The Use of Ikonos Image and Object Oriented Classification ...

The Use of Ikonos Image and Object

Oriented Classification in the cutaenous

leishmaniasis Study

Introduction

What is the cutaneous leishmaniasis?

Infectious disease caused by protozoans of leishmania gender

-Characteristics: disease with ample spatial distribution; high

incidence of cases per year; attacks the skin, in some situations,

progressing to mutilate forms

- Transmitter: phlebotomine sandfly – Insects belonged to many

species and different genders

-Transmission Standards : Sylvester, Intermediary and Peridomestic

Introduction

Peridomestic cutaneous leishmaniasis (CL)

Favorable areas to concentration of the transmitter

EcotopesContact areas

between forest and rarefy vegetation

=

Introduction

Areas with potential risk transmission of

Peridomestic CL to human population

HumanDomiciles

Ecotopes

Objective

• Application of Object Oriented Classification to

identify the variables to define the Areas with

potential risk transmission of CL:

- Forest Areas

- Areas with rarefy trees

- Areas with presence of population

Study Areas

• two scene components of satellite Ikonos II: one for the methodology development area (13/03/02) and other to the validation area (12/04/2002)

• one digital cartographic basis (1999) - scale 1/10.000

• Digitals Orthophotos - scale 1:2.000; 1999.

• Softwares: ArcGis v.9, PCI v.10, and eCognition 4.0

Materials

IKONOS II – Image Preparation:

Orthorectification

• Verification about the necessity of realize the

orthorectification:

a) Register of the IKONOS II scene components according

to the cartographic digital basis.

b) Despite the images register: displacements between 20

and 100 meters in relation to the basis.

• DEM Generation and Orthorectification

a) Altimetry data import to PCI software and DEM

generation

b) DEM + RPC file => Image orthorectification


Orthorectification

Before Orthorectification: displacement between 20 and 100 m


Orthorectification

After Orthorectification: Displacements between 1 and 2 m


Orthorectification

• Importance of the Panchromatic Band:

a) Identify isolated smalls constructions.

b) More texture detail in vegetal cover areas

• Importance of the Multispectral Bands

a) To describe different kinds of vegetation.


Fusion Process

• Fusion Process realized using Pansharp algorithm - PCI

a) Works with 11 bits, preserving the 2048 Digitals

Numbers.

b) Preserves the colors: tests realized – Statistical

Parameters Analysis; Histograms Analysis and Visual

Analysis.


Fusion Process

IKONOS II

Segmentation and Classification

• Defined Classes:

a) Dense Trees (Forest)

b) Rarefy Trees

c) Edifications

• Segmentation:

a) two levels: one for Edifications Class and other for

vegetation classes (Forest and Rarefy Trees)

• Segmentation:

a) First Level: Scale Parameter: 35; Shape: 0.3; Color: 0.7;

Compactness: 0.5 and Smoothness: 0.5

b) Second Level: Scale Parameter: 110; Shape: 0.3; Color:

0.7; Compactness: 0.5 and Smoothness: 0.5

IKONOS II


• Segmentation:

First Level Second Level

IKONOS II


• Classification:

a) Application of the Nearest Neighbor (NN) Classifier

b) First Level:

- Classes: Edifications and Not Edifications

- Describers Used: Mean Spectral Value and Compactness

c) Second Level:

- Classes: Dense Trees; Rarefy Trees; Creeping Vegetation and Others

- Describers Used: Mean Spectral Value; Compactness and Texture

IKONOS II


• Classification:

d) Application of the classification process more than one time:

necessity to add and exclude samples

e) Exportation of the objects belonged to classes of interest

(Edifications; Dense Trees and Rarefy Trees) in a SHP format

f) Editions in the ArcMap software with the purpose to exclude the

classification errors

g) Results Verification: Classification Ambiguity Analysis and

Accuracy Analysis

IKONOS II


Results - Classification

• Methodology Development Area - Original and Classified Image

• Development Area – Ambiguity Analysis

a) Dense Trees: 0,362 – “Acceptable”

b) Rarefy Trees: 0,380 – “Acceptable”

c) Edifications: 0,951 – “Very Good”

CLASSIFICAÇÃO UNACCEPTABLE Ia = 0;CLASSIFICAÇÃO AMBIGUOUS 0,01 ≤ Ia ≤ 0,30;CLASSIFICAÇÃO ACCEPTABLE 0,31 ≤ Ia ≤ 0,50;CLASSIFICAÇÃO GOOD 0,51 ≤ Ia ≤ 0,80;CLASSIFICAÇÃO VERY GOOD 0,81 ≤ Ia ≤ 1.

Linguistic Scale (ANTUNES e LINGNAUL, 2005)


• Development Area – Kappa

Linguistic Scale (LANDIS e KOCH, 1977)

Global Kappa = 0,81 => HIGH Correlation

HIGH CORRELATION = > 0.80SUBSTANTIAL CORRELATION = 0.60 a 0.79MODERATE CORRELATION = 0.40 a 0.59SMALL CORRELATION = 0.20 a 0.39LOW CORRELATION = 0.00 a 0.19NO CORRELATION = < 0.00



• Methodology Validation Area - Original and Classified Image

a) Dense Trees: 0,335 – “Acceptable”

b) Rarefy Trees: 0,506 – “Good”

c) Edifications: 0,829 – “Very Good”’


• Validation Area – Ambiguity Analysis

CLASSIFICAÇÃO UNACCEPTABLE Ia = 0;CLASSIFICAÇÃO AMBIGUOUS 0,01 ≤ Ia ≤ 0,30;CLASSIFICAÇÃO ACCEPTABLE 0,31 ≤ Ia ≤ 0,50;CLASSIFICAÇÃO GOOD 0,51 ≤ Ia ≤ 0,80;CLASSIFICAÇÃO VERY GOOD 0,81 ≤ Ia ≤ 1.

Linguistic Scale

Global Kappa = 0,76 => SUBSTANTIAL Correlation

• Development Area – Kappa

Linguistic Scale

HIGH CORRELATION = > 0.80SUBSTANTIAL CORRELATION = 0.60 a 0.79MODERATE CORRELATION = 0.40 a 0.59SMALL CORRELATION = 0.20 a 0.39LOW CORRELATION = 0.00 a 0.19NO CORRELATION = < 0.00


The Ambiguity and Accuracy analysis showed that the

IKONOS images associated with Object Oriented

Classification can be used, with security, to generate

a product which can be applied in the study of

Peridomestic CL

Conclusion

© Threetek Ltda. 2006

Rua México, 41 / 17° andar - Centro

20031-144 - Rio de Janeiro- RJ - Brasil

Tel/Fax: (21) 2524-0207

Guilherme Medina

Geógrafo- MSc. Engenharia Cartográfica, IME

[email protected]

[email protected]

OBRIGADO

The Use of Ikonos Image and Object Oriented Classification ...

Documents