An Object-Based Approach for Forest-Cover Change Detection using Multi-Temporal High-Resolution Remote Sensing Data

The increasing availability of remote-sensing images, acquired periodically by satellite sensors on the same geographical area, makes it extremely interesting to develop monitoring systems capable of automatically producing and regularly updating forest-cover maps of the considered site. In this paper, we designed and developed new object-based change detection algorithms, which are aimed at updating forest-cover maps by remote sensing images. The forest-cover change detection system includes several key modules: image segmentation, difference images processing and binary change detection model using threshold. These modules are evaluated by multi-temporal QuickBird remotely sensed data set: (1) In the image segmentation module, multi-scale segmentation algorithm was used to form the image objects. (2) In the difference image module, spectral value and NDVI (normalized difference vegetation index) were taken as input data. Correlation coefficient and t-test algorithms based on objects are used to develop difference images. (3) In the binary change detection module, change maps obtained from spectral value and NDVI are compared. Finally, experimental results carried out on multi-temporal QuickBird remotely sensed data set confirm the effectiveness of the proposed system.