main page on remote sensingarticles about remote sensingikonos orthotransformed images generating using dtm from spot spacecraft

IKONOS Orthotransformed Images Generating using DTM from SPOT Spacecraft. / November 8, 2004 /

User may instead of using costly and even unpurchaseable aerial photography have the possibility of monitoring and rather frequent updating of data on vast areas of the earth’s surface. There evolve and are developed quite new lines of activity like pointed (pinpoint) farming, high-accuracy forecasting of rock and geological structure bedding, high-accuracy engineering in building of settlements, transportation and urban infrastructure, territory development planning, and ecological monitoring.
Orthotransformed satellite images are a unique data source and correspond with topographic maps in scale and geometric accuracy. Since the orthophoto has all distinguishing advantages of a map, then it enables accurate measuring the distance between objects and the area.
On the other hand, there may be easily added some additional layers from topographic base and vice versa, may be updated existing maps using orthorectified image.
A raw image may be easily orthotrasformed by the DLT method (Direct Linear Transformation) using geometric correction. This requires only 6 ground control points following which the image co-ordinates are presented as a function of map co-ordinates for each pixel of orthrectified image reduced to the reference system identical to that of the map. Then the digitization is performed by the nearest-neighbor method (method of image digitization wherein the output image element is formed by the parameters of its nearest-neighbor pixels).
Direct Linear Transformation (DLT)
The DLT method (Direct Linear Transformation) was first used by Abdel-Aziz Al Karara in 1971. The method is used for transformation of image into ground projection.
The DLT equation is as follows:

The equation with 11 parameters enables the transformation of x and y point co-ordinates in the image into ground project. Neglecting the third coordinate, this equation takes the following form:

Where x and у are the coordinates of points in the image and (X, Y, Z) are the coordinates of points on ground (ground control points).
All 11 parameters may be derived by solving this equation. First we should determine all 4 coordinates in the map, then for each pixel of the image reduced to the reference system identical to that of the map, we determine the gray level values using the DLT function and the appropriate position of point in the source image. Then its redigitization is carried out by the nearest-neighbor method (method of image digitization wherein the output image element is formed by the parameters of its nearest-neighbor pixels) is performed. However it should be noted that we need in addition the third coordinate calculated from the completed DTM.

Digital Terrain Model (DTM) comprises a set of height measurements made with a certain step (discreteness). However DTMs may differ both in accuracy, sampling increment, and simply in quality. DTMs are valuable data source when planning gas pipelines and conduits, geological and seismic analysis, designing communication and power lines and in other areas. In this instance we use the DTM constructed after the SPOT stereo pairs processing (Fig.1).

The DTM upper left corner coordinates — Е: 533902.811, N: 3953680.726; the DTM bottom right corner — Е: 537092.811, N: 3950670.726 in UTM projection. The number of pixels in line for a file is 319301. An array pitch (sampling) of DTM is 10m, the IKONOS image was imported with the same height and parameters in plan (10×10 pixels). In order to determine the height from the DTM file, a certain processing of source file was performed. The DTM source file is 16-bit (gray scale) and has the Raw format. The further computer-based processing required the following calculation: since the file is 16-bit, that is the height is expressed by 2 bytes then the height may calculated by the equation: H=2*256+1byte.

Space Imaging Company offers 5 levels of IKONOS data processing at a different price. The data is delivered in 8-bit or 11-bit format with ASCII metadata file.

Geo Product is more fairly priced product. It has the poorest accuracy and is not orthotransformed.

Precision Product is highest-priced. It has the highest accuracy (4m). To achieve the maximum accuracy of image requires a customer to provide Space Imaging Company with ground control points for forming orthotransformed image.
Since most images have been acquired with deviation from nadir then ground control points should have the accuracy of no worse than 1m and digital terrain models – no worse than 5m. Subpixel accuracy by the IKONOS data is unattainable.
The disadvantage of the Precision Product is time delays resulted from the necessity of providing ground control points and DTM as well as high price. Nevertheless, consumers may order Level Geo data (lower in cost by a factor of 5.5 then Precision Product) and perform their processing for themselves.
A satellite image of Teheran was used in this Project (3137, 2897).
In order that the Borland software may open this file, tiff format should be transformed into jpg.

Fig.2: IKONOS raw image

Approximately 35 ground control points from the 1:2000 map were used for geocoding.

Fig.3: 1:2000 Map of Teheran

Program, Results and Analysis
This program is written in Borland C language. Images and DEM are downloaded with opening menu (Fig.4).

Fig.4: Open image and downloading of data on height from DEM file
Ground control points may be derived manually or loaded from text file (Fig.5).

Fig.5: List of parameters
Geometric correction involves 2 stages:
1. Transformation of pixel coordinates: each pixel in geolocated image is transformed by the DLT method in order to determine the position of sampling in source image.
2. Resampling: is used to determine pixel values in order to entre them into geolocated image from source image (Fig.6).

Fig.6: Resampling by the nearest neighbor method
We thus obtain orthotransformed image (Fig.7).

The image accuracy is +4 pixels along the X axis and =14 pixels along the Y axis.
To test the truth of the results obtained, let us superimpose the 1:2000 map on the image (Fig.8).

Print version