This is an old revision of the document!
Table of Contents
Proposal Structure
- Introduction: Image analysis problem, traditional techniques, multitemporal analysis, hypothesis, objective
- Review: Multitemporal approaches
- Methodology
- Schedule
- References
Fifth Attempt (15/12/2010)
from book The craft of research:
Theme | I am studying | properties of time series in remote sensing images, |
Question | because I want to find out | similar behaviors in these properties, |
Motivation | to understand/explain | their relation to changes in land use/cover. |
Trying to formulate a useful applied research problem:
If my readers want to | explain the relation of properties of time series to changes in land use/cover, |
would they think that they could do if they found out | similar behaviors in these properties? |
Fourth Attempt (08/12/2010)
Theme | I am studying | properties of time series in remote sensing images, |
Question | because I want to find out | whether they can be used to track changes in land use/cover, |
Motivation | to understand/explain whether | land changes are described by spectral changes. |
Third Attempt (18/01/2010)
- What is the hypothesis base for the work?
Classification methods based on decision trees are efficient to identify change signatures.
- What are we trying to show or explain?
?
- What is the scientific question to be answered?
How to classify change signatures from temporal and multi-dimensional remote sensing images using data mining approaches?
Theme | I am studying | multitemporal datamining, |
Question | because I want to create | a method to classify change signatures from Sattelite Image Time-Series (SITS), |
Motivation | to understand/explain how | ?. |
Second Attempt
- What is the hypothesis base for the work?
It is possible to detect spectral trajectories of certain events, using image time series.
- What are we trying to show or explain?
Region-boundary classifiers can split the feature space into detected spectral trajectories in time.
- What is the scientific question to be answered?
Multitemporal decision trees can overcome the curse of dimensionality?
Theme | I am studying | multitemporal datamining, |
Question | because I want to create | a method to recognize spectral trajectories in time, |
Motivation | to understand/explain how | certain events evolve. |
First Attempt
- What is the hypothesis base for the work?
It is possible to set a temporal signature to forest degradation.
- What are we trying to show or explain?
Forest degradation are defined by smooth (light, slim, slight?), moderate and intense levels. With a set of snapshots from remote sensing images, one can develop an algorithm to detect such levels of degradation and to model its temporal signature.
- What is the scientific question to be answered?
Forest degradation can be modelled?
Theme | I am studying | forest degradation, |
Question | because I want to create | a method to find a temporal/spectral signature using snapshots of remote sensing images, |
Motivation | to understand/explain how | this phenomena (forest degradation) triggers and develops. |
Topics for Thesis Proposal
Several and general themes.
- Temporal Data Mining
- Evolving Objects
- Change Mining
- Extend CBR
- Sustainable Computation
- Deforestation
- Urban analysis
- Application or software (?)
- Image Segmentation
- Scale-Space
- Attributes-Space
- Visualization Techniques for Geographical Databases
- Parallel algorithms
- Segmentation
- Classification