CN107679498A - A kind of airborne laser point cloud downtown roads recognition methods - Google Patents
A kind of airborne laser point cloud downtown roads recognition methods Download PDFInfo
- Publication number
- CN107679498A CN107679498A CN201710943295.XA CN201710943295A CN107679498A CN 107679498 A CN107679498 A CN 107679498A CN 201710943295 A CN201710943295 A CN 201710943295A CN 107679498 A CN107679498 A CN 107679498A
- Authority
- CN
- China
- Prior art keywords
- road
- point
- branch
- mrow
- pixel
- 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.)
- Pending
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
- G06V20/182—Network patterns, e.g. roads or rivers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
The present invention discloses a kind of airborne laser point cloud downtown roads recognition methods, mainly comprises the following steps and first changes 3D laser point clouds as 2D image models;Secondly selection road characteristic points are as seed point;The discrepancy in elevation and brightness attribute of laser point cloud are again based on, utilizes region growth method classification initial road;It is subsequently filled road blank and smooth road border;Downtown roads center line is identified using skeletonization algorithm;Gone by center line of road hot-tempered, it is vector Road to track and couple center line of road point, the less point of flexibility is deleted using evacuating algorithm, finally by curve-fitting method smooth road net.The present invention realizes the high accuracy identification of the downtown roads net under complex scene.
Description
Technical field
The present invention relates to airborne lidar field, more particularly to a kind of airborne laser point cloud downtown roads identification side
Method.
Background technology
With the development of geographic information system technology, the demand of automatic identification urban road is continuous from remote sensing images
Increase, it is different from manual digitalization mode, it can improve oneself of remote sensing image processing using the method for automatic program identification road
Dynamicization is horizontal, saves processing cost, is greatly enhanced the efficiency of data processing.Downtown roads net can be widely applied to city rule
Draw and design, downtown roads automatic Pilot and navigation, urban traffic conditions analysis, urban road management and run etc..From
Identify that the method for road network has used for reference the Intelligent Recognition algorithm of computer realm in satellite image, many methods successfully root
Road network is extracted according to satellite remote-sensing image, the road directly identified from the satellite image less than 30m resolution ratio is with discontinuous
Based on the broken string of property[1], therefore, the Study of recognition of road is carried out with stronger using the meter level even satellite image of sub-meter grade
Realistic meaning[2,3];But correlative study by city pile and street row tree, it was also found that due to being influenceed, from city height
Divide the road precision of remote sensing image extraction poor, conversely but have more preferable result in rural area[4].Airborne lidar skill
The appearance of art, downtown roads can more accurately be identified by combining the mode of aerial image and laser point cloud[5-8], both
Fusion method not only step is more but also extremely complex, this causes the recognition efficiency of road network low.
The content of the invention
A kind of the defects of it is an object of the invention to overcome above-mentioned background technology to exist, there is provided airborne laser point Yuncheng District road
Road recognition methods, so as to reduce the identification difficulty of urban road network, improve its accuracy of identification and efficiency.
Technical scheme:A kind of airborne laser point cloud downtown roads recognition methods, this method comprise the following steps:
(1) 3D laser point clouds, are changed as 2D image models;
(2) road characteristic points, are selected by hand;
(3), according to the discrepancy in elevation and monochrome information of laser point cloud, region growth method classification initial road is utilized;
(4) initial road blank and smooth road border, are filled;
(5) center line of road, is identified using skeletonization algorithm;
(6), center line of road goes hot-tempered:Short branch is deleted according to directional chain-code and length threshold, using customized route with
Track method removes unnecessary path and annular channels;
(7), track, connect centerline, evacuating algorithm deletes unnecessary flex point, and road network is formed using curve-fitting method.
The step (2) is specifically:The discrepancy in elevation and monochrome information being rich according to road laser point cloud, using region growth method
8 neighborhoods around road seeds are searched, will meet that the point of threshold value is referred to roadway area in the discrepancy in elevation and brightness value with seed point,
Complete the identification of initial road.
The step (3) is specifically:Using road characteristic points as seed point, using seed point as starting point, in 8 neighborhoods of seed point
In the range of contrast pixel and seed point, if the discrepancy in elevation and brightness value between them are respectively smaller than given threshold, just they are classified
Enter seed point region, the process that the pixel newly sorted out is searched for and merged as seed point and more than repeating again, until not having
Untill more classifiable pixels;
The road pixel of binaryzation classification, meets pixel assignment 1 during the 1st condition in formula (1), otherwise assignment 0;
Formula (1) is defined as follows:
In formula, f (x, y) represents the pixel value in image at x rows, y row, pkRepresent the kth in the neighborhood of seed point 8
It is individual, S1For the initial road point set of classification, hsAnd hkSeed point and k-th point of elevation, i are represented respectivelysAnd ikRepresent respectively
Seed point and k-th point of brightness value, ∧ are " logical AND " operator, TiAnd ThLuminance threshold and discrepancy in elevation threshold value are represented respectively.
