CN109816664A - A kind of three-dimensional point cloud dividing method and device - Google Patents
A kind of three-dimensional point cloud dividing method and device Download PDFInfo
- Publication number
- CN109816664A CN109816664A CN201811589435.9A CN201811589435A CN109816664A CN 109816664 A CN109816664 A CN 109816664A CN 201811589435 A CN201811589435 A CN 201811589435A CN 109816664 A CN109816664 A CN 109816664A
- Authority
- CN
- China
- Prior art keywords
- dimensional point
- point cloud
- dimensional
- area
- sunk area
- 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 invention discloses a kind of three-dimensional point cloud dividing method and devices to acquire three-dimensional point cloud to be parallel to the direction of plane domain；Pretreatment operation is carried out to three-dimensional point cloud, obtains pretreated three-dimensional point cloud；According to the depth value size of three-dimensional point each in three-dimensional point cloud after pretreatment, the three-dimensional point cloud of non-planar area is obtained, obtains initial recess area three-dimensional point cloud；All three-dimensional points in initial recess area three-dimensional point cloud are handled using the method for Edge Search, obtain multiple three-dimensional point compositions sunk area edge；Obtain sunk area three-dimensional point cloud；Dividing method provided by the invention is the methods of down-sampled by filtering, and eliminates mistake and redundant points cloud, simplifies the quantity of three-dimensional point in three-dimensional point cloud, effectively increase pit slot region segmentation speed；Plane domain is removed using depth threshold, can quickly the approximate region of pit slot be split by retaining non-planar area, reduced the complexity of dividing method, improved splitting speed.
Description
Technical field
The present invention relates to three-dimensional point cloud processing method and processing devices, and in particular to a kind of three-dimensional point cloud dividing method and device.
Background technique
Three-dimensional point cloud is most common, most basic threedimensional model in a kind of spatial information expression mode.Point cloud model is often
It is directly obtained by measurement, each pair of point answers a measurement point, without other methods processing, therefore includes the maximum information of target
Amount.The data that three-dimensional point cloud is mainly used for characterizing in three-dimensional space are compared with plane X-Y scheme, and three-dimensional point cloud possesses depth side
To information, natural can decouple target and background in the depth direction.In addition, for dimensional visual measurement, it is three-dimensional
Point cloud is its main support means, by the extraction to target three-dimensional point cloud, the practical object of available target in space
Manage geological information.
In practical applications, since the point cloud actually obtained includes much noise, and the three-dimensional point cloud got is not
Comprising target complete information, and in different angle, there are rotation transformations, cause same target point cloud shape multifarious, especially
In face of the not only three dimensional point cloud containing sunk area but also containing plane domain, such as the three-dimensional point cloud on the road surface comprising pit slot
When, road surface itself is a plane, but among plane again include pit slot sunk area, to this three-dimensional point cloud again
When being split, since noise is more, the cut-off rule between sunk area and plane domain can not be accurately found, causes to divide
Difficulty is larger, therefore three-dimensional point cloud dividing method in the prior art is when in face of three-dimensional point cloud comprising sunk area, can not
Accurately it is split.
Summary of the invention
The purpose of the present invention is to provide three-dimensional point cloud dividing method and devices, to solve three-dimensional point in the prior art
Cloud dividing method is when facing the three-dimensional point cloud comprising sunk area, the problem of can not being accurately split.
In order to realize above-mentioned task, the invention adopts the following technical scheme:
A kind of three-dimensional point cloud dividing method, for that will be recessed in the three-dimensional point cloud containing sunk area and plane domain
Area three-dimensional point cloud segmentation comes out, and the method executes according to the following steps:
Step 1, the direction to be parallel to the plane domain acquire the three-dimensional point containing sunk area and plane domain
Cloud；
Step 2 carries out pretreatment operation to the three-dimensional point cloud, obtains pretreated three-dimensional point cloud, described is pre-
Processing operation includes the down-sampled operation successively executed and filtering operation；
Step 3, according to the depth value size of three-dimensional point each in three-dimensional point cloud after pretreatment, the depth value of three-dimensional point is small
In all three-dimensional point deletions of depth threshold, the three-dimensional point cloud of non-planar area is obtained, the depth threshold is after pre-processing
Twice of all three-dimensional point depth value average values in three-dimensional point cloud；
Step 4, in the three-dimensional point cloud of the non-planar area, using the maximum three-dimensional point of depth value as seed point,
Initial recess area three-dimensional point cloud is obtained using the method for region growing；
Step 5, using the method for Edge Search to all three-dimensional points in the initial recess area three-dimensional point cloud at
Reason obtains multiple three-dimensional point compositions sunk area edge；
Step 6, by three-dimensional point set all within the sunk area edge, obtain the sunk area three-dimensional point
Cloud.
Further, the step 2 executes according to the following steps:
Step 21, the method that the three-dimensional point cloud is filtered using voxel grid carry out it is down-sampled, obtain it is down-sampled after
Three-dimensional point cloud；
Step 22, to it is described it is down-sampled after three-dimensional point cloud carry out straight-through filtering, filtered three-dimensional point is led directly in acquisition
Cloud；
Step 23 leads directly to filtered three-dimensional point cloud and carries out statistical filtering to described, the three-dimensional point cloud that obtains that treated.
