SlideShare a Scribd company logo
1 of 1
Download to read offline
Lidar data simplification based on Octree Structure
Project by Songrong Yang
Introduction
Laser	
  scanning	
  has	
  become	
  a	
  viable	
  
technique	
  for	
  the	
  collec5on	
  of	
  a	
  large	
  
amount	
  of	
  3D	
  point	
  data	
  densely	
  
distributed	
  on	
  the	
  scanned	
  object	
  
surface.	
  The	
  inherent	
  3D	
  nature	
  of	
  
point	
  cloud	
  provides	
  abundant	
  spa5al	
  
informa5on.	
  To	
  explore	
  valuable	
  
spa5al	
  informa5on	
  from	
  the	
  huge	
  
amount	
  of	
  3D	
  data	
  is	
  difficult	
  and	
  5me	
  
consuming.	
  Simplifica5on	
  and	
  
segmenta5on	
  are	
  generally	
  
prerequisite.	
  
	
  
We	
  propose	
  a	
  reduced	
  method	
  for	
  
point	
  cloud	
  data.	
  This	
  method	
  
transforms	
  the	
  data	
  of	
  points	
  into	
  an	
  
octree	
  based	
  on	
  the	
  spa5al	
  
informa5on,	
  such	
  as	
  area,	
  and	
  
simplifies	
  the	
  data	
  set	
  in	
  each	
  node	
  of	
  
the	
  octree.	
  
Octree-Structure-Based Simplification
The	
  principle	
  of	
  the	
  method	
  is	
  to	
  segment	
  point	
  cloud	
  into	
  3D	
  planes	
  and	
  then	
  restore	
  the	
  
points	
  of	
  the	
  boundary	
  of	
  each	
  plane.	
  
The	
  split	
  process	
  starts	
  from	
  the	
  whole	
  data	
  set	
  as	
  a	
  root	
  node.	
  The	
  data	
  set	
  space	
  will	
  be	
  
divided	
  into	
  8	
  equal	
  sub-­‐spaces.	
  The	
  split	
  generates	
  8	
  sub-­‐nodes	
  represen5ng	
  the	
  split	
  spaces.	
  
Each	
  sub-­‐node	
  will	
  be	
  split	
  con5nuously	
  un5l	
  the	
  scan	
  points	
  contained	
  in	
  the	
  split	
  space	
  of	
  the	
  
sub-­‐node	
  are	
  distributed	
  close	
  to	
  a	
  3D	
  best-­‐fit	
  plane	
  or	
  less	
  than	
  3	
  poins.	
  	
  
In	
  each	
  node,	
  we	
  apply	
  the	
  applica5on	
  of	
  least-­‐squares	
  es5ma5on	
  to	
  determine	
  whether	
  the	
  
points	
  in	
  this	
  node	
  can	
  compose	
  a	
  plane.	
  When	
  there	
  is	
  a	
  plane	
  can	
  be	
  determined,	
  the	
  
program	
  will	
  find	
  out	
  the	
  Convex	
  Hull	
  of	
  this	
  set	
  of	
  points	
  and	
  then	
  restore	
  their	
  informa5on.	
  	
  
We	
  simplify	
  the	
  data	
  set	
  by	
  filtering	
  out	
  those	
  points	
  that	
  are	
  not	
  used	
  to	
  form	
  the	
  plane.	
  
	
  
Result of Experiments
We	
  test	
  the	
  algorithm	
  with	
  some	
  laser	
  scanning	
  data	
  files.	
  Figure	
  1	
  is	
  the	
  example	
  of	
  source	
  lidar	
  points	
  cloud	
  before	
  
split.	
  Figure	
  2	
  shows	
  the	
  result	
  of	
  split	
  node	
  border.	
  And	
  Figure	
  3	
  shows	
  the	
  best-­‐fit	
  plane	
  in	
  each	
  sub	
  node.	
  	
