Per-pixel Classification of High Spatial Resolution Satellite Imagery for Urban Land-cover Mapping

Abstract

Commercial high spatial resolution satellite data now provide a synoptic and consistent source of digital imagery with detail comparable to that of aerial photography. In the work described here, per-pixel classification, image fusion, and GIS-based map refinement techniques were tailored to pan-sharpened 0.61 m QuickBird imagery to develop a six-category urban land-cover map with 89.3 percent overall accuracy (ϭ 0.87). The study area was a rapidly developing 71.5 km 2 part of suburban Raleigh, North Carolina, U.S.A., within the Neuse River basin. "Edge pixels" were a source of classification error as was spectral overlap between bare soil and impervious surfaces and among vegetated cover types. Shadows were not a significant source of classification error. These findings demonstrate that conventional spectral-based classification methods can be used to generate highly accurate maps of urban landscapes using high spatial resolution imagery.

Classification of Ultra-High Resolution Orthophotos Combined with DSM Using a Dual Morphological Top Hat Profile

“…Hester et al [12] applied the ISODATA (iterative self-organizing data analysis technique) to the spectral bands of the of the VHR images and reported 89.0% overall accuracy (OA). However, in their works, ground, buildings, roofs and roads were all specified as impervious objects, which were insufficient for urban mapping applications.…”

Mining Land Cover Information Using Multilayer Perceptron and Decision Tree from MODIS Data

“…Choice of MLC was more justified, as the time complexity involved in training the neurons in NN was very high compared to MLC and NN classification became more complex with the large image size of LISS-III MSS (5,997 rows × 6,142 columns). This conventional per-pixel spectral-based classifier (MLC) constitutes a historically dominant approach to RS-based automated land-use/land-cover (LULC) derivation (Gao et al 2004;Hester et al 2008) and in fact, this aids as "benchmark" for evaluating the performance of novel classification algorithms (Song et al 2005). The dimension of MODIS data was very less (532 rows × 546 columns) and hence it is reasonable to rigorously apply and choose appropriate classification algorithm.…”

A Comparison of Novel Optical Remote Sensing-Based Technologies for Forest-Cover/Change Monitoring

“…Moreover, dividing continuous quantitative information, such as those found in satellite images, into a finite number of discrete land classes that are considered at the outset to be exhaustively defined and mutually exclusive may lend itself to the further loss of information [60]. Such techniques may fail to accurately detect and separate "edge pixels," for example, that exist near the spectral boundaries of different classes [16,61] as well as pixels that exhibit high reflectance variability [62]. An additional constraint can be imposed when spectral signals from the land area represented by a pixel are influenced by signals from immediately surrounding pixels [63].…”

“…In fact, the CLASlite approach included similar image clean-up parameters in the form of artefact removal and pixel aggregation options (Table A2). The option of controlling the degree to which "salt-and-pepper" effects are removed as they pertain to a given study area can greatly influence the nature of the maps generated; by increasing edge densities and creating smaller blocks of any vegetation class [68], salt-and-pepper effects can contribute to the over-segmentation of an image [62], and thus reducing them through the use of filters [52,61,69] can reduce mis-registration errors [70]. However, among other parameter alterations, the CLASlite approach also critically allowed the user to define exact threshold levels for fractional photosynthetic vegetation and bare substrate values when determining forest cover (Table A2).…”