Development and Assessment of Advanced Data Fusion Algorithms for Remotely Sensed Data
No Thumbnail Available
Files
Date
2004-12-19
Authors
Journal Title
Series/Report No.
Journal ISSN
Volume Title
Publisher
Abstract
The general objective of this study is to develop and implement data fusion methodologies for remotely sensed imagery. Utilization of multisource and multidate data sources in remote sensing studies has gained popularity with recent technological advances. From increases in spatial resolution to better change detection, data fusion processes have become an important part of remote sensing. Broadly defined, data fusion is a necessary step or component of most remote sensing studies. This research combines three studies in which new data fusion methodologies are developed and implemented for remotely sensed data.
In the first study, utilization of a multiresolution approach is explored. Information gathered from the multiresolution analysis of multispectral and panchromatic images is used to increase the spatial resolution of multispectral imagery with minimal loss in the spectral integrity of the original input data. 30-meter Landsat 7 multispectral imagery is fused with 15-meter same-satellite panchromatic imagery to increase the spatial resolution of the former.
In the second study, a multivariate analysis procedure called correspondence analysis (CA) is utilized to fuse 4-meter IKONOS multispectral imagery with 1-meter IKONOS panchromatic imagery. In this method, the last component of a correspondence analysis image is replaced with 1-meter panchromatic imagery before the CA image is transformed back to the original image space. An analysis of the fused image shows that this technique keeps most of the original image variance while distortion remains minimal.
In the last study, CA is utilized to improve change detection accuracy when detecting changes between two dates of imagery. Before and after Landsat images of Raleigh, North Carolina are transformed into component space individually using CA. Then, an image-differencing method is applied to the individual components. It is concluded that using the first CA component with image differencing drastically improves the detection of change between two dates.
Overall, these studies provide new methodologies for fusing the multisource and multidate images. Two methods to increase the spatial resolution of remotely sensed data and one method to perform change detection are proposed. All the proposed methodologies are an improvement over present commonly used techniques within their respective application areas.
Description
Keywords
image transformation, change detection, data fusion, remote sensing, pan sharpening, correspondence analysis, image fusion
Citation
Degree
PhD
Discipline
Forestry
