remote sensing image data analysis system ......remote sensing 工mage data analysis system...
TRANSCRIPT
ISE-TR-86”54
・懸 .
REMOTE SENSING IMAGE DATA ANALYSIS SYSTEM{cTSUKUSYS”
AND IT’S APPLICATION
一Especially Classifing Coastal Area around a Bay Using↑M I-
by
Takashi Hoshi
Kiyoshi Torii
Tomoyuk日shida
January 6,1986
Remote Sensing 工mage Data Analysis System ”Tsukusys”
and 工tIs Appl ication
一 Espec ially ClassiEing Coastal Pllrea around a Bay Using [lhvi 一
Takashi Hoshi
工nstitute of 工nfo rma tion Sc iences and Electronics
Uni▽ersity of Tsukし止)a
Ibaraki 305 Japan
Klyoshl Torll
Depar七menし of Agricultural Engineering
コ KyOt=O UnlVerSlty
Kyoし0 666 Japan
Tomoyuk i 工shida
Depa rtment of Research Developrnent
University of Tsukuba
工baraki 305,Japan
Rem・te Sensingエmage Data袖a・ysiS SYstem ’lTSukusys・
and 工tls Application
一一一@Especially Classifing Coastal Area around a Bay Using IXvr 一一一
Takashi Hoshi
Institute of.Inforrnation Sciences and Electronics University of Tsukuba
Ibaraki・ 305 JAPAN
Kiyoshi Torii
Department of Agricuitural Kyoto University Iくyoto 606 JaDan L
Engineering
Tomoyuki Ishida
Department of Research[)evel・μnent コ
Unlversit二y of Tsukuba
Iba raki 305 JAPAN
Abstract
A system called as ”ISukusys” has been developed to
distinguish and analyze land use and land cQver patterns by
uslng remote senslr阿 1mage data r especially Landsat二
Multispectral Scgnner(IV!SS) irnage data. As, in Japanr
Landsat二 Thematic IYIapper (Tひ4) image dat二a can be acquired.
slnce last year, the appl ication of TM data by the Tsukusys
was attempted. As a result, the ad▽ant二age of the T岡data is
coinpletely vertified. This investigation is to interprete
wat二er quality condit二ion around t二he Koj ima Bay, where water
pollution becomes a social issue.
Contents
Introduction
Configuration of [Vsukusys 一一一一一一一一一一一一一一一一一一一一
2-l Configuration oE’hardware 一一.一一一一一一一一一一一一一一一一一
2-2 Configuration of software 一一’一一一一一一一一一一一一一一一一
2-3 Sorne fea tures o f the Tsukusys 一一一一一一一一一一一一一
An Exarnple of Tsukusys’s appl ication 一一一一一一一一一一・・一
3-1 0bゴective to classify coastal area 一一一一一一一一一
3-2 Description of test site 一一一一一一一一一一一一一一一一
3一一3 Hydraulic structure in test site 一一一一一一一一一一一一一一
3-4Analyticai procedure 一一一一一一一一一一一一一一一一・一一一一一一一・一一一一一一一
3-5 Results and discussion 一一一一一一一一一一一一一一一
3-6 Conclusion of appl ication 一一一一一一・一一一・一一一一一一一一一一一一
4 (lonclusion 一一一一一一一一一一一“”一....一www”.w
1
2
3
1
1
1
4
9
9
9
Acknowledgments
References
,夢9曜,劇,e-po-oo響〇一一一一一,一一一}一画岬,,,,
嘲■脚ρ●嘔幽P●■P■■■■■「刷■P■一■-騨門■■■■脚ρ口■■「,■■r騨■r■騨P■■’朝■■■■9「■■9r●■■■■■引■闘-■V燭■脚7■■’■■騨■■F■口脚q願t刷圏ρ■■PO■艶闘■■’■■F■■■■馴■P■■騨
10
12
12
13
17
18
18
18
1. 工ntroduction
an invesしigaしi・n・n regi・nal scale is indispensaple f・rしhe research・f earch
science’ for the invenヒories of natural resource and for the monition of
つ ロ
env1「o㎜ent・It ls also lM’P()rtant to invesしigate the temporal variaしion on the
「egion・da「th”・bservati・nal sens・r sys七ems・・aded in sate・・‡しes ha▽e rnet these
requirements since・97・Is・画・wぬndsat雌S data are c・㎜ercia・・Y・bしainab・e and
extensively used in Japan・T・st・re the multiしemp・ral lV6S data may be。ne。f
necessary conditions to develop t士1e apPl icaしion of remote sensing しechnique.
