CN109816682A - A kind of bracket System Partition and parameter detection method based on concavity and convexity - Google Patents
A kind of bracket System Partition and parameter detection method based on concavity and convexity Download PDFInfo
- Publication number
- CN109816682A CN109816682A CN201910058454.7A CN201910058454A CN109816682A CN 109816682 A CN109816682 A CN 109816682A CN 201910058454 A CN201910058454 A CN 201910058454A CN 109816682 A CN109816682 A CN 109816682A
- Authority
- CN
- China
- Prior art keywords
- bracket system
- linear segment
- segmentation
- concavity
- bracket
- 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)
Abstract
The invention discloses a kind of bracket System Partition and parameter detection method based on concavity and convexity, comprising the following steps: step A: contact network cantilever system three-dimensional point cloud is obtained;Step B: pre-processing source point cloud, including object extraction, filtering noise reduction;Step C: over-segmentation is carried out to bracket system using super body clustering algorithm;Step D: substep segmentation is carried out to bracket system using SC-LCCP algorithm, and provides the range of choice of partitioning parameters;Step E: obtaining each linear segment linear equation with P-RANSAC algorithm, and calculates each tie point coordinate, each linear segment points ratio and proportional positions coefficient;The present invention reduces drains on manpower and material resources, and are not influenced by the micro-judgment of weather and operating personnel;Substep segmentation has preferable noise immunity and robustness, improves the detection efficiency of contact network cantilever system, and can contact network cantilever system condition is monitored and be detected in real time.
Description
Technical field
The present invention relates to high-speed railway touching net preservation & testing fields, and in particular to a kind of bracket system based on concavity and convexity
Segmentation and parameter detection method.
Background technique
With greatly developing for electric railway, train running speed is also constantly being promoted.In order to guarantee train operation
High efficiency, the Stability and dependability flowed to pantograph just propose very high requirement, it is therefore necessary to ensure bracket system
Rigid invariance.Each coupling part of bracket system gets loose and shifts, and carrier cable and contact line on the one hand can be made to deviate solid
There is position, and phenomena such as making bow occur, on the other hand can change bracket internal system stress structure, and leads to contact net system office
The slack and undisciplined phenomenon of portion or domain type, influences the normal operation of train.Therefore, using 3D vision technique to each linear portion of bracket system
Divide effectively to be divided and calculate the intrinsic coordinate of each connecting joint point and position proportional coefficient and becomes particularly important.
Preservation & testing manually is carried out to contact network cantilever system currently, relying primarily in contact net maintenance, is connect relative to non-
Touch 3D visual pattern technology, the former will consume a large amount of manpower and material resources, interfere to driving, and can be by weather and operation people
The micro-judgment of member influences, and the latter obtains bracket system three dimensional point cloud using depth camera, and is to its each coupling part
No generation relative shift is detected, one side save the cost, passing train will not be interfered, on the other hand not by the pact of weather
The detection accuracy of beam, each parameter is higher, has preferable practicability.
Summary of the invention
Therefore, the present invention provides a kind of bracket System Partition and parameter detection method based on concavity and convexity.Specifically include with
Lower step:
Step A: contact network cantilever system three-dimensional point cloud is obtained;
Step B: pre-processing source point cloud, including object extraction, filtering noise reduction;
Step C: over-segmentation is carried out to the bracket system after pretreatment using super body clustering algorithm;
Step D: description standard bracket concavity and convexity is clear, and is connected and be packaged using the improved Local Convex based on Slope Constraint
(SC-LCCP) algorithm carries out substep segmentation to bracket system, and provides the range of choice of partitioning parameters;
Step E: using improved projection-sampling consistency (P-RANSAC) algorithm, obtain each linear segment linear equation,
And each tie point coordinate, each linear segment points are calculated than (Point-Ratio) and proportional positions coefficient (Proportional
Position Coefficients)。
Further, above-mentioned steps A obtains contact network cantilever system three-dimensional point cloud specifically: depth camera is placed in detection vehicle
Surface, to bracket system imaging, obtains bracket system three dimensional point cloud in depth camera visual range.
Further, the extraction of bracket system, filtering noise reduction are carried out to source point cloud in above-mentioned steps B, detailed process is as follows:
B1: the threshold range of x, y, z axis is respectively set using straight-through filter (Pass Through), to bracket system
Background environment is filtered out;
B2: the noise of bracket system is filtered out using statistical filtering (Statistical Outlier Removal).
