Study On Remote Sensing Image Changedetection Methods And Application

Change detection technology in remote sensing images has been developed for over 40 years since 1972. With the constant developing of modern technology, it has become a crux in updating basic geographic database and has been widely in geographic fields close to human life including land cover and use monitoring, urbanization expansion monitoring, evaluation and prediction on natural disasters and other related applications in geographic observation. Change detection technology in remote sensing images is a procedure that spatial factors and spectral information could be automatic identified and analyzed in multi-temporal images, which aiming to extract the differences in the same zone captured in different time and ultimately realize the continues dynamic monitoring of varying and developing in the selected zone. Thus, detecting and presenting changes effectively will provide reliable and valuable gist for analyzing and predicting the possible changing situations under specific cases in a certain future. The main procedure of change detection technology in remote sensing images includes selecting and importing data source, pre-processing multi-sourced images, detecting images verifications, evaluating precision and outputting result images. The current detecting methods mainly covers three categories based on pixels, features and targets and the most commonly used are Image Differencing, Image Ratioing, Change Vector Analysis, Principal Component Analysis, Post-Classification Comparison Method and some others. However, despite the methods are used in pre-processing, change detecting or evaluating results precision, these methods are mainly based on pixels. Hence, there are some disadvantages and limitations in the current change detection methods. This essay proposed a method in acquiring change threshold automatically under the basis of traditional change vector analysis by optimizing the method of determining change threshold relying on subjective experience. Furthermore, via combining data fusion by multi-temporal remote sensing images, the possibility of importing error is reduced in the aspect of data sources. The experiment results indicate the precision result PCA obtains is the highest, the results in shadows are much better by using image ratioing method, and the change vector analysis presents results closest to visual interpretation. By using modified change vector analysis method, human factors are reduced in detecting procedure. At the same time, the effectiveness and precision are improved. By introducing the image fusion idea, detection error caused by different acquiring time will be avoided. The binary images presenting information varying or not and the precision of detection results are both satisfactory.