Influence of using texture information in remote sensed data on the accuracy of forest type classification at different levels of spatial resolution

Abstract

We evaluated the influence of texture information from remote sensed data on the accuracy of forest type classification at different spatial resolutions. We used 4-m spatial resolution imagery to create five different sets of imagery with lower spatial resolutions down to 30 m. We classified forest type using spectral information alone, texture information alone, and spectral and texture information combined at each spatial resolution, and compared the classification accuracy at each resolution. The classification and regression tree method was used for classification. The accuracy of all three tests decreased slightly with lower spatial resolution. The accuracy with the combined data was generally higher than with either the spectral or texture information alone. At most resolutions, the lowest accuracy was with texture information alone. However, there was no clear difference in accuracy between the combined data and spectral data alone at 25-and 30-m spatial resolution. These results indicate that adding texture information to spatial information improves the accuracy of forest type classification from very high resolution (4-m spatial resolution) to medium resolution imagery (20-m spatial resolution), but this accuracy improvement does not appear to hold for relatively coarse resolution imagery (25-to 30-m spatial resolution).

Estimation of Forest Structural Diversity Using the Spectral and Textural Information Derived from SPOT-5 Satellite Images

“…For instance, Ota et al [65] found that the addition of textural information improved the discrimination of hinoki cypress and cool-temperate mixed forest whereas no improvement for Japanese cedar and a clear cut area was observed. Franklin [66] also documented that the addition of texture generally improved the classification accuracy of hardwood stands, more so than for softwood stands.…”

A global reference database from very high resolution commercial satellite data and methodology for application to Landsat derived 30 m continuous field tree cover data

“…In almost every case the number of derivative layers included in the image stacks exceeded the number of original VNIR bands. The utility of added dimensionality through the inclusion of layers derived from the standard VNIR bands is generally supported by common use with decision tree classifiers (Goetz, Wright, Smith, Zinecker, & Schaub, 2003;Kayitakire, Hamel, & Defourny, 2006;Lu & Weng, 2007;Ota, Mizoue, & Yoshida, 2011;Thenkabail, Enclona, Ashton, Legg, & De Dieu, 2004). It could be expected the usefulness of additional attribute layers is dependent on both land cover and data characteristics, and research has shown that gain in accuracy from adding feature layers may reach a point of diminishing return, particularly where there is a high level of co-variation among the added features (Lu & Weng, 2007;Pal & Mather, 2003).…”

Influence of Data Source and Training Size on Impervious Surface Areas Classification Using VHR Satellite and Aerial Imagery Through an Object-Based Approach

“…Moreover, a combination of different window sizes, such as those used for the feature set TextureAll, turned out to have an improvement on the accuracy results (GE1 experiment). Hence, similar to [30]- [32], the local texture indices were the only feature that led to a significant increase in the accuracy and, therefore, the feature set TextureAll was applied to both VHRsatellite orthoimages in order to obtain the corresponding ISA classification.…”