ارائه روشی جهت بهبود مدل ارتفاعی رقومی حاصل از تصاویر سنتینل1 بااستفاده از مدل رقومی srtm و تبدیل موجک دو بعدی

تولید مدل رقومی زمین با قدرت تفکیک و دقت ارتفاعی بالا همیشه یکی از مهم ‌ترین اهداف سنجش از دور ماهواره ‌‌ای بوده است.  یکی از ارکان اصلی سنجش از دور ماهواره ‌‌ای، سنجش از دور راداری می‌ باشد.  تولید مدل ارتفاعی رقومی از سطح زمین با استفاده از تداخل ‌سنجی راداری به علت ویژگی ‌های منحصر به فرد این تصاویر برای محققین جذاب است.  در سال‌های اخیر پروژه ‌های فضایی بسیاری آغاز به اخذ اطلاعات از سطح کره زمین کرده ‌اند که یکی از آخرین آنها پروژه سنتینل می‌ باشد.  سنتینل1 بخش راداری پروژه سنتینل است.  مدل ‌های رقومی حاصل از تداخل ‌سنجی راداری به علت وجود خطاهای متنوع از جمله خطا در اطلاعات فازاینترفروگرام دارای خطا و گاهی اوقات اشتباه بزرگ در نقاط ارتفاعی می ‌باشند.  از اینرو مدل ‌های رقومی حاصل از فرآیند تداخل ‌سنجی راداری پس از تولید نیاز به بهبود دارند. در این مقاله روشی برای بهبود مدل ‌رقومی ارتفاعی به دست آمده از تصاوی رسنتینل1 با استفاده از مدل رقومی ارتفاعی موجود srtm.(shuttle radar topography mission)و روشی بر اساس تبدیل موجک دو بعدی، پیشنهاد می ‌شود.  تصاویر مورد استفاده در این مقاله بخشی از شمال شهر تهران است.  مدل ارتفاعی رقومی تولید شده با استفاده از روش پیشنهادی با مدل ارتفاعی رقومی مرجع یک متر با دقت ارتفاعی بالا مورد ارزیابی قرار می ‌گیرد.  نتایج مقاله نشان می ‌دهند که روش پیشنهادی به شکل موثری در بهبود دقت مدل رقومی حاصل از تصاویر سنتینل1 عمل می ‌کند. با استفاده از این روش خطای مدل رقومی ارتفاعی به میزان قابل توجهی کاهش می ‌یابد (30% الی 82%) و این بدین معنی می ‌باشد که با حفظ قدرت تفکیک مدل رقومی حاصل از تصاویر سنتینل1 می ‌توان دقت ارتفاعی آن را به شکل محسوسی بهبود داد.

Presenting a method for the improvement of Sentinel1 generated DEM, using SRTM and 2D wavelet transform

Extended AbstractIntroductionDigital Elevation Model (DEM) is a physical representation of the earth and a way of determining its topography through a 3D digital model. DEMs with high spatial resolution and appropriate precision and accuracy of elevation are widely used in various applications, such as natural resource management, engineering, and infrastructure projects, crisis management and risk analysis, archaeology, security, aviation industry, forestry, energy management, surveying and topography, landslide monitoring, subsidence analysis, and spatial information system (Makineci&Karabörk, 2016).Satellite images are one of the main sources used to produce DEM. In satellite remote sensing, optical and radar imagery are often used to generate DEM. Compared to optical satellite images, the main advantage of using radar satellite images for DEM production is that they are available in different weather conditions and even at nights. Two strategies used to produce DEM from radar satellite images include radar interferometry and radargrammetry(Saadatseresht&Ghannadi, 2018).Phase information of the images is used in radar interferometry, whereas domain information of the images is used in radargrammetry (Ghannadi, Saadatseresht, &Eftekhary, 2014). Moreover, short baseline image pairs are used in radar interferometry, while long baseline image pairs are useful in radargrammetry. These technologies both have their own advantages and disadvantages,which were investigated in previous studies (Capaldo et al., 2015).With radar interferometry, it is possible to produce DEM forlarge areas. Sentinel is one of the recent projects in satellite remote sensing. Sentinel constellation collects multispectral imagery, radar imagery and thermal imagery from the earth. Sentinel1 is the radar satellite of the constellation.Recent studies have investigated the precision of radar interferometry using Sentinel1 imagery (YagüeMartínez et al., 2016) and the precision of DEM produced using these images(Letsios, Faraslis, &Stathakis; Nikolakopoulos &Kyriou, 2015). Generally, DEMs generated through radar interferometry needs to be improved, mainly due tothe phase errors which in many cases turn into outlier points (Zhang, Wang, Huang, Zhou, & Wu, 2012). Various methods have been used to improve DEM generated from SAR imagery, one of which use the information obtained from SRTM DEM. For instance, a previous study used SRTM DEM to improve DEM generated from ESRI/2.Using the information obtained from SRTM, the interferometric phase of areas with lower coherency were improved (Zhang et al., 2012).The present study proposed a method to improve the accuracy of DEMs generated by Sentinel1 imagery. In this method, using ascending and descending Sentinel1 image pairs from the study area, DEM is generated using radar interferometry process. Then, precision is improved using SRTM DEM and a method based on 2D wavelet transform. Wavelet transform and 2D wavelet transform methodsAs a spectral analysis tool, wavelet transform is based on expanding any function like f(t)(1) inwhichaiis the expansion coefficient and 𝜓iis the expansion function.One of the interesting characteristics of discrete wavelet transform is that it can be used as a multiresolution analysis tool. To do so, a series of scaling functions or are used along with the wavelets to determine coarse data of the signals. Signal detailsare also covered by different wavelets with different scales.Separatingcoarsedataand details of the signal isthe actual basis of discrete wavelet transform algorithm which wasintroduced by Mallat (Mallat, 1989) and improved by Beylkin et al. (Beylkin, Coifman, &Rokhlin, 1991). As a fast and simple method for discrete wavelet transform,the process is performed based on the followingrecursive relationships betweenahighpass and a lowpass filters with the impulse responses h(n) and g(n), respectively (Primer et al., 1998):(2) and(3) Where the expansion coefficients h and g are scaling filter and wavelet respectively.(4) and(5) These formulas can be expanded to calculate 2D discrete wavelet transform. Proposed MethodThis section proposes a method of enhancing DEM generated from Sentinel1 imagery using SRTM DEM and 2D wavelet transform. Considering the capability of wavelet transform as a multiresolution analysis tool which can separate coarse data from details, figure 2 shows the proposed process of improving DEM. First, using discrete 2DWT, coarse information and details of each DEMs are separated using the ascending and descending conditions of Sentinel1 images. Then, two stages are considered based on the nature of these models. First, filtering coefficients usingthresholding and considering the average as the detail or high frequency part of the enhanced model. Second, coarse information derived from wavelet transform method have a resolution of40m and thus data derived from SRTM (30m) has a higher quality. Therefore, inverse 2DWT will improve the results and reach a resolution of 20 m.  Experiments and ResultsThe study area is located in Northern areas of Tehran (Iran) at the UTM coordinates of (542450 ,3964590) northeast to (539010, 3962350) southwest. Two Sentinel1 satellite image pairs (one ascending and one descending) are used in this study.In addition, a SRTM DEM with a spatial resolution of30m is used to improve DEM generated from Sentinel1 images. Sentinel1 derived DEM is evaluated using the 1m resolutionreference DEM. RMSE values shows the effectiveness of the proposed method in enhancing the Sentinel1 derived DEM, which means that using information obtained from the SRTM and 2D wavelet transform was reasonable. RMSE values are reduced from 24.2097m to 11.1749m which shows 54% improvement. The proposed method can enhance results to 30 82 percentapproximately. ConclusionThe present study investigates methods used for generating DEM from satellite images especially Sentinel1 radar imagery. DEM derived from Sentinel1 data has a high spatial resolution.Yet, it has some outliers or errors in elevation points whichneeds to be modified.  Therefore, the present study proposes a method based on 2D wavelet transformfor deriving elevation model witha spatial resolution (20m) equal to that of Sentinel1 DEM and improved precision and accuracy. In this method, filtering the details of the model using discrete 2D wavelet transform and modifying coarse information using SRTM DEM results in an enhanced DEM with higher spatial resolution.