Therefore r a daヒabank cons i sting of a large amoしmt of the image data should be
establ ished, based on daしabase concepts. Wi t士ししhe realization of the databank in
mind, remote sensing image data analysis system ca11$d as lITsukusys闘 was developed
in1983.
This paper g ives an outl ine of the Tsukusys. Segしion 2 describes the hardware
organl zaしlon and software organization of the system. Section 3 discusses the
apρlication of the system’ esρecially t血e usefulness of Landsat TM daしa t二〇 classify
coasta]. area around tne Koゴima Bay that includes a desalted reser▽oir where water
quality has been degraded。
2. Configuration of TsuK)usys
2-i Configuration of hardware
a. Host Cornputer System
Ahost computer oE 48 M bytes memory capacity r which runs at Science
エnf・rllati。n Pr・cessing Cenしer’Uni▽ersity・・f TsukUba(Fig一・),is used as rem・te
sepsing image analysis systern. Here the most important design was how can transfer
data frqm. Iarge capacity d i sk memory to irnage rnernory at high speed. This problem was
resolved by developing an interface between thern and by connecting them onl ine
-1 一一
I
N曜
Other Faculties
ComputerNet Work
Remote Station(27 Stations )
TSS Terminal(300 Terminals )
N-1 Network (15
Universities)
Open Ba tch
Station
Science lnformation Process’ing Center
Hos t Cornpu ter
CPU Memory(48 M bytes.)
BMC
DASD DASD
ccp BMC
Optical Cable
TEK 4025
TEK 4027
× 16
×4
Computer Sys tem for Education
Hos t. Computer
CPU・Memory (24 M bytes)
のStandal.one
Termina・1・
DASD
5 GB
MTI 2)3
× 90
⑤…一……〈動
78 GB 1600/6250 BPI
OpticalDisk
g}t15ffil{fiaging’
Disk
Mass
StorageSys tem
101.7 GB
Imqge Processing System
lmage Memory
1.00 MB
Image
Memory.O.78 MB
512 ×512
ColorPisplay
×
512 ×512
ColorDisplay
Smalt Computei’
CPU Memory (O.77 MB)
1024 × 1024
ColorDisplay
1024 × 1024
ColorDisplay
×
×
×
1
1
DASD
320 MB
1
1
MT )MT Termina1 ×3
2000 × 3000
CCD Camera
Fig-1 Con.figuration of Host Copmputer Sy$tern
through central process ing units (CPU). [Ehe aux i l iary storage oE the host computer
cons i sts of 9 magnetic tape (MT) units, a d i sk systein (DASD) (78 G bytes) and a mass
storage systern (IOI.7 G bytes).
工nterdata 8/23 system is also available at the center as a sub-compu仁er syst:em
for analyzing remote sens ing image data. [[he sub-systern includes a high-density
drurn-scannerr which the Tsukusys uses for digitizing color aerial photographies.
b. mass Storage System
工mage databank of Landsat MSS data was constructed and has been arranged so far
as to .sto re 106 tandsat irnages taKen wi th in and outside Japan
〈 Hosh i et al (1983) 〉. ’IWo important features of the databank a re to bring tandsat
image data inし。 uniしy data format, that is, B工L fo rmat:and to be able to process
thern onl ine. The later fea ture i s ’pe rformed by the mass sto rage systern.’IXAio
cartridges as shown in Photo-1 (size of cartridge is 4.7 crn in diarneter and 8.8 orn
in height) can efficiently and systematically store a whole image taken by tandsat.
