GISdevelopment.net ---> AARS ---> ACRS 1995 ---> Poster Session 1

Mask Procedure for Effective Classification and Expression on Mangrove Area with TM Data of Landsat 5

Kazuhiro Sato*, Dwi Setyono* and Kenshi Takezaki**
College of Agriculture, University of the Ryukyus,
Senbaru I, Nishihara, Okinawa, 903-01 Japan
Tel. 098-895-2221 (Ex.2912),
Fax. 098-895-2864
Information System Department, Kankou Co., Ltd.,
Higasinoda 1-6-1, Miyakojimaku, Osaka, 534 Japan
Tel. 06-351-1416, Fax. 06-351-3398

Abstract
In this paper, a mask procedure as a preprocessing, preparing two sets of secondary imagery data using respective mask, separated classification of the inside and outside of a mangrove area were proposed and tried for the effective classification and expression on a mangrove area. The product took two forms of overlaying classified result of the inside to the classified result or three bands color composed image of the outside. For the mask and following procedures, manual designation and input of threshold values, and functions ,. in an imagery analysis system such as binarization with two values ( black and white) , reversing and synthesizing two images, were used. Although imagery data of Band 4 and 5 were basically used in this procedure Using the data of Band 7 with Band 5 was more effective according to circumstances.

1. Introduction
Mangrove is a unique plant society as an important ecosystem distributed along the coastal sea and river mouth soaked in brackish water in the tropics and subtropics extensively. It is composed with many woody and several herb species, has supplied materials and stuff of the traditional use from house down for the local people, has cultivated coastal marine resources such as many kinds of crab, prawn, shrimp, small fish and so ~ on, and further more has performed disaster preventive functions against the wind, waves and current from; the sea and river. Although the management, control, working and operation methods as mangrove forestry had been established for the production of charcoal, firewood and thin log in some Southeast Asian countries, ~ since the 1970s mangrove forest has been intensively cut for the chip production of pulp at each place in such : countries. cutover land.has been diligently reforested: but it must be admitted that reforestation does not I catch up with cutting. And m mangrove area, the change m land use has been proceeded on a very extensive scale than the disparity between cutting and reforestation of mangrove forest. The major changes are for digging and making fish ponds, prawn nurseries, reclamation for harbor facilities, industrial and residential sites.

In such situation the significance of mangrove ecosystem is reconsidered and many citizen' s groups begin to reforest mangrove here and there. Administrative organs also regard management and control of mangrove areas as important for the sustainable forestry and conservation of the environment and start making efforts to grasp the present circumstances of mangrove forest and collect information on it for the establishment of guideline on delicate management and control of it. But in many countries, it is not enough to collect and sum up the fundamental data such as distribution, area, species composition, tree height, trunk diameter, f; stand volume and so on because mangrove forest has been placed as the others. If the numerical values on mangrove forest in the world are asked, the distribution of species and total area should be had by fair means } 7 or foul, however it must be allowed that the source materials are derived from various survey methods and years ranged for the order of ten years.

In consideration of the situation mentioned above, many trials to extract mangrove areas using remote sensing have been reported 6) , and in most of trials it was placed at one category among many categories for land use/cover. It generally corresponded With something to be desired and had produced some degree of accuracy. By reason of that necessity of detailed information on mangrove area will increase in future, the ;if; authors have investigated fundamental problems to link up with the detailed classification and estimation of resources. The effectiveness for some by band ratio of reflectance had been suggested to separate mangrove species in Okinawa based on the measurement of reflectance from crown layer by a photometer 5). The sign ificance was investigated on the distribution of dots for CCT count of TM data of Landsat 5 and by band ratio in a three dimensional space, and the possibility was suggested to classify mangrove species and the surroundings into more detailed categories 1.2.3 In case of a trial to classify a mangrove area into some categories in a wide area including many categories, errors in classification had occurred like distributing mangrove in the mountain.

In this paper, mask procedure was proposed and tried as an effective method for the classification and ex- pression on mangrove areas. The way of thinking and procedures were described paying our attention to the features of mangrove areas recognized in Band 4 and 5 of TM data of Landsat 5.

2. The way of thinking and procedures
Mangrove areas as a type of swamp occurs distinctively between the waters of the coastal sea and brackish water in the mouth of a river, and land not to be soaked in. This matter is markedly different from the other types of swamp and becomes a key to extract mangrove areas. From this it can be derived that the inside and outside of mangrove area is separately classified after extraction and definition of mangrove areas.

Mangrove areas are covered with many terns, fern and grasses. Land hardwood forest, maritime forest, swampy grass land, mud flat and waters are generally given as the neighboring categories with mangrove areas. In an image of Band 4, a mangrove area gives similar high reflectance to the other vegetation and "'has a clear boundary with large difference of CCT count between itself and the waters or mud flat. On Band 5 it has a boundary between itself and land area not to be soaked with a tide as it's forest floor is similarly soaked with a tide to mud flat. For two mangrove areas in the mouth of the Inderagiri River, Riau,Sumatera and the Nakama River , lriomote, Okinawa, two images ( digital values from 0 to 225 correspondence from black through gray to white) of Band 4 and 5 were shown in Fig.-l with profiles of CCT count across a man- grove area. From the profiles, threshold can be easily designated for the binarization to make respective mask.


Fig. l Images and profiles of CCT count across a mangrove area of Band 4 and 5 for two mangrove areas in the mouth of Nakama River, Iriomote, Okinawa and the Inderagiri River, Riau, Sumatera

Through separate classifications of the inside and outside of mangrove areas, it becomes possible to make the error in the classification small and also to express mangrove areas effectively focusing our attention on them self. Mask and following procedures were described below.
  1. Designation of the threshold value in Band 4 and binarization: image A ( mangrove and land area to t' white of 255, waters to black of 0 )
  2. Designation of the threshold value in Band 5 and binarization: image B ( land area to white of 255, mangrove area and waters to black of 0 ) -If there are many bare fields or rock exposures in a subject area, it is more effective to add the binarized image of Band 7 with a threshold value on mangrove areas to image B-
  3. Subtraction of B from A: image C ( the inside of mangrove area to white of 255, the outside to black of 0 If there are remained white part shadows in the mountain and shallows, crests in the sea, it is effective to give them black of 0 using a software like painting.
  4. Reverse of image C: image D ( the inside to white of 255, the outside to block of 0 )
  5. Subtraction of C from images of each band: imagery data set to classify the outside ( the inside to black of 0, the outside to gray between 0 and 255 )
  6. Classification and three bands color composition of the outside: image E and F ( the inside to black of 0)
  7. Addition of D to images of each band: imagery data set to classify the inside ( the outside to white of 255, the inside to gray between 0 and 255 )
  8. Classification of the inside: image G ( the outside to black of 0 as the others )
  9. Overlaying: image H and I ( Classified the inside is separately overlaid to E or F classified and three bands color composition of the outside.)
Above mentioned mask and following procedures were shown as Fig.-2. This masking produced higher positional accuracy with the designation of appropriate threshold value than using a mouse or track ball on the CRT. Although a slightly similar method for updating land use data was reported, the NVI image and false color image were used to discriminate the waters and land or vegetation and non-vegetation area4).


Fig.2 Conception of Masking and following procedures



3. Practice of mask and following procedures
It is better to set some lines across mangrove areas in a image for designation of the thres hold. Binarized images were saved as A from Band 4 and B from Band 5. In the case of masking for the data of Iriomote Island, it was effective to use Band 7 Band 5 as there were may plowed farms, bare fields and rock exprosures around mangrove areas. A painting software was used to give black of O to remained wh1te parts With out mangrove areas in the produced image by subtraction of B from A. The modified image was saved as a : mask to cover the inside and outside of mangrove areas.

Imagery data of C was subtracted from the data of each band for the preparation of data set to classify the , outside. Each data kept its own data in the outside and had O of black in the inside. Imagery data of D was added to the data of each band for the preparation of data set to classify the inside. Each data in this set kept its Photo.-2 Comparison of Image C in lriomote between ( A -B) and ( A -B + B' ) own data in the inside and had 255 of white in the outside. Mentioned process above was a mask procedure. The preparation was completed for separate classification of the outside and inside or three bands color composition of the outside. Binarized images in this mask procedure and examples in two data sets prepared for classification of the inside and outside were shown in Photo-1, and two images were shown in Photo.2 to compare the effect using Band 7 with Band 5.


Photo.1 Example of images produced from masking and following procedures shown as Fig.1


Photo.2 Comparison of Image C in Iriomote between (A-B) and (A-B+B')

For classification of the outside, a fit classifier was applied to the prepared data set with image C and each band. Color images were composed with RGB corresponded to three bands within the set as true color, natural color or false color image. From these procedures, image E and F were produced. Classification of the in- side also was similarly done to produce image G corresponding the outside to black of 0.

Finally G was overlaid to E to produce image Has a classification map and to F to show a classification of mangrove areas emphatically.

Mask and following procedures described above was effective for more detailed classification and clear expression of mangrove areas.

4. Conclusion
Mask procedure was proposed and tried as a preprocessing to classify mangrove areas detailedly paying our attention to the features in Band 4 and Band 5. By this procedure it became possible to classify mangrove (C areas without interference from each category giving similar spectral reflectance in the inside and outside. It produced to raise accuracy and clearness of classification of mangrove areas. A system packaged this way of thinking and procedures is expected to be effective for the establishment of the management and control methods of mangrove areas.

References
  • Dwi Setyono & K. Sato : Similarities between Mandah, Sumatera and Iriomote, Okinawa on the Distribution of TM data of Land sat 5 in the Multidimensional space for Mangrove and Surroundings, Proc. of the International Symposium on Vegetation Monitoring, 298-303,(1995)
  • Dwi Setyono & K. Sato: Structure of Thematic Mapper Data of LANDSAT5 for mangrove Area in the River Mouth of Inderagiri, Riau, Sumatera,Proc. of the 15th ACRS, Vol.2,G-5-1-6,(1995)
  • Kohei Cho, H.Simoda , T.Sakata &M. Yoshimura : Use of Vegetation Index of Satellite Data for Updating Land Use Data in Japan, Proc. of the International Symposium on Vegetation Monitoring, 267-274, (1995)
  • Sato, K., D. Setyono & S. E. Dab : Structure of Satellite Data for Mangrove Area in Okinawa (1) - Structure of TM Data of LAND SAT 5 for Mangrove Area in lriomote Island -,Proc. of the 14th ACRS, E-3-1-6,(1993)
  • Sato, K., D. Setyono, O.D.Moraes & Hoshi, T.: Properties of Spectral Reflectance of Crown and Crown Structure of Mangrove Species in Okinawa, Proc. of PORSEC- 92, Vol.2,922-927,(1992)
  • S. Vibulsresth, B.K1ankamsore, S. Ratanasermpong & C.Silapathong : Mangrove Forest Ecosystem and its Land Use Zoning by Remote Sensing, Proc. of the 7th ACRS, C-10-1-12,(1986)