A Comparison of Object-Oriented Image Classification and Transect Sampling Methods for Obtaining Land Cover Information from Digital Orthophotography

作者：Tenenbaum David E^*; Yang Yun; Zhou Weiqi

来源：GIScience and Remote Sensing, 2011, 48(1): 112-129.

DOI：10.2747/1548-1603.48.1.112

Transect sampling and object-oriented classification methods are used to characterize land use/land cover in two urbanizing catchments in greater Baltimore, Maryland. The two methods are compared using a harmonized six-class scheme (woody, herbaceous, water, ground, pavement, and structure). There is higher estimation of structures in the object-oriented method. For the other five classes, results are similar, except for a difference in the proportions of the woody class. Segment-to-segment comparison shows that the resolution of the light detection and ranging (LIDAR) data used in the object-oriented method affects its accuracy. Shadows are problematic in the object-oriented method.

• 出版日期2011-3