Photo-1 Ca rtridge of itass Storage Systern
一3一
t
Using the daしabank. comp◎sed of mass storage system, it; is・ possible to select an
interested image thrdugh image ret=rieval system〈 工shii et;aユ. (1983) 〉.,、 to take ouし
the dat二a stored in mass storage syst(『m to stag ing d isk, and t len to convert the dat:a
stored in sしaging disk t二〇 disk system・ This process reduces operaしion in respect;しo
labour and time, by comparing wi th しhe operat二ion t=hat a dat二a stored in magneしic ヒape
ロ リ コ
エs d↓rectly coρied to dat;a file of DASD. 工n additlon, this mass storage system・・haS
tne merit;of smaller cost than the disk sy$tern in performing。,
c・ Color graphic disρlay
Two color graρhic displays are included in the image I/0 ρrocessing units
connected to Che host compute r. The d i splay uniしs have l.78 iyi byt二es (512 x 512 byヒes
x 4 frames and 3 frames) image memory. The reason why two dispユ.ays are adopt;ed is to
be able し。 operate them indeρendenしly and t;o compare the image before and aft;er
processing・ This is convenienし fQr image processing・
d. Co:Lor ink一一ゴさt printdr/plotter
エtis des;rable t。 preser▽e s。me images・r maps represgnted。n・c・1・r graphic
displays・ Of co]・or pr:t.nter/plotter on sale’ a color ink一ゴet printer/plot;しer made by
Applicon Comp. in U.S.A. was thought having enough capability for use. lhe maximum
picture size produced by it is more t;han four times than (重uart:o paper. 工し is larger
than an ordinary map in Japan。 Although the design ofl ha rd wa re is planed t;o process
。n・ine as menti・ned ab・▽er the printer/P・・tヒer is n・t c。n「iected・n・ine・pecause
ρ「inting and。thr「p「ocessing can p「ocee昌at甑e same time’total P「。cessing time
will be reduced’・wing t・。ffline pr。cessing・As a s・IUble ink in water.jets’ヒhe
printer/pl・tter has a demerit that c。1。r。f pictures.fades。u七as m◎istening.エt had
bette「・laminate picture串in。rdrr t。 prevent e。1・r fr・m fading「・u七・
2-21 bonfiguration of soEtware
a・ Usable inpuし image data sets . ・ 、
If only・・工andsat MSS data are concerned,the mass st;orage syst二em is enough for
pracしical use. But there are other remote sensing image data sets’ so 工/O soft壷are
is installed f・r analyzing these image daしa sets・Usable input image data sets qre
-4一
飴ndsat雌S data st・red in fl・pPy disketteダdigiしaiized aerial c・IQr ph・t・graphic
data’and airborne tv!SS data.
(1) Landsat fv!SS data
In Japan r £or rnicro-computer user, 8 inch floppy diskette (1 M bytes) in which
irnage data belonged to a portion of a irnage taken by tandsat tvlSS is storeq can be
acquired.
(2) Digitalized aerial coior pnotographic data
An aerial color photography is digitalized as 8 bits data with Red-Green一一Blue
color.coding system r using above一一mentioned drurn-scanner 〈 Hoshi et al (1981) 〉.
(3) Airborne MSS data
The airborne MSS data l i ke iY!SS data produced by Jirco Comp. in Japan or
Daedalus Cornp. in U.S.A. are also available.
Pre-processing includes 4 different cornponents, that is, speciEication of test
site, elimination of striping noiser geometrical registration and correction to
norrnal distribution.
(1) Spef iEication of test site
工n order to specify a しest siしe aしwill by using truck-ball, a whole Landsat;
LYISS dat;a or oしher image daしa are ext二racted at intervaユ of 7 ρixels, so all of しhe
selected pixels can be displyed on a coior graphic display sirnultaneously. At this
operation r a mesh at 512 pixels interval is displayed together with remote sensing
image data r so the specification of a test site is easily perforrned.
(2) Elimination of striping noise
MSS embarked on Landsat has 6 detectors peF channel. Due to the difference of
response , しhe lag of offset二and A/D con▽ersion wit二h t;hese deteqしors’ stripes t;urn
up systernatically. [[he Tsukusys has capability to correct them.
(3) Geometrical registration
As rernote sensing image data do not correspond geornetrically with ordinary map,
it is diEficult to cornbine other data with some resulting irnages or maps produced
一一 5 一一
fr。m t・he image daしa.エn・rder t。 res・1▽e this pr。blem’the Tsukusys P「epa「esξ
software especially for geornetrical registration t hrough universal transvers(
mercat。r(U珊pr・ゴecti・nくNagamine et al(1982》》・U獅pr・ゴecti。n was ad・pted b:
the reason that ordinary rnaps on the scales of 1/25,000 and 1/50,000 in Japan hav(
been produced using this projection. , ,(4) Correction to norrnal distribution
The rnernory of host computer and irnage mernory are capable of assigning a dat,
into a byte. Therefore, previous acquired Landsat rvlSS data which are forined as
/
bits data are converted. to 8 bits ones and corrected to norma1 distribution wit
taking the effect oE enhancement into account.
c. unsupervised classfication ・, [the unsupervised classfication analysis applied in this systern is t h
hierarchical clustering method which includes tne 6 different componenPs as follows
(1) Wa rd rnethod
(2) Group average rnethod
(3) Centro id rnethod
(4) Med i an rnethod
(5) Furthest ne ighbour method
(6) Nearest ne ighbour rnethod
Although it is desirable t hat all pixels within a test site are submitted to
clustering algorithm for producing a resulting map, it is not efficient in terms e
calculation time. Hence, in this systern, an approach to improve classification spee
is ad・ptedぴacc・rding t・bel・w・rnenti・ned pr・cedures〈H。shi et al(in press)〉.
