Due to the complex spatial structure of the earth surface, obtaining a detailed and accurate land use/land cover (LULC) classification results with satellite data have still been problematic. The overall goal of this research is to compare the pixel based and object oriented image classification approaches in terms of the overall accuracies and robustness of the final classification product. An Aster image, dated 4/27/2005, with 3 bands from spectral regions of VNIR is used to perform the LULC classification for 16 different LULC classes. Ground truth data are collected from field surveys, available maps and Quickbird images. In pixel-based image analysis, supervised classification is performed by using maximum-likelihood classifier in Erdas 8.7. Object-oriented image analysis is conducted by utilizing Definiens Professional 5.0: The segmentation algorithm does not solely rely on the single pixel value, but also on shape, texture, and pixel spatial continuity. During the implementation, several different sets of parameters were tested for image segmentation, 20 was selected as a scale parameter and nearest neighbor was used as the classifier. At the end, the performance of pixel based and object-oriented classifications are compared based on the accuracy assessment results.