単独シーンの高解像度 SAR 画像を用いた低層建物の高さ推定 Height

(一社)日本リモートセンシング学会 第57回学術講演会論文集 2014年11月
U3
|ÿ=SQ$ŠĴk¹ SAR ćk4ąa­ºþ$ŠÐ¤
Height estimation for low-rise buildings from a single high-resolution SAR image
NL6R67Q 1R ®°ÙŘ 1
Wen Liu and Fumio Yamazaki
Abstract : Recently, the detection of building heights as a part of inventory information draws significant attention. Owing to the
development of satellite sensors, it is possible to obtain building heights from high-resolution satellite images. In this study, the
heights of low-rise buildings were estimated from a single TerraSAR-X (TSX) intensity image. According to the geometric
characteristics of SAR, a building height is proportional to the length of its layover. An automatic extraction of building layover
was proposed for the TSX image of San Francisco, California, USA. Comparing with Lidar data, the obtained results showed a
reasonable level of accuracy.
Keywords : TerraSAR-X, building outline, layover, building height, backscattering intensity
1. %+"
Ō´ñŅþ$؁ř%ţÀç/NKSE@Q
=Q9sŎ"
1ōį!ĺş$ 1 1ºþ
Š%ţŌ´ĉ¬$ĸf-ý¥ß"
1ºþĮ¥
$ÊÓ"ãw!Îò1TÚţŃ·@Q;Éī
$ĉ¬".0ţŠĴk¹m£ćk…Æő‚PSB
(SAR)ćkÁ/21."!$+ţĬ
Þćk4ąºþŠÐ¤ƒĥ"!¯°
/ 1)%ţ0.6m Ĵk¹$ QuickBird ćk 3m Ĵk¹
$ TerraSAR-X (TSX)ćk/ċıºþŠ4Ð
¤)ţ2 ê$ SAR ćk4ąµûqĂ"
.1Š$Ф,Ī32(Bolter and Leberl 2))
åĐĕ%ţ1 =SQ$ŠĴk¹ TSX ćk/º
þ$h2ł*ę4ĦxĊ"îrţºþŠ4
Ф1Èø4Ñì1)ţTSX ćkŃß
äÔ¾2ĨĖóPS<S(Lidar)DSA4ą
ţФĠë$Ĝ¹4ĸf1
2. §ļ”–ćkDSA
åĐĕ%ţFig. 1(a)"đě‘;QHMQ=?
:´$TŊ4§ļ”–1§ļ”–o$ºþ%ţ
Y"a­(30m ^V)$‹ïÛĶ-±™12011
· 10 â 9 Ü"§ļ”–4Ô¾ WorldView-2
(WV-2)ćk4 Fig. 1(b)"đa­$ºþŸ
+ţm£ćk"
1ºþ$ܾďţ¾4ą
ŠÐ¤%Žś1
ţåĐĕ% 2011 · 10 â 13 Ü"Ô¾
2 TSX ćk4ąºþŠ$Ф4ĪTSX
ćk%ŠsĴĥ$ HighSpot (HS)KSFţVœľň
/ HH jùÔ¾2ćkXÂ"
1IJü
ij¹% 39.88°0ţIJüڇŇĪڇ"
1
Ĵk¹%,"ĝ 1m 1Ñe2DSA%
Single-look Slant-range Complex (SSC) D S A 1
ô`Š zĩ £ £Ŕ±£ĐĕĒ ºĚRŌ´Ē£¨Ö
( Ǔ” 1263-8522 zĩčzĩ´Ĕöy»ĄĆ1-33)
(ņġl Tel: 043-290-3528, E-mail: [email protected] )
1+ţ5>J?PQ>ڇ$’Ģ! $vqĂ
4ĪvqĂ¿$”Ë¾2 TSX ¼¹ć
k4 Fig. 2(a)"đm£ćk/cÆ2ºþ
ŀŋDSA4 TSX ćk"ō#1ţ¿Ú×ZgØ
$Šh2ł*ęŀŋ$@Q;ڇ"1h
2ł*$ŐºþŠõd1+ţ$ę
4îr1".0ºþŠÐ¤1
(a)
(b)
Fig. 1 The study area in the part of San Francisco, USA. (a); the
WorldView-2 image of the target area taken in Oct. 9, 2011 (b)
(a)
(b)
Fig. 2 TerraSAR-X image of the target area taken in Oct. 13,
2011, with the GIS data of building footprints (a); a close-up
of the white frame in (a) after emphasizing the high
backscattering areas that are considered as layover (b).
