Extraction of urban street trees from high resolution remote sensing image

We present a method of a hierarchical classification technique to distinguish street trees from the neighboring grasslands and roads based on high resolution remote sensing images. Street trees information is obtained based on full use of the spectral and textural characteristics of different objects with the help of available corresponding methods on the fused image of QuickBird panchromatic and multispectral data. Due to high resolution of QuickBird image, the overall regulations and internal relationships, such as their primary and secondary or causal relationship, can be summarized expeditiously. Extraction of street trees is carried out after image segmentations by a decision tree which describes the regulations and relationships among objects. This method would help much improve the accuracy of urban street trees extraction.