Abstract
Satellite remote sensing images allow us to discriminate and extract detailed topographic features of the earth. SPOT HRV multispectral images remarkably give us higher resolution for lineament extraction.
This paper is aimed to clarify the optimum local operators for image analysis using extracted lineaments from HRV multispectral data. We made an experiment in relation to various types of digital filter such as median (md3, md 5), monodirectional high frequency component enhancement (tt3, tt5, ty3, ty5), slantdirectional high frequency component enhancement (nm 3, nm 5), bidirectional high frequency compo-nent enhancement (ty3, ty5), and omnidirectional high frequency component enhancement (om3, om5) applying to enhance the selected two images of mountainous area in Tohoku district. We generated 12 images in each area to extract the lineaments that show the direction of subsurface fractures. Lineaments were analyzed in relation to detect six specific major faults.
As the results, 3×3 operator of omnidirectional high frequency component enhancement (om3) images scored the maximum matched ratio for the major faults detection. The second best scores were the 3×3 operator of bidirectional image (ty3). Different filter sizes of mask operations show that the bigger size operators were not superior properly. Monodirectional filter images (tt3, tt5, yk3, yk5) were scored the lower matched ratio. Those results suggest that omnidirectional and bidirectional images are quite adequate to utilize for the fault detection. 3×3 mask operation is enough to enhance the images. Median filters are not fit to extract the lineaments.
