吳文豪,周志偉,李 陶,龍四春
1. 湖南科技大學(xué)煤炭資源清潔利用與礦山環(huán)境保護(hù)湖南省重點(diǎn)實(shí)驗(yàn)室,湖南 湘潭 411201; 2. 武漢大學(xué)衛(wèi)星導(dǎo)航定位技術(shù)研究中心,湖北 武漢 430079
吳文豪,周志偉,李陶,等.精密軌道支持下的哨兵衛(wèi)星TOPS模式干涉處理[J].測(cè)繪學(xué)報(bào),2017,46(9):1156-1164.
10.11947/j.AGCS.2017.20160352.
WU Wenhao,ZHOU Zhiwei,LI Tao,et al.A Study of Sentinel-1 TOPS Mode Co-registration[J]. Acta Geodaetica et Cartographica Sinica,2017,46(9):1156-1164. DOI:10.11947/j.AGCS.2017.20160352.
精密軌道支持下的哨兵衛(wèi)星TOPS模式干涉處理
吳文豪1,2,周志偉2,李 陶2,龍四春1
1. 湖南科技大學(xué)煤炭資源清潔利用與礦山環(huán)境保護(hù)湖南省重點(diǎn)實(shí)驗(yàn)室,湖南 湘潭 411201; 2. 武漢大學(xué)衛(wèi)星導(dǎo)航定位技術(shù)研究中心,湖北 武漢 430079
哨兵衛(wèi)星采用TOPS模式成像,為確保影像干涉相位偏差小于3°,方位向配準(zhǔn)精確度需達(dá)到0.001個(gè)像素。本文論證了哨兵衛(wèi)星精密軌道條件下進(jìn)行幾何配準(zhǔn)即可達(dá)到較高的配準(zhǔn)精度,滿足增強(qiáng)譜分集進(jìn)行方位向殘余偏移量估計(jì)的前提條件,并基于DORIS軟件實(shí)現(xiàn)哨兵衛(wèi)星TOPS模式影像的幾何配準(zhǔn)和增強(qiáng)譜分集優(yōu)化配準(zhǔn)處理。地表方位向形變等信號(hào)可導(dǎo)致Burst干涉圖之間發(fā)生相位跳變,并對(duì)增強(qiáng)譜分集估計(jì)幾何配準(zhǔn)殘余偏移量時(shí)產(chǎn)生干擾,本文開展了相關(guān)的理論分析和數(shù)據(jù)處理試驗(yàn),提出粗差探測(cè)方法消除相位跳變對(duì)增強(qiáng)譜分集配準(zhǔn)的干擾。
合成孔徑雷達(dá)干涉;TOPS;哨兵衛(wèi)星;幾何配準(zhǔn);增強(qiáng)譜分集
哨兵衛(wèi)星(Sentinel-1)是繼ERS和EnviSat之后,歐空局(European Space Agency,ESA)發(fā)射的又一顆具有劃時(shí)代意義的C波段雷達(dá)SAR遙感衛(wèi)星。該衛(wèi)星選用TOPS(terrain observation by progressive scans)模式作為默認(rèn)成像模式,該模式繼承了ScanSAR模式的設(shè)計(jì)思路,在掃描成像過程中波束依次在各個(gè)測(cè)繪帶間進(jìn)行切換,通過犧牲方位向分辨率來增加影像覆蓋面積[1],因此衛(wèi)星可在12 d內(nèi)環(huán)繞地球運(yùn)行175周完成對(duì)地球表面的全部覆蓋[2]。TOPS模式通過壓縮目標(biāo)的多普勒歷程進(jìn)行成像,與ScanSAR模式相比,不但提高了影像質(zhì)量,更提高了干涉性能[3-4]。
歐空局在發(fā)射哨兵衛(wèi)星之前進(jìn)行了充分的星載TOPS成像模式理論和試驗(yàn)準(zhǔn)備工作,分別基于TerraSAR-X和Radarsat-2衛(wèi)星進(jìn)行了拍攝試驗(yàn),并進(jìn)行了聚焦成像處理[5-6]。哨兵衛(wèi)星在這些工作的基礎(chǔ)上進(jìn)一步提高了姿態(tài)和軌道的控制精度以提高衛(wèi)星影像干涉性能。TOPS模式影像干涉理論論證和數(shù)據(jù)測(cè)試基本上由DLR(德空局)主導(dǎo)完成。文獻(xiàn)[1]首先指出該模式影像需要較高的配準(zhǔn)精度。文獻(xiàn)[7]公布了TerraSAR-X衛(wèi)星TOPS模式干涉圖,并指出增強(qiáng)譜分集配準(zhǔn)(enhanced spectral diversity)處理方法可以滿足高精度配準(zhǔn)的要求。文獻(xiàn)[8]則論證了方位向重疊區(qū)域的增強(qiáng)譜分集算法的精度,并完成此類型影像的干涉和時(shí)序處理試驗(yàn)。隨著哨兵衛(wèi)星的成功發(fā)射,文獻(xiàn)[9]在上述研究基礎(chǔ)上完成了哨兵衛(wèi)星TOPS模式干涉處理理論分析與試驗(yàn)驗(yàn)證。與此同時(shí),其他研究團(tuán)隊(duì)針對(duì)哨兵衛(wèi)星TOP模式影像也開發(fā)了相應(yīng)的干涉處理軟件,如Gamma、GMTSAR。這些軟件各有特色,但主要算法均在幾何配準(zhǔn)和增強(qiáng)譜分集的框架中[10-11]。本文則基于開源軟件DORIS在精密軌道條件下完成哨兵衛(wèi)星TOPS模式的干涉處理。
TOPS模式在每組Burst掃描期間,波束由后向前擺動(dòng),使得波束覆蓋區(qū)的移動(dòng)速度大于衛(wèi)星速度,衛(wèi)星有足夠時(shí)間進(jìn)行多個(gè)子測(cè)繪帶的成像。TOPS模式零多普勒聚焦成像后一組Burst影像點(diǎn)目標(biāo)脈沖響應(yīng)函數(shù)表示為
exp[j·2πf0·(cosβ-1)(τ-τ0)]·
exp[j·2π·fηc·(η-η0)]
(1)
式中,r0為點(diǎn)目標(biāo)的零多普勒斜距;Vr為雷達(dá)傳感器有效速度;kψ為天線波束旋轉(zhuǎn)角速度。其中,sr和sa分別為距離向、方位向包絡(luò);τ、η分別為距離向、方位向時(shí)間;τ0、η0分別為距離向和方位向參考零多普勒時(shí)間;f0為雷達(dá)發(fā)射脈沖中心頻率;c為電磁波速度;β為直線幾何中波束中心斜視角。由式(1)可知,在沖擊響應(yīng)峰值處,即η=η0,exp[j·2π·fηc·(η-η0)]相位貢獻(xiàn)為零。但由于采樣點(diǎn)一般不在峰值處,聚焦影像存在相位斜坡。距離向相位斜坡可以選擇適當(dāng)?shù)某上駧缀斡枰韵?,然而方位向相位斜坡代表目?biāo)到散射體中心的實(shí)際斜距,無法予以消除[12]。干涉處理時(shí),方位向相位斜坡不但導(dǎo)致重采樣時(shí)需進(jìn)行方位線解斜或者內(nèi)插核調(diào)制處理,還需要較高的配準(zhǔn)精度,以降低多普勒中心變化帶來的相位偏差[13]。根據(jù)式(1)可以得到方位向配準(zhǔn)誤差Δη與干涉圖相位偏差Δφ的關(guān)系
Δφ=2π·fηc·Δη
(2)
圖1為根據(jù)式(2)計(jì)算的條帶模式影像和TOPS模式影像方位向不同位置的點(diǎn)目標(biāo)干涉相位偏差。條帶模式影像由于方位向多普勒中心相同,其干涉相位偏差(紅線)是一致的,通常配準(zhǔn)精確度只要求達(dá)到0.1個(gè)像素以確保相干性不受影響;TOPS模式影像由于多普勒中心在方位向上的變化,導(dǎo)致其不同位置的干涉相位偏差并不相同。由于哨兵衛(wèi)星TOPS模式影像多普勒中心變化范圍一般在[-2500 +2500] Hz內(nèi),為確保其干涉圖在一組Burst中干涉相位偏差不超過3°,配準(zhǔn)精確度要求達(dá)到0.001像素,相當(dāng)于方位向零多普勒時(shí)間配準(zhǔn)真誤差小于2 μs。
由于地形起伏等原因,多項(xiàng)式難以準(zhǔn)確擬合出主輔影像幾何畸變的映射,無法滿足哨兵衛(wèi)星TOPS模式影像方位向0.001個(gè)像素的配準(zhǔn)精度需要。本文則利用幾何配準(zhǔn)誤差為系統(tǒng)誤差的特點(diǎn)采用哨兵衛(wèi)星精密軌道進(jìn)行初配準(zhǔn),然后根據(jù)影像信息采用增強(qiáng)譜分集方法進(jìn)一步做精配準(zhǔn)處理,以獲取正確的干涉圖。
1.1 基于精密軌道的TOPS影像幾何配準(zhǔn)
合成孔徑雷達(dá)是通過多普勒頻域算法將原始回波信號(hào)聚焦為斜距圖像,其多普勒參數(shù)估計(jì)不僅依賴于衛(wèi)星姿態(tài)傳感器,還可以根據(jù)雷達(dá)回波多普勒頻移特性進(jìn)行估計(jì),理論上其影像幾何比光學(xué)影像更為嚴(yán)格[14-15]。歐空局對(duì)哨兵衛(wèi)星TOPS模式影像聚焦時(shí)放棄“一步一?!本劢鼓P停遣捎门c衛(wèi)星運(yùn)動(dòng)狀態(tài)相符的“持續(xù)運(yùn)動(dòng)”模型,經(jīng)過傳感器延遲、地球固體潮、大氣延遲校正后,其影像距離向絕對(duì)定位精度為1.5±0.37 m,方位向絕對(duì)定位精度為0.53±0.74 m[16-17]。SAR影像幾何配準(zhǔn)類似于光學(xué)攝影測(cè)量中物方影像匹配方法,根據(jù)DEM和衛(wèi)星軌道信息基于零多普勒幾何定位原理計(jì)算每個(gè)像元的相對(duì)偏移量。幾何配準(zhǔn)不依賴影像的相干性,可靠性較高,隨著衛(wèi)星軌道精度的提高,該方法逐步受到關(guān)注[18-20]。衛(wèi)星軌道和DEM高程精度決定了幾何配準(zhǔn)精度[21-22]。衛(wèi)星軌道對(duì)幾何配準(zhǔn)的影響通常以基線誤差的形式呈現(xiàn),其對(duì)距離向幾何配準(zhǔn)誤差δ(p)的影響如式(3)所示
(3)
式中,Δr為距離向像素采樣空間;p為影像距離向坐標(biāo);δSr為軌道距離向誤差。歐空局提供的哨兵衛(wèi)星精密軌道(precise orbit)精度約為5 cm,軌道距離向誤差δSr導(dǎo)致的幾何配準(zhǔn)誤差遠(yuǎn)遠(yuǎn)小于影像像元空間[23]。軌道誤差對(duì)方位向配準(zhǔn)誤差δ(l)的關(guān)系如下
(4)
式中,Δaz為方位向像素采樣空間;l為影像方位向坐標(biāo);δSaz為軌道切向誤差。從式(4)可以看出,軌道誤差對(duì)方位向配準(zhǔn)的影響是由軌道切向誤差δSaz引起的。哨兵衛(wèi)星方位向配準(zhǔn)精度要求為0.001個(gè)像元,相當(dāng)于幾何配準(zhǔn)時(shí)軌道切向精度達(dá)到1 cm[24]。目前哨兵精密軌道無法滿足這一條件,需要根據(jù)影像信息進(jìn)一步優(yōu)化方位向幾何配準(zhǔn)結(jié)果[25]。
除衛(wèi)星軌道外,DEM高程精度與幾何配準(zhǔn)精度直接相關(guān)。DEM誤差εh與距離向幾何配準(zhǔn)誤差δ(p)關(guān)系如式(5)所示
(5)
式中,r0為影像零多普勒斜距;B⊥為干涉影像的垂直基線;θ為影像入射角。從式(5)中可以看出垂直基線B⊥越短,距離向配準(zhǔn)精確度越高。圖2為距離向配準(zhǔn)精度0.1個(gè)像素時(shí),垂直基線、雷達(dá)側(cè)視角與DEM精度需求之間的關(guān)系。哨兵衛(wèi)星垂直基線一般在150 m以內(nèi),入射角范圍為19°~45°,如圖中白色框線所示,此時(shí)DEM誤差導(dǎo)致的距離向配準(zhǔn)誤差約為εh·10-4個(gè)像素。哨兵衛(wèi)星距離向配準(zhǔn)精度要求為0.1個(gè)像素,SRTM標(biāo)稱絕對(duì)高程精度是16 m,DEM對(duì)距離向配準(zhǔn)誤差的影響可以忽略不計(jì)。
對(duì)于雷達(dá)衛(wèi)星重復(fù)干涉測(cè)量,其重復(fù)軌道之間并不是嚴(yán)格平行的,通常存在一定夾角。哨兵衛(wèi)星軌道為準(zhǔn)回歸軌道,衛(wèi)星運(yùn)行175周后,理論上衛(wèi)星應(yīng)回到初始位置,然而受到各種攝動(dòng)因素的影響,重復(fù)軌道之間必然存衛(wèi)星位置、姿態(tài)、速率的差異,以及軌道間的不平行。衛(wèi)星軌道夾角不但引起干涉影像間的多普勒頻譜差異、時(shí)間同步問題,采用衛(wèi)星軌道進(jìn)行幾何配準(zhǔn)時(shí)還與DEM共同影響方位向幾何配準(zhǔn)精度,DEM高程誤差與方位向幾何配準(zhǔn)誤差接近于線性關(guān)系[26]。DEM誤差εh對(duì)方位向幾何配準(zhǔn)誤差δ(l)的影響如式(6)所示
(6)
式中,αg為軌道夾角,圖3為根據(jù)式(6)得到的不同入射角和軌道夾角條件下所需的DEM高程精度。哨兵衛(wèi)星多數(shù)情況下軌道夾角小于0.001°,白色框線范圍內(nèi)為哨兵影像干涉所需的DEM高程精度,可以看出SRTM可以滿足大多數(shù)地區(qū)干涉處理的需要。一般條件下,DEM高程誤差對(duì)配準(zhǔn)誤差的影響理論上優(yōu)于0.000 1個(gè)像素量級(jí),通常不對(duì)配準(zhǔn)結(jié)果產(chǎn)生影響,方位向配準(zhǔn)誤差是由軌道切向誤差引起的[27]。
幾何配準(zhǔn)的優(yōu)勢(shì)是可以充分利用影像分辨率單元(resolution cell)估計(jì)影像殘余偏移量。衛(wèi)星軌道在一個(gè)時(shí)段內(nèi)呈系統(tǒng)誤差特性,軌道切向誤差導(dǎo)致的方位向幾何配準(zhǔn)誤差可視為系統(tǒng)誤差[28-29]。哨兵衛(wèi)星軌道基線控制在150 m的范圍內(nèi),幾何配準(zhǔn)后的殘余偏移量在每組Burst中通常認(rèn)為是恒定值[30]。方位向配準(zhǔn)偏移量可由式(7)表示
F(l,p)=f(l,p)+V
(7)
式中,F(xiàn)(l,p)為主輔影像間的方位向偏移量;f(l,p)為幾何配準(zhǔn)獲得的方位向偏移量;V為幾何配準(zhǔn)殘余偏移量;它們常以像素為單位。影像殘余偏移量通常是根據(jù)影像信息進(jìn)行估計(jì),理論上影像分辨率單元數(shù)量越大,殘余偏移量估計(jì)精度越高,幾何配準(zhǔn)可以充分利用影像中的分辨率單元以提高殘余偏移量估計(jì)精度。如果采用多項(xiàng)式進(jìn)行初配準(zhǔn),每個(gè)像素配準(zhǔn)殘余偏移量都不相同,則無法利用提高窗口的像元數(shù)來提高殘余偏移量估計(jì)精度。圖4為采用不同方法估計(jì)幾何配準(zhǔn)殘余偏移量的理論精度[31],其中橫軸為影像相干系數(shù),縱軸表示配準(zhǔn)誤差的標(biāo)準(zhǔn)差。根據(jù)哨兵影像參數(shù),影像方位向過采樣率約為1.3,距離向過采樣率約為1.5,實(shí)相關(guān)(incoherent cross correlation)、復(fù)相關(guān)(coherent cross correlation)、譜分集(spectral diversity)處理窗口(一組Burst)設(shè)置為1502×21 549像素,而增強(qiáng)譜分集窗口為Burst間的重疊區(qū)域,設(shè)置為100×21 549像素。圖中黑色橫線為千分一個(gè)像素界限,可以看出增強(qiáng)譜分集配準(zhǔn)精度最高,即使在相干性很低的時(shí)候,也滿足0.001個(gè)像素精度要求。對(duì)于哨兵衛(wèi)星,采用增強(qiáng)譜分集進(jìn)行殘余偏移量估計(jì)的前提是幾何配準(zhǔn)精度優(yōu)于0.06個(gè)像素,防止增強(qiáng)譜分集相位出現(xiàn)纏繞現(xiàn)象。哨兵衛(wèi)星精密軌道切向精度優(yōu)于5 cm,幾何配準(zhǔn)時(shí)理論精度達(dá)到0.004個(gè)像素(置信度為68.3%),理論上精密軌道條件下幾何配準(zhǔn)后直接利用增強(qiáng)譜分集算法估計(jì)方位向配準(zhǔn)誤差發(fā)生相位纏繞的概率極低,精密軌道條件幾何配準(zhǔn)后無需采用相關(guān)配準(zhǔn)進(jìn)行校正,提高了干涉處理效率。當(dāng)然其缺點(diǎn)是精密軌道相對(duì)于衛(wèi)星數(shù)據(jù)發(fā)布時(shí)間較晚,無法滿足準(zhǔn)實(shí)時(shí)處理的需要。
1.2 基于增強(qiáng)譜分集TOPS精配準(zhǔn)
增強(qiáng)譜分集與傳統(tǒng)譜分集原理上是一致的,不同的是增強(qiáng)譜分集利用Burst間的重疊區(qū)域來實(shí)現(xiàn)高精度配準(zhǔn)。方位向相鄰Burst分別視為增強(qiáng)譜分集上頻帶影像和下頻帶影像,其對(duì)應(yīng)干涉圖亦可視為上頻帶干涉圖和下頻帶干涉圖,最后利用重疊區(qū)域的上頻帶干涉圖與下頻帶干涉圖的相位差獲取每組Burst的偏移量,實(shí)施高精度配準(zhǔn)。增強(qiáng)譜分集提高了上下頻帶的多普勒中心頻率基線,其對(duì)配準(zhǔn)誤差更為敏感,配準(zhǔn)精度更高[32]。文獻(xiàn)[33—34]利用增強(qiáng)譜分集估計(jì)精度較高的優(yōu)勢(shì)實(shí)現(xiàn)對(duì)哨兵衛(wèi)星不同類型軌道精度進(jìn)行了評(píng)估。
合成孔徑雷達(dá)傳感器發(fā)射的是相干脈沖,每個(gè)脈沖的起始相位和發(fā)射時(shí)間都受到精密的控制,而且雷達(dá)接收機(jī)和解調(diào)器均具有較高的時(shí)間定標(biāo)精度。盡管哨兵衛(wèi)星TOPS模式子測(cè)繪帶中的每組Burst均獨(dú)立成像,但在同一軌道面內(nèi)進(jìn)行零多普勒聚焦處理,理論上Burst方位向重疊區(qū)域幾何投影和干涉圖完全相同。圖5為哨兵衛(wèi)星TOPS模式相鄰兩Burst間的頻譜,可以看出Burst間多普勒中心分布并不連續(xù),方位向的相位斜坡致使重疊區(qū)域的多普勒中心基線ΔfESD可視為譜分集中的上下頻帶多普勒中心差異,遠(yuǎn)大于影像的方位向帶寬Baz。如果配準(zhǔn)精度不夠高,則重疊區(qū)域的干涉相位將呈現(xiàn)跳變現(xiàn)象[35]。增強(qiáng)譜分集正是利用這一原理進(jìn)行方位向殘余偏移量的估計(jì)以及配準(zhǔn)精度的優(yōu)化。
增強(qiáng)譜分集相位φESD與方位向配準(zhǔn)時(shí)間誤差Δη的關(guān)系為
(8)
將式(8)轉(zhuǎn)換到以像素為單位的殘余偏移量V,可進(jìn)一步得到
(9)
式中,F(xiàn)a為方位向采樣頻率。理論上方位向重疊區(qū)域增強(qiáng)譜分集相位值是一致的,然而受到失相干等噪聲的影響,增強(qiáng)譜分集相位分布如圖6所示,每個(gè)像元相位值均存在差異,依靠單個(gè)像元難以獲得準(zhǔn)確的偏移量估計(jì),需要優(yōu)化增強(qiáng)譜分集相位值的估計(jì)。
理想點(diǎn)目標(biāo)在SAR成像過程中被認(rèn)為存在一致性的響應(yīng),具有無限大帶寬,方位向子孔徑分解后,點(diǎn)目標(biāo)在上下頻帶影像中的響應(yīng)依然存在相關(guān)性[36]。本文則根據(jù)上下頻帶干涉圖的相干性選取點(diǎn)目標(biāo)估計(jì)增強(qiáng)譜分集相位,提高增強(qiáng)譜分集相位估計(jì)的精度,估計(jì)方法如式(10)所示,當(dāng)然增強(qiáng)譜分集準(zhǔn)確估計(jì)的前提是其相位不能出現(xiàn)纏繞現(xiàn)象
(10)
本文基于DORIS開源軟件,完成哨兵衛(wèi)星TOPS模式影像干涉處理模塊,并對(duì)墨西哥地區(qū)影像進(jìn)行干涉配準(zhǔn)處理試驗(yàn),影像拍攝時(shí)間分別為2015年2月12日和24日。本文首先對(duì)影像幾何配準(zhǔn)處理,此時(shí)方位向重疊區(qū)域差分相位分布如圖7(a)所示,相位均值為0.49°。然后采用增強(qiáng)譜分集進(jìn)行精配準(zhǔn)處理,此時(shí)方位向重疊區(qū)域差分相位分布如圖7(b)所示,相位均值為0.02°。圖8是與圖7所對(duì)應(yīng)的幾何配準(zhǔn)和增強(qiáng)譜分集配準(zhǔn)后的差分干涉圖對(duì)比圖,試驗(yàn)結(jié)果表明增強(qiáng)譜分集配準(zhǔn)消除了重疊區(qū)域內(nèi)的相位跳變,驗(yàn)證了精密軌道條件下幾何配準(zhǔn)之后采用增強(qiáng)譜分集用于精配準(zhǔn)的可行性和有效性。試驗(yàn)也說明增強(qiáng)譜分集不但可以估計(jì)殘余偏移量,也可用于配準(zhǔn)精度的檢驗(yàn)。
增強(qiáng)譜分集方法易受到地表方位向形變、電離層等因素的影響,其中地表方位向形變是最常見的現(xiàn)象[37-38]。TOPS模式天線波束視角在Burst方位向重疊區(qū)域存在較大的差異,盡管影像為零多普勒幾何投影,當(dāng)?shù)乇戆l(fā)生方位向形變時(shí),其形變相位分量在上下頻帶干涉圖中出現(xiàn)明顯的差異。以模擬地震同震形變場(chǎng)為例,如圖9所示,方位向形變易導(dǎo)致哨兵衛(wèi)星TOPS影像中方位向相鄰兩組Burst地震干涉圖條紋無法閉合,而條帶模式影像不會(huì)存在這種現(xiàn)象。方位向形變?cè)诟缮鎴D產(chǎn)生的相位分量為
(11)
(12)
由于星載SAR成像時(shí)衛(wèi)星軌道和地表都是彎曲的,而且地球獨(dú)立于衛(wèi)星軌道不停自轉(zhuǎn),其成像幾何較為復(fù)雜,式(11)和式(12)中θsq、Vs參數(shù)與式(1)中所對(duì)應(yīng)的Vr、β參數(shù)意義并不等同。為了保證影像質(zhì)量,聚焦成像時(shí)仍需對(duì)上述參數(shù)進(jìn)行嚴(yán)格區(qū)分,但是這些參數(shù)相差不大,可忽略對(duì)形變監(jiān)測(cè)結(jié)果的影響。式(13)說明方位向形變將直接影響到增強(qiáng)譜分集結(jié)果,重疊區(qū)域的方位向形變信息不但造成Burst影像拼接時(shí)干涉相位的跳變,還將干擾增強(qiáng)譜分集結(jié)果,進(jìn)一步影響多項(xiàng)式擬合結(jié)果。圖10為新疆皮山地震哨兵衛(wèi)星干涉圖,由于形變區(qū)域位于相鄰影像中,本文對(duì)其子測(cè)繪帶進(jìn)行了拼接,共采用16組Burst殘余增強(qiáng)譜分集處理,其干涉條紋在Burst 9與Burst 10結(jié)合處出現(xiàn)了明顯的相位跳變,圖11即為與之相對(duì)應(yīng)的增強(qiáng)譜分集估計(jì)殘余偏移量結(jié)果,不難看出相位跳變處所對(duì)應(yīng)的增強(qiáng)譜分集估計(jì)結(jié)果也出現(xiàn)異常。試驗(yàn)說明地表方位向形變信號(hào)類似于殘余偏移量估計(jì)中的粗差,因此進(jìn)行多項(xiàng)式擬合時(shí)應(yīng)進(jìn)行粗差探測(cè),以避免方位向形變信號(hào)干擾配準(zhǔn)精度[39]。當(dāng)增強(qiáng)譜分集獲得殘余偏移量大于0.01個(gè)像素或者干涉結(jié)果出現(xiàn)跳變時(shí),則表明增強(qiáng)譜分集中可能存在擾動(dòng)信號(hào)。
哨兵衛(wèi)星TOPS模式干涉時(shí)要求達(dá)到0.001個(gè)像元的配準(zhǔn)精度。本文利用精密軌道和DEM進(jìn)行幾何配準(zhǔn),分析幾何配準(zhǔn)的適用條件,指出哨兵衛(wèi)星精密軌道條件下幾何配準(zhǔn)精度可以確保增強(qiáng)譜分集相位不會(huì)出現(xiàn)纏繞現(xiàn)象,適合增強(qiáng)譜分集處理以實(shí)現(xiàn)系統(tǒng)性的配準(zhǔn)偏差矯正。該方法具有運(yùn)算效率較高、配準(zhǔn)精度不依賴影像相干性的優(yōu)勢(shì)。本文在開源軟件DORIS的基礎(chǔ)上,增加增強(qiáng)譜分集配準(zhǔn)模塊,實(shí)現(xiàn)哨兵衛(wèi)星TOPS數(shù)據(jù)的干涉處理。試驗(yàn)結(jié)果表明,即使是在地形起伏較大的區(qū)域,依然能夠保持較高的精度。由于方位向形變等因素往往會(huì)干擾增強(qiáng)譜分集估計(jì)結(jié)果,本文通過粗差探測(cè)方法提高哨兵數(shù)據(jù)配準(zhǔn)的可靠性。
本文提供的哨兵衛(wèi)星干涉配準(zhǔn)方案僅在精密軌道條件下使用,然而哨兵衛(wèi)星精密軌道晚于影像數(shù)據(jù)發(fā)布,如果需要對(duì)哨兵數(shù)據(jù)進(jìn)行近實(shí)時(shí)處理,還需考慮其他方法優(yōu)化幾何配準(zhǔn)精度,以避免增強(qiáng)譜分集相位出現(xiàn)纏繞現(xiàn)象。
致謝:特別感謝荷蘭代爾夫特理工大學(xué)InSAR研究團(tuán)隊(duì)的幫助和密切合作。本文相關(guān)代碼已通過DORIS軟件發(fā)布。
圖1 配準(zhǔn)誤差引起的干涉圖相位偏差Fig.1 Phase bias of interferogram caused by mis-coregistration
圖2 DEM精度需求與軌道垂直基線、雷達(dá)側(cè)視角的關(guān)系(距離向偏移量精度要求為0.1個(gè)像元)Fig.2 Required DEM accuracy as a function of the incidence angle and perpendicular baseline (an accuracy of 0.1 pixel is required for range shift)
圖3 DEM精度需求與雷達(dá)側(cè)視角、軌道夾角的關(guān)系(方位向偏移量精度要求為0.001個(gè)像元)Fig.3 Required DEM accuracy as a function of the incidence angle and orbit crossing angle (an accuracy of 0.001 pixel is required for azimuth shift)
圖4 SAR影像不同配準(zhǔn)方法精度 Fig.4 Accuracy of different methods for SAR image coregistration
圖5 增強(qiáng)譜分集原理Fig.5 Principle of enhanced spectral diversity
圖6 增強(qiáng)譜分集相位分布Fig.6 Estimated phase from enhanced spectral diversity method
圖7 增強(qiáng)譜分集校正前后的重疊區(qū)域相位差分分布圖Fig.7 Histogram of phase difference of overlap areas before and after ESD coregistration correction
圖8 增強(qiáng)譜分集校正前后的干涉圖結(jié)果對(duì)比Fig.8 Comparison of interferograms before and after ESD coregistration correction
圖9 TOPS模式地震形變模擬差分干涉圖Fig.9 Simulation differential interferogram of seismic deformation in TOPS mode
圖10 新疆皮山地震同震形變差分干涉圖(降軌)Fig.10 The TOPS interferogram of Pishan earthquake in the Xinjiang Uygur Autonomous Region(descending)
圖11 新疆皮山地震增強(qiáng)譜分集估計(jì)方位向殘余偏移量結(jié)果Fig.11 Estimated residual offset in azimuth direction between consecutive bursts by ESD in Pishan earthquake in the Xinjiang Uygur Autonomous Region
[1] DE ZAN F, GUARNIERI A M M. TOPSAR: Terrain Observation by Progressive Scans[J]. IEEE Transactions on Geoscience and Remote Sensing, 2006, 44(9): 2352-2360.
[2] ESA. Sentinel-1: ESA's Radar Observatory Mission for GMES Operational Services[M]. [S.l.]: ESA, 2012.
[3] XU Wei, HUANG Pingping, WANG R, et al. Processing of Multichannel Sliding Spotlight and Tops Synthetic Aperture Radar Data[J]. IEEE Transactions on Geoscience and Remote Sensing, 2013, 51(8): 4417-4429.
[4] HUANG Pingping, LI Shenyang, XU Wei. Investigation on Full-aperture Multichannel Azimuth Data Processing in Tops[J]. IEEE Geoscience and Remote Sensing Letters, 2014, 11(4): 728-732.
[5] PRATS P, SCHEIBER R, MITTERMAYER J, et al. A SAR Processing Algorithm for Tops Imaging Mode Based on Extended Chirp Scaling[C]∥Proceedings of 2007 IEEE International Geoscience and Remote Sensing Symposium. Barcelona: IEEE, 2007: 148-151.
[6] DAVIDSON G, MANTLE V, BERNHARD R, et al. Implementation of Tops Mode on RadarSat-2 in Support of the Sentinel-1 Mission[C]∥ESA Living Planet Symposium. London: ESA, 2013: 1-22.
[7] META A, PRATS P, STEINBRECHER U, et al. First TOPSAR Interferometry Results with TerraSAR-X[C]∥Proceedings of the ESA FRINGE Workshop. Frascati, Italy: ESA, 2007.
[8] PRATS P, MAROTTI L, WOLLSTADT S, et al. TOPS Interferometry with TerraSAR-X[C]∥Proceedings of the 8th European Conference on Synthetic Aperture Radar. Aachen, Germany: IEEE, 2010: 1-4.
[10] WEGMüLER U, WERNER C, STROZZI T, et al. Sentinel-1 Support in the Gamma Software[C]∥Geoscience and Remote Sensing Symposium. Frascati, Italy: ESA, 2015: 23-27.
[11] SANDWELL D, MELLORS R, TONG, Xiaopeng, et al. GMTSAR: An InSAR Processing System Based on Generic Mapping Tools[EB/OL]. UC San Diego: Scripps Institution of Oceanography. http:∥escholarship.org/uc/item/8zq2c02m.
[12] BARA M, SCHEIBER R, BROQUETAS A, et al. Interferometric SAR Signal Analysis in the Presence of Squint[J]. IEEE Transactions on Geoscience and Remote Sensing, 2000, 38(5): 2164-2178.
[13] Miranda N. Definition of the TOPS SLC Deramping Function for Products Generated by the S-1 IPF[R/OL]. https:∥earth.esa.int/documents/247904/165344%202/Sentinel-1-TOPS-SLC_Deramping. Paris, France: European Space Agency.
[14] 程春泉, 張繼賢, 黃國滿, 等. 考慮定向參數(shù)精度信息的TerraSAR-X和SPOT-5HRS影像RFM聯(lián)合定位[J]. 測(cè)繪學(xué)報(bào), 2017, 46(2): 179-187. DOI: 10.11947/j.AGCS.2017.20160138.
CHENG Chunquan, ZHANG Jixian, HUANG Guoman, et al. Combined Positioning of TerraSAR-X and SPOT-5 HRS Images with RFM Considering Accuracy Information of Orientation Parameters[J]. Acta Geodaetica et Cartographica Sinica, 2017, 46(2): 179-187. DOI: 10.11947/j.AGCS.2017.20160138.
[15] CURLANDER J C, MCDONOUGH R N. Synthetic Aperture Radar: Systems and Signal Processing[M]. New York: Wiley, 1991.
[16] SCHUBERT A, SMALL D, MEIER E, et al. Spaceborne SAR Product Geolocation Accuracy: A Sentinel-1 Update[C]∥Proceedings of 2014 IEEE Geoscience and Remote Sensing Symposium. Quebec City: IEEE, 2014: 2675-2678.
[17] GEUDTNER D,YAGUE-MARTINEZ N,PRATS P,et al. Sentinel-1 InSAR Performance: Results from the Sentinel-1A In-orbit Commissioning[C]∥Proceedings of the ESA FRINGE Workshop. Frascati, Italy: ESA, 2015.
[18] ARIKAN M, VAN LEIJEN F, LIU Guang, et al. Improved Image Alignment under the Influence of Elevation[C]∥Proceedings of the ESA FRINGE Workshop. Frascati, Italy: ESA, 2007.
[19] PRATS-IRAOLA P, SCHEIBER R, MAROTTI L, et al. TOPS Interferometry with TerraSAR-X[J]. IEEE Transactions on Geoscience and Remote Sensing, 2012, 50(8): 3179-3188.
[20] YAGUE-MARTINEZ N, EINEDER M, BRCIC R, et al. TanDEM-X Mission: SAR Image Coregistration Aspects[C]∥Proceedings of the 8th European Conference on Synthetic Aperture Radar. Aachen, Germany: IEEE, 2010: 1-4.
[21] SANSOSTI E, BERARDINO P, MANUNTA M, et al. Geometrical SAR Image Registration[J]. IEEE Transactions on Geoscience and Remote Sensing, 2006, 44(10): 2861-2870.
[22] PRATS-IRAOLA P, RODRIGUEZ-CASSOLA M, DE ZAN F, et al. Role of the Orbital Tube in Interferometric Spaceborne SAR Missions[J]. IEEE Geoscience and Remote Sensing Letters, 2015, 12(7): 1486-1490.
[24] BAMLER R, EINEDER M, MONTI G A, et al. Session Summaries:InSAR Theory Session[C]∥Proceedings of the ESA FRINGE Workshop.Frascati:ESA Publication, 2015.
[25] POTIN P, ROSICH B, GRIMONT P, et al. Sentinel-1 Mission Status[C]∥Proceedings of EUSAR 2016: 11th European Conference on Synthetic Aperture Radar. Hamburg: IEEE, 2016: 1-6.
[27] PRATS P,RODRIGUEZ-CASSOLA M,LEZ-DEKKER P, et al.Considerations of the Orbital Tube for Interferometric Applications[C]∥Proceedings of the ESA FRINGE Workshop. Frascati, Italy: European Space Agency, 2015: 23-27.
[28] GRANDIN R.Interferometric Processing of SLC Sentinel-1 TOPS Data[C]∥Fringe. Frascati, Italy: ESA, 2015.
[29] XU Bing, LI Zhiwei, FENG Guangcai, et al. Continent-Wide 2-D Co-seismic Deformation of the 2015 Mw 8.3 Illapel, Chile Earthquake Derived from Sentinel-1A Data: Correction of Azimuth Co-registration Error[J]. Remote Sensing, 2016, 8(5): 376.
[30] GEUDTNER D, PRATS P, YAGUE-MARTINEZ N, et al. Sentinel-1 SAR Interferometry Performance Verification[C]∥Proceedings of EUSAR 2016: 11th European Conference on Synthetic Aperture Radar. Hamburg, Germany: IEEE, 2016: 1-4.
[31] BAMLER R, EINEDER M. Accuracy of Differential Shift Estimation by Correlation and Split:Bandwidth Interferometry for Wideband and Delta-K SAR Systems[J]. IEEE Geoscience and Remote Sensing Letters, 2005, 2(2): 151-155.
[32] YAGUE-MARTINEZ N, BALSS U, BREIT H, et al. Operational Stacking OF TerraSAR-X ScanSAR and Tops Data[C]∥Proceedings of 2013 IEEE International Geoscience and Remote Sensing Symposium. Melbourne: IEEE, 2013: 888-890.
[33] MANCON S, TEBALDINI S, GUARNIERI A M, et al. Orbit Accuracy Estimation by Multi-squint Phase: First Sentinel-1 Results[C]∥Proceedings of 2015 IEEE International Geoscience and Remote Sensing Symposium. Milan: IEEE, 2015: 1276-1279.
[34] MANCON S, GUARNIERI A M, GIUDICI D, et al. On the Phase Calibration by Multisquint Analysis in TOPSAR and Strip Map Interferometry[J]. IEEE Transactions on Geoscience and Remote Sensing, 2017, 55(1): 134-147.
[35] RODRIGUEZ-CASSOLA M, PRATS-IRAOLA P, DE ZAN F, et al. Doppler-related Distortions in TOPS SAR Images[J]. IEEE Transactions on Geoscience and Remote Sensing, 2015, 53(1): 25-35.
[36] SCHNEIDER R Z,PAPATHANASSIOU K P,HAJNSEK I,et al. Polarimetric and Interferometric Characterization of Coherent Scatterers in Urban Areas[J]. IEEE Transactions on Geoscience and Remote Sensing, 2006, 44(4): 971-984.
[37] SCHEIBER R, JGER M, PRATS-IRAOLA P, et al. Speckle Tracking and Interferometric Processing of TerraSAR-X TOPS Data for Mapping Nonstationary Scenarios[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2015, 8(4): 1709-1720.
[38] PRATS-IRAOLA P, NANNINI M, SCHEIBER R, et al. Sentinel-1 Assessment of the Interferometric Wide-swath Mode[C]∥Proceedings of 2015 IEEE International Geoscience and Remote Sensing Symposium. Milan: IEEE, 2015: 5247-5251.
[39] TEUNISSEN P J G, SIMONS D G, TIBERIUS C C J M. Probability and Observation Theory[M]. Amsterdam: Delft University of Technology, 2005.
A Study of Sentinel-1 TOPS Mode Co-registration
WU Wenhao1,2,ZHOU Zhiwei2,LI Tao2,LONG Sichun1
1. Hunan Province Key Laboratory of Coal Resources Clean-utilization and Mine Environment Protection, Hunan University of Science and Technology, Xiangtan 411201, China; 2. GNSS Research Center of Wuhan University, Wuhan 430079, China
The TOPS imaging mode is the default operation mode of Sentinel-1 satellite, an overall azimuth coregistration accuracy of better than 0.001 pixels is required for this configuration in order to achieve the interferometric phase bias is less than three degrees. Based on the high accuracy of Sentinel-1 orbital information, this paper demonstrates a sufficiently high coregistration accuracy is able to be achieved by geometrical coregistration algorithm, which satisfies the pre-requirement for estimating the residual coregistration error by enhanced spectral diversity method. Based on DORIS open source software, a coregistration approach for Sentinel-A TOPS mode is implemented by an algorithm of geometric prediction of the shifts using precise orbit information, and an improvement on coregistration processing is achieved by enhanced spectral diversity. An interferometric phase jump between two Bursts could be caused by surface deformation in azimuth direction, and introduce a disturbance when estimating the residual coregistration error by enhanced spectral diversity method. In this paper, based on related theory analysis and real data processing test, an outlier detection estimation method is proposed to reduce the disturbance of enhanced spectral diversity coregistration method, which is caused by phase jumps.
interferometric synthetic aperture radar; terrain observation by progressive scans; Sentinel-1; geometric coregistration; enhanced spectral diversity
The National Natural Science Foundation of China (Nos. 41474014; 41674032)
WU Wenhao(1987—),male,PhD, lecturer, majors in InSAR data processing.
LONG Sichun
P228
A
1001-1595(2017)09-1156-09
國家自然科學(xué)基金(41474014;41674032)
(責(zé)任編輯:陳品馨)
2016-07-12
修回日期: 2017-07-20
吳文豪(1987—),男,博士,講師,研究方向?yàn)楹铣煽讖嚼走_(dá)干涉處理。
E-mail: wuwh@whu.edu.cn
龍四春
E-mail: sclong@hnust.edu.cn