On the partition of classes for specific class extraction from remote sensing imagery

Specific class extraction from remote sensing image is an essential task. The accuracy of specific class extraction is relative to the class partition in an image. This paper studies the effect of class partition on specific class extraction firstly. Based on Bayesian decision rule, specific class extraction with different partitions of classes was theoretically analyzed and an experiment was performed to verify the theory. Then, a scheme of class partition for specific class extraction from remote sensing image was proposed. The scheme of partition is based on clustering technique and an experiment verified its validity in a specific class extraction from a high resolution remote sensing image.