A New Framework of the Unsupervised Classification for High-Resolution Remote Sensing Image

Classification plays a significant role in change detection when monitoring the evolution of the Earth’s surface. This paper proposes a novel object-oriented framework for the unsupervised classification of highresolution remote sensing images based on Jenks’ optimization. The fractal net evolution approach is employed as an image segmental technique, the spectral feature of each image object is extracted, and an algorithm of Jenks’ optimization is adopted as a classifier. Two experiments with different image platforms are conducted to evaluate the performance of the proposed framework and to compare with other traditional unsupervised classification algorithms such as the iterative self-organizing data analysis technique algorithm and k-means clustering algorithms. The proposed approach is found to be feasible and valid.