Filling initial road blank in the step (4), it is specifically:
A), white space is searched for using Hole Detection method;
B), count and calculate the clear area area in bianry image, filled out for the white space less than area threshold
Fill for road.
Smooth road border in the step (4), it is specifically:
Smooth operation at road boundary is using 3 × 3 wicket as mask, to open operator to complete,
Wherein, Glycerine enema g (x, y) definition is:
In formula, f (x, y) (1≤x, y≤N) represents bianry image, and N is picture traverse;And w (x, y) (1≤x, y≤L) table
Show a structural element array, L is odd number;SymbolDilation operation is represented, symbol Θ represents erosion operation.
The step (5) is specially:
A), carry out removing the pixel of east, south bound and northwest corner for the first time during sub- iteration;
B), carry out rejecting the pixel of west, north circle and southeast corner for the second time during sub- iteration;
C), above-mentioned sub- iteration twice forms once big iteration, remaining untill not more pixels are removed
Pixel constitutes road skeleton.
The step (6) is specifically:
A), for short finger noise, from certain end point, non-zero value tie point is tracked using directional chain-code, until searching
Untill crosspoint or branch point, this section of route total length is calculated, 0 value is assigned to for the short finger noise less than length threshold;
B) whole branch points, are traveled through, if running into the situation of unnecessary path, using a certain branch point as starting point, edge respectively
Two branch directions track corresponding branch's route, untill the end points traced into is crosspoint or new branch point;
C), in the route of tracking, unnecessary path is considered longer branch's route, is removed by assigning 0 value;
D), in the route of tracking, annular channels remove again by 0 value is assigned.
The step (7) is specifically:
A), using all crosspoints for finding out and branch point as starting point, tie point is tracked along all branch directions, until
Untill searching new end points, crosspoint or branch point;While tracking, linking-up road skeleton line point is into vector quantization road
Line;
B) unnecessary road skeletal point, is reduced by vacuating method, retains the characteristic point of macrobending degree, removes road bone
Small bent portion in stringing;
C), using the smooth vector road network of curve-fitting method.
Compared with prior art, present invention has the advantage that:Directly according to the discrepancy in elevation of laser point cloud and monochrome information certainly
Dynamic identification road waypoint cloud, only with laser point cloud as data source, avoid atural object caused by trees and high building shade on image
The problem of surface properties change, reduce the cost and intractability of data acquisition;Extracted using automatic identification algorithm in road
Line simultaneously carries out vector quantization, smooth operation, solves more, more, road occlusion of broken string of bifurcated of current downtown roads identification etc. and asks
Topic, it is achieved thereby that the high accuracy identification of complicated urban road network, specific advantage are as follows:
1. the present invention only need to can complete classification just according to the discrepancy in elevation and brightness attribute of laser point cloud using region growth method
Beginning road waypoint cloud.
2. the border or central area of preliminary classification the road of the present invention often clear area of Retention area not etc., using meter
Hole Detection algorithm in calculation machine vision technique come area of finding and fill in the blanks, solve because dense trees around road,
Road vehicle, the atural object such as flower bed of road center block " blank " problem caused by laser beam, by eliminating these
The accuracy of identification of road is improved in clear area.
3. for the short-track road of unnecessary similar " burr " formula, 3 × 3 structural element is used to utilize for mask and open calculation
Son further smooths the border of road, eliminates " burr " type road.
4. the present invention extracts road axis using a kind of skeletonization algorithm, the dry of road periphery atural object is effectively reduced
Disturb, realize the function of laser point cloud road-center identification.
5. the road network in city is the system mutually to circulate, main roads both sides be usually associated with some paths,
Lane, or even some small-sized parking lots, it is substantial amounts of similar to " burr " one using occurring on turnpike road after skeletonization algorithm
The subbranch route of sample, unnecessary path and annular channels all drastically influence major trunk roads as caused by the black-out effect of road in addition
The topological structure and aesthetic measure on road;The present invention utilizes the directional chain-code tracking channel noise in computer vision, can effectively know
Not and delete them.
6. a large amount of irregular curved points that the vector center line of road being directly connected to includes have had a strong impact on the reality of center line of road
Border form, redundancy bending point, smooth road can be effectively rejected using vector point evacuating algorithm and conic fitting algorithm
Center line, realize the automatic identification of smooth and beautiful appearance road network.
Brief description of the drawings
Fig. 1 is original laser point cloud chart.
Fig. 2 is the initial road point cloud (black portions) and original laser point cloud (grey parts) of classification.
(a) is the initial road of classification in Fig. 3;(b) it is using the filling of Hole Detection method and smooth road.
Fig. 4 is the pixel distribution map of 8 neighborhoods.
Fig. 5 is the road network generated after skeletonization algorithm.
Fig. 6 is road noise schematic diagram.
Fig. 7 is directional chain-code schematic diagram.
Fig. 8 is to remove the downtown roads net after noise.
Fig. 9 is to vacuate center line of road point principle schematic:(a) before, taking out alkene;(b) after, vacuating.
Figure 10 is the design sketch after automatic identification road network (the prominent line of white) is superimposed with orthography.
Embodiment
Below by specific embodiments and the drawings, the present invention is further illustrated.Embodiments of the invention are in order to more
Those skilled in the art is more fully understood the present invention well, any limitation is not made to the present invention.
A kind of airborne laser point cloud downtown roads recognition methods, is comprised the following steps that:
1) 3D laser point clouds are changed as 2D image models:
According to the elevation distribution histogram of original laser point cloud (as shown in Figure 1), high and low make an uproar is rejected using discrepancy in elevation threshold value
Sound, and 3D laser point clouds are converted into 2D image models.Cloud data is arranged to uniformly advise according to the size of cloud data
Then grid.Select the height value and brightness value of the elevation and brightness value of the closest laser spots of grid as grid.
2) road characteristic points are selected by hand:
Road characteristic points are typically chosen at crossing and vehicle, trees nearby to avoid the black-out effect of road.According to road
The discrepancy in elevation and monochrome information that laser point cloud is rich in, 8 neighborhoods around road seeds are searched using region growth method, will be with seed
Point meets that the point of threshold value is referred to roadway area in the discrepancy in elevation and brightness value, completes the identification of initial road.
3) region growth method classification road:
Using road characteristic points as seed point, using seed point as starting point, compare the height of the pixel and seed point in 8 contiguous ranges
Difference and brightness value, if both of which is less than threshold value, then the pixel is just included in seed point class, and the pixel of new search is as new
Seed point repeats the process searched for and merged above, untill without the pixel that can more merge.
According to formula (1) to road binaryzation, meet that the pixel of the discrepancy in elevation and brightness is entered as 1, be otherwise 0, classification it is initial
Road is as shown in Figure 2;
Formula (1) is defined as follows:
In formula, f (x, y) represents the pixel value in image at x rows, y row, pkRepresent the kth in the neighborhood of seed point 8
It is individual, S1For the initial road point set of classification, hsAnd hkSeed point and k-th point of elevation, i are represented respectivelysAnd ikRepresent respectively
Seed point and k-th point of brightness value, ∧ are " logical AND " operator, TiAnd ThLuminance threshold and discrepancy in elevation threshold value are represented respectively.
4) blank is filled
The automobile on Urban Streets and periphery, flower nursery, trees can cover road surface, can be left largely not after first road filtering
The clear area of rule, these clear areas do not include any laser point cloud information, seriously constrain the identification process of road.Adopted for this
With Hole Detection algorithm searching blank area, clear area is treated as being made up of White picture element and black pixel, first looks for one
Black pixel is set as current hole number and preserved that hole number adds 1 as pixel, its value is started.Forward lookup whole pixel, is looked for
The pixel not being searched to new black, either with or without the pixel for current hole number around the forward lookup pixel, if deposited
Then assigning the value of current hole number to the pixel;Reverse search whole pixel, find the new black picture not being searched
Member, the pixel around the reverse search pixel either with or without value for current hole number, if it is present the value of current hole number is assigned
Give the pixel.It is if positive with reversely all without corresponding pixel, representing that all pixels in current hole have been traversed, going to next sky
Cave repeats above search procedure, untill searched one time until all holes.Clear area area in statistical picture, for less than face
The clear area of product threshold value is filled with road pixel value.Show that the area threshold of maximum blank will at least be set by constantly experiment
For 30 ㎡, the front and rear comparison diagram of blank filling, (a) as shown in Figure 3 is the initial road of classification, and (b) is that Hole Detection is filled simultaneously
Road after smooth.
In order to further smooth present road, a convolution is completed using the unlatching operator of 3 × 3 windows in mathematical morphology
Computing.The association that operator is erosion operator and Expanded Operators is opened, is mainly used in smooth road border burr.
Wherein, Glycerine enema g (x, y) is defined as follows:
In formula, f (x, y) (1≤x, y≤N) represents bianry image, and N is picture traverse;And w (x, y) (1≤x, y≤L) table
Show a structural element array, L is odd number;SymbolDilation operation is represented, symbol Θ represents erosion operation.
5) center line of road identifies
Urban road center line is extracted using a kind of skeletonization algorithm.Assuming that road pixel value is 1, background value 0.In mark
Imago member is p1, it is respectively p to mark its neighborhood respectively around center pel in the direction of the clock2, p3, p4…p8, p9, as shown in Figure 4.
In order to safeguard the continuity of road-center skeleton, it is divided into 2 iterative process:
During 1 subiterations, p is deleted1Pixel needs to meet following 4 conditions:
(a)2≤N(p1)≤6;
(b)A(p1)=1;
(c)p2*p4*p6=0;
(d)p4*p6*p8=0.
In formula, N (p1) it is p18 neighborhoods in nonzero value number;
A(p1) it is by p2, p3, p4…p8, p9Put in order appearance 0,1 or 1,0 number;
During 2 subiterations, meet to delete p1The condition of pixel is identical with (a) and (b), and condition (c) and (d) are changed into:
(c')p2*p4*p8=0;
(d')p2*p6*p8=0.
East, south bound and northwest corner pixel are deleted during 1 subiterations, the 2nd subiterations delete west, north circle and the southeast
The pixel at angle.This 2 subiterations forms once big iteration, untill not more pixels are removed, remaining pixel
Form road axis.Road network after skeletonizing is as shown in Figure 5.
6) center line of road denoising
Road network in city is a system mutually to circulate, and main roads both sides are usually associated with some paths, Hus
Together, or even some small-sized parking lots, using substantial amounts of similar " burr " occurs on turnpike road after skeletonization algorithm equally
Subbranch route, unnecessary path and annular channels all drastically influence turnpike road as caused by the black-out effect of road in addition
Topological structure and aesthetic measure, Fig. 6 is unnecessary path and annular channels noise schematic diagram, and (8 is adjacent by end points for the road of skeletonizing
Domain point number for 1), isolated point (8 neighborhood point numbers are 0), tie point (8 neighborhood point numbers are 2), crosspoint (8 neighborhood point numbers
4) to be formed with branch point (8 neighborhood point numbers are 3).
, can indirect assignment 0 for isolated point.For short branch's route, using end points as starting point, according to directional chain-code (such as Fig. 7
It is shown) search non-zero value tie point, until finding untill crosspoint or branch point, this section of path length is calculated, if less than length
Threshold value is spent, then is its assignment 0, if cross branch is excessive, needs to set program repeatedly to circulate to eliminate branch's route.For more
Remaining path, whole branch points are searched for, if running into unnecessary path, using some branch point as starting point, respectively along 2 branches
Direction search for branch's route untill new branch point or crosspoint is searched;Because unnecessary path-length is always more than directly
Path, it is therefore desirable to assign 0 value for longer unnecessary path.For annular channels, if the terminal of branch's route of tracking is same
A starting point, then the route is determined as annular channels, assigns 0 value, eliminates the road network after unnecessary path and annular channels such as
Shown in Fig. 8.
7) vector quantization of road
Processing in order to facilitate the later stage by grid road, it is necessary to be converted to vector road.First, all points to find out
Fulcrum and crosspoint are starting point, and search for tie point along its branch direction is up to finding new end points, crosspoint or branch point
Only, in search, tie point is road vectors line.In order to reject bending point excessive in mileage chart, choose any line segment two
End points, calculate the hanging down away from if maximum vertical away from more than threshold value, retaining the point and divide the line equally to the line segment of remaining each point in line segment
Section, to two sections of line segment said process, if maximum hang down away from less than threshold value, delete the point.Program iteration is until all the type points
Untill being deleted, algorithm principle is as shown in Figure 9.Finally, using the smooth vector road network of curve-fitting method, the road network after fitting
As shown in white route in Figure 10.
Just expansion illustrates the implementation result of the present invention exemplified by the Vaihingen cities of Germany below:
The data are using the Vaihingen cities of Lai Ka companies ALS50 airborne lidars instrument scanning Germany, airborne laser
Point cloud includes more echoes and monochrome information, and density is 9.2 Dian/㎡.The city has intensive residential block, meadow, street row
Tree, the magnitude of traffic flow are larger, and landform is flat, and road network is intensive, and path, parking lot are more, and experiment scene is complicated.
The grid size that this law is set selects 33 road characteristic points, the discrepancy in elevation and luminance threshold are respectively by hand as 0.8m
0.13m and 11, blank area threshold value are arranged to 60 ㎡, and the threshold value for identifying short branch's route is 60m, and vertical distance threshold is set as
5m。
Manual vector dissolves road axis on the luminance graph of original laser point cloud, as library track road network.Use
Percentage of head rice Cp, accuracy rate CrParameter represents the accuracy of identification of urban road network.Percentage of head rice CpRepresent in road buffering area
Reference road accounts for the ratio all referring to road.Accuracy rate CrRepresent to identify that road accounts for whole in reference road buffering area
Identify the ratio of road data.Percentage of head rice Cp, accuracy rate CrFormula is represented with following formula respectively:
In formula, TPRepresent identified road and reference road intersection;FNRepresent the road omitted;FPRepresent unnecessary knowledge
Other road.
In order to more accurately reflect the precision of identification road, set for 1/2 road actual width delaying as road
Sector width is rushed, the road Identification precision of trial zone is respectively percentage of head rice CpFor 79%, accuracy rate CrFor 81%.
It should be appreciated that embodiment and example discussed herein simply to illustrate that, to those skilled in the art
For, it can be improved or be converted, and all these modifications and variations should all belong to the protection of appended claims of the present invention
Scope.
Pertinent literature:
[1] research [J] remote sensing journals of road target recognition methods in Zhang Rong, Wang Yong, Yang Rong .TM images, 2005,
(02):220-224.
[2]Kim T,Park S R,Kim M G,et al.Tracking Road Centerlines from High
Resolution Remote Sensing Images by Least Squares Correlation Matching[J]
.Photogrammetric Engineering&Remote Sensing,2004,70(12):1417-1422.
[3]Valero S,Chanussot J,Benediktsson J A,et al.Advanced Directional
Mathematical Morphology for the Detection of the Road Network in Very High
Resolution Remote Sensing Images[J].Pattern Recognition Letters,2010,31(10):
1120-1127.
[4]MARIKHU R,Dailey M N,Makhanov S S,et al.A Family of Quadratic
Snakes for Road Extraction[J].Lecture Notes in Computer Science,2007,4843:85-
94.
[5]Youn J,Bethel J S,Mikhail E M,et al.Extracting Urban Road Networks
from High-resolution True Orthoimage and Lidar[J].Photogrammetric
Engineering&Remote Sensing,2008,74(2):227-237.
[6]Kabolizade M,Ebadi H,Ahmadi S.An Improved Snake Model for
Automatic Extraction of Buildings from Urban Aerial Images and LiDAR Data[J]
.Computers Environment&Urban Systems,2010,34(5):435-441.
[7] Li Yijing, Hu Xiangyun, Zhang Jianqing, Jiang Wanshou, Zhang Yongjun images and LiDAR data information fusion complex scene
Under road automatically extract [J] mapping journal, 2012,41 (06):870-876.
[8] usury roc, Shi Wenzhong, Lv Zhiyong, city roads of the magnificent based on airborne LiDAR and high-resolution remote sensing image
Road network extracts [J] remote sensing technologies and application, 2013,28 (04):562-568.
Claims (8)
- A kind of 1. airborne laser point cloud downtown roads recognition methods, it is characterised in that:This method comprises the following steps:(1) 3D laser point clouds, are changed as 2D image models;(2) road characteristic points, are selected by hand;(3), according to the discrepancy in elevation and monochrome information of laser point cloud, using region growth method classification initial road;(4) initial road blank and smooth road border, are filled;(5), center line of road is identified using skeletonization algorithm;(6), center line of road denoising:Short branch is deleted according to directional chain-code and length threshold, using customized route tracing Remove unnecessary path and annular channels;(7), track, connect centerline, evacuating algorithm deletes unnecessary flex point, and road network is formed using curve-fitting method.
- 2. airborne laser point cloud downtown roads recognition methods according to claim 1, it is characterised in that:The step (2) Specifically:The discrepancy in elevation and monochrome information being rich according to road laser point cloud, searched using region growth method around road seeds 8 neighborhoods, will meet that the point of threshold value is referred to roadway area in the discrepancy in elevation and brightness value with seed point, complete the knowledge of initial road Not.
- 3. airborne laser point cloud downtown roads recognition methods according to claim 1, it is characterised in that:The step (3) Specifically:Using road characteristic points as seed point, using seed point as starting point, pixel and kind are contrasted in 8 contiguous ranges of seed point It is sub-, if the discrepancy in elevation and brightness value between them are respectively smaller than given threshold, just they are classified into seed point region, new classification The process that the pixel gone out is searched for and merged as seed point and more than repeating again, until being without more classifiable pixels Only;The road pixel of binaryzation classification, when meeting formula (1) conditional, pixel is entered as 1, is otherwise entered as 0;Formula (1) is defined as follows:In formula, f (x, y) represents the pixel value in image at x rows, y row, pkK-th point in the neighborhood of seed point 8 is represented, S1For the initial road point set of classification, hsAnd hkSeed point and k-th point of elevation, i are represented respectivelysAnd ikSeed point is represented respectively With k-th point of brightness value, ∧ is " logical AND " operator, TiAnd ThLuminance threshold and elevation threshold value are represented respectively.
- 4. airborne laser point cloud downtown roads recognition methods according to claim 1, it is characterised in that:The step (4) Middle filling initial road blank, it is specifically:A), white space is searched for using Hole Detection method;B), count and calculate the clear area area in bianry image, for the white space less than area threshold be filled with for Road.
- 5. airborne laser point cloud downtown roads recognition methods according to claim 1, it is characterised in that:The step (4) Middle smooth road border, it is specifically:Smooth operation at road boundary is using 3 × 3 wicket as mask, to open operator to complete,Wherein, Glycerine enema g (x, y) definition is:<mrow> <mi>g</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>=</mo> <mo>&lsqb;</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mi>&Theta;</mi> <mi>w</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>&rsqb;</mo> <mo>&CirclePlus;</mo> <mi>w</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>2</mn> <mo>)</mo> </mrow> </mrow>In formula, f (x, y) (1≤x, y≤N) represents bianry image, and N is picture traverse;And w (x, y) (1≤x, y≤L) represents one Individual structural element array, L are odd number;SymbolDilation operation is represented, symbol Θ represents erosion operation.
- 6. airborne laser point cloud downtown roads recognition methods according to claim 1, it is characterised in that:The step (5) Specially:A), carry out removing the pixel of east, south bound and northwest corner for the first time during sub- iteration;B), carry out rejecting the pixel of west, north circle and southeast corner for the second time during sub- iteration;C), above-mentioned sub- iteration twice forms once big iteration, untill not more pixels are removed, remaining pixel Constitute road skeleton.
- 7. airborne laser point cloud downtown roads recognition methods according to claim 1, it is characterised in that:The step (6) Specifically:A), for short finger noise, from certain end point, non-zero value tie point is tracked using directional chain-code, until searching friendship Untill crunode or branch point, this section of route total length is calculated, 0 value is assigned to for the short finger noise less than length threshold;B) whole branch points, are traveled through, if running into the situation of unnecessary path, using a certain branch point as starting point, respectively along two Branch direction tracks corresponding branch's route, untill the end points traced into is crosspoint or new branch point;C), in the route of tracking, unnecessary path is considered longer branch's route, is removed by assigning 0 value;D), in the route of tracking, annular channels remove again by 0 value is assigned.
- 8. airborne laser point cloud downtown roads recognition methods according to claim 1, it is characterised in that:The step (7) Specifically:A), using all crosspoints found out and branch point as starting point, tie point is tracked along all branch directions, until searching Untill new end points, crosspoint or branch point;While tracking, linking-up road skeleton line point is into vector quantization Road;B) unnecessary road skeletal point, is reduced by vacuating method, retains the characteristic point of macrobending degree, removes road skeleton line In small bent portion;C), using the smooth vector road network of curve-fitting method.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710943295.XA CN107679498A (en) | 2017-10-11 | 2017-10-11 | A kind of airborne laser point cloud downtown roads recognition methods |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710943295.XA CN107679498A (en) | 2017-10-11 | 2017-10-11 | A kind of airborne laser point cloud downtown roads recognition methods |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107679498A true CN107679498A (en) | 2018-02-09 |
Family
ID=61140005
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710943295.XA Pending CN107679498A (en) | 2017-10-11 | 2017-10-11 | A kind of airborne laser point cloud downtown roads recognition methods |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107679498A (en) |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107689048A (en) * | 2017-09-04 | 2018-02-13 | 联想(北京)有限公司 | A kind of method of detection image characteristic point and a kind of server cluster |
CN108986122A (en) * | 2018-08-01 | 2018-12-11 | 重庆大学 | Indoor parking guidance map intelligent reconstruction method |
CN109035218A (en) * | 2018-07-09 | 2018-12-18 | 武汉武大卓越科技有限责任公司 | Pavement patching method for detecting area |
CN109493379A (en) * | 2018-11-09 | 2019-03-19 | 南京天辰礼达电子科技有限公司 | A kind of algorithm that road vacuates |
CN109766831A (en) * | 2019-01-09 | 2019-05-17 | 深圳市三宝创新智能有限公司 | A kind of road colour band recognition methods, device, computer equipment and storage medium |
CN110543885A (en) * | 2019-08-13 | 2019-12-06 | 武汉大学 | method for interactively extracting high-resolution remote sensing image road and generating road network |
CN112330604A (en) * | 2020-10-19 | 2021-02-05 | 香港理工大学深圳研究院 | Method for generating vectorized road model from point cloud data |
CN112862844A (en) * | 2021-02-20 | 2021-05-28 | 苏州工业园区测绘地理信息有限公司 | Road boundary interactive extraction method based on vehicle-mounted point cloud data |
CN112949617A (en) * | 2021-05-14 | 2021-06-11 | 江西农业大学 | Rural road type identification method, system, terminal equipment and readable storage medium |
KR20210082857A (en) * | 2019-12-26 | 2021-07-06 | 한국전자통신연구원 | Method and apparatus of generating digital surface model using satellite imagery |
CN113763550A (en) * | 2021-08-26 | 2021-12-07 | 上海市政工程设计研究总院(集团)有限公司 | Section automatic production method based on three-dimensional data |
CN114677570A (en) * | 2022-03-14 | 2022-06-28 | 北京百度网讯科技有限公司 | Road information updating method, device, electronic equipment and storage medium |
CN115171376A (en) * | 2022-06-27 | 2022-10-11 | 广州小鹏汽车科技有限公司 | Map data processing method, server and storage medium |
CN117611978A (en) * | 2024-01-23 | 2024-02-27 | 日照市自然资源和规划局 | Construction method and system of land resource mapping database |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2007073009A (en) * | 2005-09-09 | 2007-03-22 | Hitachi Software Eng Co Ltd | Automatic generation method for road network data |
WO2017020466A1 (en) * | 2015-08-04 | 2017-02-09 | 百度在线网络技术(北京)有限公司 | Urban road recognition method, apparatus, storage medium and device based on laser point cloud |
-
2017
- 2017-10-11 CN CN201710943295.XA patent/CN107679498A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2007073009A (en) * | 2005-09-09 | 2007-03-22 | Hitachi Software Eng Co Ltd | Automatic generation method for road network data |
WO2017020466A1 (en) * | 2015-08-04 | 2017-02-09 | 百度在线网络技术(北京)有限公司 | Urban road recognition method, apparatus, storage medium and device based on laser point cloud |
Non-Patent Citations (1)
Title |
---|
李峰等: "机载LiDAR点云提取城市道路网的半自动方法", 《测绘科学》 * |
Cited By (23)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107689048A (en) * | 2017-09-04 | 2018-02-13 | 联想(北京)有限公司 | A kind of method of detection image characteristic point and a kind of server cluster |
CN107689048B (en) * | 2017-09-04 | 2022-05-31 | 联想(北京)有限公司 | Method for detecting image feature points and server cluster |
CN109035218A (en) * | 2018-07-09 | 2018-12-18 | 武汉武大卓越科技有限责任公司 | Pavement patching method for detecting area |
CN109035218B (en) * | 2018-07-09 | 2021-08-31 | 武汉光谷卓越科技股份有限公司 | Pavement repair area detection method |
CN108986122A (en) * | 2018-08-01 | 2018-12-11 | 重庆大学 | Indoor parking guidance map intelligent reconstruction method |
CN108986122B (en) * | 2018-08-01 | 2021-03-02 | 重庆大学 | Intelligent reconstruction method for indoor parking guide map |
CN109493379A (en) * | 2018-11-09 | 2019-03-19 | 南京天辰礼达电子科技有限公司 | A kind of algorithm that road vacuates |
CN109766831A (en) * | 2019-01-09 | 2019-05-17 | 深圳市三宝创新智能有限公司 | A kind of road colour band recognition methods, device, computer equipment and storage medium |
CN110543885A (en) * | 2019-08-13 | 2019-12-06 | 武汉大学 | method for interactively extracting high-resolution remote sensing image road and generating road network |
CN110543885B (en) * | 2019-08-13 | 2022-03-04 | 武汉大学 | Method for interactively extracting high-resolution remote sensing image road and generating road network |
KR20210082857A (en) * | 2019-12-26 | 2021-07-06 | 한국전자통신연구원 | Method and apparatus of generating digital surface model using satellite imagery |
KR102610989B1 (en) * | 2019-12-26 | 2023-12-08 | 한국전자통신연구원 | Method and apparatus of generating digital surface model using satellite imagery |
CN112330604A (en) * | 2020-10-19 | 2021-02-05 | 香港理工大学深圳研究院 | Method for generating vectorized road model from point cloud data |
CN112862844A (en) * | 2021-02-20 | 2021-05-28 | 苏州工业园区测绘地理信息有限公司 | Road boundary interactive extraction method based on vehicle-mounted point cloud data |
CN112862844B (en) * | 2021-02-20 | 2024-01-05 | 园测信息科技股份有限公司 | Road boundary interactive extraction method based on vehicle-mounted point cloud data |
CN112949617A (en) * | 2021-05-14 | 2021-06-11 | 江西农业大学 | Rural road type identification method, system, terminal equipment and readable storage medium |
CN113763550A (en) * | 2021-08-26 | 2021-12-07 | 上海市政工程设计研究总院(集团)有限公司 | Section automatic production method based on three-dimensional data |
CN113763550B (en) * | 2021-08-26 | 2024-04-26 | 上海市政工程设计研究总院(集团)有限公司 | Automatic section production method based on three-dimensional data |
CN114677570A (en) * | 2022-03-14 | 2022-06-28 | 北京百度网讯科技有限公司 | Road information updating method, device, electronic equipment and storage medium |
CN115171376A (en) * | 2022-06-27 | 2022-10-11 | 广州小鹏汽车科技有限公司 | Map data processing method, server and storage medium |
CN115171376B (en) * | 2022-06-27 | 2024-01-05 | 肇庆小鹏新能源投资有限公司广州分公司 | Map data processing method, server, and storage medium |
CN117611978A (en) * | 2024-01-23 | 2024-02-27 | 日照市自然资源和规划局 | Construction method and system of land resource mapping database |
CN117611978B (en) * | 2024-01-23 | 2024-05-03 | 日照市自然资源和规划局 | Construction method and system of land resource mapping database |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107679498A (en) | A kind of airborne laser point cloud downtown roads recognition methods | |
CN110543885B (en) | Method for interactively extracting high-resolution remote sensing image road and generating road network | |
CN105844629A (en) | Automatic segmentation method for point cloud of facade of large scene city building | |
CN111062958B (en) | Urban road element extraction method | |
CN103679655A (en) | LiDAR point cloud filter method based on gradient and area growth | |
JP2011501301A (en) | Geospatial modeling system and related methods using multiple sources of geographic information | |
CN104376595A (en) | Method for generating three-dimensional roads on basis of airborne LiDAR (light detection and ranging) and GIS (geographic information system) cooperation | |
CN111652241B (en) | Building contour extraction method integrating image features and densely matched point cloud features | |
CN108921943A (en) | A kind of road threedimensional model modeling method based on lane grade high-precision map | |
CN109671149A (en) | Landform sketch map automatic drafting method based on DEM | |
Cheng et al. | Three-dimensional reconstruction of large multilayer interchange bridge using airborne LiDAR data | |
CN104318605A (en) | Parallel lamination rendering method of vector solid line and three-dimensional terrain | |
CN112801075B (en) | Automatic rural road boundary line extraction method based on aerial image | |
CN107657618A (en) | Regional scale erosion groove extraction method based on remote sensing image and terrain data | |
CN109903379A (en) | A kind of three-dimensional rebuilding method based on spots cloud optimization sampling | |
CN115294293B (en) | Method for automatically compiling high-precision map road reference line based on low-altitude aerial photography result | |
CN103035006A (en) | High-resolution aerial image partition method based on LEGION and under assisting of LiDAR | |
CN103500278B (en) | The similar Typhoon Analysis method in path based on shortest path first | |
CN111599007B (en) | Smart city CIM road mapping method based on unmanned aerial vehicle aerial photography | |
CN108615452B (en) | A kind of unknown method for extracting roads based on people's wheel paths point multi-resolution hierarchy | |
Hu et al. | Building modeling from LiDAR and aerial imagery | |
CN114943902A (en) | Urban vegetation unmanned aerial vehicle remote sensing classification method based on multi-scale feature perception network | |
Zhao et al. | On the topographic entity-oriented digital elevation model construction method for urban area land surface | |
Wang et al. | Seamline determination for high resolution orthoimage mosaicking using watershed segmentation | |
Lin et al. | Semi-automatic road tracking using parallel angular texture signature |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20180209 |