Comparison of artificial neural network and support vector machine methods for urban land use/cover classifications from remote sensing images A Case Study of Guangzhou, South China

Accurate land use/cover (LUC) classifications from satellite imagery are very important for eco-environment monitoring, land use planning and climatic change detection. Traditional statistical classifiers such as minimum distance (MD) have been used to extract LUC classifications in urban areas, but these classifiers rely on assumptions that may limit their utilities for many datasets. On the contrary, artificial neural network (ANN) and support vector machine (SVM) provide nonlinear and accurate ways to classify LUC from remote sensing images without having to rely on statistical assumptions. This article compared the results and accuracies of the ANN and SMV classifiers with the statistical MD classifier as a reference in extracting urban LUC from ETM+ images in Guangzhou, China. Results show that the overall accuracies of urban LUC classifications are approximately 96.03%, 94.71% and 77.79%, and the kappa coefficients reach 0.94, 0.92 and 0.67 for the ANN, SVM and MD classifiers, respectively, indicating that the ANN and SVM classifiers have greatly better accuracies than the traditional MD algorithm. It is concluded that while the ANN performs slightly better than the SVM, both ANN and SVM are effective algorithms in urban LUC extraction from ETM+ images because of their high accuracy and good performance.

[1]  Thomas Blaschke,et al.  Land cover change assessment using decision trees, support vector machines and maximum likelihood classification algorithms , 2010, Int. J. Appl. Earth Obs. Geoinformation.

[2]  Lorenzo Bruzzone,et al.  A technique for the selection of kernel-function parameters in RBF neural networks for classification of remote-sensing images , 1999, IEEE Trans. Geosci. Remote. Sens..

[3]  S. Khorram,et al.  Remotely Sensed Change Detection Based on Artificial Neural Networks , 1999 .

[4]  E. Kalnay,et al.  Impact of urbanization and land-use change on climate , 2003, Nature.

[5]  I. Kanellopoulos,et al.  Strategies and best practice for neural network image classification , 1997 .

[6]  Helmi Zulhaidi Mohd Shafri,et al.  A comparison of support vector machine and decision tree classifications using satellite data of Langkawi Island , 2009 .

[7]  Liu Xiang-nan Xiu Li-na,et al.  Current Status and Future Direction of the Study on Artificial Neural Network Classification Processing in Remote Sensing , 2011 .

[8]  P. Swain,et al.  Neural Network Approaches Versus Statistical Methods In Classification Of Multisource Remote Sensing Data , 1990 .

[9]  Richard Tay,et al.  Support vector machines for urban growth modeling , 2010, GeoInformatica.

[10]  L. S. Davis,et al.  An assessment of support vector machines for land cover classi(cid:142) cation , 2002 .

[11]  Genong Yu,et al.  Artificial Neural Networks and Remote Sensing , 2009 .

[12]  Paul M. Mather,et al.  Support vector machines for classification in remote sensing , 2005 .

[13]  Zhang Rui,et al.  State of the Art on Remotely Sensed Data Classification Based on Support Vector Machines , 2009 .

[14]  Lorenzo Bruzzone,et al.  An experimental comparison of neural and statistical non-parametric algorithms for supervised classification of remote-sensing images , 1996, Pattern Recognit. Lett..

[15]  Johannes R. Sveinsson,et al.  Feature extraction for multisource data classification with artificial neural networks , 1997 .

[16]  Jie Zhang,et al.  Accuracy assessments and uncertainty analysis of spatially explicit modeling for land use/cover change and urbanization: A case in Beijing metropolitan area , 2010 .