|
|
|
Abstract
The Analysis of Texture Features and Local Spatial Statistic in Satellite Image Data for the Mangrove Mapping Purpose – A Case Study in Malaysia
Lim Boon Leong
RS/GIS Engineer
Cilix Corporation Sdn Bhd,
Malaysia
Email: limbl@cilix.org, gnoel_blim@yahoo.com
Seng Chung Yueh
RS/GIS Engineer
Cilix Corporation Sdn Bhd
Email: scy@cilix.org
Roslinah Samad, Research Officer, MACRES, roslinah@macres.gov.my
This study experiments the texture features and local spatial statistic for the purpose of mangrove mapping. The SPOT satellite image used covers a small part of the Matang mangrove reserve forest, Malaysia which has been identified as the study area. The texture features e.g. homogeneity, entropy, energy and local spatial statistic e.g. Moran was computed from different window sizes including 3x3, 5x5 and 7x7. The Maximum Likelihood classification method was performed on the different combinations of data. The accuracy assessment was performed using the ground truth data obtained through field survey. A numerical comparison among the results is presented. The results show that local spatial statistics is highly suitable to discriminate the seven mangrove forest classes, namely Avicennia-Sonneratia forest, transitional new forest, Bruguiera cylindrica forest, Bruguiera parviflora forest, dryland forest, Rhizophora forest and inland vegetation based on the definition from forestry department.
|
|
|