Optik uydu görüntülerinin birleştirilmesinde yerel dalgacık yaklaşımı

Abstract

Bu çalışmada, 2009 yılı SPOT5 uydusuna ait, 10 m geometrik çözünürlüklü 3 bant Multispektrel verisi ile aynı alana ait 5 m geometrik çözünürlüklü tek bant Pankromatik verisi, 10 m geometrik çözünürlüğe örneklenerek birleştirilmiştir. Böylece her iki uydu verisinin üstün niteliklerini bünyesinde barındıran, hem 3 bantlı spektrel çözünürlüğü yüksek hem de geometrik çözünürlüğü yüksek yeni bir görüntü elde edilmiştir.Çalışmanın temel amacı, spektrel özellikleri olabildiğince koruyarak geometrik çözünürlüğü yüksek görüntülerin elde edilmesi ve elde edilen görüntünün hangi parametrelerle elde edildiğinin belirlenmesidir. Dört farklı görüntü birleştirme tekniği; yüksek spektrel, düşük geometrik çözünürlüklü uydu görüntüleri ile düşük spektrel, yüksek geometrik çözünürlüklü uydu görüntülerine uygulanmıştır. Kullanılan görüntü birleştirme teknikleri; Brovey dönüşümü (BT) ile görüntü birleştirme, yoğunluk renk tonu doygunluk (IHS) dönüşümü ile görüntü birleştirme, temel bileşenler analizi (PCA) ile görüntü birleştirme, dalgacık dönüşümü (WT) ile görüntü birleştirme olarak seçilmiş olup, görüntüye 3 farklı yapıda uygulanmıştır. Bu yapılar global yapı, blok yapı, pencere kaydırma yapısı olarak uygulanmıştır. Çalışmada SPOT5 uydusundan alınmış, İstanbul ili Boğaziçi bölgesine ait görüntüler kullanılmıştır.Görüntü birleştirme işleminin yöntemlere göre yapılıp, sonuçların elde edilmesi ve karşılaştırılması için yapılan hesaplamalar MATLAB platformu üzerinde geliştirilmiştir. Birbirinden farklı yöntemlerle elde edilen yüksek çözünürlüklü görüntüler; orijinal görüntüler ile birbirleri arasında karşılaştırılmıştır. Kalite değerlendirme metrikleri olarak kullanılan yöntemler; spektrel açı eşleştiricisi (SAM), ortalama karesel hata (RMSE), göreceli küresel boyutsal sentez hatası (ERGAS) ve bağıl ortalama spektral hata (RASE), referanssız kalite değerlendirmesi (QNR), evrensel görüntü kalitesi dizini değerlendirmesi (UIQI) olup, bu yöntemler kullanılan tekniklerin başarılarının değerlendirilmesinde kullanılmıştır. Böylece seçilecek parametreye bağlı olarak görüntü birleştime yönteminde başarı oranı arttırılabilinir.

Global and local scale work requires the acquisition and storage of current and accurate information of the earth and the analysis of data in different structures. Thanks to today's technology, significant tools have been developed to increase the expected success of the studies. Both software and hardware developments have supported the advancement of satellite technologies.Developments in satellite technology, especially in recent years, provide important advantages to users in order to obtain information on the earth fast, economically and with high accuracy. Most of the earth can be viewed with different spatial and spectral resolutions at different time intervals with the aid of remote sensing technologies and satellite images obtained can be used as basis in many studies.Integration of digitally recorded satellite images and different location data is done by creating geographical information systems. In this way, it is possible to produce analysis, interpretation and solution proposals of many problems based on location.In this study, the 10 m geometric resolution and 3 band Multispektrel 5 m geometric resolution and single band SPOT 5 panchromatic data of the same area with the Spot 5 data are combined by sampling at 10 m geometric resolution. Thus, a new image with high 3-band spectral resolution and high geometric resolution has been obtained that accommodates the superior qualities of both satellite data.The main purpose of the workshop is to obtain images with high geometric resolution while preserving spectral features as much as possible and determining by which parameters the resulting image is obtained. Four different image combining techniques have been applied to high geometric, low geometric resolution satellite images and high geometric and low spectral satellite images. The main purpose of the work is to obtain images with high geometric resolution while preserving spectral features as much as possible. The techniques used were selected as image combining with Brovey transformation (BT), image combining with intensity-color tone saturation (IHS) transformation, image combining with basic component analysis (PCA), image combining with wavelet transform (WT).Wavelet Transform has been started to be used in different disciplines with the researches made towards the end of 1980s and it is a method used in various analyzes such as target recognition, data compression. Wavelet transformations can effectively express data with local changes. Compared to the Fourier structure, frequency analysis was switched from frequency analysis to scale analysis. It is clear that the analysis of the measured mean fluctuations in different scales is less sensitive to noise. Instead of making general decisions about the time series, small fluctuations in the regional scale may be important. Therefore wavelet analysis is an option for users.Wavelet Transformation substitution I method; in wavelet analysis of the multispectral image, detail information expressing high frequency has been replaced with detailed information in the pan image.Wavelet analysis was performed and the reconstructed view was reconstructed with modified inverse wavelet transform.Wavelet Transformation substitution II method; wavelet analysis of the multispectral image and pan image is performed twice and the detail information expressing the high frequency in the wavelet analysis of the approximate value of the multispectral image is replaced with the detailed information in the pan image. Wavelet analysis was performed and the reconstructed view was obtained by applying an inverse wavelet transform twice to the modified data.Wavelet Transformation addition I method; in wavelet analysis of multispectral image view, detailed information expressing high frequency is collected with detailed information in pan image. Then, the combined image is obtained by the inverse wavelet transform.Wavelet Transformation addition II method; wavelet analysis of the multiple spectral image and pan image is performed twice and the wavelet analysis of the approximate value of the MS image is combined with the detail information in the pan image of the detail information expressing the high frequency.Wavelet analysis was performed and the reconstructed view was obtained by applying a reversed wavelet transform to the data with modified information.All of these are applied in 3 different structures; these structures are implemented as global structure, block structure, window shift structure. In global construction, image combining technology has been applied to the entire image at once. In the block structure, we divided the blocks into 2 blocks and applied the image merging method afterwards. Since the display is 1024 pixels, the block size can take a maximum of 29 values. In window shifting, we divide the bottom windows so that the image is 2 times, then we apply the splicing method by setting the shifting amount so that the window we selected will be 2 times and shifting the window over the image.In the study, the images of İstanbul Province taken from SPOT5 conformity were used. The main characteristics of the Istanbul test area have intensive urbanization, highway, bridge, intersection, different hydrological structures such as sea, stream, dam, intense marine traffic, less green area and less rugged area, it has fields such as park, stadium, forest.The calculations made for the image fusion process according to the methods and for obtaining and comparing the results have been developed on the MATLAB platform. High-resolution images obtained by different methods; Original images and are compared between each other. Methods used as quality evaluation metrics; (SAM), mean square error (RMSE), relative spherical dimensional error (ERGAS) and relative mean spectral error (RASE), reference quality (QNR), and universal image quality index (UIQI) Have been used in evaluating the achievements. The results obtained thus reveal the superiority of the proposed method. Thus, depending on the selected parameter, the success rate of the image combining method can be increased.