Evaluation of Aster Spectral Bands for Agricultural Land Cover Mapping Using Pixel-based and Object-based Classification Appro

The ASTER sensor onboard NASA’s Earth observing satellite Terra is an optical remote sensor comprised of 14 spectral bands ranging from the visible to thermal infrared region. With its multispectral bands the sensor bears the potential to provide data for both a detailed land use classification of heterogeneously vegetated areas. This study evaluates the potential of ASTER data for land use classification in a typical Bangladesh agricultural environment with its small-spaced fields. Land use in Bangladesh typically consists of small agricultural fields, complex vegetation covers, and scatteredly distributed residential areas, which have been problematic in terms of land cover mapping using satellite remote sensing data due to the complexity of the spatial structure. A pixel-based approach and a multi-scale object-based method are applied for agricultural land use classification using ASTER data in Sylhet district, Bangladesh. The supervised classification was performed using the Maximum Likelihood Classifier (MLC) with ERDAS Imagine software. On the other hand, objectbased image analysis was performed through the eCognition software. The results show, that the phenological stages of the cultivars are the main factors influencing the separability of agricultural classes. To identify the best method, accuracy of each method was assessed using reference data sets derived from high resolution satellite data and ground truth field investigation data. Outcome from the classification works show that the object-based approach gave more accurate results (including higher producer’s and user’s accuracy for most of the land cover classes) than those achieved by pixel-based classification algorithms.