Further, bracket is carried out using super body cluster (Supervoxel-Clustering) algorithm in above-mentioned steps C
Over-segmentation has carried out normalized to spatial component in algorithm, normalized cumulant D is by formula to calculate distance in space
(1) it provides:
λ, μ and ε control the influence of color, spatial distribution and geometric similarity to cluster respectively in formula;The value of distance D,
It will affect the systematicness on super voxel boundary;DcFor color distance;M is normaliztion constant;DsIt is a kind of maximum distance using cluster
Point carrys out standardized space length;RseedFor nucleus distance;DHiKFor histogram intersection point core;Specific step is as follows:
C1: through many experiments, determining that the value of normalized cumulant D surpasses the influence of voxel to bracket, and finally determines D;
C2: fixed normalized cumulant D specifies the systematicness on voxel boundary, determines nucleus distance RseedTo super voxel number
Influence, and finally determine parameter RseedValue;
C3: the value of setpoint color tolerance, and implement super body and cluster over-segmentation.
Further, description standard bracket concavity and convexity is clear in above-mentioned steps D, and using improved based on Slope Constraint
Local Convex connects packing algorithm, carries out substep segmentation to bracket system, and provide the range of choice of partitioning parameters, specific steps are such as
Under:
D1: the main angle theta of each linear segment and corresponding is counted | Δ α |, provide each linear segment of standard bracket system
Main angle concavity and convexity index;
|α1-α2|=| π-θ |=| Δ α | (2)
α in formula1,α2The equidirectional angle of respectively adjacent two panels point cloud center link vector and the normal vector at their centers;
D2: each parameter, respectively smoothness_threshold, β are setThresh, Minimum Segment Size,
Substep segmentations are carried out to six linear segments of bracket system: Horizontal Cantilever, inclined cantilever, cantilever support, positioning pipe support, positioning pipe,
Locator;
D3: on the basis of D2, different two o'clock M1 and N1 are arbitrarily selected to the same linear segment of bracket system being partitioned into, together
When, different two o'clock M2 and N2 are selected in coupled another linear segment, calculate their space slope with formula (2) respectively
S1, S2:
In formula: a, b, c are respectively the x of point M and N, y, z-axis coordinate;
Set space slope threshold value SThresh=0.3, if meeting formula (3), this two linear segment is same linear portion
Point, and respective point cloud Label is set as same value;
|S2-S1| < SThresh (4)
D4: the parameter setting of each point cloud data segmentation is counted, and provides the range of choice of partitioning parameters.
Further, it is obtained each linear in above-mentioned steps E with improved projection-sampling consistency (P-RANSAC) algorithm
Partial straight lines equation, and each tie point coordinate, each linear segment points ratio and proportional positions coefficient are calculated, specific steps are such as
Under:
E1: projecting to xoy plane for each linear portion branch cloud that segmentation extracts, then all the points z coordinate is 0;
E2: stochastical sampling fitting a straight line is used, absolute straight line model is obtained, there must be intersection point J (j in three-dimensional space xoy plane1,
j2, 0), export a certain fixed point h (k that each straight line model is passed through1,k2, 0) and all directions vector
E3: h (k will be pinpointed1,k2, 0) and direction vectorIt brings formula (4) into, obtains xoy plane and straight line equation:
E4: after the plane and straight line equation for obtaining each section, each junction intersecting point coordinate J (j is calculated1,j2, 0), z is sat at this time
It is designated as 0;Likewise, obtaining the corresponding former three-dimensional space linear equation of each projection straight line:
By j1, j2It brings formula (5) into, calculates the z coordinate value c of each tie point, obtain each tie point coordinate in three-dimensional space
J(j1,j2,j3);
E5: the points for each linear segment that statistics segmentation is extracted and the points of original each linear segment obtain points ratio;
E6: obtaining each linear segment both ends endpoint H and I, calculates each linear segment length P and endpoint H to J (j1,j2,
j3) length Q, calculate each tie point proportional positions coefficient W according to formula (6):
D in formula1,e1,f1For the x of endpoint H, y, z-axis coordinate;d2,e2,f2For the x of endpoint I, y, z-axis coordinate.