  
In	
  this	
  data	
  file,	
  there	
  are	
  916137	
  points	
  
and	
  aPer	
  simplifica5on	
  the	
  number	
  of	
  
points	
  is	
  reduced	
  to	
  485073.	
  The	
  
displaying	
  5me	
  in	
  OpenGL	
  is	
  also	
  
drecreased	
  by	
  47.1%.	
  There	
  are	
  more	
  
results	
  for	
  comparison	
  in	
  table	
  1.	
  
	
  
On	
  average,	
  the	
  reduced	
  data	
  set’s	
  size	
  
is	
  only	
  56.78%	
  of	
  the	
  raw	
  data	
  set’s.	
  We	
  
reload	
  the	
  reduced	
  data	
  files	
  and	
  find	
  
that	
  the	
  running	
  5me	
  is	
  averagely	
  
promoted	
  by	
  54.9%.	
  	
  
	
  
The	
  5me	
  for	
  simplifica5on	
  is	
  mainly	
  
depended	
  on	
  the	
  file	
  size.	
  Files	
  with	
  
millions	
  points	
  require	
  about	
  10	
  mins	
  for	
  
simplifica5on.	
  Along	
  with	
  the	
  size,	
  the	
  
numbers	
  of	
  level	
  and	
  leaves	
  of	
  octree	
  
affect	
  the	
  5me	
  as	
  well.	
  	
  
Supervised by Prof. David Suter
acknowledgement
Figure 1
Figure 2
Figure 3
Table 1
Future Works
Further	
  study	
  is	
  needed	
  to	
  improve	
  the	
  
proposed	
  method	
  to	
  fit	
  various	
  
applica5ons.	
  Some	
  assump5ons	
  can	
  be	
  
taken	
  into	
  account:	
  	
  
	
  
Add	
  manual	
  controls.	
  The	
  proposed	
  
algorithm	
  automa5cally	
  splits	
  the	
  
points	
  based	
  on	
  their	
  spa5al	
  
informa5on-­‐-­‐posi5ons.	
  We	
  can	
  get	
  
reasonable	
  results	
  from	
  experiments	
  
but	
  some5mes	
  it	
  splits	
  points	
  which	
  
should	
  not	
  have	
  rela5ons	
  into	
  one	
  
segmenta5on.	
  To	
  avoid	
  this,	
  we	
  can	
  
manually	
  divide	
  the	
  whole	
  data	
  set	
  into	
  
several	
  parts	
  and	
  then	
  apply	
  the	
  
proposed	
  algorithm	
  for	
  each	
  part.	
  
	
  
Consider	
  normal	
  vectors	
  of	
  points.	
  
The	
  normal	
  vectors	
  of	
  all	
  points	
  in	
  each	
  
node	
  can	
  be	
  calculated.	
  If	
  the	
  points	
  in	
  
current	
  node	
  can	
  not	
  compose	
  a	
  plane	
  
we	
  can	
  also	
  compare	
  each	
  point’s	
  
normal	
  vector	
  and	
  the	
  average	
  normal	
  
vector	
  of	
  all	
  points	
  in	
  this	
  area	
  before	
  
the	
  con5nued	
  split.	
  By	
  this	
  method,	
  we	
  
may	
  avoid	
  losing	
  some	
  important	
  
spa5al	
  informa5on.
Why Octree?
An	
  octree	
  is	
  a	
  tree	
  data	
  structure	
  in	
  
which	
  each	
  internal	
  node	
  has	
  up	
  to	
  
eight	
  children.	
  Octrees	
  are	
  most	
  oPen	
  
used	
  to	
  par55on	
  a	
  three	
  dimensional	
  
space	
  by	
  recursively	
  subdividing	
  it	
  into	
  
eight	
  octants.	
  It	
  has	
  following	
  merits:	
  	
  
	
  
Spa8al	
  index.	
  The	
  program	
  processes	
  
the	
  segmenta5on	
  automa5cally	
  by	
  
area.	
  
