IMAGE FUSION
TO IMPROVE QUALITY OF LAND COVER ANALYSIS
Ogi Gumelar
Pusat Data LAPAN, Jl. Lapan 70 Pekayon, Pasar Rebo,
Email: ogumelar@yahoo.com
A B S T R A C T
There was a successful research that has been done about image fusion where the purpose is to increase analysis of land cover quality. Even though the intensity of panchromatic will be affected by its originality, but it has been proven that fusion technique have better quality result. The result is an improvement of an image from resolution perspective with less changing any contrast differentiation for each band characteristic. A north district of Bali has been taken as an area of study because the selected area has a better morphological earth’s surface.
Keywords: Landsat 7, SPOT, Brovey , Fusion 1)
I. PREFACE
The idea of this paper is to unite data between superiority of Landsat data with SPOT data, in purpose that each images can covered its weakness, from this combination then we can have a better image. This images can be processed to get a suitable information like land use or land covering information.
II. BASIC THEORY
The meaning of “BAND” here is a wavelength that has been recorded in satellites, which each satellite have a different sensor record.
2. 1 Image Unification.
Simply, images unification can be divided into three definitions, that is :7)
2.1.1 Fusion is image unification between two images or more with a certain algorithms.
2.1.2 Merging is image unification between two images or more with some enhanced techniques and image normalization.
2.1.3 Combination is an arrangement between several bands in one single image for some purposes.
There are few steps of unification, we can illustrate as a flow diagram in picture 1.
Picture 1. Flow Diagram of Image Unification Steps
2. 2 Data
The first data that has to obtained is multi spectral image of Landsat 7 ETM in August 19, 2000 where path/row 116/066 with panchromatic image of SPOT 2 in August 10, 2007 where Knum/Jnum 303/366 also SPOT 4 image in July 29, 2007. Area of study is located in geographical boundary between 8º09’11,79” – 8 º 12’40,3” LS and 115 º 26’16,89” -115 º 30’10,3” BT.
2.3 Landsat 7
Multi spectral image of LANDSAT has spatial resolution 30m with several bands in each band has different characteristic: 3) 4)
1. 1th Band 0.45 – 0.52 mm: This blue band has high information about water, though it very suitable for land use like land and vegetation.
2. 2nd Band 0.52 - 0.60 mm: This green band has information about vegetation, though it is suitable for detecting roads and water but it can be use to discriminate vegetation or not. Where sick plantation can be acknowledged because the red light absorption by the chlorophyll is decreasing or the reflectance in red area is increasing so it causing leaf color turn to yellow.
3. 3rd Band 0.63 – 0.69 mm: This red band has information about differentiation between vegetation and non vegetation, it can be seen that the contrast differentiation between soil and vegetation especially in urban area.
4. 4th Band 0.76 – 0.90 mm: This Near infrared band has information about variety of plantation and differentiation between water element and soil element.
5. 5th Band 1.55 – 1.75 mm: This Short wave infrared band has information about color differentiation between uncultivated land with other objects. This band also very suitable for water content of soil study, water content in plantation, pebbles formation and geological in general.
6. 6.1th Band 6.1 10.40 -12.50 mm: This thermal infrared band has information about water content of soil, also it can discriminate humidity of soil and thermal phenomenon.
7. 6.2th Band 6.2 10.40 -12.50 mm, same with no 6
8. 7th Band 2.08 – 2.35 mm: This short wave infrared band has information about uncultivated land as well as band 5 but this band has more information in geological study and pebbles formation.
For the ninth band where also it can be called by “panchromatic band” has spatial resolution 15m. In this study we will use ortho image of Landsat where the multi spectral from Landsat already unite by its’ panchromatic, in other hand band combination was consist of 2nd, 4th and 7th.
2.4 SPOT
SPOT-4 image satellites has four spectral with 20 m spatial resolution and it has different wavelength, for example 1th band with interval 0.50 - 0.59 mm, the 2nd band in interval 0.61 - 0.68 mm, the 3rd Band in 0.78 - 0.89 mm, and 4th Band was Short Wave Infrared in 1.58 - 1.75 mm. SPOT-4 panchromatic image is recorded using visible wave in 0,51-0,71 mm with 10m spatial resolution. SPOT-2 image has three same spectral as SPOT-4 but its’ panchromatic is recorded by visible wave in range 0,49-0,73 mm. 6)
2.5 Unification Method
+ 
Landsat 7 Image Spot 4 Image
Landsat 7 with Spot 4
Brovey Transformation method (Color normalized) is used in joining resolution of Landsat multi spectral image with resolution of SPOT panchromatic image. Before joining process start, usually we have to correct geometrically SPOT image referred to Landsat orthorectified image. Determination of ground control points is based on image analysis, it means random withdrawal based on irremovable objects like buildings, roads, intersection, river, etc, where located in superimposed images, this points withdrawal has been doing based on visible objects in scale 1:33.204.
Due to flat area was majority chosen in this area of study then interpolation we used is polynomial linear. After we have both images then we can join them by using ER Mapper software where this software has resolution merge algorithm. This algorithm is useful in joining between 14.25m spatial resolution landsat ortho multi spectral image and 10m spatial resolution SPOT panchromatic image.
III. RESULTS
The results of fusion between SPOT and Landsat image can be used to analysis land cover area in North of Bali. The original Landsat 7 image was viewed in picture a, resolution merge between SPOT-2 with multi spectral Landsat 7 image in picture b and for the last picture (picture c) is fusion between SPOT-4 panchromatic image with Landsat 7 multi spectral image.
Picture a. Landsat 7, b. SPOT-2 fusion and c. SPOT-4 fusion
From the results it can be seen that spatial resolution of Landsat 7 has been better visually, for example like in boundary residence, agricultural field area, geological morphology, forestry, watershed, river, and other pattern objects. In fusion image resident area is rather blocky than original landsat image and also vegetation area has a contrast differentiations.
The fusion results from both image have 10m spatial resolution in each pixel, even in water area is seen differently due to small haze in SPOT-2 and SPOT 4 image, and also a different date acquisition and sensor type.
The benefits in using this fusion method are:
- We can have a better image because there are several surplus from both image
- Even though Brovey transformation methods affect color system in each band inputs from Landsat 7 but it doesn’t changes any contrast differentiation in each characteristic and also spatial resolution become improved.
- To avoid limited data or to increase economic efficiency
The weakness points using this fusion are
1. Both of images have different swath
2. Both images have different acquisition date, incident angle, nature condition and also cloud position.
3. Spectral information from Landsat becomes decrease due to intensity level from SPOT as panchromatic image.
IV. CONCLUSION
From fusion process we can conclude that:
1. The reason why using LAndsat 5 was better for this fusion method due to Landsat 5 doesn’t have high resolution band.
2. There are several fusion techniques that can be used to optimize quality of image.
3. Fusion image between high resolution (IKONOS and QUICKBIRD) with multispectral data (like SPOT, Landsat, MODIS, etc) that we can use as secondary document in land cover delineation.
4. We can also using data ALOS for alternative ways.
This conclusion only based on analysis of writer in several images and variety of image quality.
Thanks to
The writer wants to say thank to all member whose have supporting this study exclusively in remote sensing environment and
Daftar Pustaka
1) Jorge N´u˜nez, Xavier Otazu, Octavi Fors, Albert Prades, Vicen¸c Pal`a, and Rom´an Arbiol. 1999. Multiresolution-Based Image Fusion with Additive Wavelet Decomposition. IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 37, NO. 3, MAY 1999, pp:1204-1211
2) Gonzalez, Rafael C, 2001. Digital Image Processing/Richards E Woods. Prentice Hall, Inc.
3) Berita Inderaja Volume V, No.10, Desember 2006 PUSDATA LAPAN
4) Berita Inderaja Volume VI, No.11, Juli 2007 PUSDATA LAPAN
5) Penginderaan Jauh dan Interpretasi Citra. PUSDATA LAPAN dan UNNESS
6) F. Sri Hardiyanti Purwadhi, 2001. Interpretasi Citra Digital. PT Gramedia Widiasarana
7) Lillesand, T.M, Kiefer R.W., 1993. Penginderaan Jauh dan Interpretasi Citra. Yogyakarta;
8) Sutanto, 1994. Penginderaan Jauh Jilid 1 dan 2. Yogyakarta:
9) Lucien Wald, Some Terms of Reference in Data Fusion. IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 37, NO. 3, MAY 1999. pp: 1190-1193
