prasinos' work memo



ERDAS LAN ファイルの形式


さて、GIS にラスタデータを突っ込むときに、フォーマット変換ということは起こると思うのであるが、自作派にうれしい簡単なのがある。ERDAS LAN 形式である。

で、だからどうしたのかというと、左上角の座標指定という方式はどこかで見たような気もするが、とっても普遍的につかわれているということがわかる。 続きを読む


So how about Meteosat 8?

Let's see also Meteosat 8 results. 続きを読む

PC1 shows cloud excluding effect of land

I tried PCA for GOES-EAST image of 2005-04-21T17:45Z, in addition to yesterday's Meteosat 8 (2005-04-20T12:00Z). I used all available bands: IR2, IR4, VIS and WV.

PCA trial

Edgecut program is improved to fill zero out of the disk. This will eliminate nasty title bars at top and bottom of the image.
Then try PCA for new result.

Ignore Zero, Write eigenmatrix and eigenvalues, stretch to unsigned char.

One nasty thing of Erdas Imagine 8.3 is that eigensystem log is written in a text file with LF-separated text file. Are they really using Windows?

The result looks good. Even in PC1 and PC2, Sahara and low cumuli are distinguishable.

When I pick two bands from original (befor PCA) image, it is difficult to distinguish Sahara and low cumuli. Band pairs I tired are: 2-4 (lowest correlation), 3-5 (highest correlation with PC1 and PC2), 3-11 (2nd lowest correlation).

Image Size

I measured size of full disk images of Meteosat 8.

At least for 1200Z Apr 0520, the globe is a ellipse whose vertical and horizontal axis lengths are 450.47 and 450.37 pixels respectively. Center of the earth is at (450.47, 450.47) in image coordinate system in which center of the upperleft-most pixel is (0, 0).

These measurements should have 0.1 pixel precision.