1995 年 15 巻 5 号 p. 452-462
Although it will bring more to use both channel data of NOAA-APT as multi-spectral data in many fields of its applications, NOAA-APT cannot be treated as ordinary two channel multi-spectral band data. That is, the CH.A and the CH.B data of NOAA-APT cannot be laid on top of each others, because each of two channels data streams of NOAA-APT are created from consecutive AVHRR channel data.
A complex of two channels of NOAA-APT becomes a line interlacing, and shows usually a type of alternating line texture. Fog area shows especially evident line texture.
The author inspected seventy-three scenes which were received for three months from May 1994 till July 1994 by his receiving system. Sixty-three meso-scale fog area data are prepared for investigation out of forty-five scenes.
In order to characterize fog area using textural features, he has obtained co-occurrence probabilities, and made a survey of its distribution patterns.
As results, co-occurrence probabilities of cloud-free sea and fog area are clearly distributed separately. In cloud-free sea, because of small fluctuations in albedo which derived from CH.A and in Tm, which derived from CH.B, the distribution pattern shows a circle. In the fog area, fluctuations in albedo are large, the distribution pattern shows ellipse in albedo direction.
Next, in order to identify these co-occurrence probability distribution patterns, fifteen textural features included textural features proposed by Haralick et. at., are calculated.
As results, concerning textural features based on co-occurrence probabilities, ASM discriminates fog and cloud-free sea area at 89.5%, and ENTROPY at 84.2%, respectively.
Concerning textural features based on sum and difference co-occurrence probabilities, SUM AVERAGE discriminates fog and cloud-free sea area at 96.7%, DIFFERENCE AVERAGE at 93.0%, SUM ENTROPY at 91.2%, and DIFFERENCE ENTROPY at 91.2% respectively. CONTRAST textural feature shows time-variant