An Evaluation of the Utility of Two Classifiers for Mapping Woody Vegetation Using Remote Sensing

Native vegetation is vulnerable to fragmentation and loss of condition due to various environmental and anthropogenic drivers. Regrowth can vary depending upon species composition within that vegetation community. As well as maintaining biodiversity native vegetation provides a range of ecosystem services. Up-to date and reliable information on the distribution of native vegetation is essential to support decisions that minimise loss of biodiversity and maximise functionality of ecosystems. Remotely sensed data is an ideal tool for this purpose. As such, this study aims to assess the accuracy of Maximum Likelihood Classification (MLC) and Spectral Angle Mapper (SAM) algorithms to distinguished woody vegetation from non-woody vegetation in Creswick, Victoria, Australia. The use of RapidEye imagery, image fusion (spectral information combined with textual information (ALOS-PALSAR)), and the use of a filter (majority filter) was also evaluated. The classification accuracy of MLC and SAM was used as the determining factor for identifying a suitable mapping method to distinguish woody native vegetation from non-native woody vegetation (particularly, Pinus spp., Pine and Eucalyptus spp., Blue Gum) in the forest. The results demonstrated that the use of MLC on RapidEye imagery enabled the Native, Pine and Blue Gum to be accurately mapped with an overall accuracy of 88%. Kappa statistics show that there was a significant difference between MLC and SAM algorithms independent of the image type input. However, when the same algorithm was applied on each image type, no significant difference was found. The use of fused images as well as a filter, did not improve the accuracy of the classification. Considering cost and time in registering and processing the images as well as the computational time of images filtering, the use of these methods does not provide benefit in this case study.

[1]  S. Franklin.  Remote Sensing for Sustainable Forest Management , 2001 .

[2]  J. Genderen,et al.  TESTING LAND-USE MAP ACCURACY , 1977 .

[3]  Russell G. Congalton,et al.  A practical look at the sources of confusion in error matrix generation , 1993 .

[4]  Fred A. Kruse,et al.  The Spectral Image Processing System (SIPS) - Interactive visualization and analysis of imaging spectrometer data , 1993 .

[5]  Helmi Zulhaidi Mohd Shafri,et al.  The Performance of Maximum Likelihood, Spectral Angle Mapper, Neural Network and Decision Tree Classifiers in Hyperspectral Image Analysis , 2007 .

[6]  P. R. Meneses,et al.  Spectral Correlation Mapper ( SCM ) : An Improvement on the Spectral Angle Mapper ( SAM ) , 2000 .

[7]  David L. Verbyla,et al.  Conservative bias in classification accuracy assessment due to pixel-by-pixel comparison of classified images with reference grids , 1995 .

[8]  James R. Anderson,et al.  A land use and land cover classification system for use with remote sensor data , 1976 .

[9]  J. Campbell.  Introduction to remote sensing , 1987 .

[10]  J. R. Landis,et al.  The measurement of observer agreement for categorical data. , 1977, Biometrics.

[11]  C. Tucker.  Red and photographic infrared linear combinations for monitoring vegetation , 1979 .

[12]  Russell G. Congalton,et al.  A review of assessing the accuracy of classifications of remotely sensed data , 1991 .

[13]  John A. Richards,et al.  Remote Sensing Digital Image Analysis: An Introduction , 1999 .

[14]  Calvin F. Bagley,et al.  Accuracy Assessment of the Discrete Classification of Remotely-Sensed Digital Data for Landcover Mapping. , 1995 .

[15]  Masha Maltz,et al.  A new multi-spectral feature level image fusion method for human interpretation , 2009 .