The beneficial effects of the present invention are:
(1) present invention is effectively divided and is counted by each linear segment of 3D visual pattern interface differential technique net-fault bracket system
Each tie point coordinate and proportional positions coefficient are calculated, not will increase the load of contact net system;
(2) present invention by contactless 3D visual pattern technology, reduce drain on manpower and material resources, and not by weather with
The micro-judgment of operating personnel influences;
(3) present invention is connected by the improved Local Convex based on Slope Constraint and is packaged (SC-LCCP) algorithm, project-adopted
Sample consistency (P-RANSAC) algorithm is split to bracket system and calculates each tie point parameter, the good, robust with noise immunity
Strong, the higher feature of precision of property, improves the detection efficiency of contact network cantilever system.
Detailed description of the invention
Fig. 1 is overhaul flow chart in the present invention.
Fig. 2 is the detection device schematic diagram used in the present invention.
Fig. 3 is to obtain a frame bracket system point cloud schematic diagram data in the present invention.
Fig. 4 is the one frame bracket system point cloud data of scene obtained in the present invention.
Fig. 5 is the bracket system point cloud data led directly to after filtering in the present invention.
Fig. 6 is the bracket system point cloud data in the present invention after statistical filtering.
Fig. 7 is influence of the normalized cumulant D to the super voxel boundary of bracket system in the present invention.
Fig. 8 is voxel boundary systematicness schematic diagram in the present invention.
Fig. 9 is parameter R in the present inventionseedInfluence to the distribution of super voxel number.
Figure 10 is the segmentation result that super body clusters over-segmentation in the present invention.
Figure 11 is that the concavity and convexity index of Plays bracket of the present invention marks.
Figure 12 is the segmentation result that this method validity is verified in the present invention.
Figure 13 is the extraction result of each linear segment of measured data in the present invention.
Figure 14 is the straight-line detection result of each linear segment of measured data in the present invention.
Figure 15 is 4 groups of measured data segmentation results in the present invention.
Figure 16 is the points of four groups of data in the present invention than bar shaped statistical chart.
Figure 17 is four groups of data points ratio error line charts in the present invention.
Figure 18 is four groups of ratio data position parameter error line charts in the present invention.
Specific embodiment
The present invention will be further described in the following with reference to the drawings and specific embodiments.
As shown in Figure 1, a kind of bracket System Partition based on concavity and convexity and parameter detection method flow chart, including following step
It is rapid:
Step A: contact network cantilever system three-dimensional point cloud is obtained.
Depth camera is placed in right above detection vehicle, and camera is horizontal, and is tilted a certain angle, and sees that Fig. 2, moving detection vehicle make
Entire bracket system is obtained in depth camera visual range, multiframe bracket system point cloud data is obtained in real time, sees Fig. 3.Fig. 4 is to obtain
The frame bracket point cloud data taken.
Step B: pre-processing source point cloud, including object extraction, filtering noise reduction.
B1: divided using straight-through filter (Pass Through) according to bracket system in the imaging position of camera coordinates system
Not She Zhi x, y, z axis threshold range, the background environment of bracket system is filtered out, parameter reference setting is shown in Table 1, filtering knot
Fruit sees Fig. 5;
Table 1 leads directly to filter parameter setting
B2: it utilizes statistical filtering (Statistical Outlier Removal), the noise of bracket system is filtered
It removes, filter result is shown in Fig. 6.
Step C: using super body clustering algorithm, carries out over-segmentation to the bracket system point cloud after pretreatment.
In order to calculate distance in space, normalized is carried out to spatial component in algorithm, normalized cumulant D is by formula
(1) it provides:
C1: many experiments determine that the value of normalized cumulant D surpasses the influence of voxel to bracket, see Fig. 7, and finally determine
D;
C2: fixed normalized cumulant D specifies the systematicness on voxel boundary, sees Fig. 8, determine nucleus distance RseedTo super body
Prime number purpose influences, and sees Fig. 9, and finally determines parameter RseedValue.
C3: the value of setpoint color tolerance, and implement super body and cluster over-segmentation, over-segmentation the result is shown in Figure 10.
Step D: (SC-LCCP) algorithm is packaged using the improved Local Convex connection based on Slope Constraint, to bracket system
Carry out substep segmentation;
D1: the main angle theta of each linear segment and corresponding is counted | Δ α |, see Figure 11, by formula (2), provide standard
The main angle concavity and convexity index of each linear segment of bracket system.
|α1-α2|=| π-θ |=| Δ α | (2)
D2: setting parameter, respectively smoothness_threshold, βThresh, Minimum Segment Size is to wrist
Six linear segments of arm system carry out substep segmentation: Horizontal Cantilever, inclined cantilever, cantilever support, positioning pipe support, positioning pipe, positioning
Device.Parameter reference setting is shown in Table 2.
