In the work, a simple and reliable algorithm is presented to calculate the fractal dimension of single pixel for the remote sensing images, and the fractal dimension values obtained by the algorithm proposed in this work have positive correlation with the complexity of surface features. On the basis of a scene of Landsat7 ETM+ (i.e., Enhanced Thematic Mapper Plus) data and the proposed algorithm, expert classification models and fractal technique were introduced to identify the ground objects in a Chinese subtropical hilly region, where surface features are very diverse and complex. In the work, the different land use/land cover types, especially the different vegetation categories were successfully identified using the ETM+ image, and this classification has an overall accuracy of 80.25% and a K hat of 0.7738, which are higher than those of the traditional supervised classification.
Carbon stock under major land use/land cover types of Hades sub-watershed, eastern Ethiopia