	
  
Require	
  less	
  computer	
  memory.	
  
	
  
Low	
  complexity	
  and	
  simple	
  data	
  
structure.	
  
	
  

More Related Content

What's hot

MetaPerturb: Transferable Regularizer for Heterogeneous Tasks and Architectures
MetaPerturb: Transferable Regularizer for Heterogeneous Tasks and ArchitecturesMetaPerturb: Transferable Regularizer for Heterogeneous Tasks and Architectures
MetaPerturb: Transferable Regularizer for Heterogeneous Tasks and Architectures
MLAI2
 
Selection of Energy Efficient Clustering Algorithm upon Cluster Head Failure ...
Selection of Energy Efficient Clustering Algorithm upon Cluster Head Failure ...Selection of Energy Efficient Clustering Algorithm upon Cluster Head Failure ...
Selection of Energy Efficient Clustering Algorithm upon Cluster Head Failure ...
ijsrd.com
 

What's hot (20)

딥러닝 논문읽기 모임 - 송헌 Deep sets 슬라이드
딥러닝 논문읽기 모임 - 송헌 Deep sets 슬라이드딥러닝 논문읽기 모임 - 송헌 Deep sets 슬라이드
딥러닝 논문읽기 모임 - 송헌 Deep sets 슬라이드
 
sns
snssns
sns
 
J. Park, AAAI 2022, MLILAB, KAIST AI
J. Park, AAAI 2022, MLILAB, KAIST AIJ. Park, AAAI 2022, MLILAB, KAIST AI
J. Park, AAAI 2022, MLILAB, KAIST AI
 
MetaPerturb: Transferable Regularizer for Heterogeneous Tasks and Architectures
MetaPerturb: Transferable Regularizer for Heterogeneous Tasks and ArchitecturesMetaPerturb: Transferable Regularizer for Heterogeneous Tasks and Architectures
MetaPerturb: Transferable Regularizer for Heterogeneous Tasks and Architectures
 
Lalal
LalalLalal
Lalal
 
E035425030
E035425030E035425030
E035425030
 
Efficient Neural Architecture Search via Parameter Sharing
Efficient Neural Architecture Search via Parameter SharingEfficient Neural Architecture Search via Parameter Sharing
Efficient Neural Architecture Search via Parameter Sharing
 
D0931621
D0931621D0931621
D0931621
 
Enterprise Scale Topological Data Analysis Using Spark
Enterprise Scale Topological Data Analysis Using SparkEnterprise Scale Topological Data Analysis Using Spark
Enterprise Scale Topological Data Analysis Using Spark
 
A FAST FAULT TOLERANT PARTITIONING ALGORITHM FOR WIRELESS SENSOR NETWORKS
A FAST FAULT TOLERANT PARTITIONING ALGORITHM FOR WIRELESS SENSOR NETWORKSA FAST FAULT TOLERANT PARTITIONING ALGORITHM FOR WIRELESS SENSOR NETWORKS
A FAST FAULT TOLERANT PARTITIONING ALGORITHM FOR WIRELESS SENSOR NETWORKS
 
A Novel Low Complexity Histogram Algorithm for High Performance Image Process...
A Novel Low Complexity Histogram Algorithm for High Performance Image Process...A Novel Low Complexity Histogram Algorithm for High Performance Image Process...
A Novel Low Complexity Histogram Algorithm for High Performance Image Process...
 
Non-parametric probability distribution for fitting data
Non-parametric probability distribution for fitting dataNon-parametric probability distribution for fitting data
Non-parametric probability distribution for fitting data
 
Exploring Randomly Wired Neural Networks for Image Recognition
Exploring Randomly Wired Neural Networks for Image RecognitionExploring Randomly Wired Neural Networks for Image Recognition
Exploring Randomly Wired Neural Networks for Image Recognition
 
Variable neighborhood Prediction of temporal collective profiles by Keun-Woo ...
Variable neighborhood Prediction of temporal collective profiles by Keun-Woo ...Variable neighborhood Prediction of temporal collective profiles by Keun-Woo ...
Variable neighborhood Prediction of temporal collective profiles by Keun-Woo ...
 
