Eiki AIBA
Research on extraction of stand information by segmentation methods for high resolution remote sensing image
Atsushi RIKIMARU
After the Kyoto Protocol comes into effect, the grasp of forest resources information is a pressing need the inside and outside the country. But a forestry profitability of our country decreases from price hovering with the import of the outside material, now the management is a difficult situation. Because of that, the forest which is not cared increases. Moreover the forestryˇˇengineer decreasing and aging, grasp of the forest resource information by survey almost is impossible. Then, an analytical method using the remote sensing technology is variously proposed.
Recently, the high resolution image appears, the identification every of trees and shrubs it becomes possible by visual inspection. However, the analytical method of using the average luminance value which considers mixel in the former low resolution image is maintain, it could not to utilize sufficiently in the high resolution image which possesses pure pixel. And the high resolution image recognizes also the shadow and the like inside the trees and shrubs sensitively, because of that there is a fault which do not consider the trees and shrubs territory as one group. Therefore, examination of analysis method and the application which do not use average luminance value is needed.
In this research, segmentation of the covering trees area was examined making use of 3 methods as Zero Crossing Method, Region Growing Method (R. G. Method) and Watershed Method (W. Method). In addition we squeezed the intended the covering trees area by area circularity inside the segmented area. With R. G. Method and W. Method it could recognize the segmentation of the covering trees area with visual inspection. For the result of these 2 methods, the covering trees area was squeezed on the basis of form feature of the surveyed the covering trees area. As a result, removal of the non trees and shrubs area was done. We extracted independent single wooden crown by the method which combines form feature quantitative analysis to R. G. Method and W. Method. So, when it was shown agreement ratio of approx. 90% for the visual decipherment result.
Moreover, when trees position in the covering trees area which it extracts is compared with the visual decipherment result, it became the recognition rate of about 50% and 60% by R. G. Method and W. Method, respectively. The canopy was extracted both the 2 methods combining form feature quantitative analysis, which resulted in the agreement rate of about 90 %. Especially, in W. Method, the image of the resolution 1 m also demonstrated similar tendency with the 50 cm, it suggests the possibility which can extract the covering trees area from the IKONOS image that had generality than the aerophotograph. Furthermore the spatial scale and form feature parameter is able to setup semi-automatically, thus, it was suggested that it is connected to the improvement of efficiency of extraction of covering trees area.
Please use "previous botton" to return the previous page