 |
|
|
|
|
|
|
|
|
|
|
|
|
 |
|
|||||||||||
 |
 |
 |
|
|||||||||
 |
|
|||||||||||
 |
 |
 |
|
|||||||||
 |
|
|||||||||||
 |
 |
 |
|
|||||||||
 |
|
|||||||||||
 |
 |
|
||||||||||
 |
|
|||||||||||
 |
 |
 |
|
|||||||||
 |
|
|||||||||||
 |
 |
 |
|
|||||||||
 |
|
|||||||||||
 |
 |
 |
|
|||||||||
 |
|
|||||||||||
 |
|
|||||||||||
 |
|
|||||||||||
Algorithms and Methodologies of an Automated Spatial Change Informaiton Extraction System based on Artificial Neural Networks
Siamak Khorram and Xiaolong Dai
Funded by Cray Research, Inc.
Computing resources provided by NC Supercomputing Center
OBJECTIVE
To develop and implement a cost-effective, efficient, and automated technique for remotely sensed change detection with complete categorical change information based on artificial neural network algorithms.APPROACH
The algorithm for an automated land cover change detection system was developed and implemented based on the current neural network techniques for multispectral image classification. The suitability of application of neural networks in change detection and its related network design considerations unique to change detection were first investigated. A neural network-based change detection system using the backpropagation training algorithm was then developed.The experimental results using multitemporal Landsat Thematic Mapper (TM) imagery of Wilmington, North Carolina are provided. The results from the neural network method were compared to that from other change detection techniques.
RESULTS
The trained four-layered neural network is able to provide complete categorical information about the nature of changes and detect land cover changes with an overall accuracy of 95.6% for a four-class (i.e., 16 change classes) classification scheme. Findings of this study demonstrated the potential and advantages of using neural network models in change detection. This method also holds potential to provide a reliable tool for effectively integrating multisource remotely sensed data and existing geographic data. The trained neural network for change detection can perform change detection on a pixel-by-pixel basis in real-time. Therefore, this method has implications for real-time operation in local or regional applications.
The change map with complete land cover changes produced by using the trained neural network as a feed-forward network. Map Legend
Contact the Webmaster Contents Copyright © 2000 NC State University