Multiple region of interest tracking of non rigid objects using demon's algor...
Multiple region of interest tracking of non rigid objects using demon's algor...Multiple region of interest tracking of non rigid objects using demon's algor...
Multiple region of interest tracking of non rigid objects using demon's algor...
 
MULTIPLE REGION OF INTEREST TRACKING OF NON-RIGID OBJECTS USING DEMON'S ALGOR...
MULTIPLE REGION OF INTEREST TRACKING OF NON-RIGID OBJECTS USING DEMON'S ALGOR...MULTIPLE REGION OF INTEREST TRACKING OF NON-RIGID OBJECTS USING DEMON'S ALGOR...
MULTIPLE REGION OF INTEREST TRACKING OF NON-RIGID OBJECTS USING DEMON'S ALGOR...
 
A Neural Network that Understands Handwriting
A Neural Network that Understands HandwritingA Neural Network that Understands Handwriting
A Neural Network that Understands Handwriting
 
Selection of Energy Efficient Clustering Algorithm upon Cluster Head Failure ...
Selection of Energy Efficient Clustering Algorithm upon Cluster Head Failure ...Selection of Energy Efficient Clustering Algorithm upon Cluster Head Failure ...
Selection of Energy Efficient Clustering Algorithm upon Cluster Head Failure ...
 
C1804011117
C1804011117C1804011117
C1804011117
 
Proximity Detection in Distributed Simulation of Wireless Mobile Systems
Proximity Detection in Distributed Simulation of Wireless Mobile SystemsProximity Detection in Distributed Simulation of Wireless Mobile Systems
Proximity Detection in Distributed Simulation of Wireless Mobile Systems
 

Viewers also liked

Three dimensional object
Three dimensional objectThree dimensional object
Three dimensional object
Kristalina Dewi
 
3 d projections
3 d projections3 d projections
3 d projections
Mohd Arif
 

Viewers also liked (12)

Joan vinyoli diari ara
Joan vinyoli diari araJoan vinyoli diari ara
Joan vinyoli diari ara
 
Best Practices For Information Security Management 2011
Best Practices For Information Security Management 2011Best Practices For Information Security Management 2011
Best Practices For Information Security Management 2011
 
Mieszko Zielinski (Epic Games), White Nights 2015
Mieszko Zielinski  (Epic Games), White Nights 2015 Mieszko Zielinski  (Epic Games), White Nights 2015
Mieszko Zielinski (Epic Games), White Nights 2015
 
Hawaii Pacific GIS Conference 2012: Natural Resource Management - Assessing H...
Hawaii Pacific GIS Conference 2012: Natural Resource Management - Assessing H...Hawaii Pacific GIS Conference 2012: Natural Resource Management - Assessing H...
Hawaii Pacific GIS Conference 2012: Natural Resource Management - Assessing H...
 
Hawaii Pacific GIS Conference 2012: 3D GIS - Has GIS Become 3D Yet?
Hawaii Pacific GIS Conference 2012: 3D GIS - Has GIS Become 3D Yet?Hawaii Pacific GIS Conference 2012: 3D GIS - Has GIS Become 3D Yet?
Hawaii Pacific GIS Conference 2012: 3D GIS - Has GIS Become 3D Yet?
 