The 256 pixels that are randomly located are extracted frorn interested sub一一tes
site. They are regarded as rePresentative of test site. Ctr)e method chosen frorn thos
mentioned’ ≠b盾魔?@is performed within the 256 pixels. [Ehe 256 pixels are ・flnall
combined into 20 clusters. Every pixel of test site is then assigned to the ciuste
sAzhose mean vector ’is closest to the pixel vector. For display, each pixel assigne
to s珈e cluster is labeled wiしh one corresp()nding color, a color sch㎝e is Inanuall
altered so as to match well. 一一 6一
d. Supervised ciassification
ll)e well-known rnaxirnum l i kel ihood method i s implernented to ・assign the pixels to
one of pre-deterrnined classification iterns according to the analysis objectives. 2
criterions are adopted to assess the degree of success oE the classification
results.
(1) Average perforrnance
[[he average performance is obtained on the basis of correct recognition for
each i tem and is expressed as
n
S=.Z.:.1 ri/n .一一一一一一一一一一一一一一一m一一 (2-1)
Vghere r and n refer to the percentage of correct recognition for each item and
the number of i terns respectively.
(2) Equivocation quantification
.1he equivocation・qunatification is ’based on the concept of entropy by
information theory and considers the infuluence of rejection items. 〔【『be fしmction is
denoted by [V(G,G) as expressed in equation (2-2).
T(G,召) = 工(G,ご)’/H(G》
= 1 一 { H(Gr ’G一’) 一 H(G一) }/ H(G) ’一”一一一一” 〈2‘”’2)
Vthere 1(G,G)r H(GrG), H(G) and H(G) refer to rnutual inforination, joint entropy,
entropy of a prior itern and entropy of the posterior item respectively.
T(G’G) varies from o’七〇 1・ エf the value is obtained to be higher than o.65, the
classi[ication result is generally considered as good.
e. EValuation of・distance and area
-The estimation in distance and area can be carried out on the basis of counting
the nuMbeF of line and colume in interested image data. A distance can be evaluated
-7 一一
by specifying two ρoints on CRT wit士1しruck・一ball・Also, an area can be sized up from
sum total of pixels belonged し。 same classification item not only in a rectangular
region but also in an irregular region. A disしance or an area evaluaしed in this way
will be compared wit血 bygone data acquired by ground survey or image dat二a.
f. 工mage descripしion
Ouしput devices for describipg image data include color graphic display, coユ.or
ink一ゴeし prinしer/plotter and X-Y plotter.
(1) Display on color graphic display
trh6 subr。utines f・r pr。▽iding a funとti。n t・dispユay。n CRT were devel・ped.
These subroutines include the data ponversion from one「CRT to ot二her‘CRT and the
representat二ion of image data depend ing on some formats. ’,
(2) Display on color printer/plotter
Although the color ink一つet printer/plotter installed aし ーヒhe center adopts
Lyragent;a, Yellow, and Cyan (M.Y.C) color coordinate, Red r Green, and Blue (R.G.B)
c・1・rc・ding systern can beしransf・rmed t。 LYi.Y.C acc・rd ingし・an expressi・n bel・w’
M al「bl cl R
Y = a2 b2 c2 . G 一一。一一一一一一一一一一一一一一(2-3》
C a3 b3 c3 B
Where the matrix elements are experimentally determined.
Hisしograms for each of M, Y and C are construc=ヒed for each image. The M-Y・一C
levels are transforrned to color plotter patch which is composed of l6 dots either in
equal disしance basis・r in equal pr・babiliしy basis.16 d・ts patch can represenしi7
Cし止×∋ color le▽eユS.
(3) Display on X-Y plotter
工し 1s lmportant;to observe features of training data utilized by suρervised
classification analysis. The Tsukusys has capabilit;y to represent these data on X-Y
pl・tter’using.1’c。nstellaしi・n meth・d”by which rnulti▽ariate data can be displayed。n
plane or it 4-axis methodi1 by which 4 efficient channel da ta can be disρlayed on
ort血ogonal coordinaしe system, so distribution of these data becomes clear。
一8一
h
Z2-3 Sorne features of the Tsukusys
Landsat tvlSS data cover all over the werld and are . represantive of rernote
sensingネmade da ta・エt is imp。rtant as a system f。r educati・n and training t・
analyze the popular irnage data by using the authorized unsupervised and supervised
cl.assiEication analysis. ln designing digital image data analysis system,. most of
designers only have an inclination to construct hardware systern and to develop
software system, not including image databank. The reason is that the databank is
too expensive to be werthy of study obj’ective. However, in future, it may be
indispensable Eor.remote sensing irnage data analysis systern to contain an image
databank and the retrieval systern.