Further, in the step 3, according to the depth value size of three-dimensional point each in three-dimensional point cloud after pretreatment,
The depth value of three-dimensional point is less than to all three-dimensional point deletions of depth threshold using RANSAC algorithm, obtains the three of non-planar area
Dimension point cloud.
Further, the method using Edge Search is to all three-dimensionals in the initial recess area three-dimensional point cloud
When point is handled, using a vertex of the initial recess area three-dimensional point cloud as start boundary point, most using boundary point
Big angle principle successively obtains multiple three-dimensional point compositions sunk area edge.
A kind of three-dimensional point cloud segmenting device, for realizing the three-dimensional point cloud dividing method, the device includes three
Dimension point cloud obtains module, three-dimensional point cloud preprocessing module, plane domain three-dimensional point cloud removal module and sunk area three-dimensional point
Cloud divides module；
The three-dimensional point cloud obtains module and connect with three-dimensional point cloud preprocessing module, for be parallel to the plane area
The direction in domain acquires the three-dimensional point cloud containing sunk area and plane domain；
The three-dimensional point cloud preprocessing module is connect with the plane domain three-dimensional point cloud removal module, for institute
The three-dimensional point cloud stated carries out pretreatment operation, obtains pretreated three-dimensional point cloud, and the pretreatment operation includes successively holding
Capable down-sampled operation and filtering operation；
The plane domain three-dimensional point cloud removal module is connect with the sunk area three-dimensional point cloud segmentation module, is used
In the three-dimensional point cloud in the non-planar area, using the maximum three-dimensional point of depth value as seed point, using region growing
Method obtain initial recess area three-dimensional point cloud；
Be also used to the method using Edge Search to all three-dimensional points in the initial recess area three-dimensional point cloud at
Reason obtains multiple three-dimensional point compositions sunk area edge；
It is also used to obtain three-dimensional point set all within the sunk area edge sunk area three-dimensional point
Cloud.
Further, the three-dimensional point cloud preprocessing module, plane domain three-dimensional point cloud removal module and depressed area
It is Intelligent mobile equipment that domain three-dimensional point cloud, which divides module,.
Further, the Intelligent mobile equipment is mobile phone.
The present invention has following technical effect that compared with prior art
1, three-dimensional point cloud dividing method and device provided by the invention are dropped when being split to sunk area by filtering
The methods of sampling eliminates mistake and redundant points cloud, simplifies the quantity of three-dimensional point in three-dimensional point cloud, effectively increase sunk area
Splitting speed；
2, three-dimensional point cloud dividing method and device provided by the invention utilize depth threshold when being split to sunk area
Value removal plane domain, can quickly the approximate region of sunk area be split by retaining non-planar area, reduce segmentation
The complexity of method, improves splitting speed；
3, three-dimensional point cloud dividing method and device provided by the invention use area when being split to sunk area first
Domain generation method obtains initial recess region, reuses the method for Edge Search on this basis to obtain complete sunk area
Edge, improve sunk area segmentation integrality and accuracy；
4, three-dimensional point cloud dividing method and device provided by the invention utilize boundary point when being split to sunk area
Maximum angular principle successively obtains multiple three-dimensional point compositions sunk area edge, further increases the accurate of sunk area segmentation
Property.
Detailed description of the invention
Fig. 1 is the flow chart of three-dimensional point cloud dividing method provided by the invention；
Fig. 2 is in one embodiment of the present of invention for rebuilding the left and right camera image when three-dimensional point cloud of road surface；
Fig. 3 is the road surface three-dimensional point cloud acquired in one embodiment of the present of invention；
Fig. 4 is the pit slot area three-dimensional point cloud for dividing acquisition in one embodiment of the present of invention；
Fig. 5 is the three-dimensional point cloud segmenting device structural schematic diagram provided in one embodiment of the present of invention.
Specific embodiment
Three-dimensional point cloud: being made of multiple three-dimensional points, and each three-dimensional point has abscissa value, ordinate value and depth value,
Using the left camera lens of camera as coordinate origin, in general abscissa value, ordinate value and depth value are all positive values.
Boundary point maximum angular principle: in multiple neighbor points of a certain boundary point, boundary point is deposited with any one neighbor point
In angle (angle does not include other neighbor points), therefore there can be multiple angles, and difference pair between maximum two angles
Two point of proximity answered must be boundary point.
Embodiment one
A kind of three-dimensional point cloud dividing method, for that will be recessed in the three-dimensional point cloud containing sunk area and plane domain
Area three-dimensional point cloud segmentation comes out, as shown in Figure 1, dividing method executes according to the following steps:
Step 1, the direction to be parallel to the plane domain acquire the three-dimensional point containing sunk area and plane domain
Cloud；
It is obtained first using three-dimensional imaging devices such as binocular solid cameras in this step and contains sunk area and plane
The three-dimensional point cloud in region, using the three-dimensional point cloud as the input of the measurement method provided in the present embodiment.
Notice that imaging device face plane is shot when shooting, and covers complete sunk area range.
In the present embodiment, to the left and right camera image of binocular camera as shown in Figure 2, carry out what three-dimensional reconstruction acquired
Road surface three-dimensional point cloud is as shown in Figure 3.
In this step, when being split to three-dimensional point cloud, existing method have Euclidean algorithm, kdTree algorithm with
And OcTree algorithm, but above several algorithms are when facing road surface three-dimensional point cloud, it may appear that the not high problem of segmentation accuracy rate.
Since the data volume that collected three-dimensional point cloud includes is larger, and contain noise, if directly handled, meeting
There is splitting speed is slow and segmentation effect is poor phenomenon, therefore in this step, three-dimensional point cloud is carried out after dimensionality reduction again first
Filtering.
Step 2 carries out pretreatment operation to the three-dimensional point cloud, obtains pretreated three-dimensional point cloud, described is pre-
Processing operation includes the down-sampled operation successively executed and filtering operation；
It is down-sampled to three-dimensional point cloud progress first, it is down-sampled to three-dimensional point cloud progress using the method for voxel grid filtering,
To reduce amount of redundant data, points cloud processing speed is improved, when filtering can keep details in the depth direction, reduce depth direction
Loss of detail.
Detailed process are as follows: a three-dimensional voxel grid is created by the three-dimensional point cloud of input, voxel grid is equivalent to space
The set of three-dimensional cube, then in each voxel, with the center of gravity of all the points in voxel come other points in approximate display voxel,
All the points finally just indicate that all the points cloud is dropped after all handling by voxel grid with a focus point in this way in the voxel grid
Sampled point cloud obtains the expression that accurate sampled point corresponds to curved surface.
Step 22, to it is described it is down-sampled after three-dimensional point cloud carry out straight-through filtering, filtered three-dimensional point is led directly in acquisition
Cloud；
In this step, the three-dimensional point cloud after down-sampled is filtered, using the method for straight-through filtering, particular by
Three-dimensional point cloud depth direction data area is set, such as between 0.8-1.2m, the point cloud except this range all will be as interference
Point removal.
Step 23 leads directly to filtered three-dimensional point cloud and carries out statistical filtering to described, the three-dimensional point cloud that obtains that treated.
In this step, to filtered three-dimensional point cloud is led directly to using statistical filtering method, each neighborhood of a point is carried out
One statistical analysis, and some non-compliant clouds are removed, reduce isolated redundant points cloud, neighbour of the statistical filtering to each point
Domain carries out a statistical analysis, and trims those points for not meeting established standards.Specifically, it uses in the present embodiment dilute
Outlier removing method is dredged, based on the calculating in input three-dimensional point cloud to the range distribution of point to neighbor point.To each three
Tie up point, calculate the average distance that it arrives itself all neighbor point, it is obtaining the result is that a Gaussian Profile, shape by mean value and
Standard deviation determines that three-dimensional point of the average distance except critical field (being defined by global distance average and variance) is to peel off
Point, and got rid of from three-dimensional point cloud.
When being split in the present embodiment to sunk area, plane domain is removed using depth threshold, is retained non-planar
Region quickly can substantially split sunk area, reduce the complexity of dividing method, improve splitting speed.
Step 3, according to the depth value size of three-dimensional point each in three-dimensional point cloud after pretreatment, the depth value of three-dimensional point is small
In all three-dimensional point deletions of depth threshold, the three-dimensional point cloud of non-planar area is obtained, the depth threshold is after pre-processing
Twice of all three-dimensional point depth value average values in three-dimensional point cloud；
In this step, plane domain acquisition, plane domain three are parallel to when obtaining due to three-dimensional point cloud
Depth value should be the same in dimension point cloud, by this feature, plane three-dimensional point cloud can be found.
It preferably, will using RANSAC algorithm according to the depth value size of three-dimensional point each in three-dimensional point cloud after pretreatment
The depth value of three-dimensional point is less than all three-dimensional point deletions of depth threshold, obtains the three-dimensional point cloud of non-planar area.
In this step, when extracting plane using RANSAC algorithm, depth threshold is set, depth threshold is according to plane
Thickness setting, such as when road pavement pit slot region is split, the road plane of 2mm depth is partitioned into, in the present embodiment
In the measurement method of offer will using all the points in 2mm depth near the plane of road as road plane three-dimensional point, the step for phase
When in cutting three-dimensional point cloud in the depth direction, road plane is reamed.
It in this step, (include depression points by other non-flat portion point clouds while extracting planar point cloud
Cloud) it splits.Under normal circumstances, the interference due to noise and without matching error point cloud point can not propose complete at this time
The accurate not partial dot cloud that is recessed, needs further to handle.
Therefore, optionally, before executing the step 23, the three-dimensional point cloud of the non-planar area is adopted first
Secondary filtering is carried out with the mode of statistical filtering, obtains the three-dimensional point cloud of filtered non-planar area, after the filtering
The three-dimensional point cloud of non-planar area be input in the step 23.
In this step, when secondary filtering again by the way of statistical filtering, filtering parameter is set, useless model is removed
Lesser isolated point cloud is enclosed, the three-dimensional point sum for including in certain distance by setting each cloud, if it is less than setting
Minimum three-dimensional point sum threshold value, then it is assumed that be isolated point cloud, if it is greater than the maximum three-dimensional point sum threshold value of setting, be considered as
Invalid three-dimensional point cloud.Such as three-dimensional point quantity around some three-dimensional point in 5 centimetres is counted, if statistical result is less than 500
It is a, then all three-dimensional points in 5 centimetres around the three-dimensional point and three-dimensional point are considered as isolated point cloud, if statistical result is greater than
50000 are considered as the Null Spot cloud cluster more than theoretical recess data volume, will also remove, such as the pit slot region in Fig. 3, maximum
Pit slot region beside these small pit slot regions are removed in this step there are also some small pit slot regions, these
Small pit slot region is exactly isolated point cloud.
Step 4, in the three-dimensional point cloud of the non-planar area, using the maximum three-dimensional point of depth value as seed point,
Initial recess area three-dimensional point cloud is obtained using the method for region growing；
In this step, according to it is down-sampled when the depth threshold that sets as the distance threshold in the method for region growing,
Using the maximum three-dimensional point of depth value as starting point, the point within the scope of distance threshold is searched, is risen with what the point iteration newly searched was newly
Point, continues to search for, and is partitioned into sunk area edge three-dimensional point until searching, and obtains initial recess area three-dimensional point cloud.
In order to improve the accuracy rate of segmentation, therefore initial recess area three-dimensional point cloud is carried out again in the present embodiment
Edge Search, to obtain the edge of complete sunk area.
Step 5, using the method for Edge Search to all three-dimensional points in the initial recess area three-dimensional point cloud at
Reason obtains multiple three-dimensional point compositions sunk area edge；
In this step, the method for Edge Search can be depth threshold detection, method of gradient difference value threshold test etc.,
But these methods are when in face of sunk area in plane in the prior art, especially when the edge of recess has gradient, therefore
The edge that sunk area good can not be all found using the method for existing threshold test causes sunk area to be divided inaccurate
Really.
Therefore a kind of iterative edge searching method constantly approached is provided in the present embodiment, by the initial recess area
One vertex of domain three-dimensional point cloud successively obtains multiple three-dimensional points using boundary point maximum angular principle and forms as start boundary point
Sunk area edge.
Specifically, it executes according to the following steps:
Step I, boundary three-dimensional point set is emptied；
Step II, multiple neighbouring three-dimensional points around the start boundary point are found using k nearest neighbor algorithm；
In this step, specifically using the start boundary point as the centre of sphere, using distance threshold as the radius of ball, selection is established
Sphere selects described three-dimensional point all in sphere as neighbouring three-dimensional point, obtains multiple neighbouring three-dimensional points, the distance
Threshold value is after secondary filtering twice of all three-dimensional point depth value average values in three-dimensional point cloud；
Step III, the angle each between neighbouring three-dimensional point and the start boundary point is calculated, multiple angle values are obtained；
Step IV, all angle values are arranged according to size descending, calculate the differential seat angle between two neighboring angle value,
Obtain multiple differential seat angles；
Step V, select maximum angle poor from all differential seat angles, the corresponding two angle values of maximum angle difference is corresponding
Two neighbouring three-dimensional points as two boundary three-dimensional points；
If step VI, boundary three-dimensional point set is combined into sky at this time, VIII is thened follow the steps；Otherwise VII is executed；
Step VII, the two boundary three-dimensional points obtained in judgment step V whether have one in the boundary three-dimensional point set
In conjunction；
If having one in the boundary three-dimensional point set, by another not side in the three-dimensional point set of boundary
After boundary's three-dimensional point is as new start boundary point, executed in step IX after placing into boundary three-dimensional point set；
If two boundary three-dimensional points all in the boundary three-dimensional point set, select distance starting three-dimensional point distance most
A remote point is as new start boundary point；
Otherwise, two boundary three-dimensional points are put into the three-dimensional point set of boundary first, then calculate separately two boundary three-dimensionals
Angle value between point and the start boundary point, the maximum boundary three-dimensional point of selected angle value is as new start boundary point
Afterwards, step IX is executed；
Step IX, using new start boundary point as input, repeat step II to step IX, until this obtain it is new
Starting three-dimensional point is execution step behind a vertex of the initial recess area three-dimensional point cloud chosen when executing step II for the first time
X；
In the present embodiment, multiple three-dimensional points are successively obtained using boundary point maximum angular principle and form sunk area edge,
Further increase the accuracy of recess segmentation.
Step X, all boundaries three in the boundary three-dimensional point set are extracted from the initial recess area three-dimensional point cloud
Point is tieed up, multiple three-dimensional point compositions sunk area edge is obtained.
In the step X, institute in the boundary three-dimensional point set is extracted from the initial recess area three-dimensional point cloud
After having boundary three-dimensional point, multiple boundary three-dimensional points at this time are sparse boundary three-dimensional points, are directly obtained by sparse boundary three-dimensional point
If obtaining sunk area three-dimensional point cloud, sunk area three-dimensional point cloud can have noise, that is, divide the situation of inaccuracy, therefore at this
Denseization is carried out to all boundary three-dimensional points in embodiment, obtains multiple three-dimensional point compositions sunk area edge.
In this step, denseization being carried out to all boundary three-dimensional points, the method for denseization can be the methods of interpolation,
But in the present embodiment, in order to improve the accuracy rate of denseization, enable the sunk area after segmentation more complete and
Accurately, boundary denseization is that sparse boundary point is evolved into dense point cloud, and the boundary point on sparse line is also added
Into boundary point cloud.
Specifically, denseization is carried out to all boundary three-dimensional points according to the following steps in the present embodiment:
Step A, two adjacent boundary three-dimensional points are found from all boundary three-dimensional points respectively as endpoint and terminal；
Step B, using endpoint as the centre of sphere, using distance threshold as the radius of ball, selection sphere is established, will be owned in selection sphere
Three-dimensional point is used as neighbouring boundary three-dimensional point, obtains multiple neighbouring boundary three-dimensional points；
Distance threshold in this step is identical as the distance threshold in step II.
Step C, all neighbouring boundary three-dimensional points are calculated to the distance between the straight line being made of endpoint and terminal, are obtained more
The distance value of a neighbouring boundary three-dimensional point；
Step D, using the corresponding neighbouring boundary three-dimensional point of the distance value for being less than dense distance threshold as dense boundary three-dimensional point,
Multiple dense boundary three-dimensional points are obtained, dense distance threshold is the density value of three-dimensional point cloud after secondary filtering；
Step E, step A to step E is repeated until all boundary three-dimensional points are used as endpoint mistake, the multiple dense sides of acquisition
Boundary's three-dimensional point；
Step F, by all dense boundary three-dimensional points in conjunction with all boundary three-dimensional points, multiple three-dimensional point composition recess are obtained
Edges of regions.
Step 25, by three-dimensional point set all within the sunk area edge, it is three-dimensional to obtain the sunk area
Point cloud.
Therefore, sunk area edge is being made of dense boundary point and all boundary three-dimensional points of obtaining in step F
Edge takes out all three-dimensional points within the edge, including the edge, has obtained sunk area three-dimensional point cloud.
In the present embodiment, it to road surface three-dimensional point cloud image as shown in Figure 3, is searched using the edge provided in the present embodiment
The method of rope and edge denseization obtains pit slot region as shown in Figure 4, and the edge in the pit slot region is i.e. by dense three-dimensional
What boundary point and three-dimensional boundaries point were constituted.
When being split in the present embodiment to sunk area, initial recess area is obtained using Area generation method first
Domain reuses the method for Edge Search on this basis to obtain the edge of complete sunk area, improves sunk area point
The integrality and accuracy cut, and the method to denseization of sparse edge is provided, to obtain more accurate recess
Region.
Embodiment two
A kind of three-dimensional point cloud segmenting device is disclosed in the present embodiment, for the three-dimensional point cloud segmentation side in embodiment one
Method, as shown in figure 5, the device includes that three-dimensional point cloud obtains module, three-dimensional point cloud preprocessing module, plane domain three-dimensional point
Cloud removes module and sunk area three-dimensional point cloud divides module；
The three-dimensional point cloud obtains module and connect with three-dimensional point cloud preprocessing module, for be parallel to the plane area
The direction in domain acquires the three-dimensional point cloud containing sunk area and plane domain；
Three-dimensional point cloud acquisition module, which can be binocular camera etc., in the present embodiment can acquire setting for three dimensional point cloud
It is standby.
The three-dimensional point cloud preprocessing module is connect with the plane domain three-dimensional point cloud removal module, for institute
The three-dimensional point cloud stated carries out pretreatment operation, obtains pretreated three-dimensional point cloud, and the pretreatment operation includes successively holding
Capable down-sampled operation and filtering operation；
The plane domain three-dimensional point cloud removal module is connect with the sunk area three-dimensional point cloud segmentation module, is used
In the three-dimensional point cloud in the non-planar area, using the maximum three-dimensional point of depth value as seed point, using region growing
Method obtain initial recess area three-dimensional point cloud；
Be also used to the method using Edge Search to all three-dimensional points in the initial recess area three-dimensional point cloud at
Reason obtains multiple three-dimensional point compositions sunk area edge；
It is also used to obtain three-dimensional point set all within the sunk area edge sunk area three-dimensional point
Cloud.
Optionally, the three-dimensional point cloud preprocessing module, plane domain three-dimensional point cloud removal module and sunk area
It is Intelligent mobile equipment that three-dimensional point cloud, which divides module,.
In the present embodiment, it in order to acquire sunk area three-dimensional point cloud in real time, and can handle in real time recessed
Fall into area three-dimensional point cloud to be split to it, by three-dimensional point cloud preprocessing module, plane domain three-dimensional point cloud removal module with
And sunk area three-dimensional point cloud segmentation module is arranged on Intelligent mobile equipment, Intelligent mobile equipment can be moveable electricity
The hardware devices such as brain, mobile phone.
Optionally, the Intelligent mobile equipment is mobile phone.
In the present embodiment, for more convenient actual use, Intelligent mobile equipment is mobile phone, such as is cheated in road pavement
When slot region is split, maintenance of surface personnel directly can obtain module with three-dimensional point cloud with its mobile phone, and (binocular camera etc. is set
It is standby) be attached, the collected road surface three-dimensional point cloud of three-dimensional point cloud acquisition module is transmitted in mobile phone, using cell phone application into
Row processing, the pit slot region after being divided carry out subsequent processing.
Claims (7)
1. a kind of three-dimensional point cloud dividing method, in the three-dimensional point cloud containing sunk area and plane domain by depressed area
Domain three-dimensional point cloud is split, which is characterized in that the method executes according to the following steps:
Step 1, the direction to be parallel to the plane domain acquire the three-dimensional point cloud containing sunk area and plane domain；
Step 2 carries out pretreatment operation to the three-dimensional point cloud, obtains pretreated three-dimensional point cloud, the pretreatment
Operation includes the down-sampled operation successively executed and filtering operation；
Step 3, according to the depth value size of three-dimensional point each in three-dimensional point cloud after pretreatment, the depth value of three-dimensional point is less than deep
All three-dimensional point deletions for spending threshold value, obtain the three-dimensional point cloud of non-planar area, and the depth threshold is three-dimensional after pretreatment
Twice of all three-dimensional point depth value average values in point cloud；
Step 4, in the three-dimensional point cloud of the non-planar area, using the maximum three-dimensional point of depth value as seed point, use
The method of region growing obtains initial recess area three-dimensional point cloud；
Step 5 is handled all three-dimensional points in the initial recess area three-dimensional point cloud using the method for Edge Search, is obtained
It obtains multiple three-dimensional points and forms sunk area edge；
Step 6, by three-dimensional point set all within the sunk area edge, obtain the sunk area three-dimensional point cloud.
2. three-dimensional point cloud dividing method as described in claim 1, which is characterized in that the step 2 is held according to the following steps
Row:
Step 21 carries out down-sampled, after acquisition is down-sampled three using the method that voxel grid filters to the three-dimensional point cloud
Dimension point cloud；
Step 22, to it is described it is down-sampled after three-dimensional point cloud carry out straight-through filtering, filtered three-dimensional point cloud is led directly in acquisition；
Step 23 leads directly to filtered three-dimensional point cloud and carries out statistical filtering to described, the three-dimensional point cloud that obtains that treated.
3. three-dimensional point cloud dividing method as claimed in claim 2, which is characterized in that in the step 3, after pretreatment
The depth value of three-dimensional point is less than depth threshold using RANSAC algorithm by the depth value size of each three-dimensional point in three-dimensional point cloud
All three-dimensional point deletions, obtain the three-dimensional point cloud of non-planar area.
4. three-dimensional point cloud dividing method as described in claim 1, which is characterized in that the method pair using Edge Search
When all three-dimensional points are handled in the initial recess area three-dimensional point cloud, by the one of the initial recess area three-dimensional point cloud
A vertex successively obtains multiple three-dimensional points using boundary point maximum angular principle and forms sunk area edge as start boundary point.
5. a kind of three-dimensional point cloud segmenting device, for realizing three-dimensional point cloud described in any one of claim 1-4 claim
Dividing method, the device include that three-dimensional point cloud obtains module, three-dimensional point cloud preprocessing module, plane domain three-dimensional point cloud
Except module and sunk area three-dimensional point cloud divide module；
The three-dimensional point cloud obtains module and connect with three-dimensional point cloud preprocessing module, for be parallel to the plane domain
Direction acquires the three-dimensional point cloud containing sunk area and plane domain；
The three-dimensional point cloud preprocessing module is connect with the plane domain three-dimensional point cloud removal module, for described
Three-dimensional point cloud carries out pretreatment operation, obtains pretreated three-dimensional point cloud, the pretreatment operation includes successively executing
Down-sampled operation and filtering operation；
The plane domain three-dimensional point cloud removal module is connect with the sunk area three-dimensional point cloud segmentation module, is used for
In the three-dimensional point cloud of the non-planar area, using the maximum three-dimensional point of depth value as seed point, using the side of region growing
Method obtains initial recess area three-dimensional point cloud；
It is also used to handle all three-dimensional points in the initial recess area three-dimensional point cloud using the method for Edge Search, obtain
It obtains multiple three-dimensional points and forms sunk area edge；
It is also used to obtain three-dimensional point set all within the sunk area edge sunk area three-dimensional point cloud.
6. three-dimensional point cloud segmenting device as claimed in claim 5, which is characterized in that the three-dimensional point cloud preprocessing module,
It is Intelligent mobile equipment that plane domain three-dimensional point cloud, which removes module and sunk area three-dimensional point cloud segmentation module,.
