic description model |
| CN106157309A (en) * | 2016-07-04 | 2016-11-23 | 南京大学 | A kind of airborne LiDAR ground point cloud filtering method based on virtual Seed Points |
| CN106529469A (en) * | 2016-11-08 | 2017-03-22 | 华北水利水电大学 | Unmanned aerial vehicle airborne LiDAR point cloud filtering method based on adaptive gradient |
| CN108109139A (en) * | 2017-12-18 | 2018-06-01 | 辽宁工程技术大学 | Airborne LIDAR three-dimensional building object detecting method based on gray scale volume element model |
| US10002407B1 (en) * | 2017-08-11 | 2018-06-19 | Intermap Technologies Inc. | Method and apparatus for enhancing 3D model resolution |
| CN108805154A (en) * | 2017-08-23 | 2018-11-13 | 辽宁工程技术大学 | A kind of geological fault recognition methods based on space clustering |
-
2019
- 2019-01-11 CN CN201910027077.0A patent/CN109785261B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130101175A1 (en) * | 2011-10-21 | 2013-04-25 | James D. Lynch | Reimaging Based on Depthmap Information |
| CN103679655A (en) * | 2013-12-02 | 2014-03-26 | 河海大学 | LiDAR point cloud filter method based on gradient and area growth |
| CN105844602A (en) * | 2016-04-01 | 2016-08-10 | 辽宁工程技术大学 | Airborne LIDAR point cloud 3D filtering method based on volume elements |
| CN106022259A (en) * | 2016-05-20 | 2016-10-12 | 江苏得得空间信息科技有限公司 | Laser-point cloud based method for extracting mountainous road by use of three-dimensional characteristic description model |
| CN106157309A (en) * | 2016-07-04 | 2016-11-23 | 南京大学 | A kind of airborne LiDAR ground point cloud filtering method based on virtual Seed Points |
| CN106529469A (en) * | 2016-11-08 | 2017-03-22 | 华北水利水电大学 | Unmanned aerial vehicle airborne LiDAR point cloud filtering method based on adaptive gradient |
| US10002407B1 (en) * | 2017-08-11 | 2018-06-19 | Intermap Technologies Inc. | Method and apparatus for enhancing 3D model resolution |
| CN108805154A (en) * | 2017-08-23 | 2018-11-13 | 辽宁工程技术大学 | A kind of geological fault recognition methods based on space clustering |
| CN108109139A (en) * | 2017-12-18 | 2018-06-01 | 辽宁工程技术大学 | Airborne LIDAR three-dimensional building object detecting method based on gray scale volume element model |
Non-Patent Citations (6)
| Title |
|---|
| LIYING WANG 等: "A greyscale voxel model for airborne lidar data applied to building detection", 《THE PHOTOGRAMMETRIC RECORD》 * |
| SAGI FILIN 等: "Segmentation of airborne laser scanning data using a slope adaptive neighborhood" * |
| 史建青: "机载LiDAR在省级基础测绘中若干关键技术研究" * |
| 王丽英 等: "强度体元基元下的机载LiDAR 3D滤波" * |
| 王丽英 等: "强度体元基元下的机载LiDAR 3D滤波", 《地球信息科学学报》 * |
| 王丽英 等: "机载LiDAR点云体元化及其在3D滤波中的应用", 《仪器仪表学报》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110517220A (en) * | 2019-06-10 | 2019-11-29 | 长安大学 | A kind of surface of aggregate quantity detection method based on laser three-D data |
| CN112099046A (en) * | 2020-09-16 | 2020-12-18 | 辽宁工程技术大学 | Airborne LIDAR three-dimensional plane detection method based on multi-valued voxel model |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109785261B (en) | 2023-03-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111091105B (en) | Remote sensing image target detection method based on new frame regression loss function | |
| CN110264468B (en) | Point cloud data mark, parted pattern determination, object detection method and relevant device | |
| CN111832655B (en) | Multi-scale three-dimensional target detection method based on characteristic pyramid network | |
| CN106199557B (en) | A kind of airborne laser radar data vegetation extracting method | |
| CN109146948B (en) | Crop growth phenotype parameter quantification and yield correlation analysis method based on vision | |
| CN105844602B (en) | A kind of airborne LIDAR point cloud three-dimensional filtering method based on volume elements | |
| CN104331699B (en) | A kind of method that three-dimensional point cloud planarization fast search compares | |
| CN108109139B (en) | Airborne LIDAR three-dimensional building detection method based on gray voxel model | |
| CN104240251B (en) | Multi-scale point cloud noise detection method based on density analysis | |
| CN109829399A (en) | A kind of vehicle mounted road scene point cloud automatic classification method based on deep learning | |
| CN107316058A (en) | Improve the method for target detection performance by improving target classification and positional accuracy | |
| Li et al. | A leaf segmentation and phenotypic feature extraction framework for multiview stereo plant point clouds | |
| CN111986322B (en) | Point cloud indoor scene layout reconstruction method based on structural analysis | |
| CN106529469A (en) | Unmanned aerial vehicle airborne LiDAR point cloud filtering method based on adaptive gradient | |
| CN105469098A (en) | Precise LINDAR data ground object classification method based on adaptive characteristic weight synthesis | |
| CN108074232B (en) | Voxel segmentation-based airborne LIDAR building detection method | |
| CN105574527A (en) | Quick object detection method based on local feature learning | |
| Cheng et al. | Building boundary extraction from high resolution imagery and lidar data | |
| CN109766824B (en) | Active and passive remote sensing data fusion classification method based on fuzzy evidence theory | |
| CN106780551A (en) | A kind of Three-Dimensional Moving Targets detection method and system | |
| CN112099046A (en) | Airborne LIDAR three-dimensional plane detection method based on multi-valued voxel model | |
| CN108765446B (en) | Power line point cloud segmentation method and system based on random field and random forest | |
| CN109785261A (en) | A kind of airborne LIDAR three-dimensional filtering method based on gray scale volume element model | |
| CN117495891B (en) | Point cloud edge detection method and device and electronic equipment | |
| CN117765006A (en) | Multi-level dense crown segmentation method based on unmanned aerial vehicle image and laser point cloud |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |