The Effect of the Size of Training Sample on Classification Accuracy in Object-oriented Image Analysis

As opposed to per-pixel classification,the selection of training samples is different in object-oriented method.Based on statistical theory,the number of training samples required in object-oriented classification is studied in this paper.First,feature space analysis of images is implemented in object-oriented classification,which shows that the number of training samples needed for object-oriented classification is much less than that in per-pixel classification.Then,an experiment of remote sensing image classification is carried out to verify the authenticity based on the relations between samples and bands.