Development of satellite sensor systems capable of producing high spatial resolution digital images has led to the emergence of various alternative methods beside the more established per-pixel multispectral classifications. One alternative method is object-based image analysis (OBIA). At the beginning of its development, OBIA was primarily used for high spatial resolution images. However, the OBIA is now widely applied to images with medium- and even low-spatial resolutions. This study aimed to compare the effects of the OBIA and per-pixel classifications using using Landsat-8 OLI medium-spatial resolution image. Since the per-pixel classification relies solely on spectral aspects on various spectral bands, while the OBIA classification made use of spatial aspects as the main criteria, this study also made use of two land-cover/land-use classification schemes as references, i.e. spectral-oriented and spatial-oriented classification systems. The spectral-oriented classification scheme specifies categories from spectral perspective, i.e. pixel values in n-dimensional feature space; while the spatial-oriented one specifies categories with respect to their spatial characteristics. By using Kulon Progo region as a test area, the results showed that the OBIA classification was able to provide higher accuracy than that of per-pixel classification, both by referring to the spectral and spatial dimension classification schemes. The increase in accuracy provided by the OBIA classification proved to be greater when applied with a spatial dimension classification scheme, which was more than 10%, as compared to the improvement obtained by the spectral dimension classification scheme, i.e. 7%. This study also recommends the need for comparison studies using higher-spatial resolution imagery.
Multispectral classification is one of the main methods in the analysis and processing of digital remotely sensed imagery, which until now is still widely used to generate land-cover/ land-use information. Technically, pixel-based classification methods rely on conventional approaches, as compared to GeoBIA, and it can be implemented using either supervised or unsupervised methods. The classification methods are supported by the rapid development of various image processing software, which provide a wide variety of algorithm options, so that the classification process can be carried out easily. Although relatively simple, an appropriate selection of multispectral classification algorithm may provide highly accurate land-cover maps. However, the highly accurate land-cover/land-use maps may also be influenced by image types and classification schemes that are used in the study. This study aimed to compare the results of the multispectral classification using maximum likelihood algorithm, for generating land-cover maps based on Landsat-8 OLI images (30 meters) and Pleiades imagery (2 meters). The classification referred to two different classification schemes relating to spectral and spatial dimensions. The results showed that the multispectral classification with spectral-related classification scheme applied to Pleiades imagery gave higher overall accuracy as compared to that of Landsat-8 OLI. It was also found that the highest overall accuracy achieved in this study was 81.7%, obtained using Pleiades imagery and referring to spectral dimension classification scheme. On the other hand, the lowest overall accuracy was obtained by the same imagery applied using spatial-related dimension. The relatively similar values of low overall accuracy for spatial-related dimension was also gained by Landsat-8 OLI imagery, proving that multispectral classification does not work well for spatial-related land cover classification scheme.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.