Remote Sensing Image Classification By A Gis Guided Spatial Analysis

Pattern recognition techniques currently used in remote sensing analysis are supervised and unsupervised classification methods. However, remotely sensed imagery can be distorted by many factors, for instance the surface slope, illumination and atmospheric effects, which will cause the classification errors. In order to improve classification accuracy, other types of information are needed such as the interrelationships between pixels or regions, previous classification results or existing map data, photometric or geometric properties on the area. This paper introduces an approach in which spatial information commonly stored in geographic information systems (GIS) is incorporated to assist the remote sensing image classification. The experimental results show that it is advantageous to use the spatial information in remote sensing analysis.