4.2 Difficult Detection
Detection and dentification was very difficult for the kakegawa CR because of nearby cliffs and parking lots, for the Hamaoka CR because of a number of objects surrounding the athletics field, and for the Mikura CR because it is located with a mountainous region.

The Kakegawa CR was installed in a small empty lot between the parking lot of the Kakegawa Civic Hospital and a cliff. Kakegawa Civic Hospital was built on land that was reclaimed from a hilly region southeast of Kakegawa City. With a large parking lot and an artificial cliff created from the reclamation on the east side, it forms a very prominent scene in the area. However, it is not a feature that stands out in a SAR image (Fig. 3). First, we searched for the hospital using as landmarks the dark Tomei Highway and Japan Railways line running through the bright city, finding the shaped inter change on the highway, and following that and the cliff, which appears as a somewhat bright line, on the east side of the hospital. The hospital is a large building with walls facing east, so we expected a bihedral corner reflection from the ground and side of the building. But the area appears to be blocked by some low-lying affiliated facilities just above the cliffs, and the building does not stand out in the image. The single strong point target (marked H in small subimage (1998/12) in Fig. 4(a) associated with the hospital is conjectured to come from the air conditioning unit on the rooftop. There is a large Parking lot on the north and west sides of the hospital, but area appears bright or dark depending on whether there are cars parked there or not. In the image taken in may 1998 (Fig.3), the observation data coincided with a holiday, so there were almost no cars, and the parking lot appears dark. The artificial natural cliffs on the east side of the hospital appear on the whole as a somewhat bright line, with occasion bright point targets.


Fig 3. Kakegawa CR and surrounding region

Fig. 4(a) shows a series of 120 x 120 pixel subimages centered approximately on the Kakegawa CR. The appearance of the parking lot and the bright points along the cliffs vary with each image; the area around the after they have been despeckled with the Enhanced SFP filter.


Fig 4. Area around Kakegawa CR.

With speckle reduction, it became significantly easier to locate the CR. Even then, point targets in the area immediately surrounding the CR vary with each observation, so the CR could not be identified and Confirmed just from its relation to the surrounding features. So instead, we checked its position relative to a prominent point target that was further away. The positions of the CR and other prominent targets in the surrounding area relative to the point H on the hospital are summarized in Table 2. The positions of the CR candidate relative to the hospital target on December 1997, May 1998, and July 1998, and in turn the position of other surrounding targets, are practically constant. Based on this result we confirmed that the candidate point target as the Kakegawa CR.

Table 2. Positions of Kakegawa CR and nearby objects relative to point H at Kakegawa Civic Hospital [refer to a subimage (1997/12) in Fig.4(a) for locations of points CR, A, B, and C] )

	CR	A	B	C
1997/09	NA	(-46, -18)	(-9, -93)	(52.-56
1997/11	NA	(-46, -20)	(-9, -95)	(52.-58)
1997/12	(2,-28)	(-45, -19)	(-10, -94)	(52.-58)
1997/02	NA	(-46, -19)	(-9, -93)	(52.-58)
1997/03	NA	(-46, -18)	(-9, -93)	(52.-56)
1997/05	(2,-28)	(-45, -20)	(-10, -95)	(52.-59)
1997/07	(2,-28)	(-45, -19)	(-10, -94)	(52.-57)

Coordinates are pixel coordinates(Ra,Az): Right Ra>0, Left Ra<0, Down Az>0, Up Az<0

In the case of the Hamaoka CR, shown in Fig, 5 the athletics field is relatively small, surrounded by tall metal framed wire fences, and contains confounding movable equipment such as goals. The field itself could be easily identified from its relation to other nearby objects, since the national highway, drawing a gentle curve nearby, is very visible, and the nearby river also appears clearly in the image. The CR candidate was determined from its relation to surrounding objects, although we have not been able to confirm it as yet because we still have only observation containing the CR.


Fig 5. Hamaoka CR

The Mikura CR is located in the corner of a public space in front the public hall next to the shrine in a small hamlet halfway up the side of a mountain. This area, around the southern tip of the Chubu Mountain Range, is located within layers upon layers of mountain ranges that look very similar to each other (see Figure 6), which made the identification of specific regions very difficult. By comparing the vertically compressed image with a topographic map, we very roughly extracted the area shown in Figure 6 as the target area. In a mountainous area like this, creeks, and roads are too thin to act as landmarks. Initially, we tried to locate the target area from the pattern of cluster of point target that may correspond to the scattered hamlets in the region. However, we found that such cluster did not match the location of the hamlets in the map very well. Therefore, we first located the 548 meter Mt. Sanmaru from the shape of the mountains, and then Determined the location of the shrine from the pattern made by scattered hamlets and the shape of the small ridges on the side of mountain. We wee then finally able to detect the CR candidate as a newly appearing point in the area.

Images of the area from before and after the installation of the CR are shown in Fig. 7. We have determined a CR candidate but not yet positively identified the CR, since it was installed this July, and we still have only the July SAR data. Fig. 7 (a) is the single -look image of the area around the CR, and (b) is the image after speckle reduction.


Fig 6. Mikura and surrounding region (800x1000)


Fig 7. Mikura CR

5. Summary
Japan is mostly mountainous, and the flat area have very little open space, se we are forced to place small corner reflectors in areas surrounded by confounding objects. We have described the methods used to visually detect and identify L-band CRs from SAR images in such circumstances.

6. References

• G.T. Ruck, D.E. Barrick, W.D. Stuar, and C.K. Krichbaum, "Radar Cross Section Handbook", Section 8.22 : Radar Reflectors, Plenum Press, NY-London, 1970.
• T. Sakuari-Amano, and J. Iisaka, "Speckle Reduction for Automatic Detection f Small Features, manuscript in preparation.

Page 2 of 2
| Previous |