Damaged area by pine beetle have been gradually expanding in Japan. The Forestry Agency of Japan have been promoting various undertakings against the damage such as scattering chemicals and investigation of damaged area and volume of damaged pines.
A large area of about 110 km by 110 km including a southern part of Ibaragi Prefecture and a northern part of Chiba Prefecture was selected as a study area. Landsat MSS images of the study area were generated from two temporal Landsat data taken on November 26, 1972 and November 11, 1980 and those images were geometrically corrected to be exactly overlapped. Pine area (A) was extracted by a supervised maximum likelihood classification using multispectral data of the overlapped images of the study area. On the other hand, the area (B) where the value of Band 5 increased and that of Band 6 decreased was extracted using the difference images between the two overlapped images. Finally the damaged area by the pine beetle was extracted by taking logical AND operation between the above area A and B. The damaged area was classified into three classes as heavy, middle and slight damage according to the extent of the increase of the value of Band 5 and the decrease of that of Band 6.
Volume of damaged pines surveyed on the ground by each local government office, about 120 local governments being included in the study area, was gathered. It was found that there was a high correlation, correlation coefficient of 0.78, between the damaged volume surveyed on the ground and the damaged area of each local governmental area extracted from Landsat MSS data. Landsat MSS data are considered to be useful for extracting the damaged area by the pine beetle.

View full abstract

A newly developed image sharpening method is presented for the enhancement of LANDSAT MSS data. Since the lineaments as terrain features are essential in the landform analysis, the images with linear features enhanced are expected to assist a geologist in interpretation.
A popular image sharpening operation is to subtract a Laplacian (the second derivative) from its original image, which is equivalent to the unsharp masking well-known in photography. However, this method often produces heavy noise on LANDSAT images especially in false-color representation. It is due to the characteristics of the discrete Laplacian operators ; i. e. isolated points are much more enhanced than linear features.
To overcome this undesirable effect, we have developed a fairly new concept of filtering a pair of Laplacian values by an output of a sophisticatedline detector before subtracting it from its input. This approach can sharpen linear features selectively but suppress noise considerably. In this study the VanderBrug operator is adopted as a line detector suitable to Landsat linear features. The effectiveness of our new method is verified in the lineament extraction from LANDSAT-2 MSS images at the southern part of Kyushu.

View full abstract