2 SC-LCCP parameter setting of table
D3: on the basis of D2, arbitrarily selecting different two o'clock M1 and N1 to the same linear segment of bracket system being partitioned into,
Meanwhile different two o'clock M2 and N2 are selected in coupled another linear segment, their Space Oblique is calculated with formula (3) respectively
Rate S1, S2:
In formula: a, b, c are respectively the x of point M and N, y, z-axis coordinate.Set space slope threshold value SThresh=0.3, if meeting
Formula (4), then this two linear segment is same linear segment, and Label is set as same value.
|S2-S1| < SThresh (4)
Segmentation result is shown in Figure 12.
Step E: improved projection-sampling consistency (P-RANSAC) algorithm is used, each linear segment straight line side is obtained
Journey, and calculate each tie point coordinate, each linear segment points ratio and proportional positions coefficient.
E1: projecting to xoy plane for each linear portion branch cloud that segmentation extracts, then all the points z coordinate is 0, sees Figure 13.
E2: E1 basis under, with stochastical sampling fitting a straight line, obtain absolute straight line model, three-dimensional space xoy plane must
There is intersection point J (j1,j2, 0), export a certain fixed point h (k that each straight line model is passed through1,k2, 0) and all directions vector
E3: h (k will be pinpointed1,k2, 0) and direction vectorIt brings formula (5) into, obtains xoy plane and straight line equation:
E4: after the plane and straight line equation for obtaining each section, each junction intersecting point coordinate J (j is calculated1,j2, 0), z is sat at this time
It is designated as 0.Likewise, obtaining the corresponding former three-dimensional space linear equation of each projection straight line:
It brings j1, j2 into formula (6), calculates the z coordinate of each tie point, obtain each tie point coordinate J in three-dimensional space
(j1,j2,j3)。
E5: the points for each linear segment that statistics segmentation is extracted and the points of original each linear segment obtain points ratio
(P-R)。
E6: obtaining each linear segment both ends endpoint H and I, calculates each linear segment length P and endpoint H to J (j1,j2,
j3) length Q, calculate each tie point proportional positions coefficient W according to formula (7):
Fitting a straight line is shown in Figure 14.
4 groups of bracket system point cloud datas that the present invention obtains depth camera at this carry out effective segmentation of each linear segment
With parameter detecting, 4 groups of data segmentation results are shown in Figure 15.And the points of four groups of data are provided than bar shaped statistical chart, and see Figure 16, four groups
Data points ratio error line chart, is shown in that Figure 17, a group parameter setting table are shown in Table 3, bracket System Partition parameter selection range table is shown in
Table 4, each coupling part artis three-dimensional coordinate of a group point cloud and errors table, are shown in Table 5, a group point cloud proportional positions coefficient and error
Table, is shown in Table 6, and four groups of ratio data position parameter error line charts are shown in Figure 18.
Table 3 surveys bracket system point cloud a parameter setting
4 bracket System Partition parameter selection range of table
Table 5 surveys bracket system point cloud a each section connecting joint point three-dimensional coordinate and error
Table 6a group ratio data position parameter and error
The present invention by 3D visual pattern technology, to contact network cantilever system carry out each linear segment it is effective divide and
The parameter detecting of each connecting joint point.This contactless bracket system maintenance detection method: contact net system will not be increased
Add additional load, and driving will not be interfered;Reduce drain on manpower and material resources;Not by the constraint of weather and operating personnel
Micro-judgment influences;Each parameter detecting precision is higher, and has preferable noise immunity, higher robustness.A kind of base of the present invention
There is preferable prospect of the application in the bracket System Partition and parameter detection method of concavity and convexity.
Claims (6)
1. a kind of bracket System Partition and parameter detection method based on concavity and convexity, which comprises the following steps:
Step A: contact network cantilever system three-dimensional point cloud is obtained;
Step B: pre-processing source point cloud, including object extraction, filtering noise reduction;
Step C: over-segmentation is carried out to the bracket system after pretreatment using super body clustering algorithm;
Step D: description standard bracket concavity and convexity is clear, and connects packing algorithm using the improved Local Convex based on Slope Constraint
Substep segmentation is carried out to bracket system, and provides the range of choice of partitioning parameters;
Step E: improved projection-sampling consistency algorithm is used, obtains each linear segment linear equation, and calculate each tie point
Coordinate, each linear segment points ratio and proportional positions coefficient.
2. a kind of bracket System Partition and parameter detection method, feature based on concavity and convexity according to claim 1 exists
In it is specific as follows that the step A obtains contact network cantilever system three-dimensional point cloud:
Depth camera is placed in right above detection vehicle, in depth camera visual range, to bracket system imaging, obtains bracket system
Three dimensional point cloud.
3. a kind of bracket System Partition and parameter detection method, feature based on concavity and convexity according to claim 1 exists
In to source point cloud progress bracket system extraction, filtering noise reduction in the step B, detailed process is as follows:
B1: the threshold range of x, y, z axis being respectively set using straight-through filter, filters out to the background environment of bracket system;
B2: it is filtered out using noise of the statistical filtering to bracket system.
4. a kind of bracket System Partition and parameter detection method, feature based on concavity and convexity according to claim 1 exists
In over-segmentation being carried out to bracket using super body clustering algorithm in the step C, in order to calculate distance in space, to sky in algorithm
Between component carried out normalized, normalized cumulant D is provided by formula (1):
λ, μ and ε control the influence of color, spatial distribution and geometric similarity to cluster respectively in formula;The value of distance D, by shadow
Ring the systematicness on super voxel boundary;DcFor color distance;M is normaliztion constant;DsIt is that a kind of maximum distance point using cluster is next
Standardized space length;RseedFor nucleus distance;DHiKFor histogram intersection point core;Specific step is as follows:
C1: through many experiments, determining that the value of normalized cumulant D surpasses the influence of voxel to bracket, and finally determines D;
C2: fixed normalized cumulant D specifies the systematicness on voxel boundary, determines nucleus distance RseedTo super number of voxels purpose shadow
It rings, and finally determines parameter RseedValue;
C3: the value of setpoint color tolerance, and implement super body and cluster over-segmentation.
5. a kind of bracket System Partition and parameter detection method, feature based on concavity and convexity according to claim 1 exists
In description standard bracket concavity and convexity is clear in the step D, and is connected and be packaged using the improved Local Convex based on Slope Constraint
Algorithm carries out substep segmentation to bracket system, and provides the range of choice of partitioning parameters, the specific steps are as follows:
D1: the main angle theta of each linear segment and corresponding is counted | Δ α |, it is main to provide each linear segment of standard bracket system
Angle concavity and convexity index;
D2: being arranged each parameter, carries out substep segmentations to six linear segments of bracket system: Horizontal Cantilever, inclined cantilever, cantilever support,
Positioning pipe support, positioning pipe, locator;
D3: on the basis of D2, arbitrarily selecting different two o'clock M1 and N1 to the same linear segment of bracket system being partitioned into, meanwhile,
Select different two o'clock M2 and N2 in coupled another linear segment, respectively with formula (3) calculate they space slope S1,
S2:
In formula: a, b, c are respectively the x of point M and N, y, z-axis coordinate;
Set space slope threshold value SThresh=0.3, if meeting formula (4), this two linear segment is same linear segment, and
Respective point cloud Label is set as same value;
|S2-S1| < SThresh (4)
D4: the parameter setting of each point cloud data segmentation is counted, and provides the range of choice of partitioning parameters.
