Remotely sensed change detection based on artificial neural networks

A new method for remotely sensed change detection based on artificial neura l networks is presented. The algorithm for an automated land-cover change-d etection system was developed and implemented based on the current neural n etwork techniques for multispectral image classification. The suitability o f application of neural networks in change defection and its related networ k design considerations unique to change detection were first investigated. A neural-network-based change-detection system using the backpropagation t raining algorithm was then developed. The trained four-layered neural netwo rk was able to provide complete categorical information about the nature of changes and detect land-cover changes with an overall accuracy of 95.6 per cent for a four-class (i.e., 16 change classes) classification scheme. Usin g the same training data, a maximum-likelihood supervised classification pr oduced an accuracy of 86.5 percent. The experimental results using multitem poral Landsat Thematic Mapper imagery of Wilmington, North Carolina are pro vided. Findings of this study demonstrated the potential and advantages of using neural network in multitemporal change analysis.