7. three-dimensional point cloud segmenting device as claimed in claim 6, which is characterized in that the Intelligent mobile equipment is mobile phone.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811589435.9A CN109816664B (en) | 2018-12-25 | 2018-12-25 | Three-dimensional point cloud segmentation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811589435.9A CN109816664B (en) | 2018-12-25 | 2018-12-25 | Three-dimensional point cloud segmentation method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109816664A true CN109816664A (en) | 2019-05-28 |
| CN109816664B CN109816664B (en) | 2023-05-09 |
Family
ID=66602476
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811589435.9A Active CN109816664B (en) | 2018-12-25 | 2018-12-25 | Three-dimensional point cloud segmentation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109816664B (en) |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110443836A (en) * | 2019-06-24 | 2019-11-12 | 中国人民解放军战略支援部队信息工程大学 | A kind of point cloud data autoegistration method and device based on plane characteristic |
| CN110728688A (en) * | 2019-10-17 | 2020-01-24 | 杭州电子科技大学上虞科学与工程研究院有限公司 | Energy optimization-based three-dimensional mesh model segmentation method and system |
| CN110793437A (en) * | 2019-10-23 | 2020-02-14 | 珠海格力智能装备有限公司 | Positioning method and device of manual operator, storage medium and electronic equipment |
| CN112166457A (en) * | 2019-08-29 | 2021-01-01 | 深圳市大疆创新科技有限公司 | Point cloud segmentation method, system and movable platform |
| CN112365582A (en) * | 2020-11-17 | 2021-02-12 | 电子科技大学 | Countermeasure point cloud generation method, storage medium and terminal |
| CN113723468A (en) * | 2021-08-06 | 2021-11-30 | 西南科技大学 | Object detection method of three-dimensional point cloud |
| CN114596331A (en) * | 2020-12-04 | 2022-06-07 | 中国移动通信有限公司研究院 | Point cloud background generation method and device and communication equipment |
| CN115375699A (en) * | 2022-10-25 | 2022-11-22 | 杭州华橙软件技术有限公司 | Point cloud segmentation method, mobile robot and computer-readable storage medium |
| CN115903777A (en) * | 2022-10-11 | 2023-04-04 | 深圳优地科技有限公司 | Method, device and equipment for determining robot drop environment |
| CN116716781A (en) * | 2023-06-06 | 2023-09-08 | 山东大学 | Intelligent repair device and method for pavement diseases based on 3D printing |
| CN117115391A (en) * | 2023-10-24 | 2023-11-24 | 中科云谷科技有限公司 | Model updating method, device, computer equipment and computer readable storage medium |
| CN117994272A (en) * | 2024-04-07 | 2024-05-07 | 武汉科技大学 | A point cloud segmentation method and system for industrial disorderly placed soft package piles |
| CN119693442A (en) * | 2024-12-03 | 2025-03-25 | 合肥工业大学 | A method for rapid identification and detection of pothole volume for road repair |
| US12293593B2 (en) | 2019-12-12 | 2025-05-06 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | Object detection method, object detection device, terminal device, and medium |
| US12315165B2 (en) | 2019-12-12 | 2025-05-27 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | Object detection method, object detection device, terminal device, and medium |
| US12333830B2 (en) | 2019-12-12 | 2025-06-17 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | Target detection method, device, terminal device, and medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2011070927A1 (en) * | 2009-12-11 | 2011-06-16 | 株式会社トプコン | Point group data processing device, point group data processing method, and point group data processing program |
| CN105989604A (en) * | 2016-02-18 | 2016-10-05 | 合肥工业大学 | Target object three-dimensional color point cloud generation method based on KINECT |
-
2018
- 2018-12-25 CN CN201811589435.9A patent/CN109816664B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2011070927A1 (en) * | 2009-12-11 | 2011-06-16 | 株式会社トプコン | Point group data processing device, point group data processing method, and point group data processing program |
| CN105989604A (en) * | 2016-02-18 | 2016-10-05 | 合肥工业大学 | Target object three-dimensional color point cloud generation method based on KINECT |
Non-Patent Citations (1)
| Title |
|---|
| 赵凯等: "一种城市环境三维点云配准的预处理方法", 《光电工程》 * |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110443836A (en) * | 2019-06-24 | 2019-11-12 | 中国人民解放军战略支援部队信息工程大学 | A kind of point cloud data autoegistration method and device based on plane characteristic |
| CN110443836B (en) * | 2019-06-24 | 2022-03-18 | 中国人民解放军战略支援部队信息工程大学 | A method and device for automatic registration of point cloud data based on plane features |
| CN112166457A (en) * | 2019-08-29 | 2021-01-01 | 深圳市大疆创新科技有限公司 | Point cloud segmentation method, system and movable platform |
| CN110728688B (en) * | 2019-10-17 | 2022-02-22 | 杭州电子科技大学上虞科学与工程研究院有限公司 | Energy optimization-based three-dimensional mesh model segmentation method and system |
| CN110728688A (en) * | 2019-10-17 | 2020-01-24 | 杭州电子科技大学上虞科学与工程研究院有限公司 | Energy optimization-based three-dimensional mesh model segmentation method and system |
| CN110793437A (en) * | 2019-10-23 | 2020-02-14 | 珠海格力智能装备有限公司 | Positioning method and device of manual operator, storage medium and electronic equipment |
| US12293593B2 (en) | 2019-12-12 | 2025-05-06 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | Object detection method, object detection device, terminal device, and medium |
| US12333830B2 (en) | 2019-12-12 | 2025-06-17 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | Target detection method, device, terminal device, and medium |
