Keywords: DEM, SAR, RADARSAT . Stereo Matching, Mt. Tsukuba
Abstract :
Geographical Survey Institute (GSI) developed a technology to make a DEM using RADARSET high-resolution data. We're going to show you an examlple of Mt. Tsukuba, kanto district , Japan. We used visualized RADARSAT Fine Mode ( path image plus) images and an originally developed program. We used to stereo matching method to make a DEM. The accuracy of the DEM around Mt. Tsukuba acrea compared to existing basic map was 40m in the height ( standard deviation ). It was not so far from the accuracy which was expected by the resolution of RADARSAT data and satellite positions.
1. Introduction
SAR ( synthetic aperture radar ) is a very useful method to collect information in the tropics, because optical sensors cannot get information from the earth surface as a result of thick clouds. In addition, volcanic eruptions might interrupt information from the surface. Microwaves are scarcely influenced by theses barriers . Not only are they usable to make a basic map, but they are useful to estimate degree of topographical change and to know the situation of disaster. But former existing satellite like JERS -1 and ERS-1 cannot supply sufficient data because their resolution aren't high enough. RADARSET can improve this situation. The resolution of FINE DEAM MODE OF RADARSAT is nearly 8m to both azimuth and range directions. If we use PATH IMAGE PLUS data, we can get 3.12m square grand range data. this high resolutions and high -density data might provide us very accurate and useful topographical information. We will be also able to get higher resolution data in the future. Thus, is is proper to think that making a DEM ( digital elevation model ) using SAR data will play a very important role in the future.
2. Purpose
The purpose is to develop the technology to make an accurate DEM using RADARSAT data. there are mainly tow methods to make an elevation model. First, we can make a DEM using complex data. we use the interferometry method in this case. It can produce a detailed DEM. But it takes a long time to calculate the height of each grid, because we have to consider several complex processes visualize process ( including range compression, azimuth compression ), coherency check phase unwrapping and so on. The other way is the stereo matching method . Although it can't make a detailed DEM, we don't have to consider those complex processes. The program just considers correlation of features between tow images, and these features on the ground. A positional relationship between these two images shows the height of the feature. In this case, we use visualized satellite data instead of complex data. we originally developed programs to deal with these data.
3. Contents
3.1 Observation Area and Data
This Time, we decided to make a DEM around Mt. Tsukuba ( Figure 1) which is located in northern Kanto plain. We adopted this are because this are involves both flat area and mountainous area. We used two RADARSAT visualized images
Figure 1 Observation area and topographical feature of Mt. Tsukuba ( 1:25,000 topographical map " Tsukuba ", Published by GSI Japan )
| Date | 98.12.22 | 1998.12.29 |
| Orbit No. | 16347 | 16447 |
| Orbit Direction | Ascending |
| Beam Mode | FINE 4 NEAR | FINE 1FAR |
| Product Style | PATH IMAGE PLUS (SGX ) |
| Pixel size | 3.125X3.125M |
| Resolution | =8X8m |
| Coordinates for image center | 36° 13' N | 36° 12' N |
| 140° 11' E | 140° 09' E |
| Incidence Angle | =44° | =390 |
Table 1 Information for RADARSAT images
Figure 2. RADARSAT images around Mt. Tsukuba
( left : 1998.12.22, Right : 1998.12.29)
( Table 1 , Figure 2) to make a stereo pair. By matching these images, we calculated the height of each; grid in this area. A distance from the satellite to a certain feature corresponds with a projected position of the feature on the SAR image. with this idea, we can calculate the height of the feature. Since SAR data contain a lot of noises, we made correction to the data. we smoothed the data by median filter and standardized them.
