Image fusion methods for land cover classification and its potential for slope failure detection on a mountainous terrain

Land cover classification has become routine works done based on satellite images. However, the resolution size of the available satellite images as well as the degree of expected accuracy classification results have been the major concern for many researchers. In many cases, classification results accuracy positively correlated with image resolution; the higher the image resolution, the higher the degree of classification result accuracy. For the purpose of land cover classification of Hitotsuse area which is located on a mountainous terrain in Miyazaki Prefecture, Kyushu, Japan, SPOT5/HRG and ALOS satellite images have been utilized. The research methodology have been done by applying different method of image fusion such as IHS, HPF, Wavelet, Subtractive and Multiplicative. The high resolution image for fusing the multispectral images are SPOT panchromatic and ALOS PRISM image with resolution of 2.5 m. Supervised and unsupervised classification was then conducted on the fused images, as well as the results assessment. The results show that even though there is spectral distortion on the fused images, the spatial information is much improved. The classification result accuracy from the fused images also increases about 3.89% and 2.79% for SPOT5 and ALOS respectively. The slope failure area detection is hindered by the similar spectral values of bare-land, slope failure and some vegetated area. Hence, slope failure detection task could not give satisfactory results. However, if it is combined with further scenario, the fused images show its potential for detecting slope failure in the study area.