| US12315165B2 (en) | 2019-12-12 | 2025-05-27 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | Object detection method, object detection device, terminal device, and medium |
| CN112365582A (en) * | 2020-11-17 | 2021-02-12 | 电子科技大学 | Countermeasure point cloud generation method, storage medium and terminal |
| CN114596331A (en) * | 2020-12-04 | 2022-06-07 | 中国移动通信有限公司研究院 | Point cloud background generation method and device and communication equipment |
| CN113723468A (en) * | 2021-08-06 | 2021-11-30 | 西南科技大学 | Object detection method of three-dimensional point cloud |
| CN113723468B (en) * | 2021-08-06 | 2023-08-04 | 西南科技大学 | Object detection method of three-dimensional point cloud |
| CN115903777A (en) * | 2022-10-11 | 2023-04-04 | 深圳优地科技有限公司 | Method, device and equipment for determining robot drop environment |
| CN115375699A (en) * | 2022-10-25 | 2022-11-22 | 杭州华橙软件技术有限公司 | Point cloud segmentation method, mobile robot and computer-readable storage medium |
| CN116716781A (en) * | 2023-06-06 | 2023-09-08 | 山东大学 | Intelligent repair device and method for pavement diseases based on 3D printing |
| CN117115391B (en) * | 2023-10-24 | 2024-01-12 | 中科云谷科技有限公司 | Model updating method, device, computer equipment and computer readable storage medium |
| CN117115391A (en) * | 2023-10-24 | 2023-11-24 | 中科云谷科技有限公司 | Model updating method, device, computer equipment and computer readable storage medium |
| CN117994272A (en) * | 2024-04-07 | 2024-05-07 | 武汉科技大学 | A point cloud segmentation method and system for industrial disorderly placed soft package piles |
| CN119693442A (en) * | 2024-12-03 | 2025-03-25 | 合肥工业大学 | A method for rapid identification and detection of pothole volume for road repair |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109816664B (en) | 2023-05-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109816664A (en) | A kind of three-dimensional point cloud dividing method and device | |
| CN110717983B (en) | Building elevation three-dimensional reconstruction method based on knapsack type three-dimensional laser point cloud data | |
| CN111932688B (en) | Indoor plane element extraction method, system and equipment based on three-dimensional point cloud | |
| CN109900338B (en) | Method and device for measuring volume of pavement pit | |
| CN111724433B (en) | Crop phenotype parameter extraction method and system based on multi-view vision | |
| CN106970375B (en) | A method for automatically extracting building information from airborne lidar point clouds | |
| CN106529469B (en) | Unmanned aerial vehicle-mounted LiDAR point cloud filtering method based on self-adaptive gradient | |
| CN103017739B (en) | Manufacturing method of true digital ortho map (TDOM) based on light detection and ranging (LiDAR) point cloud and aerial image | |
| CN105354883B (en) | The quick subtle three-dimensional modeling methods of 3ds Max and system based on a cloud | |
| CN105488770B (en) | A kind of airborne laser radar point cloud filtering method of object-oriented | |
| CN104463871B (en) | The extraction of streetscape dough sheet and optimization method based on vehicle-mounted LiDAR point cloud data | |
| CN108734728A (en) | A kind of extraterrestrial target three-dimensional reconstruction method based on high-resolution sequence image | |
| CN102607569B (en) | Method for automatically generating data of multiple small scales by using large scale water system in navigation | |
| CN107123164A (en) | Keep the three-dimensional rebuilding method and system of sharp features | |
| CN105069843A (en) | Rapid extraction method for dense point cloud oriented toward city three-dimensional modeling | |
| CN108765568A (en) | A kind of multi-level building quick three-dimensional reconstructing method based on laser radar point cloud | |
| CN112365534B (en) | Large coal pile volume measurement method based on monocular camera three-dimensional reconstruction | |
| CN113537141B (en) | A rapid detection method and system for dam piping and landslide diseases | |
| CN107392954A (en) | A kind of gross error point elimination method based on sequence image | |
| CN113409332A (en) | Building plane segmentation method based on three-dimensional point cloud | |
| CN109191484B (en) | A method for fast extraction of planar slices from airborne lidar point clouds | |
| CN114119902A (en) | Building extraction method based on unmanned aerial vehicle inclined three-dimensional model | |
| CN116758217A (en) | A three-dimensional reconstruction method of object surface based on fixed sensors | |
| CN118887539A (en) | A method for mining field wheat population characteristics based on three-dimensional point cloud | |
| CN111932574A (en) | Building facade point cloud extraction system and method based on multi-level semantic features |
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 | ||
| CB02 | Change of applicant information |
Address after: 710000 No. 605, west section, 6th floor, block B, operation and maintenance international headquarters building, No. 501, East Chang'an Street, national civil aerospace industry base, Xi'an, Shaanxi Province Applicant after: XI'AN ZHONGKE TIANTA TECHNOLOGY Co.,Ltd. Address before: Room 311, Zutong Building, No. 17 Information Avenue, New Industrial Park, High tech Zone, Xi'an City, Shaanxi Province, 710000 Applicant before: XI'AN ZHONGKE TIANTA TECHNOLOGY Co.,Ltd. |
|
| CB02 | Change of applicant information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |