CN104143194A - Point cloud partition method and device - Google Patents
Point cloud partition method and device Download PDFInfo
- Publication number
- CN104143194A CN104143194A CN201410410469.2A CN201410410469A CN104143194A CN 104143194 A CN104143194 A CN 104143194A CN 201410410469 A CN201410410469 A CN 201410410469A CN 104143194 A CN104143194 A CN 104143194A
- Authority
- CN
- China
- Prior art keywords
- cluster
- point
- ground
- segmentation
- point cloud
- 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 Analysis (AREA)
- Measurement Of Velocity Or Position Using Acoustic Or Ultrasonic Waves (AREA)
Abstract
The invention discloses a point cloud partition method and device, and relates to the technical field of three-dimensional reconstruction. The point cloud partition method includes the following steps: (S1) line-by-line scanning is carried out on a measured scene through a depth transducer to obtain depth information of the measured scene, and coordinate transformation is carried out on the depth information of the measured scene to obtain three-dimensional information of the measured scene under a local coordinate system; (S2) ground point clouds are partitioned from the three-dimensional information; (S3) clustering partition is carried out on non-ground point clouds in the self-adaptation threshold value clustering partition mode, wherein the non-ground point clouds are the other point clouds except the ground point clouds in the three-dimensional information. By means of the point cloud partition method and device, clustering partition is carried out on the non-ground point clouds in the self-adaptation threshold value clustering partition mode, and insufficient partition and excessive partition caused by clustering partition on the non-ground point clouds with fixed threshold values are effectively avoided.
Description
Technical field
The present invention relates to three-dimensional reconstruction field, particularly a kind of some cloud dividing method and device.
Background technology
For the three-dimensional reconstruction of large scene, because it is in the important application of the aspects such as three-dimensional city map, roadupkeep, city planning, receive very big concern.Utilize depth transducer and the position and attitude sensor three-dimensional information based on fixed station or mobile platform collection surrounding environment, because its efficient, real-time, high-precision characteristic is widely adopted.Because the scene of scanning comprises dissimilar object, such as ground, buildings, trees, vehicle etc., before carrying out three-dimensional reconstruction, need to cut apart cloud data corresponding dissimilar object is separated each other by a cloud, to each object is carried out respectively to a cloud modeling.
Current great majority point cloud dividing method is that unordered, discrete some cloud is processed, and does not use the order that in data acquisition, depth transducer scans by column successively.In some cloud dividing method, the especially cluster segmentation method based on distance of cluster segmentation method, because its method complexity is low, be easy to realization, is most commonly used to cutting apart of spatial point cloud.But due to the existence of ground point cloud in outdoor large scene, and use separately the clustering method based on distance to be difficult to effectively ground and non-ground object be cut apart, the method is combined with other method conventionally, first extracts and belongs to the some cloud on ground, more non-ground point cloud is carried out to cluster segmentation.
In non-ground point cloud cluster segmentation part, the cluster segmentation method based on fixed threshold radius of current more employing, threshold value to choose the impact of result on cutting apart very large, excessive if threshold value is chosen, the nearer wisp of spacing distance may not can by separately (less divided); If threshold value is chosen too small, the object (as buildings) that spacing distance is larger may be divided into multiple clusters (over-segmentation).
Summary of the invention
For fear of adopting fixed threshold to carry out to non-ground point cloud less divided and the over-segmentation problem that cluster segmentation causes, the invention provides a kind of some cloud dividing method, said method comprising the steps of:
S1: by depth transducer, tested scene is scanned by column, to obtain the depth information of described tested scene, the depth information of described tested scene is carried out to coordinate conversion, to obtain the three-dimensional information of described tested scene under local coordinate;
S2: be partitioned into ground point cloud from described three-dimensional information;
S3: the cluster segmentation mode by adaptive threshold is carried out cluster segmentation to non-ground point cloud, described non-ground point cloud is other clouds except described ground point cloud in described three-dimensional information.
Wherein, in step S2, from described three-dimensional information, be partitioned into ground point cloud and specifically comprise:
S201: travel through described three-dimensional information with the unit of classifying as, and using row that traverse as working as prostatitis;
S202: according to the described each point when prostatitis of direction traversal from bottom to up;
S203: the absolute value of the Z axis coordinate of current point that calculating traverses and the difference of the floor level of described depth transducer position, if described absolute value is less than height threshold, using described current point as first initial point Ps (1), execution step S204, otherwise return to step 202;
S204: detect described initial point Ps (k) cut off Pe (k) afterwards, and using described initial point Ps (k) and cut off Pe (k) as a pair of;
S205: detect cut off Pe (k) initial point Ps (k+1) afterwards, if initial point Ps (k+1) detected, make k=k+1, and return to step S204, otherwise the point between every couple of initial point Ps (k) and cut off Pe (k) and every couple of initial point Ps (k) and cut off Pe (k) adds described ground point cloud as ground point, and performs step S206;
S206: whether each row that judges described three-dimensional information are all traversed, and if so, perform step S3, otherwise return to step S202.
Wherein, in step S3, by the cluster segmentation mode of adaptive threshold, non-ground point cloud is carried out to cluster segmentation and specifically comprises:
S301: adopt fixing initial threshold dth (0) to carry out cluster segmentation to described non-ground point cloud;
S302: each cluster that traversal cluster segmentation obtains;
S303: the adaptive threshold dth (w) that calculates the current cluster traversing according to the cluster scale of construction, judge whether described adaptive threshold dth (w) is less than described initial threshold radius dth (0), if so, based on described adaptive threshold dth (w), described current cluster is further cut apart;
S304: judge that whether each cluster that described cluster segmentation acquires is all traversed, and if not, returns to step S303.
Wherein, in step S303, the adaptive threshold dth (w) that calculates the current cluster traversing according to the cluster scale of construction specifically comprises:
S3031: travel through the each point in described current cluster, current in conjunction with its neighborhood point fit Plane by what traverse, and matching is obtained to the normal vector of plane as the normal vector of described current point;
S3032: calculate in described current cluster the mean value of normal vector a little, and normal vector using described mean value as described current cluster;
S3033: the institute in described current cluster is a little all projected on projection plane, chooses the minimum rectangle that comprises all subpoints on described projection plane, described projection plane is the plane vertical with the normal vector of described current cluster;
S3034: calculate the adaptive threshold dth (w) of described current cluster by following formula,
dth(w)=k
1·L(w)+k
2
Wherein, the bond length that L (w) is described minimum rectangle, k
1and k
2for constant.
Wherein, in step S3, by the cluster segmentation mode of adaptive threshold, non-ground point cloud is carried out to cluster segmentation and specifically comprises:
S311: travel through the each point in described non-ground point cloud, calculate the adaptive threshold of the current point traversing;
S312: described non-ground point cloud is carried out to cluster segmentation according to the adaptive threshold of each point in described non-ground point cloud.
Wherein, in step S311, calculate the adaptive threshold dth of the current point traversing by following formula,
Wherein, L is the depth information of described current point, A is the angular resolution of described depth transducer, v is described depth transducer translational speed in the horizontal direction, f is the sweep frequency of described depth transducer, α is the angle between sweep trace place plane and the direction of v, the plane at the central point that described sweep trace place plane is described depth transducer and the T row place of described tested scene, described T classifies row that comprise described current point as, β is the angle on testee surface in sweep trace place plane and tested scene, θ is that the T of described tested scene is listed as the angle between horizontal direction, A
wfor the effective angle of depth transducer scanning, the error amount that Δ L is L, the error amount that Δ v is v, the error amount that Δ α is α.
Wherein, after step S3, also comprise:
S4: merged by the cluster of over-segmentation in the cluster that step S3 is obtained.
Wherein, step S4 specifically comprises:
S401: each cluster that traversal cluster segmentation acquires;
S402: the each point in the current cluster q traversing is traveled through, current in conjunction with its neighborhood point fit Plane by what traverse, and matching is obtained to the normal vector of plane as the normal vector of described current point;
S403: calculate in described current cluster q the mean value of normal vector a little, and normal vector using described mean value as described current cluster;
S404: the institute in described current cluster q is a little all projected on projection plane, calculates the projected area of described current cluster q, described projection plane is the plane vertical with the normal vector of described current cluster q;
S405: judge whether described projected area is greater than preset area, if so, using described current cluster q as large scale of construction cluster;
S406: judge whether each cluster that cluster segmentation acquires is all traversed, and if so, performs step S407, otherwise return to step S402;
S407: if the spacing distance between two large scale of construction clusters is less than predeterminable range, be a cluster by two of correspondence large scale of construction Cluster mergings, until the spacing distance between any two large scale of construction clusters is all greater than described predeterminable range, the spacing distance between described two large scale of construction clusters is the distance between nearest two points in two clusters.
Wherein, after step S407, also comprise:
Each cluster that S408: traversal step S407 acquires;
S409: the difference DELTA H between height and the floor level of the current cluster r that calculating traverses, the height of described current cluster r is described current cluster r a little in the minimum value of Z axis coordinate, the floor level of described current cluster r is the maximal value of Z axis coordinate in first initial point of each some institute's respective column in described current cluster r;
S410: judge whether described difference DELTA H is greater than preset difference value, if so, using described current cluster r as suspension cluster;
S411: whether each cluster that determining step S407 acquires is all traversed, if so, performs step S412, otherwise returns to step S409;
S412: by the extremely cluster nearest with it of each suspension Cluster merging, until each suspension cluster is all incorporated in a non-suspension cluster.
The invention also discloses a kind of some cloud segmenting device, described device comprises:
Scan conversion module, for tested scene being scanned by column by depth transducer, to obtain the depth information of described tested scene, the depth information of described tested scene is carried out to coordinate conversion, to obtain the three-dimensional information of described tested scene under local coordinate;
Module is cut apart on ground, for being partitioned into ground point cloud from described three-dimensional information;
Module is cut apart on non-ground, for the cluster segmentation mode by adaptive threshold, non-ground point cloud is carried out to cluster segmentation, and described non-ground point cloud is other clouds except described ground point cloud in described three-dimensional information.
The present invention carries out cluster segmentation by the cluster segmentation mode of adaptive threshold to non-ground point cloud, has effectively avoided adopting fixed threshold to carry out to non-ground point cloud less divided and the over-segmentation problem that cluster segmentation causes.
Brief description of the drawings
When Fig. 1 is the present invention to tested scene scanning, based on the structured flowchart of scanning system;
Fig. 2 is the depth information of the tested scene that collects according to the scanning system of Fig. 1;
Fig. 3 is the process flow diagram of the some cloud dividing method of one embodiment of the present invention;
Fig. 4 is the schematic diagram of an embodiment of the present invention at a sweep trace of depth transducer collection;
Fig. 5 is the schematic diagram of the some cloud on the sweep trace shown in Fig. 4;
Fig. 6 is the schematic diagram of an embodiment of the present invention while cutting apart on ground;
Fig. 7 a be an embodiment of the present invention sweep trace from bottom to up first point of direction belong to topocentric schematic diagram;
Fig. 7 b be an embodiment of the present invention sweep trace from bottom to up first point of direction belong to non-topocentric schematic diagram;
Fig. 8 is that the bond length of the minimum rectangle of an embodiment of the present invention is obtained process flow diagram;
Fig. 9 is the schematic diagram of the cluster projected area of an embodiment of the present invention;
Figure 10 is the design sketch that is partitioned into ground point cloud of an embodiment of the present invention;
Figure 11 be an embodiment of the present invention by the design sketch after being merged by the cluster of over-segmentation;
Figure 12 is the neighbourhood model of the impact point A1 of an embodiment of the present invention;
Figure 13 is the structured flowchart of the some cloud segmenting device of one embodiment of the present invention.
Embodiment
Below in conjunction with drawings and Examples, the specific embodiment of the present invention is described in further detail.Following examples are used for illustrating the present invention, but are not used for limiting the scope of the invention.
With reference to Fig. 1, during to the scanning of tested scene, institute based on scanning system formed by storage control module, depth transducer and position and attitude sensor, described depth transducer and position and attitude sensor are all placed on above mobile platform.Storage control module provides order, make the depth information (with reference to Fig. 2) of the tested scene that depth transducer acquisition scans arrives, meanwhile, position and attitude sensor record is in information such as position, attitude and the mobile platform gait of march of each moment mobile platform.By the depth information obtaining under depth transducer coordinate system, can be by coordinate transform, obtain the three-dimensional information of tested scene under the attitude sensor coordinate system of position, obtain again the three-dimensional information of tested scene under local coordinate by coordinate transform, because the outdoor scene gathering is generally made up of a large amount of spatial point, the set of these spatial point is also referred to as a cloud.
In present embodiment, described depth transducer is selected laser radar, and position and attitude sensor is selected GPS, and described position and attitude sensor also can be selected GPS/IMU (inertial navigation) integrated navigation system.
Fig. 3 is the process flow diagram of the some cloud dividing method of one embodiment of the present invention; With reference to Fig. 3, said method comprising the steps of:
S1: by depth transducer, tested scene is scanned by column, to obtain the depth information of described tested scene, the depth information of described tested scene is carried out to coordinate conversion, to obtain the three-dimensional information of described tested scene under local coordinate;
S2: be partitioned into ground point cloud from described three-dimensional information;
S3: the cluster segmentation mode by adaptive threshold is carried out cluster segmentation to non-ground point cloud, described non-ground point cloud is other clouds except described ground point cloud in described three-dimensional information.
In step S1, when tested scene being scanned by column by depth transducer, can adopt direction of scanning from bottom to up or from top to bottom to scan by column, first point that scanning obtains is sweep trace starting point, and last point is sweep trace terminating point.
In the process that depth transducer moves at mobile platform, scan by column with the angular resolution of fixing, obtain the depth information of corresponding environment, every column data of scanning is also referred to as sweep trace.Every column data has fixing acquisition order and counts, and the depth information of the final tested scene obtaining is by forming by all sweep traces of showing order, and the data that each testee is corresponding are also made up of jointly the part point in some sweep traces.Because depth transducer is taked the mode gathering by column, counting of every row is fixing with order, and the depth information therefore obtaining has by the characteristic of showing order.Shown in Fig. 4, be under a given scenario, the schematic diagram of the sweep trace that the depth transducer on mobile platform gathers in the course of the work, depth transducer scans from bottom to top with the angular resolution of fixing, scan successively ground, wisp, ground and buildings, obtain depth information corresponding to each point.Triangle pair should be the ground data that collect, and circle corresponds to the data on the non-ground object collecting.The information of the depth information binding site attitude sensor that depth transducer is collected, through coordinate transform, finally obtains the coordinate of each analyzing spot under local coordinate, is the three-dimensional information of described tested scene under local coordinate.
For accurately determining initially face amount, utilize continuous feature extraction ground in segmentation that the ground point on each column scan line has, section, to improve to accurately the cutting apart of ground point cloud, preferably simultaneously, in step S2, from described three-dimensional information, be partitioned into ground point cloud and specifically comprise:
S201: travel through described three-dimensional information with the unit of classifying as, and using row that traverse as working as prostatitis;
S202: according to the described each point when prostatitis of direction traversal from bottom to up;
S203: the absolute value of the Z axis coordinate of current point that calculating traverses and the difference of the floor level of described depth transducer position, if described absolute value is less than height threshold, using described current point as first initial point Ps (1), execution step S204, otherwise return to step 202;
S204: detect described initial point Ps (k) cut off Pe (k) afterwards, and using described initial point Ps (k) and cut off Pe (k) as a pair of;
S205: detect cut off Pe (k) initial point Ps (k+1) afterwards, if initial point Ps (k+1) detected, make k=k+1, and return to step S204, otherwise the point between every couple of initial point Ps (k) and cut off Pe (k) and every couple of initial point Ps (k) and cut off Pe (k) adds described ground point cloud as ground point, and performs step S206;
S206: whether each row that judges described three-dimensional information are all traversed, and if so, perform step S3, otherwise return to step S202.
Realizing the cluster segmentation of the non-ground point cloud of adaptive threshold can adopt in two ways: a kind of is the adaptive threshold partitioning scheme based on the scale of construction, and another kind is the adaptive threshold partitioning scheme based on multiparameter.
For ease of realizing the adaptive threshold partitioning scheme based on the scale of construction, preferably, in step S3, by the cluster segmentation mode of adaptive threshold, non-ground point cloud is carried out to cluster segmentation and specifically comprises:
S301: adopt fixing initial threshold dth (0) to carry out cluster segmentation to described non-ground point cloud;
S302: each cluster that traversal cluster segmentation obtains;
S303: the adaptive threshold dth (w) that calculates the current cluster traversing according to the cluster scale of construction, judge whether described adaptive threshold dth (w) is less than described initial threshold radius dth (0), if so, based on described adaptive threshold dth (w), described current cluster is further cut apart;
S304: judge that whether each cluster that described cluster segmentation acquires is all traversed, and if not, returns to step S303.
For ease of determining adaptive threshold according to the scale of construction, in step S303, the adaptive threshold dth (w) that calculates the current cluster traversing according to the cluster scale of construction specifically comprises:
S3031: travel through the each point in described current cluster, current in conjunction with its neighborhood point fit Plane by what traverse, and matching is obtained to the normal vector of plane as the normal vector of described current point;
S3032: calculate in described current cluster the mean value of normal vector a little, and normal vector using described mean value as described current cluster;
S3033: the institute in described current cluster is a little all projected on projection plane, chooses the minimum rectangle that comprises all subpoints on described projection plane, described projection plane is the plane vertical with the normal vector of described current cluster;
S3034: calculate the adaptive threshold dth (w) of described current cluster by following formula,
dth(w)=k
1·L(w)+k
2
Wherein, the bond length that L (w) is described minimum rectangle, k
1and k
2for constant.
For ease of realizing the adaptive threshold partitioning scheme based on multiparameter, preferably, in step S3, by the cluster segmentation mode of adaptive threshold, non-ground point cloud is carried out to cluster segmentation and specifically comprises:
S311: travel through the each point in described non-ground point cloud, calculate the adaptive threshold of the current point traversing;
S312: described non-ground point cloud is carried out to cluster segmentation according to the adaptive threshold of each point in described non-ground point cloud.
For ease of determining adaptive threshold according to multiparameter, in step S311, calculate the adaptive threshold dth of the current point traversing by following formula,
Wherein, L is the depth information of described current point, A is the angular resolution of described depth transducer, v is described depth transducer translational speed in the horizontal direction, f is the sweep frequency of described depth transducer, α is the angle between sweep trace place plane and the direction of v, the plane at the central point that described sweep trace place plane is described depth transducer and the T row place of described tested scene, described T classifies row that comprise described current point as, β is the angle on testee surface in sweep trace place plane and tested scene, θ is that the T of described tested scene is listed as the angle between horizontal direction, A
wfor the effective angle of depth transducer scanning, the error amount that Δ L is L, the error amount that Δ v is v, the error amount that Δ α is α.
For solving the problem of over-segmentation, after step S3, also comprise:
S4: merged by the cluster of over-segmentation in the cluster that step S3 is obtained.
For solving large scale of construction cluster by the problem of over-segmentation, preferably, step S4 specifically comprises:
S401: each cluster that traversal cluster segmentation acquires;
S402: the each point in the current cluster q traversing is traveled through, current in conjunction with its neighborhood point fit Plane by what traverse, and matching is obtained to the normal vector of plane as the normal vector of described current point;
S403: calculate in described current cluster q the mean value of normal vector a little, and normal vector using described mean value as described current cluster;
S404: the institute in described current cluster q is a little all projected on projection plane, calculates the projected area of described current cluster q, described projection plane is the plane vertical with the normal vector of described current cluster q;
S405: judge whether described projected area is greater than preset area, if so, using described current cluster q as large scale of construction cluster;
S406: judge whether each cluster that cluster segmentation acquires is all traversed, and if so, performs step S407, otherwise return to step S402;
S407: if the distance between the central point of two large scale of construction clusters is less than predeterminable range, be a cluster by two of correspondence large scale of construction Cluster mergings, until the distance between the central point of any two large scale of construction clusters is all greater than described predeterminable range.
For solving suspension cluster by the problem of over-segmentation, preferably, after step S407, also comprise:
Each cluster that S408: traversal step S407 acquires;
S409: the difference DELTA H between height and the floor level of the current cluster r that calculating traverses, the height of described current cluster r is described current cluster r a little in the minimum value of Z axis coordinate, the floor level of described current cluster r is the maximal value of Z axis coordinate in first initial point of each some institute's respective column in described current cluster r;
S410: judge whether described difference DELTA H is greater than preset difference value, if so, using described current cluster r as suspension cluster;
S411: whether each cluster that determining step S407 acquires is all traversed, if so, performs step S412, otherwise returns to step S409;
S412: by the extremely cluster nearest with it of each suspension Cluster merging, until each suspension cluster is all incorporated in a non-suspension cluster.
Embodiment 1
Data on analysis scan line, as shown in Figure 5, can find out, some p
1to a p
a1, some p
a4to a p
a5for ground cloud data, some p
a2to a p
a3with a p
a6to a p
nfor the cloud data on non-ground object.Can obtain in conjunction with most scene analysis, on a sweep trace, the most segmentation existence of cloud data and the section that belong to ground are continuously interior, as Fig. 5 mid point p
1to a p
a1some cloud tract, some p
a4to a p
a5some cloud tract be two sections of some cloud sections that belong to ground; Same section belongs to the some cloud on ground, and local relief is less between points, and elevation difference is also less.Analyze in conjunction with above, for effectively extracting ground point cloud, detect by column the some cloud tract that belongs to ground.In every section of some cloud sequence that belongs to ground, defining first point is initial point Ps, and last point is cut off Pe, and initial point and cut off be the some p in corresponding diagram 5 respectively
1, p
a4with a p
a1, p
a5.
With a specific embodiment, the present invention is described below, but does not limit protection scope of the present invention.On the basis of above-mentioned embodiment, from described three-dimensional information, be partitioned into ground point cloud and specifically comprise:
Process by row, extract by column the some cloud section that belongs to ground in every sweep trace.It is Ps (k) that the initial point that k section in one article of processed sweep trace belongs to the some cloud tract on ground is set, cut off is Pe (k), initial point corresponding to some cloud tract that first paragraph belongs to ground is Ps (1), cut off is Pe (1), the method that the data of a sweep trace are carried out cutting apart on ground is mainly made up of three links, be respectively to detect first initial point Ps (1), detect cut off Pe (k) and detect the initial point Ps (k+1) after cut off.The flow process that the data of one row are carried out cutting apart on ground is as Fig. 6, and concrete steps comprise:
Step S2-1: detect first initial point Ps (1);
As shown in Fig. 7 a and 7b, triangle represents ground point, round dot represents non-ground point (wherein, Ps1 is with above-mentioned Ps (1), Pe1 is with above-mentioned Pe (1), the expression way of other mark is the same, again do not repeating), article one, sweep trace from bottom to up first point of direction both can belong to ground point, also can be non-ground point, in order to obtain accurately cutting apart of a sweep trace upper ground surface part and non-above ground portion, first must guarantee the correctness of first ground reference point.The detection of first initial point Ps (1) on sweep trace can utilize the known elements based on mobile platform, due to the position (X of depth transducer scanning center under local coordinate
lidar, Y
lidar, Z
lidar) be real-time storage, depth transducer scanning center is to the height (abbreviation podium level) of platform bottom, as shown in H in Fig. 4, can be used as systematic parameter uses, on mobile platform after erecting equipment, can draw podium level H by experiment under broad smooth environment, in system, not again dismounting in the situation that, podium level H is constant.Utilize position and the podium level of depth transducer scanning center, in conjunction with the elevation information of tested measuring point on sweep trace, can detect first initial point Ps (1) that belongs to ground on sweep trace, above-mentioned steps S2-1, comprises the steps:
(1): to processed J row depth transducer depth information, set-point sequence number i (i=0,1 ..., n) with i the corresponding local coordinate system (X putting
i, Y
i, Z
i), arranging that in this row depth information, first puts corresponding i=0, corresponding local coordinate system is (X
0, Y
0, Z
0);
(2): the Z axis coordinate Z that calculates i point under local coordinate system
ifloor level (Z with depth transducer position
lidar-H) difference DELTA h
i, computing formula is: Δ h
i=| (Z
lidar-H)-Z
i|;
(3): judge corresponding difference in height Δ h
iwhether be less than threshold value Zth, if Δ h
i<Zth, thinks that i point is for belonging to first initial point Ps (1) on ground in this column scan point; Otherwise make i=i+1, return to step (2);
Threshold value Zth considers situations such as measuring noise, ground local relief (comprising deceleration strip, traffic stud etc.), patient difference in height between two ground points on sweep trace.China national standard path regulation, height of projection on the road surfaces such as deceleration strip can not be higher than 5 centimetres, consider the smooth degree on ground and the impact of measuring noise, in the present embodiment, between two ground points, patient difference in height Zth is set to 15 centimetres to guarantee enough allowances again.
Step S2-2: detect initial point Ps (k) cut off Pe (k) afterwards;
Detect the initial point Ps (k) of k section ground point cloud tract afterwards, need further to detect the cut off Pe (k) of this section of some cloud sequence, and then k section ground point cloud section could be extracted.In the method proposing in the present invention, judge after the initial point Ps (k) of i point as k section ground point cloud tract, utilize initial point Ps (k) the angle θ that the line of adjacent 2 and XOY plane form afterwards as topocentric examination criteria, judge successively whether the point after i point belongs to k section ground point cloud tract, until the cut off Pe (k) of k section ground point cloud tract detected, think thus and started a little all to belong to ground to the institute of a Pe (k) cut-off by a Ps (k).
Under local coordinate, for 2 u of difference and the v that belong to ground on same sweep trace, the angle θ that their line and XOY plane form
u, vthe fluctuating quantity that is used for characterizing ground between u, v at 2, this angle θ is as topocentric examination criteria.θ specific formula for calculation is as follows:
For certain section of ground point set G={Ps (k) being partitioned into ..., Pe (k) }, because the fluctuating between adjacent 2 is less, this section of ground point concentrated 2 angle θ corresponding to j, j+1 of arbitrary neighborhood
j, j+1should be less than certain threshold value, should meet θ
j, j+1< θ
th.Thus can by initial point Ps (k) afterwards adjacent 2 corresponding angle information θ further detect ground sequence, until find cut off Pe (k), i.e. above-mentioned steps S2-2, specifically comprises the following steps:
(1) initial point Ps (k) (i point on corresponding sweep trace) first point being afterwards set is j point, j=i+1;
(2) under local coordinate, the line that calculating j point and j+1 are ordered and the angle θ of XOY plane
j, j+1, specific formula for calculation is as follows:
Due to the noise that exists radar and GPS to produce, only by with adjacent a bit corresponding angle θ
j, j+1the method that detects cut off Pe (k) is too sensitive, for avoiding noise effect, in the present embodiment, by the next point of j point consecutive point (being j+2 point), also takes in simultaneously, also calculates angle θ simultaneously
j, j+2;
(3) judge θ
j, j+1, θ
j, j+2whether be all greater than threshold value θ
th, if so, think that the fluctuating of consecutive point is excessive, judge that j point is as cut off Pe (k); Otherwise, make j=j+1, enter step (2);
Threshold value θ
thfluctuating quantity and system noise by local ground determine jointly, consider the above-mentioned various influence factor that may exist, threshold value θ in the present embodiment
thbe set as 17 ° to retain enough allowances.
Step S2-3: detect cut off Pe (k) initial point Ps (k+1) afterwards;
Obtain the cut off Pe (k) of k section ground point cloud tract afterwards, all be judged as to the point of some Pe (k) point that belongs to ground from a Ps (k), the local coordinate system of set-point Pe (k) (being j point corresponding in step S2-2) is (X
j, Y
j, Z
j).Belong on the basis that the cut off Pe (k) of the some cloud sequence on ground obtained at the preceding paragraph, need to detect next section of topocentric initial point Ps (k+1), and then continue to detect the some cloud sequence that next section belongs to ground.
For cut off Pe (k) on this column scan line (i.e. j point of these row) some m afterwards, use j, 2 corresponding angle θ of m
j, mand elevation Z
j, mconstraint carry out judging point m and whether belong to next initial point Ps (k+1), if meet θ
j, mbe less than threshold value θ
thand Z
j, mbe less than threshold value Z
th, decision-point m is next initial point Ps (k+1), otherwise thinks that a m is not initial point, continues the point after m point to judge.
Wherein j, 2 corresponding angle θ of m
j, mto give directions the line of j, m and the angle that XOY plane forms, angle θ in definition and computing formula and step S2-2
u, videntical; Elevation Z
j, mbe a j and the height difference of some m under local coordinate, computing formula is: Z
j, m=| Z
j-Z
m|; Threshold value θ
thdefinition and value and step S2-2 in threshold value θ
thidentical; The definition of threshold value Zth is identical with the threshold value Zth in step S2-1 with value.
The concrete implementation step of step S2-3 is as follows:
(1) cut off Pe (k) (i.e. j point) afterwards first being set is m point, i.e. m=j+1;
(2), under local coordinate, calculate the difference in height Z of cut off Pe (k) and some m
j, m, Z
j, m=| Z
j-Z
m|; The line that calculating cut off Pe (k) and m are ordered and the angle θ of XOY plane
j, m;
(3) judge whether Z
j, mbe less than threshold value Zth and θ
j, mbe less than threshold value θ
th, if so, think that m point is the initial point Ps (k+1) of next section of ground sequence; Otherwise make m=m+1, enter step (2);
Perform step S2-3 until cut off Pe (k) next initial point Ps (k+1) afterwards detected, or this row mid point Pe (k) institute afterwards is a little all judged, carry out step S2-4.
Step S2-4: judge whether cut off Pe (k) detects new initial point Ps (k+1) afterwards, if new initial point Ps (k+1) detected afterwards at cut off Pe (k), can make k=k+1 return to step S2-2 and continue to carry out; If new initial point Ps (k+1) do not detected afterwards at cut off Pe (k), on this sweep trace, cut off Pe (k) institute is afterwards a little all non-ground points, the institute that can judge this row cloud data is a little all processed, and all ground point cloud tracts of these row are all extracted.
For next sweep trace, carry out same step S2-1 to the identical processing of S2-4, thus by column, every row piecemeal ground-to-ground millet cake cloud extract, can complete cutting apart of ground point cloud.
Embodiment 2
According to the ground distributor segmentation method based on sweep trace, after ground point cloud and non-ground point cloud are cut apart, need to carry out cluster segmentation to non-ground point cloud, the some cloud that belongs to different objects in non-ground point cloud is separated from each other.
The cluster over-segmentation there will be for fear of the same fixed threshold of above-mentioned employing and less divided phenomenon, the present embodiment adopts the adaptive threshold partitioning scheme based on the scale of construction, the i.e. scale of construction based on object and the relation between required threshold value while cutting apart, as a rule, spacing distance between the object (as buildings) of the large scale of construction is larger, while cutting apart the object of the large scale of construction, threshold value should obtain larger, and spacing distance between the object of small volume (as automobile) can be less, while cutting apart small volume object, threshold value should obtain less, namely, the threshold value radius of cluster segmentation and the scale of construction of object exist and increase progressively relation.
Described adaptive threshold partitioning scheme concrete steps are as follows:
Step S3-1: use fixing initial threshold dth0, non-ground point cloud is carried out cutting apart based on the basis of European clustering method;
After arranging non-ground point being finished based on fixed threshold dth0 cluster, obtain W cluster, cluster numbering w, the corresponding w=1 of first cluster;
S3-2: for carrying out based on fixed threshold dth0 the cluster w that cluster obtains, calculate the adaptive threshold dth (w) based on the scale of construction of w;
For carrying out based on fixed threshold dth0 the cluster w that cluster obtains, need to calculate its corresponding adaptive threshold dth (w).As above-mentioned analysis, between the object (as buildings) of the large scale of construction, spacing distance is larger, and between the object of small volume (as automobile), spacing distance can be less.For characterizing the relation that increases progressively between threshold size and the object scale of construction, while is for the consideration of computation complexity aspect, self-adaption cluster radial design is the linear relationship being directly proportional to cluster size, and the computing formula of the adaptive threshold based on the scale of construction of w cluster is shown below:
dth(w)=k
1·L(w)+k
2
Characterize in the parameter of linear relationship at above formula, coefficient k 1, by relation acquisition between distance and the object scale of construction between the general familiar object of analysis, as the normal object occurring in the scenes such as buildings, automobile, bicycle, street lamp, is got k1=0.01m in the present embodiment; Coefficient k 2 is distinguishable distances minimum between two articles, in the present embodiment, is determined by the resolution of vehicle-mounted mobile platform depth transducer image data, gets k2=0.18m in the present embodiment.
Length L (w) in formula is the length of cluster w, and its concrete operation process as shown in Figure 8.Point in figure represents the point in cluster w.As shown in (a1) in Fig. 8, for the point in cluster w, by the some fit Plane within the scope of neighborhood around each point, obtain the normal vector of this plane, i.e. the normal vector of this point.According to identical method, obtain after the normal vector of each point, using in w normal vector a little average as the normal vector n of cluster w, as shown in (a2) in Fig. 8.Elect the plane vertical with the normal vector of this cluster as projection plane, by all spot projections in w, to this projecting plane, in Fig. 8, in (a3), the point on projecting plane is the subpoint obtaining after cluster projection.In this projection plane, choose the minimum rectangle that all subpoints can be included, in the both sides of this rectangle, shorter length of side L is defined as to the length L (w) of current cluster w, in (a4) in Fig. 8, rectangle is compared with as shown in the long L of minor face.
S3-3: whether the adaptive threshold dth (w) that judges cluster w is less than initial threshold radius dth0, if, based on adaptive threshold dth (w), cluster w is further cut apart, so that the object segmentation not separating during by initial segmentation is abundant; Otherwise think divided completing of cluster w, no longer carry out cutting apart based on adaptive threshold.
S3-4: judge whether all W cluster all to handle, if so, the cluster segmentation based on scale of construction adaptive threshold finishes; Otherwise, make w=w+1, return to step S3-2;
Embodiment 3
Based on adaptive threshold radius, non-ground points all in tested scene is carried out after cluster segmentation, due to the reason such as scanning is blocked, mobile platform swerves and testee is irregular, cluster segmentation result can exist object by the phenomenon of over-segmentation, mainly contain two class situations, one is for irregularly shaped objects such as trees, taking trees as example, because the associations such as trunk, each branch are less, be easy to be divided into multiple clusters, wherein only having trunk is cluster on the ground, and other cluster suspends on the ground, be referred to as the cluster that suspends.Also having one is exactly for large scale of construction objects such as buildingss, because the setting of threshold value radius is less, swerves or scans to exist under circumstance of occlusion can cause large scale of construction cluster to be too slit into the situation of multiple clusters at vehicle-mounted mobile platform.Therefore need the cluster of over-segmentation to merge processing.
When the large scale of construction cluster of over-segmentation is merged to processing, tested cluster is projected to the plane vertical with the normal vector of this cluster, the profile that then extracts projecting plane calculates the projected area of tested cluster.If the projected area of a cluster exceedes empirical value S
u, this cluster is judged as a large scale of construction cluster.If the spacing distance between any two large scale of construction clusters is less than the distance l requiring between buildings in general
u, these two clusters are considered to belong to same building thing and need to be merged into a cluster, and wherein the spacing distance between two large scale of construction clusters is the distance between two points nearest in these two clusters.The combining step of the large scale of construction cluster of over-segmentation is specific as follows:
(1) arrange carry out after cluster segmentation based on adaptive threshold, non-ground point cloud is divided into Q cluster, arrange cluster sequence number q (q=1,2 ..., Q), the corresponding q=1 of first cluster in segmentation result is set;
(2) for the point in cluster q, be similar in S2-2 the process shown in (a1) to (a3) in Fig. 8, by the some fit Plane within the scope of neighborhood around each point, obtain the normal vector of this plane, the i.e. normal vector of this point, according to identical method, obtains after the normal vector of each point, using in q normal vector a little average as the normal vector n of cluster q, by all spot projections in q to the plane vertical with n.Calculate afterwards the projected area S (q) of tested cluster q by convex hull method extraction cluster q profile of projection on projecting plane, as shown in Figure 9;
(3) whether the projected area S (q) of judgement cluster q is greater than threshold value S
u, if so, think that q is large scale of construction cluster; Otherwise think that q does not belong to large scale of construction cluster;
(4) determine whether all Q cluster is processed, if so, enter step (5); Otherwise make q=q+1, enter step (2);
(5) the large scale of construction cluster of over-segmentation is merged to processing, if the distance between any two large scale of construction clusters is less than the distance l requiring between buildings in general
u, these two clusters are considered to belong to same building thing, and these two Cluster mergings are become to a cluster.Until the distance between any two large scale of construction clusters is all greater than l
utime, the large scale of construction Cluster merging of over-segmentation finishes; Otherwise continue execution step (5);
In the present embodiment, consider tree and blocking of causing of buildings itself, determine whether the area threshold S of large scale of construction cluster
ube made as empirical value 30m
2, the general spacing l of two buildingss
ube made as empirical value 10m.
When the levitated object of over-segmentation is merged to processing, first need to judge whether cluster is suspension cluster.Cut apart link at ground point cloud, obtain by column ground point cloud and non-ground point cloud on every sweep trace, and then can mate corresponding floor level Zg for each non-ground point, to each the non-ground point on any surface sweeping line j, all corresponding floor level Zg of j row position, first of all non-ground point in the present embodiment, j being listed as and j row be the Z axis coordinate figure coupling of millet cake Ps (1) under local coordinate initially, thinks that the Z axis coordinate figure of a Ps (1) has reflected the floor level of the non-ground point position of j row;
Arrange the large scale of construction Cluster merging of over-segmentation is finished to rear total R cluster, to any one cluster r wherein, need to calculate the difference DELTA H between height and position and the corresponding floor level of this cluster r, and then judge whether it suspends.All mate a corresponding floor level Zg (t) owing to putting arbitrarily t in cluster r, get in cluster r the minimum value of Z axis coordinate be a little min{Z (t) be the height and position of cluster r, get the maximal value max{Zg (t) in the floor level Zg that in cluster r, all-pair is answered } for floor level corresponding to cluster r, think difference DELTA H=|min{Z (t) between height and position and the corresponding floor level of cluster r }-max{Zg (t) } |.For any cluster r, if corresponding Δ H is greater than threshold value Fth, think that cluster r is suspension cluster.For the cluster suspending, calculate the spacing distance between cluster, and this Cluster merging is arrived to its nearest cluster, until all finally merging to one, suspension cluster belongs in ground non-suspension cluster.Concrete steps are as follows:
(1) arrange after large scale of construction Cluster merging is processed, total R cluster, arrange cluster sequence number r (r=1,2 ..., R), the corresponding r=1 of first cluster;
(2) for cluster r, calculate respectively the height min{Z (t) of this cluster } and corresponding floor level max{Zg (t), and then calculate the difference DELTA H between height and the corresponding floor level of cluster r, i.e. Δ H=|min{Z (t) }-max{Zg (t) } |;
(3) judge the no threshold value Fth that is greater than of Δ H corresponding to cluster r, if so, think that r is suspension cluster; Otherwise think that r does not belong to suspension cluster;
In the present embodiment, consider the impact of error and stay certain allowance, setting threshold Fth is 1m.
(4) determine whether all R cluster has all been carried out to the judgement that suspends, if so, enter step (5); Otherwise make r=r+1, enter step (2);
(5) for any one suspension cluster e to be combined, calculate the spacing distance between other clusters in this cluster e and R, and this cluster e is merged to nearest cluster with interval.Belong in ground non-suspension cluster if suspension cluster finally all merges to one, the merging of the suspension cluster of over-segmentation is finished dealing with, otherwise continues to carry out (5); Wherein, the spacing distance between two suspension clusters refers to the distance of the central point of two clusters, and the coordinate of the central point of cluster is this cluster, and coordinate figure is a little averaging.
Figure 10 shows that the design sketch being partitioned into after ground point cloud.The ground distributor segmentation method based on sweep trace that the present invention is proposed is while carrying out Performance Evaluation, for the accuracy of method from two class error evaluations: a) originally belong to ground point but be mistaken for non-ground point, b) originally belonging to non-ground point and be mistaken for ground point.For assessing above-mentioned error, we select special scenes manually to add up that counting in ground and counts in non-ground, and carry out comparing calculation with the result that experiment obtains.Ground distributor segmentation method (the R.B.Rusu based on random sampling coherence method that the ground distributor segmentation method based on sweep trace that the present invention is proposed and R.B.Rusu propose, " Semantic3D object maps for everyday manipulation in human living environments, " KI-K ü nstliche Intelligenz, vol.24, no.4, pp.345-348, Nov.2010.) compare, by relevant experiment and calculating, show that the ground distributor segmentation method error rate based on sweep trace in this paper is 0.674%, the error rate of the method for R.B.Rusu is 1.52%.Aspect method time complexity, two kinds of methods all operate in the E7400 of Intel, 2.8GHz, the computer platform of 2GB internal memory, the time of the method that the present invention proposes and the operation of the method for R.B.Rusu is respectively 0.028 second and 4.820 seconds, the method that the present invention's proposition is described is faster, and can process nearly 22358179 data average p.s..Concrete performance comparison result is as following table:
Method | Error a (%) | Error b (%) | Total error (%) | Working time (s) |
The inventive method | 0.401 | 0.273 | 0.674 | 0.028 |
R.B.Rusu method | 0.919 | 0.623 | 1.52 | 4.820 |
Do as one likes energy comparing result is visible, and the overall performance of the inventive method is higher than the method for R.B.Rusu.
Figure 11 shows that non-ground point cloud is carried out cutting apart based on scale of construction adaptive threshold, result after again the cluster of over-segmentation being merged, cloud datas corresponding to vehicle different in Figure 11 have different gray-scale values, the method that represents has determined different vehicles and has been separated, can find out, method can be cut apart accurately to the large scene three-dimensional point cloud obtaining, and even can cut apart accurately for slope pavement, automobile etc. in scene.
Embodiment 4
The cluster over-segmentation there will be for fear of the same fixed threshold of above-mentioned employing and less divided phenomenon, the present embodiment adopts the adaptive threshold partitioning scheme based on multiparameter, and the described adaptive threshold partitioning scheme based on multiparameter specifically comprises:
First, set up the distance model of consecutive point in some cloud: based on mobile platform operation logic, along with platform moves forward, depth transducer scans by column successively to tested scene, suppose in subrange, a certain impact point point of proximity place face is around similar to plane, gathers schematic diagram as shown in figure 12, to impact point A
1and the modeling of neighborhood point, neighborhood point is chosen the some A as marked in Figure 12
2, B
1, B
2deng, respective distances is followed successively by impact point A
1distance A between its neighbouring point on the sweep trace of place
1a
2, distance A on its adjacent two sweep traces and between its nearest point and this point
1b
1, the some B nearest with it on its adjacent two sweep traces
1up and down 2 with this distance A
1b
2.Based on consecutive point distance model, derivation consecutive point spacing A
1a
2, A
1b
2, A
1b
1calculation expression.Due to A
1b
2>A
1b
1, and A
1b
2>A
1a
2, each impact point consecutive point distance around can be passed through A
1b
2with reference to determining.Consider the distance L of radar scanning center to measured point, mobile platform pace v, between radar scanning line place plane and car travel direction there is respectively error delta L in angle α, Δ v and Δ α, the cluster segmentation threshold value radius of each point is shown below:
Wherein, L is the depth information of described current point, A is the angular resolution of described depth transducer, v is described depth transducer translational speed in the horizontal direction, f is the sweep frequency of described depth transducer, α is the angle between sweep trace place plane and the direction of v, the plane at the central point that described sweep trace place plane is described depth transducer and the T row place of described tested scene, described T classifies row that comprise described current point as, β is that the surperficial angle of sweep trace place plane and testee is (because this parameter is a uncertain value, so the value of the present embodiment using a default angle as this parameter), θ is that the T of described tested scene is listed as the angle between horizontal direction, A
wfor the effective angle of depth transducer scanning, the error amount that Δ L is L, the error amount that Δ v is v, the error amount that Δ α is α.
In the parameter relating at above-mentioned formula, L, A, A
w, α, f has been defined as constant before collection, constant in image data process, and in the present embodiment, depth transducer range error Δ L is 0.03 meter, and angular resolution A is 0.5 °, i.e. π/360 radian, scanning effective range A
wbe 90 °, i.e. pi/2 radian, α is 45 ° and 135 °, sweep frequency is 75 hertz.β value is a preset value, gets β=15 °.L, v, Δ v and Δ α are the real time data of obtaining in scanning process, wherein radar scanning center is obtained by depth transducer to the distance L between being scanned a little, mobile platform travel speed v, velocity error Δ v, sweep trace angular error Δ α is obtained by position and attitude sensor.
For N non-ground point of j row, set-point sequence number I (I=1,2,, N), according to the above-mentioned cluster segmentation threshold expression based on multiparameter, calculate respectively the each self-corresponding cluster segmentation threshold value of each non-ground point, it is d that I threshold value corresponding to point is set
thI; The corresponding I=1 of first non-ground point of j row is set, for impact point I, the threshold value radius d corresponding according to this point
thI, judge whether the non-ground point that j-1 row and j list is less than threshold value d to the distance of this impact point I
thI, if so, thinking that corresponding point is the neighbor point of impact point I, it is a class that neighbor point and impact point I are gathered; Repeat above-mentioned steps, until all non-ground points have all been carried out to the cluster segmentation processing based on multiparameter adaptive threshold, to non-ground point, the cluster segmentation based on multiparameter adaptive threshold finishes.
Even if also do not carry out the merging processing of over-segmentation cluster, the segmentation result that cluster segmentation method based on multiparameter adaptive threshold obtains is very fully and very close to the result of final cluster, simultaneously, because the method is directly all non-ground point clouds to be carried out to the once cluster segmentation based on threshold value radius corresponding to each point oneself, and the cluster segmentation method of adaptive threshold based on the scale of construction will first be carried out once basis and cut apart, again basic segmentation result is carried out to the cluster segmentation based on adaptive threshold, therefore the cluster segmentation method time complexity of the adaptive threshold based on multiparameter is lower.
Embodiment 5
Be different from depth transducer in embodiment 1-4 and adopt fixing effective angle scope A
w(A
wbe less than 360 degree) scan by column, in the present embodiment, depth transducer adopts the mode of 360 degree scannings, be in the process that moves in system platform of depth transducer with the angular resolution of fixing by the 360 degree rotation sweep collections depth information of scene around, depth transducer often revolves to turn around and correspondingly obtains one and enclose depth information.Under normal circumstances, adopt the mode of 360 degree scannings can more efficiently, more fully obtain the depth information of tested scene, for example, in the time that system platform moves forward into line scanning on road; If adopt the scan mode in embodiment 1-4, i.e. the effective angle scope A of depth transducer to fix
w(A
wbeing less than 360 degree) when tested scene is scanned by column, depth transducer often can not collect the depth information corresponding to scene of road both sides simultaneously; If depth transducer adopts the mode of 360 degree scannings, because the angle of scanning is enough large, depth transducer can scan the depth information of the tested scene within the scope of 360 degree simultaneously.
Depth transducer adopts the mode of 360 degree scannings, is actually the effective angle scope A of depth transducer in embodiment 1-4
w360 degree are increased to, in the corresponding embodiment 1-4 of each circle depth information collecting, scan by column each the row depth information obtaining, if think, the circle of one in embodiment 5 depth information is " row " data, the feature of the data that 360 degree scan modes obtain is consistent with the data characteristics of processing in embodiment 1-4, and some cloud dividing method proposed by the invention is applicable to the situation of depth transducer 360 degree scannings.
The invention also discloses a kind of some cloud segmenting device, with reference to Figure 13, described device comprises:
Scan conversion module, for tested scene being scanned by column by depth transducer, to obtain the depth information of described tested scene, the depth information of described tested scene is carried out to coordinate conversion, to obtain the three-dimensional information of described tested scene under local coordinate;
Module is cut apart on ground, for being partitioned into ground point cloud from described three-dimensional information;
Module is cut apart on non-ground, for the cluster segmentation mode by adaptive threshold, non-ground point cloud is carried out to cluster segmentation, and described non-ground point cloud is other clouds except described ground point cloud in described three-dimensional information.
Described system is corresponding with said method, also comprises that other are the module that realizes said method, submodule, unit, subelement etc., for avoiding redundancy, so do not illustrate herein.
Above embodiment is only for illustrating the present invention; and be not limitation of the present invention; the those of ordinary skill in relevant technologies field; without departing from the spirit and scope of the present invention; can also make a variety of changes and modification; therefore all technical schemes that are equal to also belong to category of the present invention, and scope of patent protection of the present invention should be defined by the claims.
Claims (10)
1. a some cloud dividing method, is characterized in that, said method comprising the steps of:
S1: by depth transducer, tested scene is scanned by column, to obtain the depth information of described tested scene, the depth information of described tested scene is carried out to coordinate conversion, to obtain the three-dimensional information of described tested scene under local coordinate;
S2: be partitioned into ground point cloud from described three-dimensional information;
S3: the cluster segmentation mode by adaptive threshold is carried out cluster segmentation to non-ground point cloud, described non-ground point cloud is other clouds except described ground point cloud in described three-dimensional information.
2. the method for claim 1, is characterized in that, in step S2, is partitioned into ground point cloud and specifically comprises from described three-dimensional information:
S201: travel through described three-dimensional information with the unit of classifying as, and using row that traverse as working as prostatitis;
S202: according to the described each point when prostatitis of direction traversal from bottom to up;
S203: the absolute value of the Z axis coordinate of current point that calculating traverses and the difference of the floor level of described depth transducer position, if described absolute value is less than height threshold, using described current point as first initial point Ps (1), execution step S204, otherwise return to step 202;
S204: detect described initial point Ps (k) cut off Pe (k) afterwards, and using described initial point Ps (k) and cut off Pe (k) as a pair of;
S205: detect cut off Pe (k) initial point Ps (k+1) afterwards, if initial point Ps (k+1) detected, make k=k+1, and return to step S204, otherwise the point between every couple of initial point Ps (k) and cut off Pe (k) and every couple of initial point Ps (k) and cut off Pe (k) adds described ground point cloud as ground point, and performs step S206;
S206: whether each row that judges described three-dimensional information are all traversed, and if so, perform step S3, otherwise return to step S202.
3. the method for claim 1, is characterized in that, in step S3, by the cluster segmentation mode of adaptive threshold, non-ground point cloud is carried out to cluster segmentation and specifically comprises:
S301: adopt fixing initial threshold dth (0) to carry out cluster segmentation to described non-ground point cloud;
S302: each cluster that traversal cluster segmentation obtains;
S303: the adaptive threshold dth (w) that calculates the current cluster traversing according to the cluster scale of construction, judge whether described adaptive threshold dth (w) is less than described initial threshold radius dth (0), if so, based on described adaptive threshold dth (w), described current cluster is further cut apart;
S304: judge that whether each cluster that described cluster segmentation acquires is all traversed, and if not, returns to step S303.
