Geometric accuracy assessment of high resolution satellite imagery

Yüksek çözünürlüklü uydu görüntülerinin geometrik doğruluk yönünden incelenmesi

Abstract

The high quality level that high resolution satellite images have reached in the last years has proved that these images could be a useful data source for the production of orthophoto images and different mapping products. Specifically, it is very important in areas where aerial photogrammetry is not possible or feasible due to political reasons / restrictions and mapping frequency constrained by the limits of flight planning. Research activities about high resolution satellite images have been mainly focusing on the geometric accuracy aspects in ortho-image generation and Digital Elevation Model (DEM) extraction. In addition to these activities, some researches are increasingly concentrating on feature detection, recognition and reconstruction studies like automated mapping of roads, extraction of 3D buildings and cadastral mapping. There are two important criteria for quality assessment of the images; geometric accuracy and object definition. When we take the mapping processes into consideration, these criteria are basically dependent on the relation between pixel size and the map scale, contrast information (spectral range and colour), atmosphere and the sun elevation, the printing technology and the resolution of the human eye in a usual reading distance. A study has been implemented in Gölbaşı near Ankara, which has 10 km X 10 km area and a height difference about 440 m, for geometric accuracy assessment of high resolution satellite images. 12 ground control points were revised and signalized before the recording of images. Stereo-pair Ikonos (dated August 4th, 2002) and mono Quickbird (dated May 26th, 2002) images and 1/16000 scaled aerial photographs (dated August 29th, 2002) have been provided by Inta (Turkey), Eurimage (Italy) and General Command of Mapping (Turkey) free of charge.For geometric accuracy assessment; different digital elevation models (DEM’s) have been created using photogrammetric and cartographic methods. Photogrammetric DEM has been produced using1/16000 scaled stereo aerial photographs and has accuracy about 1 m. From the other side, cartographic DEM has been produced digitizing map contours and has accuracy about 5 m. And lastly, Ikonos DEM has been produced from stereo-pair Ikonos images automatically and has accuracy about 2 m. With these models, various orthophoto and mosaic images have been produced using ground control points. Used ground control points have had different distributions and number of points between 0 and 11. Well-recognized 27 checkpoints such as road junctions and field corners have been selected and measured with GPS. Planimetric coordinates of the checkpoints were measured on all orthophoto, mosaic and stereo images. These coordinates have been compared with GPS coordinates. From the comparison of these measurements, the differences and also the errors have been obtained. Afterwards the gross error positions have been evaluated according to ground control points’ network. The geometric accuracy results show that IKONOS - DEM can be used instead of photogrammetric DEM, the quality of GCP’s is as important as the number of GCP’s, systematic errors have occured in the easting / north easting (across track) direction and IKONOS images have better accuracy than Quickbird images when using direct sensor orientation parameters. Also we can say that the high resolution satellite images can be used in mid / large also from 1/6000 to 1/10000 scale map productions and these results are harmonious with the examples given in the literature.Actually there is nearly a consensus among photogrammetry and remote sensing society that some mapping applications from high resolution space imagery can be realized in mid / large scale. In addition, in spite of increasingly concentrating by some researches on feature detection, recognition and reconstruction studies in high resolution satellite images, it can not be said that these images couldn’t reach to the level of aerial images in determining and identifying of small features in big scale maps (e.g. 1/5000) yet. Therefore the discussions about usage of high resolution space imagery in big scaled map production are still lasting.

Özet

Yüksek çözünürlüklü uydu görüntülerinin son yıllarda eriştiği üstün kalite düzeyi, bu verilerin ortofoto görüntüler ve farklı haritacılık üretimleri konusunda iyi bir veri kaynağı olabileceğini kanıtlamıştır. Ancak günümüzde halen, yüksek çözünürlüklü uydu görüntülerinin büyük ölçekli harita üretim çalışmalarında kullanılabilmesine ilişkin tartışmalar sürmektedir. Bu nedenle, yüksek çözünürlüklü uydu görüntülerinin geometrik doğruluk yönünden incelenmesi amacıyla, 10 km x 10 km boyutlarında bir alana ve 440 m yükseklik farkına sahip olan Ankara yakınlarındaki Gölbaşı bölgesinde bir çalışma gerçekleştirilmiştir. Bu çalışma için 12 adet Yer Kontrol Noktası (YKN) tespit edilmiş ve görüntü alımından önce bu noktalar işaretlenmiştir. Daha sonra stereo Ikonos (4 Ağustos 2002 tarihli) ve mono Quickbird (26 Mayıs 2002 tarihli) görüntüleri ile 1/16000 ölçekli stereo hava fotoğrafları (29 Ağustos 2002 tarihli) temin edilmiştir. Geometrik doğruluk çalışması için öncelikle fotogrametrik ve kartografik yöntemler ile Ikonos uydu görüntülerinden Sayısal Yükseklik Modelleri (SYM) elde edilmiş ve farklı sayı ve dağılımda bulunan YKN kullanılarak değişik özelliklere sahip ortofotolar üretilmiştir. Arazide belirgin olarak tespit edilebilen 27 adet bağımsız kontrol noktası (test noktası) seçilmiş ve bunların koordinatları GPS ile arazide ölçülmüştür. Bağımsız kontrol noktalarının GPS koordinatları, aynı noktaların tüm ortofoto görüntüler üzerinde okunan koordinatları ile karşılaştırılmış ve aradaki farklar belirlenmiştir. Geometrik doğruluk araştırmaları sonucunda yapılan değerlendirmeler, yüksek çözünürlüklü uydu görüntülerinin orta/büyük ölçekli (1/6000–1/10000) harita üretim çalışmalarında kullanılabileceğini göstermektedir.