Out-of-Core Construction of Sparse Voxel Octrees
Out-of-Core Construction of Sparse Voxel OctreesOut-of-Core Construction of Sparse Voxel Octrees
Out-of-Core Construction of Sparse Voxel Octrees
 
Three dimensional object
Three dimensional objectThree dimensional object
Three dimensional object
 
object 3d(1)
object 3d(1)object 3d(1)
object 3d(1)
 
Object representations
Object representationsObject representations
Object representations
 
Three dimensional (3D) GIS
Three dimensional (3D) GISThree dimensional (3D) GIS
Three dimensional (3D) GIS
 
3 d projections
3 d projections3 d projections
3 d projections
 
Projection In Computer Graphics
Projection In Computer GraphicsProjection In Computer Graphics
Projection In Computer Graphics
 

Similar to SamYang's Project Poster

ModelingOfUnsegmentedCloudPointData-RP-SanjayShukla
ModelingOfUnsegmentedCloudPointData-RP-SanjayShuklaModelingOfUnsegmentedCloudPointData-RP-SanjayShukla
ModelingOfUnsegmentedCloudPointData-RP-SanjayShukla
Sanjay Shukla
 
LogicProgrammingShortestPathEfficiency
LogicProgrammingShortestPathEfficiencyLogicProgrammingShortestPathEfficiency
LogicProgrammingShortestPathEfficiency
Suraj Nair
 
A Novel Technique to Enhance the Lifetime of Wireless Sensor Networks through...
A Novel Technique to Enhance the Lifetime of Wireless Sensor Networks through...A Novel Technique to Enhance the Lifetime of Wireless Sensor Networks through...
A Novel Technique to Enhance the Lifetime of Wireless Sensor Networks through...
IJECEIAES
 
Poster_Reseau_Neurones_Journees_2013
Poster_Reseau_Neurones_Journees_2013Poster_Reseau_Neurones_Journees_2013
Poster_Reseau_Neurones_Journees_2013
Pedro Lopes
 

Similar to SamYang's Project Poster (20)

ModelingOfUnsegmentedCloudPointData-RP-SanjayShukla
ModelingOfUnsegmentedCloudPointData-RP-SanjayShuklaModelingOfUnsegmentedCloudPointData-RP-SanjayShukla
ModelingOfUnsegmentedCloudPointData-RP-SanjayShukla
 
Experimental study of Data clustering using k- Means and modified algorithms
Experimental study of Data clustering using k- Means and modified algorithmsExperimental study of Data clustering using k- Means and modified algorithms
Experimental study of Data clustering using k- Means and modified algorithms
 
IRJET - Implementation of Neural Network on FPGA
IRJET - Implementation of Neural Network on FPGAIRJET - Implementation of Neural Network on FPGA
IRJET - Implementation of Neural Network on FPGA
 
IRJET- Significant Neural Networks for Classification of Product Images
IRJET- Significant Neural Networks for Classification of Product ImagesIRJET- Significant Neural Networks for Classification of Product Images
IRJET- Significant Neural Networks for Classification of Product Images
 
LogicProgrammingShortestPathEfficiency
LogicProgrammingShortestPathEfficiencyLogicProgrammingShortestPathEfficiency
LogicProgrammingShortestPathEfficiency
 
Predicting rainfall using ensemble of ensembles
Predicting rainfall using ensemble of ensemblesPredicting rainfall using ensemble of ensembles
Predicting rainfall using ensemble of ensembles
 
A Novel Technique to Enhance the Lifetime of Wireless Sensor Networks through...
A Novel Technique to Enhance the Lifetime of Wireless Sensor Networks through...A Novel Technique to Enhance the Lifetime of Wireless Sensor Networks through...
A Novel Technique to Enhance the Lifetime of Wireless Sensor Networks through...
 
TOWARDS REDUCTION OF DATA FLOW IN A DISTRIBUTED NETWORK USING PRINCIPAL COMPO...
TOWARDS REDUCTION OF DATA FLOW IN A DISTRIBUTED NETWORK USING PRINCIPAL COMPO...TOWARDS REDUCTION OF DATA FLOW IN A DISTRIBUTED NETWORK USING PRINCIPAL COMPO...
TOWARDS REDUCTION OF DATA FLOW IN A DISTRIBUTED NETWORK USING PRINCIPAL COMPO...
 
Comparative Study of Pre-Trained Neural Network Models in Detection of Glaucoma
Comparative Study of Pre-Trained Neural Network Models in Detection of GlaucomaComparative Study of Pre-Trained Neural Network Models in Detection of Glaucoma
Comparative Study of Pre-Trained Neural Network Models in Detection of Glaucoma
 
Poster_Reseau_Neurones_Journees_2013
Poster_Reseau_Neurones_Journees_2013Poster_Reseau_Neurones_Journees_2013
Poster_Reseau_Neurones_Journees_2013
 
IEEE_SMC_2011
IEEE_SMC_2011IEEE_SMC_2011
IEEE_SMC_2011
 
COMPARATIVE ANALYSIS OF SIMULATION TECHNIQUES: SCAN COMPRESSION AND INTERNAL ...
COMPARATIVE ANALYSIS OF SIMULATION TECHNIQUES: SCAN COMPRESSION AND INTERNAL ...COMPARATIVE ANALYSIS OF SIMULATION TECHNIQUES: SCAN COMPRESSION AND INTERNAL ...
COMPARATIVE ANALYSIS OF SIMULATION TECHNIQUES: SCAN COMPRESSION AND INTERNAL ...
 
IRJET - Comparative Study of Flight Delay Prediction using Back Propagati...
IRJET -  	  Comparative Study of Flight Delay Prediction using Back Propagati...IRJET -  	  Comparative Study of Flight Delay Prediction using Back Propagati...
IRJET - Comparative Study of Flight Delay Prediction using Back Propagati...
 
A Dependent Set Based Approach for Large Graph Analysis
A Dependent Set Based Approach for Large Graph AnalysisA Dependent Set Based Approach for Large Graph Analysis
A Dependent Set Based Approach for Large Graph Analysis
 
Big data Clustering Algorithms And Strategies
Big data Clustering Algorithms And StrategiesBig data Clustering Algorithms And Strategies
Big data Clustering Algorithms And Strategies
 
Multimode system condition monitoring using sparsity reconstruction for quali...
Multimode system condition monitoring using sparsity reconstruction for quali...Multimode system condition monitoring using sparsity reconstruction for quali...
Multimode system condition monitoring using sparsity reconstruction for quali...
 
Large Scale Kernel Learning using Block Coordinate Descent
Large Scale Kernel Learning using Block Coordinate DescentLarge Scale Kernel Learning using Block Coordinate Descent
Large Scale Kernel Learning using Block Coordinate Descent
 
DECISION TREE CLUSTERING: A COLUMNSTORES TUPLE RECONSTRUCTION
DECISION TREE CLUSTERING: A COLUMNSTORES TUPLE RECONSTRUCTIONDECISION TREE CLUSTERING: A COLUMNSTORES TUPLE RECONSTRUCTION
DECISION TREE CLUSTERING: A COLUMNSTORES TUPLE RECONSTRUCTION
 
Why is my_oracle_e-biz_database_slow_a_million_dollar_question
Why is my_oracle_e-biz_database_slow_a_million_dollar_questionWhy is my_oracle_e-biz_database_slow_a_million_dollar_question
Why is my_oracle_e-biz_database_slow_a_million_dollar_question
 
Algorithm
AlgorithmAlgorithm
Algorithm
 

SamYang's Project Poster

  • 1. Lidar data simplification based on Octree Structure Project by Songrong Yang Introduction Laser  scanning  has  become  a  viable   technique  for  the  collec5on  of  a  large   amount  of  3D  point  data  densely   distributed  on  the  scanned  object   surface.  The  inherent  3D  nature  of   point  cloud  provides  abundant  spa5al   informa5on.  To  explore  valuable   spa5al  informa5on  from  the  huge   amount  of  3D  data  is  difficult  and  5me   consuming.  Simplifica5on  and   segmenta5on  are  generally   prerequisite.     We  propose  a  reduced  method  for   point  cloud  data.  This  method   transforms  the  data  of  points  into  an   octree  based  on  the  spa5al   informa5on,  such  as  area,  and   simplifies  the  data  set  in  each  node  of   the  octree.   Octree-Structure-Based Simplification The  principle  of  the  method  is  to  segment  point  cloud  into  3D  planes  and  then  restore  the   points  of  the  boundary  of  each  plane.   The  split  process  starts  from  the  whole  data  set  as  a  root  node.  The  data  set  space  will  be   divided  into  8  equal  sub-­‐spaces.  The  split  generates  8  sub-­‐nodes  represen5ng  the  split  spaces.   Each  sub-­‐node  will  be  split  con5nuously  un5l  the  scan  points  contained  in  the  split  space  of  the   sub-­‐node  are  distributed  close  to  a  3D  best-­‐fit  plane  or  less  than  3  poins.     In  each  node,  we  apply  the  applica5on  of  least-­‐squares  es5ma5on  to  determine  whether  the   points  in  this  node  can  compose  a  plane.  When  there  is  a  plane  can  be  determined,  the   program  will  find  out  the  Convex  Hull  of  this  set  of  points  and  then  restore  their  informa5on.     We  simplify  the  data  set  by  filtering  out  those  points  that  are  not  used  to  form  the  plane.     Result of Experiments We  test  the  algorithm  with  some  laser  scanning  data  files.  Figure  1  is  the  example  of  source  lidar  points  cloud  before   split.  Figure  2  shows  the  result  of  split  node  border.  And  Figure  3  shows  the  best-­‐fit  plane  in  each  sub  node.     In  this  data  file,  there  are  916137  points   and  aPer  simplifica5on  the  number  of   points  is  reduced  to  485073.  The   displaying  5me  in  OpenGL  is  also   drecreased  by  47.1%.  There  are  more   results  for  comparison  in  table  1.     On  average,  the  reduced  data  set’s  size   is  only  56.78%  of  the  raw  data  set’s.  We   reload  the  reduced  data  files  and  find   that  the  running  5me  is  averagely   promoted  by  54.9%.       The  5me  for  simplifica5on  is  mainly   depended  on  the  file  size.  Files  with   millions  points  require  about  10  mins  for   simplifica5on.  Along  with  the  size,  the   numbers  of  level  and  leaves  of  octree   affect  the  5me  as  well.     Supervised by Prof. David Suter acknowledgement Figure 1 Figure 2 Figure 3 Table 1 Future Works Further  study  is  needed  to  improve  the   proposed  method  to  fit  various   applica5ons.  Some  assump5ons  can  be   taken  into  account:       Add  manual  controls.  The  proposed   algorithm  automa5cally  splits  the   points  based  on  their  spa5al   informa5on-­‐-­‐posi5ons.  We  can  get   reasonable  results  from  experiments   but  some5mes  it  splits  points  which   should  not  have  rela5ons  into  one   segmenta5on.  To  avoid  this,  we  can   manually  divide  the  whole  data  set  into   several  parts  and  then  apply  the   proposed  algorithm  for  each  part.     Consider  normal  vectors  of  points.   The  normal  vectors  of  all  points  in  each   node  can  be  calculated.  If  the  points  in   current  node  can  not  compose  a  plane   we  can  also  compare  each  point’s   normal  vector  and  the  average  normal   vector  of  all  points  in  this  area  before   the  con5nued  split.  By  this  method,  we   may  avoid  losing  some  important   spa5al  informa5on. Why Octree? An  octree  is  a  tree  data  structure  in   which  each  internal  node  has  up  to   eight  children.  Octrees  are  most  oPen   used  to  par55on  a  three  dimensional   space  by  recursively  subdividing  it  into   eight  octants.  It  has  following  merits:       Spa8al  index.  The  program  processes   the  segmenta5on  automa5cally  by   area.     Require  less  computer  memory.     Low  complexity  and  simple  data   structure.