4. method as claimed in claim 3, is characterized in that, in step S303, the adaptive threshold dth (w) that calculates the current cluster traversing according to the cluster scale of construction specifically comprises:
S3031: travel through the each point in described current cluster, current in conjunction with its neighborhood point fit Plane by what traverse, and matching is obtained to the normal vector of plane as the normal vector of described current point;
S3032: calculate in described current cluster the mean value of normal vector a little, and normal vector using described mean value as described current cluster;
S3033: the institute in described current cluster is a little all projected on projection plane, chooses the minimum rectangle that comprises all subpoints on described projection plane, described projection plane is the plane vertical with the normal vector of described current cluster;
S3034: calculate the adaptive threshold dth (w) of described current cluster by following formula,
dth(w)=k
1·L(w)+k
2
Wherein, the bond length that L (w) is described minimum rectangle, k
1and k
2for constant.
5. the method for claim 1, is characterized in that, in step S3, by the cluster segmentation mode of adaptive threshold, non-ground point cloud is carried out to cluster segmentation and specifically comprises:
S311: travel through the each point in described non-ground point cloud, calculate the adaptive threshold of the current point traversing;
S312: described non-ground point cloud is carried out to cluster segmentation according to the adaptive threshold of each point in described non-ground point cloud.
6. method as claimed in claim 5, is characterized in that, in step S311, calculates the adaptive threshold dth of the current point traversing by following formula,
Wherein, L is the depth information of described current point, A is the angular resolution of described depth transducer, v is described depth transducer translational speed in the horizontal direction, f is the sweep frequency of described depth transducer, α is the angle between sweep trace place plane and the direction of v, the plane at the central point that described sweep trace place plane is described depth transducer and the T row place of described tested scene, described T classifies row that comprise described current point as, β is the angle on testee surface in sweep trace place plane and tested scene, θ is that the T of described tested scene is listed as the angle between horizontal direction, A
wfor the effective angle of depth transducer scanning, the error amount that Δ L is L, the error amount that Δ v is v, the error amount that Δ α is α.
7. method as claimed in claim 2, is characterized in that, after step S3, also comprises:
S4: merged by the cluster of over-segmentation in the cluster that step S3 is obtained.
8. method as claimed in claim 7, is characterized in that, step S4 specifically comprises:
S401: each cluster that traversal cluster segmentation acquires;
S402: the each point in the current cluster q traversing is traveled through, current in conjunction with its neighborhood point fit Plane by what traverse, and matching is obtained to the normal vector of plane as the normal vector of described current point;
S403: calculate in described current cluster q the mean value of normal vector a little, and normal vector using described mean value as described current cluster;
S404: the institute in described current cluster q is a little all projected on projection plane, calculates the projected area of described current cluster q, described projection plane is the plane vertical with the normal vector of described current cluster q;
S405: judge whether described projected area is greater than preset area, if so, using described current cluster q as large scale of construction cluster;
S406: judge whether each cluster that cluster segmentation acquires is all traversed, and if so, performs step S407, otherwise return to step S402;
S407: if the spacing distance between two large scale of construction clusters is less than predeterminable range, be a cluster by two of correspondence large scale of construction Cluster mergings, until the spacing distance between any two large scale of construction clusters is all greater than described predeterminable range, the spacing distance between described two large scale of construction clusters is the distance between nearest two points in two clusters.
9. method as claimed in claim 8, is characterized in that, after step S407, also comprises:
Each cluster that S408: traversal step S407 acquires;
S409: the difference DELTA H between height and the floor level of the current cluster r that calculating traverses, the height of described current cluster r is described current cluster r a little in the minimum value of Z axis coordinate, the floor level of described current cluster r is the maximal value of Z axis coordinate in first initial point of each some institute's respective column in described current cluster r;
S410: judge whether described difference DELTA H is greater than preset difference value, if so, using described current cluster r as suspension cluster;
S411: whether each cluster that determining step S407 acquires is all traversed, if so, performs step S412, otherwise returns to step S409;
S412: by the extremely cluster nearest with it of each suspension Cluster merging, until each suspension cluster is all incorporated in a non-suspension cluster.
10. a some cloud segmenting device, is characterized in that, described device comprises:
Scan conversion module, for tested scene being scanned by column by depth transducer, to obtain the depth information of described tested scene, the depth information of described tested scene is carried out to coordinate conversion, to obtain the three-dimensional information of described tested scene under local coordinate;
Module is cut apart on ground, for being partitioned into ground point cloud from described three-dimensional information;
Module is cut apart on non-ground, for the cluster segmentation mode by adaptive threshold, non-ground point cloud is carried out to cluster segmentation, and described non-ground point cloud is other clouds except described ground point cloud in described three-dimensional information.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410410469.2A CN104143194B (en) | 2014-08-20 | 2014-08-20 | A kind of point cloud segmentation method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410410469.2A CN104143194B (en) | 2014-08-20 | 2014-08-20 | A kind of point cloud segmentation method and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104143194A true CN104143194A (en) | 2014-11-12 |
CN104143194B CN104143194B (en) | 2017-09-08 |
Family
ID=51852362
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410410469.2A Active CN104143194B (en) | 2014-08-20 | 2014-08-20 | A kind of point cloud segmentation method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104143194B (en) |
Cited By (47)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105139376A (en) * | 2015-07-16 | 2015-12-09 | 武汉体育学院 | Shooting counting method based on shooting counting device |
CN107369161A (en) * | 2017-07-19 | 2017-11-21 | 无锡信捷电气股份有限公司 | A kind of workpiece point cloud segmentation method at random based on the European cluster of improvement |
CN107657621A (en) * | 2017-10-20 | 2018-02-02 | 南京林业大学 | Two-dimensional laser point cloud sequence real time method for segmenting based on range of linearity growth |
CN108596860A (en) * | 2018-05-10 | 2018-09-28 | 芜湖航飞科技股份有限公司 | A kind of ground point cloud dividing method based on three-dimensional laser radar |
CN108710845A (en) * | 2018-05-11 | 2018-10-26 | 北京旷视科技有限公司 | The correlating method of target object and article, apparatus and system |
CN108734120A (en) * | 2018-05-15 | 2018-11-02 | 百度在线网络技术(北京)有限公司 | Mark method, apparatus, equipment and the computer readable storage medium of image |
CN108885791A (en) * | 2018-07-06 | 2018-11-23 | 深圳前海达闼云端智能科技有限公司 | ground detection method, related device and computer readable storage medium |
CN108986162A (en) * | 2018-06-28 | 2018-12-11 | 四川斐讯信息技术有限公司 | Vegetable and background segment method based on Inertial Measurement Unit and visual information |
CN109033920A (en) * | 2017-06-08 | 2018-12-18 | 株式会社理光 | A kind of recognition methods grabbing target, device and computer readable storage medium |
CN109033989A (en) * | 2018-07-02 | 2018-12-18 | 深圳辰视智能科技有限公司 | Target identification method, device and storage medium based on three-dimensional point cloud |
CN109100741A (en) * | 2018-06-11 | 2018-12-28 | 长安大学 | A kind of object detection method based on 3D laser radar and image data |
CN109118500A (en) * | 2018-07-16 | 2019-01-01 | 重庆大学产业技术研究院 | A kind of dividing method of the Point Cloud Data from Three Dimension Laser Scanning based on image |
CN109190573A (en) * | 2018-09-12 | 2019-01-11 | 百度在线网络技术(北京)有限公司 | A kind of ground detection method, apparatus, electronic equipment, vehicle and storage medium |
CN109242855A (en) * | 2018-07-19 | 2019-01-18 | 中国科学院自动化研究所 | Roof dividing method, system and equipment based on Three-dimensional Multi-resolution statistical information |
CN109359614A (en) * | 2018-10-30 | 2019-02-19 | 百度在线网络技术(北京)有限公司 | A kind of plane recognition methods, device, equipment and the medium of laser point cloud |
CN109613553A (en) * | 2018-12-18 | 2019-04-12 | 歌尔股份有限公司 | The method, apparatus and system of physical quantities in scene are determined based on laser radar |
CN109872329A (en) * | 2019-01-28 | 2019-06-11 | 重庆邮电大学 | A kind of ground point cloud fast partition method based on three-dimensional laser radar |
CN109961440A (en) * | 2019-03-11 | 2019-07-02 | 重庆邮电大学 | A kind of three-dimensional laser radar point cloud Target Segmentation method based on depth map |
CN110021040A (en) * | 2017-12-21 | 2019-07-16 | 福特全球技术公司 | Depth data segmentation |
CN110033457A (en) * | 2019-03-11 | 2019-07-19 | 北京理工大学 | A kind of target point cloud dividing method |
CN110148144A (en) * | 2018-08-27 | 2019-08-20 | 腾讯大地通途(北京)科技有限公司 | Dividing method and device, storage medium, the electronic device of point cloud data |
CN110349158A (en) * | 2018-04-04 | 2019-10-18 | 北京京东尚科信息技术有限公司 | A kind of method and apparatus handling point cloud data |
CN110363834A (en) * | 2018-04-10 | 2019-10-22 | 北京京东尚科信息技术有限公司 | A kind of dividing method and device of point cloud data |
CN110389359A (en) * | 2018-04-19 | 2019-10-29 | 法拉第未来公司 | System and method for ground level detection |
CN110858415A (en) * | 2018-08-24 | 2020-03-03 | 北京图森未来科技有限公司 | Method and device for labeling object in 3D point cloud data |
CN110909569A (en) * | 2018-09-17 | 2020-03-24 | 深圳市优必选科技有限公司 | Road condition information identification method and terminal equipment |
CN111192284A (en) * | 2019-12-27 | 2020-05-22 | 吉林大学 | Vehicle-mounted laser point cloud segmentation method and system |
CN111227444A (en) * | 2020-01-17 | 2020-06-05 | 泉州装备制造研究所 | 3D sole glue spraying path planning method based on k nearest neighbor |
CN111352106A (en) * | 2018-12-24 | 2020-06-30 | 珠海市一微半导体有限公司 | Sweeping robot slope identification method and device, chip and sweeping robot |
CN111462141A (en) * | 2020-05-19 | 2020-07-28 | 北京爱笔科技有限公司 | Method and device for acquiring point cloud plane, equipment and computer readable storage medium |
CN111724429A (en) * | 2019-03-21 | 2020-09-29 | 北京京东尚科信息技术有限公司 | Ground feature extraction method and device |
CN111722249A (en) * | 2019-03-22 | 2020-09-29 | 丰田自动车株式会社 | Object recognition device and vehicle control system |
CN111716340A (en) * | 2019-03-22 | 2020-09-29 | 达明机器人股份有限公司 | Correcting device and method for coordinate system of 3D camera and mechanical arm |
CN112213735A (en) * | 2020-08-25 | 2021-01-12 | 上海主线科技有限公司 | Laser point cloud noise reduction method for rainy and snowy weather |
CN112508970A (en) * | 2020-12-16 | 2021-03-16 | 北京超星未来科技有限公司 | Point cloud data segmentation method and device |
CN112634181A (en) * | 2019-09-24 | 2021-04-09 | 北京百度网讯科技有限公司 | Method and apparatus for detecting ground point cloud points |
CN112785596A (en) * | 2021-02-01 | 2021-05-11 | 中国铁建电气化局集团有限公司 | Dot cloud picture bolt segmentation and height measurement method based on DBSCAN clustering |
CN113064179A (en) * | 2021-03-22 | 2021-07-02 | 上海商汤临港智能科技有限公司 | Point cloud data screening method and vehicle control method and device |
WO2021134339A1 (en) * | 2019-12-30 | 2021-07-08 | 深圳元戎启行科技有限公司 | Point cloud-based segmentation processing method and apparatus, computer device, and storage medium |
CN113450315A (en) * | 2021-06-08 | 2021-09-28 | 北京伟景智能科技有限公司 | Bar counting method and device and steel separating system |
CN113537311A (en) * | 2021-06-30 | 2021-10-22 | 北京百度网讯科技有限公司 | Spatial point clustering method and device and electronic equipment |
CN113607166A (en) * | 2021-10-08 | 2021-11-05 | 广东省科学院智能制造研究所 | Indoor and outdoor positioning method and device for autonomous mobile robot based on multi-sensor fusion |
US20210374400A1 (en) * | 2020-05-29 | 2021-12-02 | Robert Bosch Gmbh | Method for classifying measuring points of a point cloud |
CN114766042A (en) * | 2019-12-12 | 2022-07-19 | Oppo广东移动通信有限公司 | Target detection method, device, terminal equipment and medium |
CN114782467A (en) * | 2022-04-14 | 2022-07-22 | 电子科技大学 | Point cloud ground segmentation method based on region division and self-adaptive threshold |
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 |
US20230116869A1 (en) * | 2021-10-08 | 2023-04-13 | Institute Of Intelligent Manufacturing, Gdas | Multi-sensor-fusion-based autonomous mobile robot indoor and outdoor navigation method and robot |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN201218726Y (en) * | 2008-04-23 | 2009-04-08 | 哈尔滨理工大学 | Cultural relics three-dimensional reconstruction apparatus based on colorful structured light |
CN101738172A (en) * | 2008-11-05 | 2010-06-16 | 哈尔滨理工大学 | Method for three-dimensional measurement of high sampling density color structured light based on green stripe segmentation |
CN102419794A (en) * | 2011-10-31 | 2012-04-18 | 武汉大学 | Method for quickly filtering airborne laser point cloud data |
CN102881047A (en) * | 2012-08-01 | 2013-01-16 | 桂林电子科技大学 | Automatic non-closed implicit curved surface reconstruction method |
CN103020637A (en) * | 2012-11-22 | 2013-04-03 | 北京航空航天大学 | Point cloud data segmentation method on top surface of building based on K-plane algorithm |
CN103268609A (en) * | 2013-05-17 | 2013-08-28 | 清华大学 | Partition method for orderly extracting point cloud on ground |
CN103514598A (en) * | 2013-07-20 | 2014-01-15 | 北京航空航天大学 | Automatic partitioning method of building SFM point clouds |
CN103886312A (en) * | 2014-03-21 | 2014-06-25 | 华东师范大学 | Method for rapidly selecting and displaying interested blood vessel in magnetic resonance blood vessel image |
-
2014
- 2014-08-20 CN CN201410410469.2A patent/CN104143194B/en active Active
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN201218726Y (en) * | 2008-04-23 | 2009-04-08 | 哈尔滨理工大学 | Cultural relics three-dimensional reconstruction apparatus based on colorful structured light |
CN101738172A (en) * | 2008-11-05 | 2010-06-16 | 哈尔滨理工大学 | Method for three-dimensional measurement of high sampling density color structured light based on green stripe segmentation |
CN102419794A (en) * | 2011-10-31 | 2012-04-18 | 武汉大学 | Method for quickly filtering airborne laser point cloud data |
CN102881047A (en) * | 2012-08-01 | 2013-01-16 | 桂林电子科技大学 | Automatic non-closed implicit curved surface reconstruction method |
CN103020637A (en) * | 2012-11-22 | 2013-04-03 | 北京航空航天大学 | Point cloud data segmentation method on top surface of building based on K-plane algorithm |
CN103268609A (en) * | 2013-05-17 | 2013-08-28 | 清华大学 | Partition method for orderly extracting point cloud on ground |
CN103514598A (en) * | 2013-07-20 | 2014-01-15 | 北京航空航天大学 | Automatic partitioning method of building SFM point clouds |
CN103886312A (en) * | 2014-03-21 | 2014-06-25 | 华东师范大学 | Method for rapidly selecting and displaying interested blood vessel in magnetic resonance blood vessel image |
Non-Patent Citations (3)
Title |
---|
QINGMING ZHAN等: "A Point Cloud Segmentation Method based on Vector Estimation and Color Clustering", 《THE 2ND INTERNATIONAL CONFERENCE ON INFORMATION SCIENCE AND ENGINEERING》 * |
郭臻 等: "图像阈值分割算法研究", 《中国传媒大学学报自然科学版》 * |
闫利 等: "一种新的点云平面混合分割方法", 《武汉大学学报·信息科学版》 * |
Cited By (74)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105139376B (en) * | 2015-07-16 | 2018-02-02 | 武汉体育学院 | A kind of shooting method of counting based on shooting counting device |
CN105139376A (en) * | 2015-07-16 | 2015-12-09 | 武汉体育学院 | Shooting counting method based on shooting counting device |
CN109033920A (en) * | 2017-06-08 | 2018-12-18 | 株式会社理光 | A kind of recognition methods grabbing target, device and computer readable storage medium |
CN109033920B (en) * | 2017-06-08 | 2021-10-08 | 株式会社理光 | Recognition method and device for grabbed target and computer readable storage medium |
CN107369161A (en) * | 2017-07-19 | 2017-11-21 | 无锡信捷电气股份有限公司 | A kind of workpiece point cloud segmentation method at random based on the European cluster of improvement |
CN107369161B (en) * | 2017-07-19 | 2020-09-11 | 无锡信捷电气股份有限公司 | Scattered workpiece point cloud segmentation method based on improved Euclidean clustering |
CN107657621A (en) * | 2017-10-20 | 2018-02-02 | 南京林业大学 | Two-dimensional laser point cloud sequence real time method for segmenting based on range of linearity growth |
CN110021040A (en) * | 2017-12-21 | 2019-07-16 | 福特全球技术公司 | Depth data segmentation |
CN110349158A (en) * | 2018-04-04 | 2019-10-18 | 北京京东尚科信息技术有限公司 | A kind of method and apparatus handling point cloud data |
CN110363834A (en) * | 2018-04-10 | 2019-10-22 | 北京京东尚科信息技术有限公司 | A kind of dividing method and device of point cloud data |
CN110389359A (en) * | 2018-04-19 | 2019-10-29 | 法拉第未来公司 | System and method for ground level detection |
CN108596860A (en) * | 2018-05-10 | 2018-09-28 | 芜湖航飞科技股份有限公司 | A kind of ground point cloud dividing method based on three-dimensional laser radar |
CN108710845A (en) * | 2018-05-11 | 2018-10-26 | 北京旷视科技有限公司 | The correlating method of target object and article, apparatus and system |
CN108734120B (en) * | 2018-05-15 | 2022-05-10 | 百度在线网络技术(北京)有限公司 | Method, device and equipment for labeling image and computer readable storage medium |
CN108734120A (en) * | 2018-05-15 | 2018-11-02 | 百度在线网络技术(北京)有限公司 | Mark method, apparatus, equipment and the computer readable storage medium of image |
CN109100741B (en) * | 2018-06-11 | 2020-11-20 | 长安大学 | Target detection method based on 3D laser radar and image data |
CN109100741A (en) * | 2018-06-11 | 2018-12-28 | 长安大学 | A kind of object detection method based on 3D laser radar and image data |
CN108986162B (en) * | 2018-06-28 | 2022-02-22 | 杭州吉吉知识产权运营有限公司 | Dish and background segmentation method based on inertial measurement unit and visual information |
CN108986162A (en) * | 2018-06-28 | 2018-12-11 | 四川斐讯信息技术有限公司 | Vegetable and background segment method based on Inertial Measurement Unit and visual information |
CN109033989A (en) * | 2018-07-02 | 2018-12-18 | 深圳辰视智能科技有限公司 | Target identification method, device and storage medium based on three-dimensional point cloud |
CN108885791A (en) * | 2018-07-06 | 2018-11-23 | 深圳前海达闼云端智能科技有限公司 | ground detection method, related device and computer readable storage medium |
CN109118500A (en) * | 2018-07-16 | 2019-01-01 | 重庆大学产业技术研究院 | A kind of dividing method of the Point Cloud Data from Three Dimension Laser Scanning based on image |
CN109118500B (en) * | 2018-07-16 | 2022-05-10 | 重庆大学产业技术研究院 | Image-based three-dimensional laser scanning point cloud data segmentation method |