A Color graphic dispiay plays an important role for observing image data. So,
the Tsukusys efficiently ㎝ployes the 2 displays.工t is also important to shorten
the time for displaying image data. Here, high-speed processing is realized by
developing the inte r face between host cornputer and image rnernory.
Citing other features, the way how to extract the data appl ied in cluster
a19。砒㎞with unsupe「vised classificati・n’new criteri・n with accgracy。f
supervised classification (equivocation quantification) and so on are included.
3. An example of Tsukusys’s appl ication
Here, the attempt to classifY coastal area around a bay using [ITvl is introduced
as one of applications 〈’Ishida et al (1985) 〉.
3一一i Qbjective to classify coastal area
[[he objective oE this study is to investigate the usefulness of Tlvr data for
lnte「p「etユng職e「quality・Especially’the study was E・cused・n classifying water
zones and extracting flow patterns.
一9一
Phoし。-2 Natural Color 工mage of Test Site and Training Classes
3-2 Description of test site
The area stud ied includes the Koゴima Bay in Japan. 工t i s encompassed by t」ae
K・jima Peninsula Wnich is 1・cated in the cenヒralρart。f Seし・エnland Sea and several
Kiiorneters away from Okayama City in the southern direction・A reclamation proゴect
aroしmd the bay by drainage has been begun more than one hundred years before.
However e because there is no sufficient water for the reclairned agricultural land,
crops frequently sustained drought damage・ 工n order し。 sol▽e this problem’ a levee
across the middle of the bay was constructed in l962・ 工し ρrevents sea wat二er from
flowing into tne bayr and a desalted reservoir with an area of 1,088 ha was formed.
As a resultr suf£icient irrigation water supply got guarantee and reclaimed land
became available.
一 10 一
tately, Kurashiki and Sasagase River catchrnents proceeded to be urbanized, wi th
the inc rease of population on the outskirts of ekayama City. Accord ingly,
urbanizaVion caused water pollution in the reservoir. Because both industry waste
water and dornestic sewage of cities are ’poured ’into” rivers and water from the
yoshii, Asahi, Sasagase and Kurashiki River Elows into the bay (Fig-2)i So various
chernical materials move into the reservoir and deposit there. Water quality of the
reservo i r comes to be deg raded. :t i s thought that i t is irnpo rtant to construct a
reg ional sewerage net to draw clear watet from the upper reach and to drain water
flowing in lower stratum by desalination underdrain, as to irnprove the water
Sea ef Japan
D
、欝t確書
Se
’Kyes.hLu
Di st ri ct
Shikoku
District
Pacfic Ocean
ioo KmN一一,.一一..一,.一一dve
Fig-2 Geographical ILocation of Test Site
一 11 一一
quality. The present study was attempted to extract the information about; flow
pat;しerns aroしind the bay from TM data 〈 Torii et al (1985) 〉’ in order to pro▽ide
scientific e▽idence for the proゴect of water purification・
3一・3 Hydraulic structure in test site
The level of sluice gaしe for drainage in Koゴima desalted reservoir is usually
adゴusted acc。rding t。 the rise and fall・f tide・[[he sill。f the gaしe is f・unded at
the elevation of 3 meters below the sea-1e▽el and’ some spots are deeper しhan・ 3
rneters. Fig」3 shows the isobat血 of desalted reservoir at present二・Most area of tt)e
reser▽oir are.shallow, aboUt 2.5 m in depth. There are only two deeper spots along
しhe levee and two deeper spots nea亡 ’].efし and right shore respectively.
ウ Fig-4 shows the distribution of ▽elocit二y of plan flow obtained by もol▽ing・two
dilnens ional sha1ユow water equation 〈Kawachi et al (1983) 〉, using depth
disしribution・data as boしmdary condition. The resuit is acquried on t士ie calculation
condiしion that much water water flows int二〇 tne reservoir, on t士1e ot血er hand r much
water emits from tne gate, and so the stream patterns、 can l be completely understood・
3-4Analytical procedure
Of softwares. mentioned in configuration of software, pre一一processing,
uni3up6r▽ised classification, supervised classification and image descript二ion are
adopted and t血e image processing proceeds in turn. T()beg in wi th, the TM digiしal
daしa t血at were acqui red on 8th May l984, are con▽erted for handling a§ MSS data and
pre・一precesSing is performed on t血e convert二ed data set二s.
エnthis c昌se, pre-pr・ceSsing d・es n。t include ge。metrical registraしi・n and
correction of striping noise, because the registration is not necessary judging from
the obゴective of the ρresent study and striping no i se can not be foしmd. 4 bands TM
daしa corresp◎nding し。 wave range of 瞭∋Sグ that is, Band 2,3, 4 and 5. were selecしed
for analysis. The しest site is disp].ayed by composing bands data in Photo一・2. And
t血en, unsupervised classification is performed using centroid method. The result is
utilized as reference[data for super▽ised classificat二ion・ Super▽ised classification
is performed using maximum likelihood method・
一12一
3-5 Results and discussion
Phot二〇一3 is t二be classificat二ion map on the case that しraining siヒes amount二 t二〇 14
sites shown in Table-1 and Table-2. Water bQdies and land area were classiEied into
8 and 6 types respectively. Judg ing frorn the con;ingency table (Table-2), average
accuracy is high at 83.9 percentage of correct classiEication and at O.86
equivocaしion quantification ・ エヒ is considered as high as t;he results acquired from
airborne MSS data 〈Hoshi (1979) and Torii et a1 (1979) 〉. However, the accuracy of
classiEication item 4t 5, 12 and 13 are extremly low. Lirniting nbtice to the types
concerning water body.e altnough 8 types are correctly classified, the types do not
lead t士1e resultanし map representing some features of filow. 4 combined t二ypes in
respect of land and 3 coinbined types of water body are displayed in PhQto-4 so as to
disしinguish definiしely the st二aしus of flow.
Photo-3 Supervised Classification Map (No. of Training Classes is 14)
一 13 一一
Table-1 Characteristic of Training Classes
1
心i
Classification it㎝ No.of pixels Band 2 Band 3 Band 4 Band 5
1. Wa ters in the reservo i r 216 105.6±1.2 42.2±0.7 41.4±0.8 28.8±0.5
2. Waters in the Koj ima Bay 200 115.2±1.5 49.1±0.7 45.7±0.9 27.3±0.8
3. Wa ters in the I n l and Sea 416 109.5±1.3 ’40.9±0.7 36.8±1.0 21.3±0.8
4. Wa ters in Sasagase River 120 117.3±3.2「 49.3±2.4 ’51.0±4。7 44.6±14.8
5. Waters in Kurashiki River 104 112.7±L4 47.3±1.4 48.2±1.9 49.1±20.9
6. Wa ters in reclaimed area 脳 110.8±1.5 45.8±0.8 44.1±1.2 33.2‡4.8
7. Waters in Yoshino River 192 119.3±1.5 49.6±1.1 50.4±1.9 32.0±3.3
8. Wa ters in Asahi River ㎜ 119.1±1.6 49乙5±1.2 48.6±3.1 33.3±7.6
9. Paddyfields in planting season 132 111.6±2.5 47.9±1.6 47.3±3.3 122.4±7.4
10.Urban areas 娚 137.6±4.1 60.5±2.1. 71.8±4.3 65.8±4.8
11.Trees 謝 103.0±2。0 46.7±1.9 43.6±1.8 107.3±15.0
12.Bare fields 45 140.9±6.9 .〔違.0±4.1 87.9±7.6 76.9±6.5
13. Concre te s t1Mc tures around coas t 80 126.3±4.7 56.7±2.9 65.5±5.5’ 69.3±8.5
14.Shad㎝s 140 122.6±4.6 55。7±2.6 63.4±4.6 84.5±9.6
1
Aor l
Table-2 Contingency Table of Supervised Classification
Classification it㎝ No.of pixels 1 2 3 4 5 6 7 8 9 10 11 12 13 14
1. Wa ters ■1n the reservoir 216 98.6 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.5
2. Wa ters 01n the Koj ima Bay 200 0.0 84.5 0.0 2.5 0.0 7.5 0.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0
3. Wa ters 01n the Inland Sea 416 0.0 0.0 99.5 0.5 0.0 0.0 0.0 0.0 0.0 0.0 「0.0 0.0 0.0 0.0
4. Wa ters ■1n Sasagase River 120 0.0 0.0 0.0 35.0 45.8 11.7 0.O 2.5 0.0 0.0 0.00.O P
0.0 0.0
5. Wa ters ■1n Kurashiki River 104 0.0 σ.0 0.0 1.9 2510 66.3 0.℃ 0.0 0.0 0.0 0.0 0.0 0.0 0.0
6. Wa ters ■1n rec l a i med area 204 0.0 0.0 0.0 0.0 0.0 10010 0.0 0.0 0.0 0.0 0.0 0.0 0.0 010
7. Wa ters ■1n Yoshino River 192 0.0 0.0 0.0 1.0 0.0 0.0 87.5 5.7 0.0 2.1 0.0 0.0 2.7 0.0
8. Wa ters ■1n Asahi River ㎜ O.0 1.0 0.0 0.5 0.0 0.0 22’.5 76.0 0.0 0.0 0.0 0.0 0.0 0.0
9. Paddyf ields in planting seaSOn 132 0.0 0.0 0.0 0.0 3.0 0.0 0.0 0.0 93.9 0.0 0.8 0.0 0.0 0.8
10. Urban areas 螂 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 85.5 0.0 8.1 4.8 0.0
11. Trees 304 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.7 0.0 98.0 0.0 0.0 0.3
12. Bare fields 45 0.0 0.0 0.0 2.2 0.0 0.0 0.0 0.0 0.0 6.7 0.0 22.2 66.7 0.0
13. Concre te struc tures around (x沿s t 80 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 62.5 35.0
14. Shadows 140 0.0 0.0 0.0 0.0 2.9 0.0 0.0 0.0 1.4 0.0 0.0 0.0 22.9 69.3
’Average performance=83.9% Equivbcation quantification = O.as
rf7NE5i7 濫
’ }zi!
・Nl.5
!
N
q
奪
の
憾ミf一
瞥蕎 N
1.5
2Z
’
O,5
Fig・一3 工sobath (m) in the Reservo1「
VELOCITY(MIS) ト峰ト0.400 e■
工く)㎜「1盟N O.0100M/S
ノ る 1二,’一、・、
、こご二乞楽藁 \ 一 ノーく「」二、 ヘ ノ
ノ
ー、 A 、 ! / //\_ \ イ / /
\く、『/ ! ノ\ ’ \ 噸 ’, ’ ,
、 、 ’ ! 1 ’ 、 ,
鳴 ” ● 覧
、 ●
・ ○
○ ■ .
. o
. , ,
・ . ● . ・
. .
… ■ ・ ●
. ,
o ● φ
‘ ● ●
く1.
蝿 ’、 3跡」層●印の一」
驚募= ノ’t ”_,の一
知
..ッ
髪纏..’、、” @ミミ1
糠’沁\ 瓢\\ xx N N N N
.
%
Fig ・一 4 Caluculated Velocities .in the Reservolr
一 16 一
It is very interesting to compare、the map wi th the distribution of calcu:Lated
velocity (Fig 一4)● Because the map represenしs waしer flowing out二side、t士lrough water
gate, so iヒ is supposed that二water from Sasagase and Kurashiki River flows into the
reserVOIr●
3rr6 Conclusion of appl ication
From this test example, following conclusions can be madeラ
(1) The accuracy o:E the TM data is as well as airborne MSS data.
(2) 工f successi▽e TM digital data of a test;site are st二〇red, it may be’possible to
detect exactly future transition on the enviro㎜ent around theヒest site.
.(3) 工naddition, the どemote sensing’techniques may bring their capabilities into
full play’ taking togeth〔学r with simulation model in respect of water and chemical
maしerial flow.
Photo-4 Supervised ClassiEication Dvlap (No. of Training Classes is 7)
e’ 17 ’
4. Conclusion
工し has been shown that;the Tsukusys was deVeloped mainly for しhe ana:Lysis of
Landsat tvlSS data r could cope wi th the Landsat [“vr da ta. HOwever, in o rder to expand
the utilization of Landsat Tlvr data, it is necessary that the TsuKusys can sele¢t
spectral bands wi th anaiysis or rnore than 4 bands data can be processed, while the
TsukusYs presently only deals wi th 4 bands data.
The second systern is being designed r taking notice of the selection and
eniargernent of the bands. The systern i s scheduled to process tandsat Trvl data, MOS-1
tv!ESSR data .and SpCrr XS data. When the new systern i s accornpl i shed, t he sys tern wi l l
aiso be open for education and training as well as the TsuKusys.
AcKnowledgments
The authors would like to thank Mrs. Zhang Shaoxing Eor her enl ightening
comments and suggestions.
References
1) Hoshi,T. (1979) ; Processing System ”USASi, of MSS Dat;a from an Aircraft;and 工し。s
Application, Proc. oE Japanese Societ’ 凵@gf.Civil Eng., ’No.285r pp.69-83. (in
Japanese)
2)H。sh・,T. andエs・hama,K.(1981);lnput/Output$・ftware・fエrnag・e Pr。cessing’JAFSA
Rep. 811012, pp.207-220. (in Japanese)
一 18 一
3) Hosh i,T. and Nakayarna,K. (1983).; A 1!andsa t rmage Databank Us i ng iY!ass Sto rage
System-et TrsUKUSYSie 工lnage-bank一・一’r Inter Graphics I 83, An 工nternational Conf. &
Exhibition on Computer Graphic Technical SessionS, D8-i, pp.1-15.
4) Hosh i,T. and Sato,K. ( in press ) ; Analys i s of iYlang rove Forest in OKinawa Us ing
Airborne Remote Sensing Data r Proc. of 6Vh Asian Conf. on Rernote Sensig.
5) lshida,T. and Hoshi rT. (1985) ; Estirnation of Regional Evapotranspiration Using
Land Classification wap Obtained frorn Landsat D4SS Data r Proc. of Annual Conf.’85r
Soc iety of Photog rammetry and Remote Sensing, pp.95-100. (in Japanese)
6) 工shii’Y・’Nakayama’K・’and Hoshi’T・ (1983); The Analysis of Landsat (MSS) 工mage
Dat『亀 Using 工nしeractive Information Retrieval L尋nguage SOAR, 工nter Graphics曹83, An
エnternati・nal C・nf・&Exhibiti・n・n’ C。mputer Graphic Technical Sessi。ns’D8・一4,
pp.1一一i4.
7) Kawachi,T. and Kooutangkulvong r S. (1983) ; Numerical Investigation iWo
Dimensi。na・Backwaしer by㎞e Finite Element Me伽d,」.・fエrri. hng. and Rural
Planning, No.3, pp.32・一51.
8)Nagamine’K・and H・shi’T・(1982);pre-pr・cessing。f tandsat lYISSエmage Data by
Using Graphic Display, Proc. of Annual Conf.’82, Society of Photogrammetry and
Remote Sens ing, pp.61一一64. (in Japanese)
9)[DoriirK. OkagawarN.r Kltarnura,T. and Hoshi rT. (1979) ; Analysis of Water Area
around the Suwa Lake Using CCCT, Proc. of Annual Conf.’79, Soc i ety o f lnstrument and
Control Eng.t pp..111-114. (in Japanese)
IO) ToriirK. HoShi,T., and lshida,T. (1985);Flow Patterns around the Kojima
Reservo i r Us ing [IM r Proc. o£ Annual Conf.’85, Society of Photog rammetry and Rernote
Sensing r pp.IOI-104. (in Japanese)
一 19 一一
INSTITUTE OF INFORIvrATION SCIENCE AND ELECTRONICS UNIVERSITY OF TSUKUBA
SAKURA-MURA, NIIHARI-GUN, I BARAKI, JAPAN
REPORT NUMBERREPORT DOCU凹ENTATION PAGE
ISE-TR-86-54
TITLERemo・ヒe Sensing工mage Data Analysis Sys’ヒem 町sukusys”
and工t 1 s Application
● ・ ● und aBa. Usin MT一AUTHOR(S)
Takashi Hbshi
1くiyoshi Torii
TQmoydd工shida
REPORT DATE NUMBER OF PAGES
January 6, 1986 19
MAIN CATEGORY CR CATEGORIES
Remote Sensin9 4.0
KEY WORDS1血㎝atic Map鉾r, Stream Pattem ,Super▽ised Classification
ABSTRACT
A system ca】」. ed as“TSukusys.。 has been(皐eveloped to
distinguish arKヨanalyze 1and use and 1and cover patterns’by
・ .
tSlrg remote SenSlng image δata, especially Landsat
Mul tispectral Scanner (MSS) imag e data. As」 in Japan,
Landsat Th㎝atiC Mapper (別) image data can be acquireq .
SinCe laSt year, the apPIication of TM dat二a by the Tsuku寧ys
was attempted. AS a reSult, the ad▽antage of t二he TM data is
c㎝Pletely vertified・ This in▽estigation is し。 interprete
water quality condition around the K・」ima臼ay油e「e mter
pollution becomes a social ●1ssue●
SUPPしEMENTARY NOTES