6. a kind of bracket System Partition and parameter detection method, feature based on concavity and convexity according to claim 1 exists
In with improved projection-sampling consistency algorithm in the step E, obtaining each linear segment linear equation, and calculate each company
Contact coordinate, each linear segment points ratio and proportional positions coefficient, the specific steps are as follows:
E1: projecting to xoy plane for each linear portion branch cloud that segmentation extracts, then all the points z coordinate is 0;
E2: stochastical sampling fitting a straight line is used, absolute straight line model is obtained, there must be intersection point J (j in three-dimensional space xoy plane1,j2,
0) a certain fixed point h (k that each straight line model is passed through, is exported1,k2, 0) and all directions vector
E3: h (k will be pinpointed1,k2, 0) and direction vectorIt brings formula (5) into, obtains xoy plane and straight line equation:
E4: after the plane and straight line equation for obtaining each section, each junction intersecting point coordinate J (j is calculated1,j2, 0), z coordinate is at this time
0;Likewise, obtaining the corresponding former three-dimensional space linear equation of each projection straight line:
By j1, j2It brings formula (6) into, calculates the z coordinate value c of each tie point, obtain each tie point coordinate J in three-dimensional space
(j1,j2,j3);
E5: the points for each linear segment that statistics segmentation is extracted and the points of original each linear segment obtain points ratio;
E6: obtaining each linear segment both ends endpoint H and I, calculates each linear segment length P and endpoint H to J (j1,j2,j3)
Length Q calculates each tie point proportional positions coefficient W according to formula (7):
D in formula1,e1,f1For the x of endpoint H, y, z-axis coordinate;d2,e2,f2For the x of endpoint I, y, z-axis coordinate.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910058454.7A CN109816682B (en) | 2019-01-22 | 2019-01-22 | Wrist-arm system segmentation and parameter detection method based on concavity and convexity |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910058454.7A CN109816682B (en) | 2019-01-22 | 2019-01-22 | Wrist-arm system segmentation and parameter detection method based on concavity and convexity |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109816682A true CN109816682A (en) | 2019-05-28 |
CN109816682B CN109816682B (en) | 2022-12-06 |
Family
ID=66603537
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910058454.7A Active CN109816682B (en) | 2019-01-22 | 2019-01-22 | Wrist-arm system segmentation and parameter detection method based on concavity and convexity |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109816682B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110473223A (en) * | 2019-08-15 | 2019-11-19 | 西南交通大学 | Two dimensional image auxiliary partition method based on contact network cantilever system three-dimensional point cloud |
CN110647939A (en) * | 2019-09-24 | 2020-01-03 | 广州大学 | Semi-supervised intelligent classification method and device, storage medium and terminal equipment |
CN112710241A (en) * | 2020-12-17 | 2021-04-27 | 中铁电气化局集团有限公司 | Method and device for detecting cantilever offset |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107463933A (en) * | 2017-07-24 | 2017-12-12 | 宗晖(上海)机器人有限公司 | A kind of image object detection method |
CN107578400A (en) * | 2017-07-26 | 2018-01-12 | 西南交通大学 | A kind of contact net device parameter detection method of BIM and three-dimensional point cloud fusion |
US20180144458A1 (en) * | 2016-11-21 | 2018-05-24 | Seiko Epson Corporation | Multiple Hypotheses Segmentation-Guided 3D Object Detection and Pose Estimation |
US20180225515A1 (en) * | 2015-08-04 | 2018-08-09 | Baidu Online Network Technology (Beijing) Co. Ltd. | Method and apparatus for urban road recognition based on laser point cloud, storage medium, and device |
CN108564575A (en) * | 2018-04-11 | 2018-09-21 | 西南交通大学 | A kind of contactless catenary's parameters detection method based on three dimensional point cloud |
CN108921164A (en) * | 2018-06-15 | 2018-11-30 | 西南交通大学 | A kind of contact net positioner slope detection method based on three-dimensional point cloud segmentation |
CN108961271A (en) * | 2018-06-27 | 2018-12-07 | 西安工程大学 | A kind of super voxel bumps partitioning algorithm of colour 3D point cloud |
CN109166145A (en) * | 2018-08-10 | 2019-01-08 | 中国农业大学 | A kind of fruit tree leaf growth parameter(s) extracting method and system based on cluster segmentation |
-
2019
- 2019-01-22 CN CN201910058454.7A patent/CN109816682B/en active Active
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20180225515A1 (en) * | 2015-08-04 | 2018-08-09 | Baidu Online Network Technology (Beijing) Co. Ltd. | Method and apparatus for urban road recognition based on laser point cloud, storage medium, and device |
