Spatio-temporal segmentation of Satellite image time series - PowerPoint PPT Presentation

Loading...

PPT – Spatio-temporal segmentation of Satellite image time series PowerPoint presentation | free to download - id: 163480-ZDc1Z



Loading


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation
Title:

Spatio-temporal segmentation of Satellite image time series

Description:

and Image Understanding for Earth Observation. Outline. I Graph representation of the SITS ... and Image Understanding for Earth Observation. For each node n ... – PowerPoint PPT presentation

Number of Views:22
Avg rating:3.0/5.0
Slides: 37
Provided by: dlr7
Learn more at: http://www.dlr.de
Category:

less

Write a Comment
User Comments (0)
Transcript and Presenter's Notes

Title: Spatio-temporal segmentation of Satellite image time series


1
Spatio-temporal segmentation of Satellite image
time series
Camille Le Men Ph. D. 2nd year
2
  • Outline

3
Graph Building
Each region corresponds to a node
Regions adjacency graph
4
Graph Building
Each region corresponds to a node
Regions adjacency graph
5
Graph Building
For each node n
For each order o
6
Graph Building
For each node n
For each order o
7
Graph Building
Result Subgraph of the intersection graph
8
Finding Clouds
BAD MATCHING
GOOD MATCHING
9
Finding Clouds
Definition Matching measure
10
Finding clouds
K-means separation
11
Finding clouds
K-means separation
12
Finding clouds
K-means separation
13
Finding Clouds
Cloud inlayed in a segmentation over a
temporally constant background
14
Finding clouds
cloud inlayed in a segmentation chosen among
the multisegmentations of a series.
15
Finding radiometric evolution models
Segmentation
16
Finding radiometric evolution models
Building of a matrix for each image containing
the mean over each region of each chanel value.
17
Finding radiometric evolution models
Clustering
18
Finding radiometric evolution models
Clusters coassociation matrices
19
Finding radiometric evolution models
Optimization of the cluster number using MDL
principle
20
Finding radiometric evolution models
Result regions in a cluster at a certain time
remain all in a common cluster at each time
unless they become too different one from another.


Result regions in a cluster at a certain time
remain all in a common cluster at each time
unless they become too different one from another.
Result regions in a cluster at a certain time
remain all in a common cluster at each time
unless they become too different one from another.
21
Finding radiometric evolution models
(
), (
)
Initialization M
T 1
M1
For each non visited node of the image T One
agglomerate to its model the cluster number of
its next node.
22
Finding radiometric evolution models
(
), (
)
Initialization M
T 1
M1
For each non visited node of the image T One
agglomerate to its model the cluster number of
its next node.
23
Finding radiometric evolution models
(
), (
)
Initialization M
T 1
M1
For each non visited node of the image T One
agglomerate to its model the cluster number of
its next node.
24
Finding radiometric evolution models
(
), (
)
Initialization M
T 1
M1
For each non visited node of the image T One
agglomerate to its model the cluster number of
its next node.
25
Finding radiometric evolution models
(
), (
)
Initialization M
T 1
M1
For each non visited node of the image T One
agglomerate to its model the cluster number of
its next node.
26
Finding radiometric evolution models
(
), (
)
Initialization M
T 1
M1
For each non visited node of the image T One
agglomerate to its model the cluster number of
its next node.
27
Finding radiometric evolution models
(
), (
)
Initialization M
T 1
M1
For each non visited node of the image T One
agglomerate to its model the cluster number of
its next node.
28
Finding radiometric evolution models
(
), (
)
Initialization M
T 1
M1
For each non visited node of the image T One
agglomerate to its model the cluster number of
its next node.
29
Finding radiometric evolution models
(
), (
)
Initialization M
T 1
M1
For each non visited node of the image T One
agglomerate to its model the cluster number of
its next node.
30
Finding radiometric evolution models
(
), (
)
Initialization M
T 1
M1
For each non visited node of the image T One
agglomerate to its model the cluster number of
its next node.
31
Finding radiometric evolution models
(
), (
)
Initialization M
T 1
M1
For each non visited node of the image T One
agglomerate to its model the cluster number of
its next node.
32
Finding radiometric evolution models
(
), (
)
Initialization M
T 1
M1
For each non visited node of the image T One
agglomerate to its model the cluster number of
its next node.
(
, ),
M1
( , ),( , ),( , ), ( ,
),( )
NbSample (2,1,2,1,4,1)
33
Finding radiometric evolution models
(
( , ),( , )
, ),
M1
34
Finding radiometric evolution models
35
Conclusion and future work
  • Achieved to find clouds on simulated data
  • Cloud detection
  • For real data - use statistics over time for
    a region. - improve the multisegmentation.
  • Finding radiometric evolution models
  • Do a better selection for the choice of the
    regions following a region.

36
Questions?
About PowerShow.com