CN105184250B - A kind of terrain classification method of electric power corridor airborne LiDAR point cloud data - Google Patents

A kind of terrain classification method of electric power corridor airborne LiDAR point cloud data Download PDF

Info

Publication number
CN105184250B
CN105184250B CN201510547324.1A CN201510547324A CN105184250B CN 105184250 B CN105184250 B CN 105184250B CN 201510547324 A CN201510547324 A CN 201510547324A CN 105184250 B CN105184250 B CN 105184250B
Authority
CN
China
Prior art keywords
point cloud
point
cloud data
feature
exemplar
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.)
Active
Application number
CN201510547324.1A
Other languages
Chinese (zh)
Other versions
CN105184250A (en
Inventor
谷延锋
解冰谦
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Harbin Institute of Technology
Original Assignee
Harbin Institute of Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Harbin Institute of Technology filed Critical Harbin Institute of Technology
Priority to CN201510547324.1A priority Critical patent/CN105184250B/en
Publication of CN105184250A publication Critical patent/CN105184250A/en
Application granted granted Critical
Publication of CN105184250B publication Critical patent/CN105184250B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/10—Terrestrial scenes
    • G06V20/13—Satellite images
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
    • G06F18/24133—Distances to prototypes
    • G06F18/24143—Distances to neighbourhood prototypes, e.g. restricted Coulomb energy networks [RCEN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/254—Fusion techniques of classification results, e.g. of results related to same input data
    • G06F18/256—Fusion techniques of classification results, e.g. of results related to same input data of results relating to different input data, e.g. multimodal recognition

Abstract

The present invention is a kind of terrain classification method of electric power corridor airborne LiDAR point cloud data, is related to airborne LiDAR point cloud data processing field；The problem of use of information included to airborne LiDAR point cloud data in electric power corridor the invention solves the prior art is insufficient, and then propose a kind of terrain classification method of electric power corridor airborne LiDAR point cloud data characteristics；Specific steps include：The acquisition of airborne LiDAR point cloud data；Point cloud data rough error point removes；Point cloud data feature extraction and processing, point cloud data feature includes single-point feature and neighborhood characteristics, and the feature of extraction is normalized；Selection has exemplar from LiDAR point cloud data acquisition system, the target informations such as power line, vegetation, building and earth's surface are isolated from LiDAR point cloud data using there is exemplar to classify data set, obtain the classification informations such as power line, vegetation, building and earth's surface.

Description

A kind of terrain classification method of electric power corridor airborne LiDAR point cloud data
Technical field
The present invention relates to airborne LiDAR point cloud data processing fields；More particularly to a kind of electric power corridor airborne LiDAR point cloud The terrain classification method of data.
Background technology
Traditional electric power corridor inspection generally relies on the visual interpretation of human eye, other than expending a large amount of manpower and materials, is obtained The data precision taken is not also high, and for complicated landform, and vehicle, manpower are difficult to reach, and is brought to the inspection of electric power corridor larger It is difficult.And the equipment such as helicopter carrying digital camera, infrared video camera are used, acquired data deficiency three-dimensional information, not It can accurately judge the atural object status information in electric power corridor.And airborne LiDAR (Light Detection and Ranging) system System can have the characteristics that high efficiency, low-loss, high-precision, therefore in recent years with the three-dimensional spatial information of quick obtaining ground scene To start more in the inspection of electric power corridor.
Inspection is carried out to electric power corridor currently with airborne LiDAR, acquired LiDAR point cloud data electric power multipair greatly is set It applies (including electric wire and pylon) and carries out three-dimensional modeling.And it is not only set in the point cloud data that airborne LiDAR systems obtain comprising electric power Information is applied, other terrestrial object informations in electric power corridor are further included, including vegetation, building etc..If being concerned only with electric power facility point cloud, So that the utilization ratio of LiDAR point cloud data substantially reduces.The hair of publication number CN102590823A disclosed in 18 days July in 2012 Bright patent application《A kind of airborne LIDAR data power line rapid extraction and reconstructing method》In disclose it is a kind of based on a point cloud level journey The method of threshold filter and density filtering extracts power line and pylon point cloud from LiDAR point cloud, and then cloud is carried out Layering fitting obtains single power line vector, realizes the power line between multiple spans and automatically extracts and pylon extraction.2013 1 The application for a patent for invention of Publication No. CN102879788A disclosed in the moon 16《A kind of power transmission line based on on-board LiDAR data The power line extraction method on road》In disclose and a kind of be fitted based on the electric power cluster center clustered to cross section sampled point cloud Power line extraction method, the center line after fitting is replicated and offset obtains multiple-loop line road electric wire.In November, 2013 The application for a patent for invention of Publication No. CN103412296A disclosed in 27 days《Power tower is automatically extracted in random laser point cloud data Method》In disclose it is a kind of using LiDAR point cloud carry out pylon extraction method.These technologies are to utilize airborne LiDAR point Cloud data carry out the extraction and reconstruction of power line or pylon in electric power corridor, ignore big included in non-electricity facility point cloud Measure useful information.
Invention content
The present invention is only limitted to carry power line to solve the prior art to airborne LiDAR point cloud data in electric power corridor The problem of taking, making included use of information insufficient, therefore propose a kind of atural object of electric power corridor airborne LiDAR point cloud data Sorting technique.
Airborne LiDAR systems are by three points of GPS positioning system, INS inertial navigation systems and laser scanning and ranging system System forms.Wherein, GPS positioning system realizes the positioning to airborne platform；The posture letter of INS inertial navigation system measuring tables Breath, including roll angle, pitch angle and azimuth；The scanning angle of laser scanning and ranging system recording laser and flight time, Distance of the culture point away from laser receiver is obtained by measuring the flight time.It is as follows：
Step 1: obtaining LiDAR point cloud data, LiDAR point cloud data progress three-dimensional visualization is shown, sets elevation threshold Value removal rough error point；
Step 2: feature extraction and characteristic processing are carried out to LiDAR point cloud data；
Step 3: random selection has exemplar from LiDAR point cloud data；
Step 4: using there is exemplar to be treated to LiDAR point cloud data by grader i.e. k neighbours criterion sorting technique point Class sample is classified, and obtains electric power facility, vegetation, building, earth's surface point cloud data；
Step 1 calculates sample to be sorted and all theorem in Euclid space distances for having exemplar Y represents sample to be sorted, xiIndicate exemplar, niRepresent that the i-th class has the quantity of exemplar；
Step 2 classifies sample to be sorted according to k neighbours criterion, by sample to be sorted and has exemplar distance Minimum is divided into one kind, i.e., the sample size that k arest neighbors of sample to be sorted has the i-th class in exemplar is mi, then classification Label ω=argmaxmi。
Wherein step 1 detailed process is：
Point cloud data is read workspace, and carry out visualization and show by step 1 one using MATLAB softwares；
Step 1 two, basis, not higher than ground object target point highest in scene, determine that atural object target point is deposited not less than ground Elevation section [L, H], wherein L represent elevation threshold value minimum point, H represent elevation threshold value peak, by elevation threshold value be less than L It is weeded out with point of the elevation threshold value more than H.
Step 2 detailed process is：
Step 2 one, extraction single-point feature；
Step 2 two, extraction neighborhood characteristics；
Step 2 three, feature normalization and feature selecting.
Step 211, the height using in LiDAR point cloud, intensity, echo times, echo position information are as current point A feature；
LiDAR point cloud is carried out piecemeal by step 2 one or two, is chosen per minimum point in the block as ground point progress earth's surface mould Type is fitted, and original point cloud level degree is subtracted relative altitude information that ground level put as a feature；By relative altitude Point less than 1m is set as ground point.
Step 221 calculates distance between two points in point cloud, finds out and current point P0K closest point forms it Neighbour's point set { Pi, i=1,2 ..., k, the 3 d space coordinate of every are expressed as P with vectori=(xi,yi,zi)T；Wherein, Pi Represent i-th of Neighbor Points vector, xi,yi,ziRepresentation space three-dimensional coordinate, T represent transposition.
Step 2 two or two calculates the covariance matrix that each point concentrates all the points three-dimensional coordinate Wherein, Represent k Neighbor Points average value, P0Represent current point；Solve three eigenvalue λs of Σ1>λ2 >λ3And its corresponding three feature vector v1,v2,v3。
Using characteristic value as three features of current point, other characteristic quantities are defined on the basis of characteristic value：
Assuming that normal vector is v3=(v31,v32,v33)T, then angle beBy angle angle Spend a feature of the information as current point.
Step 2 two or three, the height variance for calculating the interior point of neighborhood, intensity variance, angle variance, three as current point Feature；
Step 2 two or four searches the point that distance is farthest in neighborhood point set, according to the sphere using this maximum distance as radius Volume calculates the dot density feature of neighborhood point set.
Step 231, the feature point cloud data extracted, using formula f=(f-min)/(max-min), f ∈ [0, 1], wherein max is characterized the maximum value of f, and min is characterized the minimum value of f；
Step 2 three or two, the character numerical value after normalization are concentrated near 0 or 1, and dynamic range is smaller；And through normalization Residue character numeric distribution afterwards is more discrete, and dynamic range is larger, then passes through y=ln (x+1) or y=ex+1To dynamic range It is adjusted.
Step 3 detailed process is：
Step 3 one, the classification information by determining certain point or certain point set, being added to corresponding has exemplar collection In conjunction；
Step 3 two selected has that exemplar should be uniformly distributed and sorted Different categories of samples number is equal.
Advantageous effect of the present invention：
The present invention solves prior art electric power corridor scene airborne LiDAR point cloud data processing technique and cannot obtain rationally It utilizes, using electric power corridor area airborne LiDAR point cloud data as research object, it is proposed that a kind of new electric power corridor is airborne The terrain classification method of LiDAR point cloud data.Pass through the acquisition of airborne LiDAR point cloud data；Point cloud data rough error point removes；Point Cloud data characteristics is extracted and processing, and point cloud data feature includes single-point feature and neighborhood characteristics, and to the feature of extraction into Row normalized；Selection has exemplar from LiDAR point cloud data acquisition system, and power line will be isolated in LiDAR point cloud, is planted The target informations such as quilt, building and earth's surface obtain power line, vegetation, building using there is exemplar to classify data set With the classification informations such as earth's surface.Information is fully efficiently used in LiDAR point cloud data.The present invention is applied to three-dimensional modeling and leads Domain.
Description of the drawings
Fig. 1 is the flow chart element diagram of the terrain classification of scene airborne LiDAR point cloud data in electric power corridor of the present invention It is intended to；
Fig. 2 is that airborne LiDAR systems obtain electric power corridor scene point cloud datagram in embodiment, and Fig. 2 (a) is initial three-dimensional Point cloud chart, Fig. 2 (b) are three-dimensional point cloud side views, and Fig. 2 (c) is point cloud data figure used in embodiment；
Fig. 3 is rough error point schematic diagram in LiDAR point cloud in embodiment：Fig. 3 (a) is 3-D view, and Fig. 3 (b) is side view, Wherein, zero flying spot is represented, represents low spot；
Fig. 4 is that LiDAR point cloud rough error point rejects design sketch in embodiment：Fig. 4 (a) is original point cloud three-dimensional effect diagram, Fig. 4 (b) it is that rough error point rejects result three-dimensional effect diagram, Fig. 4 (c) is original point cloud section design sketch, and Fig. 4 (d) is that rough error point rejects knot Fruit section design sketch, Fig. 4 (e) are original point cloud vertical views, and Fig. 4 (f) rough errors point rejects result vertical view；
Fig. 5 is single-point feature information extraction result in embodiment：Fig. 5 (a) is altitude feature, and Fig. 5 (b) is strength characteristic, Fig. 5 (c) is echo times feature；
Fig. 6 is embodiment mesorelief modeling schematic diagram：Fig. 6 (a) is original point cloud, and Fig. 6 (b) is that ground seed point chooses knot Fruit, Fig. 6 (c) are terrain model modeling results；
Fig. 7 is neighborhood choice schematic diagram in embodiment：Fig. 7 (a) represents spheric neighbo(u)rhood N3DWith cylindricality neighborhood N2D, Fig. 7 (b) Represent k neighbour's point set neighborhoods；
Fig. 8 is LiDAR point cloud geometry feature extraction schematic diagram in embodiment：Fig. 8 (a) represents original point cloud, Fig. 8 (b) planar structure feature extraction is represented as a result, Fig. 8 (c) represents linear structural feature extraction result；
Fig. 9 is characteristic dynamic range Tuning function schematic diagram in embodiment：Fig. 9 (a) suitable for character numerical value concentrate on compared with The situation of small value, Fig. 9 (b) concentrate on the situation of higher value suitable for character numerical value；
Figure 10 be in embodiment characteristic dynamic range adjustment before and after comparison diagram：Figure 10 (a) represents certain primitive character, Figure 10 (b) feature after dynamic range adjustment is represented, Figure 10 (c) represents the plan view of primitive character, and Figure 10 (d) represents dynamic range Characteristic plane view after adjustment；
Figure 11 is embodiment point cloud classifying quality figure：Figure 11 (a) represents original point cloud, and Figure 11 (b) represents power line, figure 11 (c) represents vegetation, and Figure 11 (d) represents building, and Figure 11 (e) represents Ground Point.
Specific embodiment
Specific embodiment one：The present embodiment will be described with reference to Fig. 1, the airborne LiDAR numbers in electric power corridor of present embodiment According to terrain classification method include the following steps：
Step 1: obtaining LiDAR point cloud data, LiDAR point cloud data progress three-dimensional visualization is shown, it is high according to setting Journey threshold value removes rough error point；
Step 2: feature extraction and characteristic processing are carried out to LiDAR point cloud data；
Step 3: random selection has exemplar from LiDAR point cloud data；
Step 4: using there is exemplar to be treated to LiDAR point cloud data by grader i.e. k neighbours criterion sorting technique point Class sample is classified, and obtains electric power facility, vegetation, building, earth's surface point cloud data；
Step 1 calculates sample to be sorted and all theorem in Euclid space distances for having exemplar Y represents sample to be sorted, xiIndicate exemplar, niRepresent that the i-th class has the quantity of exemplar；
Step 2 classifies sample to be sorted according to k neighbours criterion, by sample to be sorted and has exemplar distance Minimum is divided into one kind, i.e., the sample size that k arest neighbors of sample to be sorted has the i-th class in exemplar is mi, then classification Label ω=argmaxmi。
The advantageous effect of present embodiment：
The present invention solves prior art electric power corridor scene airborne LiDAR point cloud data processing technique and cannot obtain rationally It utilizes, using electric power corridor area airborne LiDAR point cloud data as research object, it is proposed that a kind of new electric power corridor is airborne The terrain classification method of LiDAR point cloud data.
Specific embodiment two：The difference of present embodiment and specific embodiment one is：Present embodiment is in step Removal rough error point described in one, is as follows：
Point cloud data is read workspace, and carry out visualization and show by step 1 one using MATLAB softwares；
Step 1 two, basis, not higher than ground object target point highest in scene, determine that atural object target point is deposited not less than ground Elevation section [L, H], wherein L represent elevation threshold value minimum point, H represent elevation threshold value peak, by elevation threshold value be less than L It is weeded out with point of the elevation threshold value more than H.
Specific embodiment two：The difference of present embodiment and specific embodiment one is：Present embodiment is in step Carry out feature extraction and characteristic processing described in two, are as follows：
Step 2 one, extraction single-point feature；
Step 2 two, extraction neighborhood characteristics；
Step 2 three, feature normalization and feature selecting.
Specific embodiment four：The difference of present embodiment and specific embodiment three is：Present embodiment is in step Extraction single-point feature described in 21, is as follows：
Step 211, the height using in LiDAR point cloud data, intensity, echo times, echo position information are as point One feature；
LiDAR point cloud data are carried out piecemeal by step 2 one or two, are chosen per minimum point in the block as ground point progress ground Table model is fitted, and original point cloud level degree is subtracted relative altitude information that ground level put as a feature；It will be opposite Highly it is set as ground point less than 1m.
Specific embodiment five：The difference of present embodiment and specific embodiment three is：Present embodiment is in step Extraction neighborhood characteristics described in two or two, are as follows：
Step 221 calculates distance between two points in point cloud, finds out and current point P0K closest point forms it Neighbour's point set { Pi, i=1,2 ..., k, the 3 d space coordinate of every are expressed as P with vectori=(xi,yi,zi)T, wherein, Pi Represent i-th of Neighbor Points vector, xi,yi,ziRepresentation space three-dimensional coordinate, T represent transposition；
Step 2 two or two calculates the covariance matrix that each point concentrates all the points three-dimensional coordinate Wherein, Represent k Neighbor Points, P0Represent current point；Solve three eigenvalue λs of Σ1>λ2>λ3And Its corresponding three feature vector v1,v2,v3；
Using characteristic value as three features of current point, other characteristic quantities are defined on the basis of characteristic value：
According to characteristic value, the physical significance of feature vector, third feature vector v3The normal vector of corresponding fit Plane, and The calculating method vector v on the basis of normal vector3With the angle of vertical direction, using angle angle information as current point P0One Feature；Assuming that normal vector is v3=(v31,v32,v33)T, then angle be
Step 2 two or three, the height variance for calculating the interior point of neighborhood, intensity variance, angle variance, three as current point Feature；
Step 2 two or four searches the point that distance is farthest in neighborhood point set, according to the sphere using this maximum distance as radius Volume calculates the dot density feature of neighborhood point set.
Specific embodiment six：The difference of present embodiment and specific embodiment three is：Present embodiment is in step During feature normalization and feature selecting described in two or three, it is as follows：
Step 231, the feature point cloud data extracted, using formula f=(f-min)/(max-min), f ∈ [0, 1], wherein max is characterized the maximum value of f, and min is characterized the minimum value of f；
Step 2 three or two, the character numerical value after normalization are concentrated near 0 or 1, and dynamic range is smaller；And through normalization Residue character numeric distribution afterwards is more discrete, and dynamic range is larger, then passes through y=ln (x+1) or y=ex+1To dynamic range It is adjusted.
Specific embodiment seven：The difference of present embodiment and specific embodiment one is：Present embodiment is in step When selection described in three has exemplar, it is as follows：
Step 3 one, the classification information by determining certain point or certain point set, being added to corresponding has exemplar collection In conjunction；
Step 3 two selected has that exemplar should be uniformly distributed and sorted Different categories of samples number is equal.
Embodiment
Illustrate the present embodiment with reference to Fig. 1-11, the present embodiment is a kind of electric power corridor airborne LiDAR point cloud data characteristics Terrain classification method, mainly using electric power corridor airborne LiDAR point cloud data shown in Fig. 2 as a cloud feature extraction and classifying Initial data, this initial data is to intercept the point cloud data in Sonoma areas provided from Univ Maryland-Coll Park USA (to come from " Sonoma County Vegetation Mapping and LiDAR Program " project, by the Carbon of NASA Monitoring System are subsidized), as shown in Fig. 2 (c), point cloud sum about 5,800,000, point cloud density is 13.75 points/m2, interception About 100,000 points of used cloud quantity.
The present embodiment airborne LiDAR point cloud data characteristics, which is extracted with the realization classified, to be carried out on Matlab platforms , specific implementation flow is as shown in Figure 1, specific implementation step is as follows：
1. digital independent --- the LiDAR point cloud data downloaded are txt forms, and original document is larger (about 470M), i.e., point cloud data is read in workspace using MATLAB softwares, is converted to mat data formats, wherein each columns According to information such as the X-coordinate of representation space point successively, Y coordinate, height value, echo strength, echo times.
2. cloud elimination of rough difference --- after reading in point cloud data, it is shown in the form of three dimensions, such as Fig. 2 (a), according to Its elevation distribution situation, as shown in Fig. 2 (b), it is [135m, 308m] artificially to determine elevation section, will height higher than 308m or Highly it is less than the rough error point deletion of 135m；Embodiment has intercepted one piece of region in the point cloud data after excluding gross error point and has carried out Processing, as shown in Fig. 2 (c)；Shown in Fig. 3, rough error point schematic diagram；As shown in figure 4, characteristic rejects front and rear pair in rough error point Than：Fig. 4 (a) and Fig. 4 (b) is that original point cloud three-dimensional effect diagram rejects the comparison of result three-dimensional effect diagram with rough error point；Fig. 4 (c) and Fig. 4 (d) is that original point cloud section design sketch rejects result section design sketch with rough error point；Fig. 4 (e) and Fig. 4 (f) is original point cloud Vertical view and rough error point reject result vertical view.
3. single-point feature extracts --- not only include the three dimensions of every in the point cloud data that airborne LiDAR systems obtain Coordinate information further includes the information such as the intensity in relation to return laser beam, number, position, by the elevation information of cloud, strength information, Echo times information is stored as single-point feature information, these three single-point feature information are respectively such as Fig. 5 (a), 5 (b) and 5 (c) shown in.
Elevation information represents that the data that GPS, INS in LiDAR systems, laser scanning system are obtained are calculated by coordinate Three-dimensional point height, the return laser beam intensity that strength information is obtained by LiDAR systems obtains by quantization, and echo times pass through LiDAR receives the return laser beam wave crest quantity that system receives and determines.
It for relative altitude feature, needs to establish ground level model, and contains in original point cloud such as Fig. 6 (a) scenes and build Object is built, therefore when choosing ground seed point, sets sizing grid as 30m, ground point can be included in each mesh scale, 144 ground seed points finally are obtained, are evenly distributed in the plane of about 360m × 360m, ground seed point such as Fig. 6 (b).All seed points model ground level using surface fitting function griddata, terrain model such as Fig. 6 (c) institutes Show.Ground level model, which is subtracted, with test area original point cloud level degree obtains relative altitude feature.
4. neighborhood characteristics are extracted --- point cloud variable density so that the point quantity in fixed dimension neighborhood differs greatly, even The lazy weight of point is to carry out local space structural analysis, therefore the present embodiment substitutes fixed dimension neighborhood using k neighbours point set Neighborhood characteristics extraction is carried out, selected Neighbor Points quantity is 10, as shown in Figure 7.Calculate the three-dimensional coordinate covariance matrix of point set(wherein,P0Represent current point) and its characteristic value, feature vector, planar process Vector angle.Partial structurtes information is calculated using characteristic value, as shown in Fig. 8 (b), (c).Calculate echo strength Variance featureWith height variance featureWherein, I0、h0Represent current point echo strength information and Elevation information, hi、Represent the height of every and the average height of all the points in neighborhood point set, Ii、Represent every in neighborhood point set The echo strength of point and the mean echo intensity of all the points.
5. feature pre-processes --- every one-dimensional characteristic f that step 4 is extracted utilizes formula f=(f-min)/(max- Min it) is transformed in [0,1] section, wherein max is characterized the maximum value of f, and min is characterized the minimum value of f.By normalizing Most of characteristic value set after change is near 0 or 1, by taking maximal eigenvector as an example, is distributed such as Figure 10 (a) institutes Show, most of characteristic point is concentrated near 0 or 1, and dynamic range is smaller, and remaining characteristic numeric distribution compared with from Dissipate, dynamic range is larger, then by function shown in Fig. 9 (a) to the function y=ln (x+1) that one unit of left obtains to dynamic State range is adjusted, and expands the dynamic range near numerical value 0, compresses the dynamic range near 1 value, obtained result such as Figure 10 (b) shown in, Figure 10 (c) and 10 (d) are its plane comparison diagrams.
6. point cloud classifications --- 50 are selected in data has exemplar as one kind, calculates sample to be sorted and 4 classes The theorem in Euclid space distance of the feature vector of totally 200 exemplars finds out 20 closest points, which belongs in this 20 points The quantity of one kind point is more, then just sample to be sorted is assigned in such, classification results are as shown in figure 11.

Claims (5)

  1. A kind of 1. terrain classification method of electric power corridor airborne LiDAR point cloud data, it is characterised in that：One kind is based on airborne LiDAR systems include the following steps the terrain classification method of electric power corridor LiDAR point cloud data：
    Step 1: obtaining LiDAR point cloud data, LiDAR point cloud data progress three-dimensional visualization is shown that setting elevation threshold value is gone Except rough error point；
    Step 2: feature extraction and characteristic processing are carried out to LiDAR point cloud data；
    Step 2 one, extraction single-point feature；
    Step 211, the height using in LiDAR point cloud data, intensity, echo times, echo position information are as one put Feature；
    LiDAR point cloud data are carried out piecemeal by step 2 one or two, are chosen per minimum point in the block as ground point progress earth's surface mould Type is fitted, and original point cloud level degree is subtracted relative altitude information that ground level put as a feature；By relative altitude It is set as ground point less than 1m；
    Step 2 two, extraction neighborhood characteristics；
    Step 2 three, feature normalization and feature selecting；
    Step 3: random selection has exemplar from LiDAR point cloud data；
    Step 4: using there is exemplar by grader i.e. k neighbours criterion sorting technique to LiDAR point cloud data sample to be sorted This is classified, and obtains electric power facility, vegetation, building, earth's surface point cloud data；
    Step 4 one calculates sample to be sorted and all theorem in Euclid space distances for having exemplar Y represents sample to be sorted, xiIndicate exemplar, niRepresent that the i-th class has the quantity of exemplar；
    Step 4 two classifies sample to be sorted according to k neighbours criterion, by sample to be sorted and has exemplar distance most Small is divided into one kind, i.e., the sample size that k arest neighbors of sample to be sorted has the i-th class in exemplar is mi, then classification mark Sign ω=argmaxmi。
  2. 2. a kind of terrain classification method of electric power corridor airborne LiDAR point cloud data according to claim 1, feature exist In：The step 1 is as follows：
    Point cloud data is read workspace by step 1 one using MATLAB softwares, is carried out visualization and is shown；
    Step 1 two, basis, not higher than the ground object target point of peak in scene, determine that atural object target point is deposited not less than ground point Elevation section [L, H], wherein L represent elevation threshold value minimum point, H represent elevation threshold value peak, by elevation threshold value be less than L It is weeded out with point of the elevation threshold value more than H, completes removal rough error point.
  3. 3. a kind of terrain classification method of electric power corridor airborne LiDAR point cloud data according to claim 2, feature exist In：The step 2 two is extracted neighborhood characteristics and is as follows：
    Step 221 calculates distance between two points in point cloud, finds out and current point P0K closest point forms its neighbour Point set { Pi, i=1,2 ..., k, the 3 d space coordinate of every are expressed as P with vectori=(xi,yi,zi)T, wherein, PiIt represents I-th of Neighbor Points vector, xi,yi,ziRepresentation space three-dimensional coordinate, T represent transposition；
    Step 2 two or two calculates the covariance matrix that each point concentrates all the points three-dimensional coordinateIts In,Represent k Neighbor Points average value, P0Represent current point；Solve three eigenvalue λs of Σ1>λ2> λ3And its corresponding three feature vector v1,v2,v3；
    Using characteristic value as three features of current point, other characteristic quantities are defined on the basis of characteristic value：
    Linear linearityInvariance omnivariance
    Face shape planarityAnisotropy anisotropy
    Spherical sphericityIntrinsic entropy eigenentropy
    According to characteristic value, the physical significance of feature vector, third feature vector v3The normal vector of corresponding fit Plane, and in method Calculating method vector v on the basis of vector3With the angle of vertical direction, using angle angle information as current point P0A feature； Assuming that normal vector is v3=(v31,v32,v33)T, then angle be
    Step 2 two or three calculates the height variance put in neighborhood, intensity variance, angle variance, three features as current point；
    Step 2 two or four searches the point that distance is farthest in neighborhood point set, according to the sphere volume using this maximum distance as radius Calculate the dot density feature of neighborhood point set.
  4. 4. a kind of terrain classification method of electric power corridor airborne LiDAR point cloud data according to claim 3, feature exist In：Three feature normalization of step 2 and feature selecting are as follows：
    Step 231, the feature point cloud data extracted, using formula f1=(f-min)/(max-min), f1 ∈ [0,1], Wherein max is characterized the maximum value of f, and min is characterized the minimum value of f；
    Step 2 three or two, the character numerical value after normalization are concentrated near 0 or 1, and dynamic range is smaller；And after normalization Residue character numeric distribution is more discrete, and dynamic range is larger, then passes through y=ln (x+1) or y=ex+1Dynamic range is carried out Adjustment.
  5. 5. a kind of terrain classification method of electric power corridor airborne LiDAR point cloud data according to claim 4, feature exist In：Random selection has exemplar to be as follows in the step 3 LiDAR point cloud data：
    Step 3 one, the classification information by determining certain point or certain point set, being added to corresponding has in exemplar set；
    Step 3 two selected has that exemplar should be uniformly distributed and sorted Different categories of samples number is equal.
CN201510547324.1A 2015-08-31 2015-08-31 A kind of terrain classification method of electric power corridor airborne LiDAR point cloud data Active CN105184250B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201510547324.1A CN105184250B (en) 2015-08-31 2015-08-31 A kind of terrain classification method of electric power corridor airborne LiDAR point cloud data

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201510547324.1A CN105184250B (en) 2015-08-31 2015-08-31 A kind of terrain classification method of electric power corridor airborne LiDAR point cloud data

Publications (2)

Publication Number Publication Date
CN105184250A CN105184250A (en) 2015-12-23
CN105184250B true CN105184250B (en) 2018-06-22

Family

ID=54906319

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201510547324.1A Active CN105184250B (en) 2015-08-31 2015-08-31 A kind of terrain classification method of electric power corridor airborne LiDAR point cloud data

Country Status (1)

Country Link
CN (1) CN105184250B (en)

Families Citing this family (25)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105701856B (en) * 2016-01-15 2018-10-19 厦门大学 A kind of vegetation extracting method and system
CN105701478B (en) * 2016-02-24 2019-03-26 腾讯科技（深圳）有限公司 The method and apparatus of rod-shaped Objects extraction
CN106199557B (en) * 2016-06-24 2018-07-10 南京林业大学 A kind of airborne laser radar data vegetation extracting method
CN106526610B (en) * 2016-11-04 2019-04-09 广东电网有限责任公司电力科学研究院 A kind of pylon automatic positioning method and device based on unmanned plane laser point cloud
CN106897686A (en) * 2017-02-19 2017-06-27 北京林业大学 A kind of airborne LIDAR electric inspection process point cloud classifications method
CN107273902B (en) * 2017-05-19 2019-10-18 中国科学院遥感与数字地球研究所 A method of pylon point cloud is automatically extracted from on-board LiDAR data
CN107767389A (en) * 2017-10-27 2018-03-06 云南电网有限责任公司电力科学研究院 A kind of power transmission line extraction method based on airborne laser point cloud data
CN108564650B (en) * 2018-01-08 2022-03-04 南京林业大学 Lane tree target identification method based on vehicle-mounted 2D LiDAR point cloud data
CN108562885B (en) * 2018-03-29 2021-12-31 同济大学 High-voltage transmission line airborne LiDAR point cloud extraction method
CN108647607A (en) * 2018-04-28 2018-10-12 国网湖南省电力有限公司 Objects recognition method for project of transmitting and converting electricity
CN109492658A (en) * 2018-09-21 2019-03-19 北京车和家信息技术有限公司 A kind of point cloud classifications method and terminal
CN109492699A (en) * 2018-11-21 2019-03-19 国网江苏省电力有限公司扬州供电分公司 Passway for transmitting electricity method for three-dimensional measurement and device
CN109741329A (en) * 2018-11-27 2019-05-10 广东工业大学 A kind of point cloud segmentation method towards electric power corridor scene
CN109948414A (en) * 2018-12-29 2019-06-28 中国科学院遥感与数字地球研究所 Electric power corridor scene classification method based on LiDAR point cloud feature
CN109948683B (en) * 2019-03-12 2021-07-27 百度在线网络技术（北京）有限公司 Method and device for dividing difficulty degree of point cloud data and related equipment thereof
CN109901618A (en) * 2019-03-29 2019-06-18 国网山东省电力公司梁山县供电公司 Power-line patrolling system and method
CN110782457B (en) * 2019-09-06 2023-05-23 广东工业大学 Point cloud data segmentation method based on unmanned aerial vehicle power inspection
CN111060922B (en) * 2019-12-11 2023-04-18 电子科技大学 Tree point cloud extraction method based on airborne laser radar point cloud spatial distribution characteristics
CN111369610B (en) * 2020-03-05 2022-09-06 山东交通学院 Point cloud data gross error positioning and eliminating method based on credibility information
CN111553946B (en) * 2020-04-17 2023-04-18 中联重科股份有限公司 Method and device for removing ground point cloud and method and device for detecting obstacle
CN111859772B (en) * 2020-07-07 2023-11-17 河南工程学院 Power line extraction method and system based on cloth simulation algorithm
CN112767352B (en) * 2021-01-19 2023-07-07 国网电力空间技术有限公司 Automatic refined extraction method for high-voltage line tower for airborne LiDAR power inspection
CN113344956B (en) * 2021-06-21 2022-02-01 深圳市武测空间信息有限公司 Ground feature contour extraction and classification method based on unmanned aerial vehicle aerial photography three-dimensional modeling
CN113759947B (en) * 2021-09-10 2023-08-08 中航空管系统装备有限公司 Airborne flight obstacle avoidance assisting method, device and system based on laser radar
CN114462493A (en) * 2021-12-29 2022-05-10 浙江大华技术股份有限公司 Clustering method, clustering device and computer readable storage medium

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102590823A (en) * 2012-01-06 2012-07-18 中国测绘科学研究院 Rapid extraction and reconstruction method for data power line of airborne LIDAR
CN102103202B (en) * 2010-12-01 2013-01-09 武汉大学 Semi-supervised classification method for airborne laser radar data fusing images
CN102879788A (en) * 2012-09-04 2013-01-16 广州建通测绘技术开发有限公司 Power line extraction method of electric transmission line based on on-board LiDAR data
CN103412296A (en) * 2013-06-28 2013-11-27 广东电网公司电力科学研究院 Automatic extraction method of power tower in random laser point cloud data

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP5161936B2 (en) * 2010-08-11 2013-03-13 株式会社パスコ Data analysis apparatus, data analysis method, and program

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102103202B (en) * 2010-12-01 2013-01-09 武汉大学 Semi-supervised classification method for airborne laser radar data fusing images
CN102590823A (en) * 2012-01-06 2012-07-18 中国测绘科学研究院 Rapid extraction and reconstruction method for data power line of airborne LIDAR
CN102879788A (en) * 2012-09-04 2013-01-16 广州建通测绘技术开发有限公司 Power line extraction method of electric transmission line based on on-board LiDAR data
CN103412296A (en) * 2013-06-28 2013-11-27 广东电网公司电力科学研究院 Automatic extraction method of power tower in random laser point cloud data

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
"基于机载LiDAR点云数据的建筑物重建技术研究";陈永枫;《中国优秀硕士学位论文全文数据库信息科技辑》;20140215(第2期);正文第10页第2-3段,第11页第1段,第12页第4-6段,第13页第1-2段 *
"机载LiDAR点云与遥感影像融合的地物分类技术研究";董保根;《中国博士学位论文全文数据库信息科技辑》;20140115(第1期);正文第36页图3-1,第37页第2段,第42页第5段,第61页第1段,第3-4段，第64页第3段,70页第6段,第71页第1-2段,表4-1,第75页图4-11 *

Also Published As

Publication number Publication date
CN105184250A (en) 2015-12-23

Similar Documents

Publication Publication Date Title
CN105184250B (en) A kind of terrain classification method of electric power corridor airborne LiDAR point cloud data
CN110570428B (en) Method and system for dividing building roof sheet from large-scale image dense matching point cloud
CN112381861B (en) Forest land point cloud data registration and segmentation method based on foundation laser radar
CN106199557B (en) A kind of airborne laser radar data vegetation extracting method
Kim et al. 3D classification of power-line scene from airborne laser scanning data using random forests
Carlberg et al. Classifying urban landscape in aerial LiDAR using 3D shape analysis
CN104463164B (en) It is a kind of based on umbrella frame method and crown height than trees canopy structure information extracting method
CN103218787B (en) Multi-source heterogeneous remote sensing image reference mark automatic acquiring method
CN102103202A (en) Semi-supervised classification method for airborne laser radar data fusing images
CN109446986B (en) Effective feature extraction and tree species identification method for tree laser point cloud
CN110796694A (en) Fruit three-dimensional point cloud real-time acquisition method based on KinectV2
Belton et al. Processing tree point clouds using Gaussian Mixture Models
CN113470090A (en) Multi-solid-state laser radar external reference calibration method based on SIFT-SHOT characteristics
CN107067405A (en) Based on the preferred Remote Sensing Image Segmentation of yardstick
CN104597449B (en) A kind of airborne many scanning weather radar target vertical contour reconstruction methods
CN109961470B (en) Living standing tree She Shuxing accurate estimation method based on laser point cloud
CN114764871B (en) Urban building attribute extraction method based on airborne laser point cloud
CN109492606A (en) Multispectral vector picture capturing method and system, three dimensional monolithic method and system
Hu et al. A fast and simple method of building detection from LiDAR data based on scan line analysis
WO2023060632A1 (en) Street view ground object multi-dimensional extraction method and system based on point cloud data
CN111323788A (en) Building change monitoring method and device and computer equipment
CN115761303A (en) Ground object classification method based on airborne laser radar point cloud and remote sensing image data
CN116563466A (en) Deep learning-based three-dimensional Shan Mudian cloud completion method
CN115390040A (en) Tree point cloud branch and leaf separation method based on segmentation geometric features
CN107705344A (en) Plant canopy model extracting method in laser scanning environment cloud data

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