3. ºþŠ$ФÈø
h2ł*ę4Ìr1+"ţćknb$¿Ú
×ZgØ$¶•i1-10.5dB ^U$G8@O4
h2ł*ęİ! 1 "Ò1$]$G
8@O4 0 1Fig. 2(a)$TŊ4Í ţ0-1
̶ 95 ̶
ćk"Ò4(b)"đ2.0ţ º
þ$h2ł*ę%¦Ý"ĵü1ţªº
þ$h2ł*$Ő%ċıü1Žś
1ţºþ$ŀŋDSA4ąţh2ł*
ę$Ő4Ħx"ĵü1Èø4Ñì1
)ţºþ$Š 1.5m 2m $ß"
1h
2ł*$Ő4ćk$Ô¾æ_/Ęr1ºþ
ŀŋ4@Q;ڇ'ţÁ/2Ős4ēx1
ēx 2 $ŀŋ$Œ4täCQIPSE
cÆ1"ţºþŠ 0.1m Č½$h2ł
*Ő4Ęrţ1 s$ēxĽŚ1täC
QIPSE4@Q;ڇ'‡ţÁ/2ēx
ĽŚs4ēx1ēx"ţCQIPSE
"
1h2ł*ę$}ãā4ĵĘ1CQI
PSEh2ł*ꏞ2&ţēx4ğ[ţ
áğēxĽŚ/ºþŠ4Ф1
åĐĕ%ţCQIPSEo$h2ł*ęā
40%^V"!2&ţîĞ4ğ[1)ţºþ
à 2m ^U1+ţtäCQIPSEo"h
2ł*ę¢“!2&ţēx4ő¡!
ĹФ4ª1+ţºþŀŋo"1G8@
O%(h2ł*ę/ŕ~$Úø"
.1h2ł*ę$Ìrø4 Fig. 3(a)"đ
Fig. 3(b)"đ."ţŠĈ!1 2 í$ºþ
ŗÏ1™…ţ†ß"CQIPSE$ēx4
ĪţaºþŠºþ$h2ł*ę4îĞ
ŠÐ¤21$+ţºþ%@Q;/
$ĽŚ$Ş"îĞ4Ī@Q;"ŃÚ$ºþ
/l"Ф4ĪţîĞ2ę4J?81
".0ţ^œîĞ2!!1
4. Ð ¤Ġëîķ
Fig. 2(a)"đ TSX ćk4ąţºþŠ4Ð
¤Ġë4 Fig. 4(a)"đ§ļ”–o"1 116
í$ºþ$ţ86 í$ŠÐ¤2Ф
!ºþ%ţY"@Q;ڇ"u$ºþŗ
ÏtäCQIPSE4cÆ!Ф
Ġë$Ĝ¹4îķ1+ţLidar DSAÁ/2
D>AO”ĭKDO(DSM) DEM $²4ºþŠ
÷+ţFig. 4(b)"đ$WĤ".1
ºþŠ4õĿ1ţð#Tħ1
/"ţ@Q;ڇšè{Ţ"
1 Lidar DS
A$ºþŠ$¶•i4ĘrţФĠëõĿ
,$4 Fig. 5(a)"đФĠë$¶•Ĺ²%
0.4m ţRMS Ĺ²% 2.5m 1àń$ºþ$Ŗ
Š1 3m .0ª+ţåÈø4ąºþ
ŖØ$Ф"ãw1"đĹ²á, 2 í$ºþ$pĎ4 Fig. 5(b)"đºþ a
%«U"¸ˆé1+ţ2".1h2ł*ę
ºþİ!2ţŠÐ¤2)ţºþ
(a)
(b)
Fig.3 Simulation of the template for layover searching (a), and
the definition of boundaries for two connected buildings with
different heights (b).
(a)
(b)
Fig.4 Detected heights by the proposed method (a); the
reference heights obtained from the Lidar data (b).
(a)
(b)
Fig.5 Comparison of the detected heights from the TSX
image and the reference one from the Lidar data (a); two
ground photos of the buildings that have the largest errors,
cited from Google Earth (b).
b %ţ@Q;ڇ"ŗÏ1›$€©h2ł*
ęĹÌr2+ţŠÐ¤2
5. )+
åĐĕ%ţě‘;QHMQ=?:´$a­ºþ
4Ô¾ 1 =SQ$ TSX ćk/ţºþŀŋ4
ą 3m ^V$Ĺ²ºþŠ4ФФ
Ġë$Ĝ¹%§ļ”–$‰Łăš"¾ŝ21
+ţ\¿%ućk'$ʼnąÈø$Ռ4Ī
ģÙĀ
1) ¯°úÄţNL6R67Qţ®°ÙŘŤm£@Q;ćk"
1Ü¾Ő SAR ¼¹ćk"
1h2ł*ŏ "—
ºþŠ”Ŝ".1hœ$îrţÜå”Ŝ±£`ĻÙř
ė 12 ³ţė 6 „ţpp.133-146, 2012.
2) R. Bolter, and F. Leberl, Detection and reconstruction of human
scale features from high resolution interferometric SAR data, In:
Pro. 15th Int. Conf. Pattern Recog., Vol. 4, 2000, pp. 91–294.
̶ 96 ̶