The availability of spatial-contextual information for classification of multispectral image data is discussed and the experimental evaluation of the important classifiers is presented. The contextual classifiers treated in this paper are clustered into three groups according to the way of utilizing spatial-contextual information. The first approach is preprocessing of image data such as moving average filters, median filters and edge preserving filters. The second approach is postprocessing of classified results accoring to some rules such as majority filters. The third approach combines the spectral and spatial-contextual information at the same time, and in practice compound decision methods, extended adaptive classifiers and probabilistic relaxations with forced convergence are evaluated. Especially, the last two are the modified classifiers which have been proposed by the authors. The experimental results demonstrate that the highest accuracy of classification is achieved by the classifiers belonging to the third approach and in order they are probabilistic relaxa-tion with forced convergence, extended adaptive classifiers and compound decision methods. Putting together with their expending memory spaces and CPU times, it is concluded that extended adaptive classifiers are best for practical usage.
For an atmospheric correction on satellite data, validation of single scattering approximation is examined at 0.4, 0.5, and 0.6μm, using the model atmosphere bounded by the Lambert surface. The LOWTRAN 5 model atmosphere is adopted in mid-latitude in summer. The results of single scattering approximation are compared with those of multiple scattering in homogeneous and inhomogeneous atmosheres. It was found from the computations that in a clear condition, the upward radiance in an inhomogeneous atmosphere using multiple scattering approximation method would be estimated by multiplying some empirical factors on that derived by single scattering approximation method. Furthermore, the upward radiance with single scattering approximation method considering upto four scattering modes in radiative transfer processes shows small differences in comparison with that of single scattering method considering all modes until convergence. Thus the upward radiance would simply be evaluated using the single scattering approximation method upto four modes. Whereas in a hazy condition, the above factors exhibit a strong dependence on observational nadir angles.They are affected by the surface albedo as well. Furthermore, an optical characteristics of aerosols is one of the essential factors to perform atmospheric correction. Therefore single scattering approximation method would be quite crude.
The purpose of this report is to examine effects of band combination and color sequence on visual interpretation and to select some suitable combinations and sequences. TM data used is WRS 111-36 (Setouchi) scene received at NASDA EOC on May 8, 1984. Color composites used are 210 kinds of 240 × 240mm (1 : 1, 000, 000) images which are made of three primary colors and three bands from TM seven bands in the previous report. Aptitude for visual interpretation of the images is investigated as follows. Because of relatively low definition of the thermal band (band 6), interpretation is made on the color composites excluding and including band 6 separately. On the former, aptitude for interpretation is examined on eight kinds of large scale objects and on twelve kinds of small scale objects separately. On the latter, the aptitude is examined on eight kinds of large scale objects and four kinds of relatively large scale objects, i. e. twelve kinds of objects. At the first step, the aptitude for the each object is examined with a method of five grades including the first grade always on one group of 20 kinds of images excluding band 6 and another group of 15 kinds including band 6 in one block of 35 kinds to each color sequence separately. Then the mean of scores of each band combination to 6 kinds of all color sequences for the each object is calculated and the best band combinations are selected. At the second step, the best scored images are picked up and aptitude for the each object is examined on them elaborately. The each of them is ranked in order of aptitude for the each object. On the basis of these results, the best band combination and color sequence is selected. Finally frequency of each color sequence in the best images is counted from the results of the second step and the best for the all objects is selected. In case of band combinations excluding band 6, the best is (257) to the large scale objects or (345) to the small scale objects. In case of band combinations including band 6, good combinations are (376) and (356). The best color sequence is (BGR) or (BRG) in all of the cases.
The authors investigated the distribution of the surface temperature of Unzen volcano and development of cold air lakes at Unzen hot-spring resort and in its vicinity, west of Mt.Fugendake of Shimabara peninsula, Nagasaki prefecture by airborne MSS thermal IR data obtained early morning on March 16, 1982. Meteorological data obtained at Kinugasayama (846 m) and Unzendake (677m) meteorogical station showed temperrature inversion of the study area. Surface temperature of the summits and convex upward mountainsides was higher than that of the valley and basin floors and concave upward mountainsides. The result was confirmed by temperature inversion indicated by the meteorological data. Cold air was produced mainly not only on the basin foor but also on the surface of the gentle concave upward and straight slopes of the mountain. Cold surfaces were also recognized on the valley plains and gentle slopes upstrem from barriers such as hills and narrow valley walls. Surface temperature of the bare soil or thin vagetated surface was colder than that of vegetated or forested surfaces. Therefore, the forest makes no contribution to the production of the cold air of the mountain.
This report describes the results of oil-slick observation experiment using an X-band SLAR (Side-looking Airborne Radar) developed in our research program from 1984 to 1986. The purposes of the experiment are to verify the ability of oil slick detection by the SLAR and to clarify the properties of Δσ°(the difference of backscattering cross-sections between clean and slick ocean surfaces) which is the most important parameter for the SLAR observation of ocean oil slicks. In the experiment, two kinds of oil slick were observed : artificial and natural oil slicks. From the artificial slick observation, it was found that even a small slick made with only 18-liter oleyl-alcohol can be clearly detected. The following properties of Δσ°were also clarified : (1) Δσ° shows clear azimuth-angle dependence for the incident angle range from 30°to 80°, (2) incident angle dependence of Δσ° is not large for the same range, (3) Δσ° appears to decrease gradually at incidence angles larger then 80°. From the natural slick observation, it was found that very small (about 50m×50m) slicks can be detected and that Δσ°was approximately the same as the value for the artifical slick. These results indicate that the SLAR has a very high sensitivity in oil slick detection over a wide incident angle range. Based on the results of our experiment and those of other investigators, SLAR observation parameters suitable for oil slick detection are also discussed.
This report reviews the present state of the research and study on remote sensing in the environmental engneering field of the Architectural Institute of Japan. Members of the working group individually introduce their activites and give their impressions of remote sensing. We also included a list of the documents at the end, which will help you to know the actual conditions and keywords of our research. These are the latest products of our working group for a literature survey on the research related to remote sensing in the Architectural Institute of Japen.