CN109242855B (en) * | 2018-07-19 | 2020-08-11 | 中国科学院自动化研究所 | Multi-resolution three-dimensional statistical information-based roof segmentation method, system and equipment |
CN109242855A (en) * | 2018-07-19 | 2019-01-18 | 中国科学院自动化研究所 | Roof dividing method, system and equipment based on Three-dimensional Multi-resolution statistical information |
CN110858415B (en) * | 2018-08-24 | 2021-04-02 | 北京图森智途科技有限公司 | Method and device for labeling object in 3D point cloud data |
CN110858415A (en) * | 2018-08-24 | 2020-03-03 | 北京图森未来科技有限公司 | Method and device for labeling object in 3D point cloud data |
CN110148144B (en) * | 2018-08-27 | 2024-02-13 | 腾讯大地通途(北京)科技有限公司 | Point cloud data segmentation method and device, storage medium and electronic device |
CN110148144A (en) * | 2018-08-27 | 2019-08-20 | 腾讯大地通途(北京)科技有限公司 | Dividing method and device, storage medium, the electronic device of point cloud data |
US11313951B2 (en) | 2018-09-12 | 2022-04-26 | Baidu Online Network Technology (Beijing) Co., Ltd. | Ground detection method, electronic device, and vehicle |
CN109190573A (en) * | 2018-09-12 | 2019-01-11 | 百度在线网络技术(北京)有限公司 | A kind of ground detection method, apparatus, electronic equipment, vehicle and storage medium |
CN110909569A (en) * | 2018-09-17 | 2020-03-24 | 深圳市优必选科技有限公司 | Road condition information identification method and terminal equipment |
CN110909569B (en) * | 2018-09-17 | 2022-09-23 | 深圳市优必选科技有限公司 | Road condition information identification method and terminal equipment |
CN109359614A (en) * | 2018-10-30 | 2019-02-19 | 百度在线网络技术(北京)有限公司 | A kind of plane recognition methods, device, equipment and the medium of laser point cloud |
CN109613553A (en) * | 2018-12-18 | 2019-04-12 | 歌尔股份有限公司 | The method, apparatus and system of physical quantities in scene are determined based on laser radar |
CN109613553B (en) * | 2018-12-18 | 2021-05-18 | 歌尔光学科技有限公司 | Method, device and system for determining number of objects in scene based on laser radar |
CN111352106B (en) * | 2018-12-24 | 2022-06-14 | 珠海一微半导体股份有限公司 | Sweeping robot slope identification method and device, chip and sweeping robot |
CN111352106A (en) * | 2018-12-24 | 2020-06-30 | 珠海市一微半导体有限公司 | Sweeping robot slope identification method and device, chip and sweeping robot |
CN109872329A (en) * | 2019-01-28 | 2019-06-11 | 重庆邮电大学 | A kind of ground point cloud fast partition method based on three-dimensional laser radar |
CN110033457B (en) * | 2019-03-11 | 2021-11-30 | 北京理工大学 | Target point cloud segmentation method |
CN109961440A (en) * | 2019-03-11 | 2019-07-02 | 重庆邮电大学 | A kind of three-dimensional laser radar point cloud Target Segmentation method based on depth map |
CN110033457A (en) * | 2019-03-11 | 2019-07-19 | 北京理工大学 | A kind of target point cloud dividing method |
CN109961440B (en) * | 2019-03-11 | 2021-06-18 | 重庆邮电大学 | Three-dimensional laser radar point cloud target segmentation method based on depth map |
CN111724429B (en) * | 2019-03-21 | 2024-03-01 | 北京京东乾石科技有限公司 | Ground characteristic extraction method and device |
CN111724429A (en) * | 2019-03-21 | 2020-09-29 | 北京京东尚科信息技术有限公司 | Ground feature extraction method and device |
CN111716340A (en) * | 2019-03-22 | 2020-09-29 | 达明机器人股份有限公司 | Correcting device and method for coordinate system of 3D camera and mechanical arm |
CN111722249B (en) * | 2019-03-22 | 2023-12-01 | 丰田自动车株式会社 | Object recognition device and vehicle control system |
CN111722249A (en) * | 2019-03-22 | 2020-09-29 | 丰田自动车株式会社 | Object recognition device and vehicle control system |
CN112634181B (en) * | 2019-09-24 | 2024-06-14 | 阿波罗智能技术(北京)有限公司 | Method and device for detecting ground point cloud points |
CN112634181A (en) * | 2019-09-24 | 2021-04-09 | 北京百度网讯科技有限公司 | Method and apparatus for detecting ground point cloud points |
CN114766042A (en) * | 2019-12-12 | 2022-07-19 | Oppo广东移动通信有限公司 | Target detection method, device, terminal equipment and medium |
CN111192284A (en) * | 2019-12-27 | 2020-05-22 | 吉林大学 | Vehicle-mounted laser point cloud segmentation method and system |
CN111192284B (en) * | 2019-12-27 | 2022-04-05 | 吉林大学 | Vehicle-mounted laser point cloud segmentation method and system |
WO2021134339A1 (en) * | 2019-12-30 | 2021-07-08 | 深圳元戎启行科技有限公司 | Point cloud-based segmentation processing method and apparatus, computer device, and storage medium |
CN111227444A (en) * | 2020-01-17 | 2020-06-05 | 泉州装备制造研究所 | 3D sole glue spraying path planning method based on k nearest neighbor |
CN111462141B (en) * | 2020-05-19 | 2024-01-12 | 北京爱笔科技有限公司 | Point cloud plane acquisition method and device, equipment and computer readable storage medium |
CN111462141A (en) * | 2020-05-19 | 2020-07-28 | 北京爱笔科技有限公司 | Method and device for acquiring point cloud plane, equipment and computer readable storage medium |
US11640703B2 (en) * | 2020-05-29 | 2023-05-02 | Robert Bosch Gmbh | Method for classifying measuring points of a point cloud |
US20210374400A1 (en) * | 2020-05-29 | 2021-12-02 | Robert Bosch Gmbh | Method for classifying measuring points of a point cloud |
CN112213735A (en) * | 2020-08-25 | 2021-01-12 | 上海主线科技有限公司 | Laser point cloud noise reduction method for rainy and snowy weather |
CN112508970B (en) * | 2020-12-16 | 2023-08-25 | 北京超星未来科技有限公司 | Point cloud data segmentation method and device |
CN112508970A (en) * | 2020-12-16 | 2021-03-16 | 北京超星未来科技有限公司 | Point cloud data segmentation method and device |
CN112785596A (en) * | 2021-02-01 | 2021-05-11 | 中国铁建电气化局集团有限公司 | Dot cloud picture bolt segmentation and height measurement method based on DBSCAN clustering |
CN113064179A (en) * | 2021-03-22 | 2021-07-02 | 上海商汤临港智能科技有限公司 | Point cloud data screening method and vehicle control method and device |
CN113450315A (en) * | 2021-06-08 | 2021-09-28 | 北京伟景智能科技有限公司 | Bar counting method and device and steel separating system |
CN113537311B (en) * | 2021-06-30 | 2023-08-04 | 北京百度网讯科技有限公司 | Spatial point clustering method and device and electronic equipment |
CN113537311A (en) * | 2021-06-30 | 2021-10-22 | 北京百度网讯科技有限公司 | Spatial point clustering method and device and electronic equipment |
US20230116869A1 (en) * | 2021-10-08 | 2023-04-13 | Institute Of Intelligent Manufacturing, Gdas | Multi-sensor-fusion-based autonomous mobile robot indoor and outdoor navigation method and robot |
CN113607166B (en) * | 2021-10-08 | 2022-01-07 | 广东省科学院智能制造研究所 | Indoor and outdoor positioning method and device for autonomous mobile robot based on multi-sensor fusion |
CN113607166A (en) * | 2021-10-08 | 2021-11-05 | 广东省科学院智能制造研究所 | Indoor and outdoor positioning method and device for autonomous mobile robot based on multi-sensor fusion |
US20230111122A1 (en) * | 2021-10-08 | 2023-04-13 | Institute Of Intelligent Manufacturing, Gdas | Multi-sensor-fusion-based autonomous mobile robot indoor and outdoor positioning method and robot |
US12019453B2 (en) * | 2021-10-08 | 2024-06-25 | Institute Of Intelligent Manufacturing, Gdas | Multi-sensor-fusion-based autonomous mobile robot indoor and outdoor positioning method and robot |
CN114782467A (en) * | 2022-04-14 | 2022-07-22 | 电子科技大学 | Point cloud ground segmentation method based on region division and self-adaptive threshold |
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 |
Also Published As
Publication number | Publication date |
---|---|
CN104143194B (en) | 2017-09-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104143194A (en) | Point cloud partition method and device | |
CN106997049B (en) | Method and device for detecting barrier based on laser point cloud data | |
US11709058B2 (en) | Path planning method and device and mobile device | |
CN109684921B (en) | Road boundary detection and tracking method based on three-dimensional laser radar | |
CN105404844B (en) | A kind of Method for Road Boundary Detection based on multi-line laser radar | |
CN103226833B (en) | A kind of point cloud data segmentation method based on three-dimensional laser radar | |
CN105184852B (en) | A kind of urban road recognition methods and device based on laser point cloud | |
CN106570454B (en) | Pedestrian traffic parameter extracting method based on mobile laser scanning | |
CN103148804B (en) | Indoor unknown structure identification method based on laser scanning | |
WO2018133851A1 (en) | Point cloud data processing method and apparatus, and computer storage medium | |
CN103390169B (en) | A kind of city terrain classification method of Vehicle-borne Laser Scanning cloud data | |
CN106199558A (en) | Barrier method for quick | |
CN112184736B (en) | Multi-plane extraction method based on European clustering | |
CN106842231A (en) | A kind of road edge identification and tracking | |
CN103487034B (en) | Method for measuring distance and height by vehicle-mounted monocular camera based on vertical type target | |
CN103136393B (en) | A kind of areal coverage computing method based on stress and strain model | |
CN106887020A (en) | A kind of road vertical and horizontal section acquisition methods based on LiDAR point cloud | |
CN107657636A (en) | A kind of method that route topography figure elevational point is automatically extracted based on mobile lidar data | |
CN106097755A (en) | For identifying the method parked place and/or vacate place | |
CN103499337B (en) | Vehicle-mounted monocular camera distance and height measuring device based on vertical target | |
CN101915570B (en) | Vanishing point based method for automatically extracting and classifying ground movement measurement image line segments | |
CN104376595A (en) | Method for generating three-dimensional roads on basis of airborne LiDAR (light detection and ranging) and GIS (geographic information system) cooperation | |
CN112346463B (en) | Unmanned vehicle path planning method based on speed sampling | |
CN105740505A (en) | GPS-RTK technology based road space line shape recovery method | |
CN111717244A (en) | Train automatic driving sensing method and system |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |