Journal of Geomatics Science and Technology

fa بهبود الگوریتم تداخل‌سنجی PSInSAR با استفاده از بهینه‌سازی شاخص پراکندگی دامنه داده‌های پلاریمتریک دوگانه سنجنده Sentinel1-A PSInSAR Algorithm Improvement Using Amplitude Dispersion Index Optimization of Sentinel1-A Dual-Polarimetric فتوگرامتری و سنجش از دور Photo&RS پژوهشي Research <div style="text-align: justify;"><span style="font-family:b nazanin;"><span style="font-size:11.0pt;">روش‌های تداخل‌سنجی راداری پلاریمتریک مبتنی بر پراکنش‌گرهای دائمی یک تکنیک موثر در افزایش تراکم و کیفیت فاز پیکسل‌های  پراکنش‌گر دائمی می‌باشند. این روش</span></span><span dir="LTR">‌</span><span style="font-family:b nazanin;"><span style="font-size:11.0pt;">ها با ترکیب خطی کانال‌های پلاریمتریک بر اساس بهینه</span></span><span dir="LTR">‌</span><span style="font-family:b nazanin;"><span style="font-size:11.0pt;">سازی پلاریمتریک کانال بهینه</span></span><span dir="LTR">‌</span><span style="font-family:b nazanin;"><span style="font-size:11.0pt;">ای را جستجو می</span></span><span dir="LTR">‌</span><span style="font-family:b nazanin;"><span style="font-size:11.0pt;">کنند که در آن تراکم و کیفیت فاز پیکسل‌های پراکنش‌گر دائمی نسبت به کانال</span></span><span dir="LTR">‌</span><span style="font-family:b nazanin;"><span style="font-size:11.0pt;">های خطی افزایش پیدا کند. در همین راستا، هدف اصلی این مقاله توسعه الگوریتم تداخل‌سنجی </span></span><span dir="LTR">PSInSAR</span><span style="font-family:b nazanin;"><span style="font-size:11.0pt;">، که تاکنون تنها بر روی داده</span></span><span dir="LTR">‌</span><span style="font-family:b nazanin;"><span style="font-size:11.0pt;">های تک قطبی بکار گرفته شده است، جهت بکارگیری داده</span></span><span dir="LTR">‌</span><span style="font-family:b nazanin;"><span style="font-size:11.0pt;">های چندزمانه پلاریمتریک دوگانه با هدف بهبود این الگوریتم در شناسایی پیکسل‌های  پراکنش‌گر دائمی قابل </span></span><span style="font-family:b nazanin;"><span style="font-size:11.0pt;">اطمینان </span></span><span style="font-family:b nazanin;"><span style="font-size:11.0pt;">می</span></span><span dir="LTR">‌</span><span style="font-family:b nazanin;"><span style="font-size:11.0pt;">باشد. این بهبود بر اساس بهینه</span></span><span dir="LTR">‌</span><span style="font-family:b nazanin;"><span style="font-size:11.0pt;">سازی پلاریمتریک با تابع هدف شاخص پراکندگی دامنه (</span></span><span dir="LTR">ADI</span><span style="font-family:b nazanin;"><span style="font-size:11.0pt;">) بر روی 17 تصویر پلاریمتریک دوگانه (</span></span><a name="_Hlk507370290"><span dir="LTR">VV/VH</span></a><span style="font-family:b nazanin;"><span style="font-size:11.0pt;">) سنجنده </span></span><span dir="LTR">Sentinel1-A</span><span style="font-family:b nazanin;"><span style="font-size:11.0pt;"> انجام گرفت. روش بهینه</span></span><span dir="LTR">‌</span><span style="font-family:b nazanin;"><span style="font-size:11.0pt;">سازی مورد استفاده در این تحقیق روش </span></span><span dir="LTR">ESPO</span><span style="font-family:b nazanin;"><span style="font-size:11.0pt;"> می</span></span><span dir="LTR">‌</span><span style="font-family:b nazanin;"><span style="font-size:11.0pt;">باشد. نتایج نشان می</span></span><span dir="LTR">‌</span><span style="font-family:b nazanin;"><span style="font-size:11.0pt;">دهد که تعداد پیکسل‌های  پراکنش‌گر کاندید و نهایی کانال بهینه در مقایسه با کانال </span></span><span dir="LTR">VV</span><span style="font-family:b nazanin;"><span style="font-size:11.0pt;"> به ترتیب حدود 2.6 و 2 برابر افزایش پیدا کرد. همچنین در این مقاله مکانیزم‌های پراکنشی که در بهینه‌سازی پلاریمتریک با داده‌های پلاریمتریک دوگانه قابل استخراج هستند، مورد بررسی و تفسیر فیزیکی قرار گرفتند. در نهایت نقشه فرونشست جنوب غربی تهران با پردازش سری زمانی هر دو الگوریتم  </span></span><span dir="LTR">PSInSAR</span><span style="font-family:b nazanin;"><span style="font-size:11.0pt;"> معمولی و الگوریتم </span></span><span dir="LTR">PSInSAR</span><span style="font-family:b nazanin;"><span style="font-size:11.0pt;"> بهبودیافته، بدست آمد. </span></span><br>  </div> <ol> <li><strong>Introduction</strong></li> </ol> Persistent Scatterer Interferometry (PSI) is a technique to detect and analysis of a network of coherent pixels referred as Permanent/Persistent scatterer (PS) which are stable throughout time-series of SAR images. This technique has been applied to monitor and measure phenomena such as earth subsidence fault movements and earthquake volcanic activity and other geological and environmental studies. In all PSI techniques, the processing is carried out only on the PS pixels. Therefore, high density and phase quality of these pixels are the most effective factors on the results of these techniques. The main challenge of this technique is to detect the coherent pixels over non-urban areas which suffer from the temporal decorrelation.<br> Nowdays and with the development of polarimetric SAR sensors, polarimetric radar data are available. Polarimetric data consist of several conventional SAR acquisitions, usually addressed as channels. Each channel in a PolSAR acquisition is sensitive to different geometric characteristics of the scene. This additional redundancy over the scene may allow to increase both quality and density of the PS pixels. Therefore, the combination of polarimetry and interferometry enables to improve the effectiveness of PSI techniques, especially in non-urban areas.<br> In this paper, we employ a method to improve the conventional PSInSAR algorithm for detecting PSC by using polarimetric optimization method on dual-pol SAR data. The improvement of this research is based on minimizing ADI criterion by means of an Exhaustive Search Polarimetric Optimization method to increase the number of PSCs. <ol> <li value="2"><strong>Materials & Methods</strong></li> </ol> <strong><em>2.1 Dataset Description</em></strong><br>  <br> The proposed method is tested using a dataset of 17 dual-pol SAR data (VV/VH) acquired by Sentinel1-A satellite March 2017 and October 2017.<br>  <br> <strong><em> 2.2 Polarimetric SAR Interferometry</em></strong><br>  <br> A general formulation for polarimetric SAR interferometry was proposed in (Cloude & Papathanassiou, 1997). The scattering matrix S represents the polarimetric information associated to each pixel of the scene. Considering a monostatic configuration, the scattering matrix S is defined as follows: <table align="center" border="0" cellpadding="0" cellspacing="0"> <tbody> <tr> <td style="width:170px;height:1px;"><a name="_Hlk502997397"></a><img alt="" height="25" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image002.png" width="68" ></td> <td style="width:58px;height:1px;"> <img alt="" height="25" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image002.png" width="68" ></td> <td style="width:96px;height:1px;">(1)</td> </tr> </tbody> </table> <div style="clear:both;"></div>where <img alt="" height="19" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image004.png" width="19" >, <img alt="" height="19" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image006.png" width="23" >are copolar terms, <img alt="" height="19" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image008.png" width="19" > is the cross polar term. This can be represented with the target scattering vector <img alt="" height="20" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image010.png" width="7" > using the Pauli basis as: <table align="center" border="0" cellpadding="0" cellspacing="0"> <tbody> <tr> <td style="width:162px;height:30px;"><a name="_Hlk502995636"></a><img alt="" height="27" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image012.png" width="145" ></td> <td style="width:54px;height:30px;">(2)</td> </tr> </tbody> </table> <div style="clear:both;"></div>where <img alt="" height="19" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image014.png" width="11" > is the transpose operator. In Sentinel-1configuration (VV/VH), where there is no knowledge of <img alt="" height="19" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image016.png" width="19" > , scattering vector <img alt="" height="20" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image010.png" width="7" > can be written as: <table align="center" border="0" cellpadding="0" cellspacing="0"> <tbody> <tr> <td style="width:162px;height:10px;"><img alt="" height="15" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image018.png" width="68" ></td> <td style="width:162px;height:10px;">(3)</td> </tr> </tbody> </table> <div style="clear:both;"></div>In order to generate an interferogram, each target vector <img alt="" height="20" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image010.png" width="7" > can be projected onto a unitary complex vector <img alt="" height="22" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image020.png" width="13" > . Result of this step obtains the <a name="_Hlk516178398">scattering coefficient μ</a> defined as: <table align="center" border="0" cellpadding="0" cellspacing="0"> <tbody> <tr> <td style="width:162px;height:1px;"><a name="_Hlk502997294"></a><img alt="" height="15" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image022.png" width="102" ></td> <td style="width:29px;height:1px;">(4)</td> </tr> </tbody> </table> <div style="clear:both;"></div>where i corresponds to two images, and * represents the conjugate operator. The scattering coefficient <em>μ</em> is a new channel or polarization state which is a linear combination of the Pauli vector elements. In this regard, all interferometry techniques can be extended from single-pol configuration to a desired polarization state by using (4) and (5). The projection vector <img alt="" height="22" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image024.png" width="17" >for dual-pol data defined as: <table align="center" border="0" cellpadding="0" cellspacing="0"> <tbody> <tr> <td style="width:212px;height:10px;"><img alt="" height="30" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image026.png" width="161" > </td> <td style="width:82px;height:10px;">(5)</td> </tr> </tbody> </table> <div style="clear:both;"></div>where <a name="_Hlk516178829"></a><img alt="" height="19" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image028.png" width="8" > and <img alt="" height="19" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image030.png" width="11" > are two real parameters whose ranges are finite and known and are related to the geometrical and electromagnetic properties of the targets. The parameter <img alt="" height="19" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image028.png" width="8" > represents the type of scattering mechanism and <img alt="" height="19" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image030.png" width="11" > corresponds to orientation of scattering. In our research, the main purpose of polarimetric optimization is to search in a two-dimensional space, <img alt="" height="19" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image028.png" width="8" > and <img alt="" height="19" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image030.png" width="11" >, to find an optimum projection vector, <img alt="" height="22" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image032.png" width="14" >.<br>  <br> <strong><em>2.3 Amplitude Dispersion Index Optimization</em></strong><br>  <br> In order to generate a polarimetric form of <img alt="" height="19" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image034.png" width="16" >, it is sufficient to replace scattering coefficient, <img alt="" height="19" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image036.png" width="8" >, in (6) by <img alt="" height="21" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image038.png" width="78" > as define in (4):<br>   <table align="center" border="0" cellpadding="0" cellspacing="0"> <tbody> <tr> <td style="width:170px;height:29px;"><img alt="" height="55" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image040.png" width="148" ></td> <td style="width:76px;height:29px;"> </td> <td style="width:33px;height:29px;"> </td> <td style="width:33px;height:29px;">(6)</td> </tr> </tbody> </table> <div style="clear:both;"></div> <br>  <br>   <table align="center" border="0" cellpadding="0" cellspacing="0"> <tbody> <tr> <td style="width:208px;height:79px;"><a name="_Hlk502997450"></a><img alt="" height="69" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image042.png" width="180" ></td> <td style="width:117px;height:79px;">(7)</td> </tr> </tbody> </table> <div style="clear:both;"></div> <br> The main issue in the ADI optimization is finding a <a name="_Hlk516183283">projection vector </a><img alt="" height="22" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image032.png" width="14" > for each pixel, which leads to minimize the <img alt="" height="19" src="file:///C:UsersDaViDAppDataLocalTempmsohtmlclip11clip_image044.png" width="37" > value. <ol> <li value="3"><strong>Results & Discussion</strong></li> </ol> Our results confirm that the algorithm substantially improves the PSInSAR performance, increasing the number of PS pixels with respect to standard PSI, and increasing the phase quality of selected pixels. The results reveal that using the optimum scattering mechanism increases the number of PSC about 2.6 times and PS density about 2 times than using single channel datasets. Also, the effectiveness of the method is evaluated in urban and non-urban regions. The experimental results showed that the method was successful to increase the final set of PS pixels in both regions significantly. <ol> <li value="4"><strong>Conclusion</strong></li> </ol> In summary, it can be inferred that the polarimetric optimization method is successful to increase the number of the final set of PS pixels in different regions, significantly.<br>   تداخل‌سنجی راداری پلاریمتریک, پیکسل‌های پراکنش‌گر دائمی, شاخص پراکندگی دامنه, الگوریتم تداخل‌سنجی PSInSAR, نقشه فرونشست Interferometry, PSInSAR, Subsidence, Sentinel-1A, TerraSAR-X, Optimization 45 57 http://jgst.issgeac.ir/browse.php?a_code=A-10-697-1&slc_lang=fa&sid=1 S. Azadnezhad سعید آزادنژاد saeedazadnezhad@email.kntu.ac.ir 10031947532846005886 10031947532846005886 No K. N. Toosi University of Technology دانشگاه صنعتی خواجه نصیرالدین طوسی Y. Maghsoudi یاسر مقصودی ymaghsoudi@kntu.ac.ir 10031947532846005887 10031947532846005887 Yes K. N. Toosi University of Technology دانشگاه صنعتی خواجه نصیرالدین طوسی