Change Detection Methodology For Land Cover Data Updating

Land cover data play an import role in global change research, geographic conditions monitoring and ecological resource management. Because of continuously changing, accurate and up-to-date land cover data are needed. Change detection with remotely sensed imagery is a cost-effective manner for land cover updating. Euclidean distance, correlation and other mathematic metrics between spectral curves have been used to calculate change magnitude in most change detection methods. However, many pseudo changes would also be detected because of inter-class spectral variance, which remains a significant challenge for operational remote sensing applications. In general, different land cover types have their own spectral/NDVI curves characterized by typical spectral values and shapes. These spectral/NDVI values are widely used for designing change detection algorithms. However, the shape of curves has not yet been fully considered. This paper proposes to use gradient difference to quantitatively describe the spectral/NDVI shapes and the differences in shape, which the projects the change information from the original spectral space to gradient space. The main researches include the following parts.(1) In the first part, analyzing the global and regional land cover products, the challenge of land cover updating is how to design an effective methodology for change detection. By reviewing related literatures are reviewed, this part proposed the main problems of change detection are pseudo changes, which including random disturbance factors and seasonal difference factors. For solving the problems, the idea of gradient difference change detection is described, and it consists of spectral gradient difference (SGD) change detection and NDVI gradient difference (NDVI-GD) change detection. Meanwhile, a comprehensive change detection based on SGD and NDVI-GD is developed for land cover data updating in large area.(2) For random disturbance factors, SGD change detection method is developed. A spectral gradient is introduced to illustrate the shape of a spectral curve. The difference of the spectral gradient between the two spectral curves is calculated as the change magnitude and is used to determine change areas with a certain threshold. A chain model is developed to represent the knowledgebase of SGD patterns of typical change types on the basis of the reference spectra, and then land cover change type is identified by pattern matching. The effectiveness of this SGD-based change detection approach was verified by a simulation experiment and a case study of Landsat data. The results indicated that the SGD-based approach was superior to the traditional methods. (3) For seasonal difference factors, based on Landsat and MODIS data fusion, the change detection of NDVI-GD change detection is developed. The30m resolution NDVI gradient data are generated using linear mixed model based on Landsat and MODIS data. NDVI gradients describe the trend of NDVI curves, and reveal features of the growth cycle. The change detection areas are detected by NDVI gradient differences, and NDVI chain model is build from gradient and NDVI differences to determine the change types. The experiment of NDVI-GD bsed on two images from different seasons indicates that this method is effective for pseudo changes caused by seasonal difference factors.(4) Based on the above researches, Shanxi province of China is selected as study area. Integrating SGD and NDVI-GD, the land cover data of Shanxi province are updated from circa2000to2010.