Identification scales for urban vegetation classification using high spatial resolution satellite data

The scale identification is an important issue for the vegetation classification in the same urban landscape. In this paper, a method of identification scale and the determination criterion for urban vegetation image segmentation using high spatial resolution remotely sensed imagery was proposed. The criterion of the relevant deviation with two parameters, area and number of object, was used to optimize the scale of urban objects. The effect of the optimizing scales was examined. A hierarchy classification was performed for six vegetation types using the fuzzy k-means classifier. The results showed that overall accuracy is 85.5% for our approach, and 69.7% and 65.5% for k- mean classifier with single scale and MLC (Maximum Likelihood Classifier), respectively. The improvement is achieved by the proposed method of determination scale, in which the criterion and the multi-scales classification for urban vegetation types are of the most critical values.