US20180144458A1 (en) * | 2016-11-21 | 2018-05-24 | Seiko Epson Corporation | Multiple Hypotheses Segmentation-Guided 3D Object Detection and Pose Estimation |
CN107463933A (en) * | 2017-07-24 | 2017-12-12 | 宗晖(上海)机器人有限公司 | A kind of image object detection method |
CN107578400A (en) * | 2017-07-26 | 2018-01-12 | 西南交通大学 | A kind of contact net device parameter detection method of BIM and three-dimensional point cloud fusion |
CN108564575A (en) * | 2018-04-11 | 2018-09-21 | 西南交通大学 | A kind of contactless catenary's parameters detection method based on three dimensional point cloud |
CN108921164A (en) * | 2018-06-15 | 2018-11-30 | 西南交通大学 | A kind of contact net positioner slope detection method based on three-dimensional point cloud segmentation |
CN108961271A (en) * | 2018-06-27 | 2018-12-07 | 西安工程大学 | A kind of super voxel bumps partitioning algorithm of colour 3D point cloud |
CN109166145A (en) * | 2018-08-10 | 2019-01-08 | 中国农业大学 | A kind of fruit tree leaf growth parameter(s) extracting method and system based on cluster segmentation |
Non-Patent Citations (3)
Title |
---|
周靖松等: "基于三维点云的接触网几何参数检测方法", 《仪器仪表学报》 * |
徐建芳: "基于三维模型的接触网检测技术研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 * |
徐建芳等: "基于SIFT和LBP点云配准的接触网零部件三维重建研究", 《铁道学报》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110473223A (en) * | 2019-08-15 | 2019-11-19 | 西南交通大学 | Two dimensional image auxiliary partition method based on contact network cantilever system three-dimensional point cloud |
CN110473223B (en) * | 2019-08-15 | 2023-05-05 | 西南交通大学 | Two-dimensional image auxiliary segmentation method based on three-dimensional point cloud of catenary cantilever system |
CN110647939A (en) * | 2019-09-24 | 2020-01-03 | 广州大学 | Semi-supervised intelligent classification method and device, storage medium and terminal equipment |
CN112710241A (en) * | 2020-12-17 | 2021-04-27 | 中铁电气化局集团有限公司 | Method and device for detecting cantilever offset |
Also Published As
Publication number | Publication date |
---|---|
CN109816682B (en) | 2022-12-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110570428B (en) | Method and system for dividing building roof sheet from large-scale image dense matching point cloud | |
Huang et al. | Automatic data segmentation for geometric feature extraction from unorganized 3-D coordinate points | |
Yang et al. | An individual tree segmentation method based on watershed algorithm and three-dimensional spatial distribution analysis from airborne LiDAR point clouds | |
CN104809689B (en) | A kind of building point cloud model base map method for registering based on profile | |
CN109816682A (en) | A kind of bracket System Partition and parameter detection method based on concavity and convexity | |
CN112070769B (en) | Layered point cloud segmentation method based on DBSCAN | |
CN108875595A (en) | A kind of Driving Scene object detection method merged based on deep learning and multilayer feature | |
CN107958183A (en) | A kind of city road network information automation extraction method of high-resolution remote sensing image | |
CN104376595B (en) | A kind of three-dimensional road generation method cooperateed with based on airborne LiDAR and GIS | |
CN109544456A (en) | The panorama environment perception method merged based on two dimensional image and three dimensional point cloud | |
CN106600690A (en) | Complex building three-dimensional modeling method based on point cloud data | |
CN106780524A (en) | A kind of three-dimensional point cloud road boundary extraction method | |
CN103234969B (en) | Method for measuring fabric weft density based on machine vision | |
CN106202728A (en) | Based on Micro CT D braided composites non-homogeneous Voxel grid discrete method | |
CN108389256A (en) | Two three-dimensional interactive unmanned plane electric force pole tower inspection householder methods | |
CN102938066A (en) | Method for reconstructing outer outline polygon of building based on multivariate data | |
CN110532602A (en) | A kind of indoor autodraft and modeling method based on plan view image | |
CN109615654B (en) | Method for measuring corrosion depth and area of inner surface of drainage pipeline based on binocular vision | |
CN103698553A (en) | Novel surface flow field velocity measurement system and velocity measurement method on basis of infrared image pickup | |
CN108596165A (en) | Road traffic marking detection method based on unmanned plane low latitude Aerial Images and system | |
CN106327576B (en) | A kind of City scenarios method for reconstructing and system | |
CN102222357A (en) | Foot-shaped three-dimensional surface reconstruction method based on image segmentation and grid subdivision | |
CN108985143A (en) | A method of based on unmanned plane image recognition iron towers of overhead power transmission lines structure | |
CN114119863A (en) | Method for automatically extracting street tree target and forest attribute thereof based on vehicle-mounted laser radar data | |
Zhang et al. | Updating of cartographic road databases by image analysis |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |