Home Page       Sat Dec 6 19:31:03 JST 1997

Star detection and noises

As [2] says, matching process in the PIXY system consists of two steps, star detection from an image and then matching two sets of points, that of detected stars and star data in a catalog. Matching itself succeeds for many images as mentioned before. However the process of star detection is still preliminary and quite incomplete.

I show you an example. This image is the plot of detected stars from the image of Comet Hyakutake (hy960327.jpg), a successful example in the previous section.

As you see in this plotted image, the current process of star detection has these faults.

• It cannot detect faint stars.
• It cannot separate near double stars.
• It cannot detect stars on a diffused object such as coma or tail of a comet.
• Faint part of a diffused object appears as minute noises and recognized as many faint stars.
• Trailed stars cannot be recognized as those of one star.

Why it cannot detect faint stars is the determination of the threshold is still preliminary at star detection. Why it cannot separate near double stars and why it cannot detect stars on a diffused object is the current system does not consider the point spread function of each object. Furthermore, the field flattening [3] at star detection assumes that the background of the image should be originally flat. Therefore it cannot flatten such images properly as that taken in a morning or evening glow or that influenced strongly by light pollution which causes the background incline. Sometimes many noises appear after wrong flattening for such images.

Among the sample images listed up in Section 2, the two: hb970117.jpg and hb970309b.jpg were failed to be matched. The reason in both cases is because many noises appeared in the flattening process and they were all recognized as faint stars. These images were taken in a morning or evening glow and the background level is very different depending on the position on the image. That is, the flattening method mentioned in [3] cannot be applied to them. So the flattening caused an adverse effect in both cases. However, there is a successful case of an image taken in a morning glow (hb970103.jpg). I do not know why matching sometimes succeeded and sometimes failed for such images.