趙?苓,費(fèi)?融,李醒飛,拓衛(wèi)曉,邢偉達(dá)
基于MHD角速度傳感器的面陣相機(jī)模糊圖像復(fù)原
趙?苓1, 2,費(fèi)?融1,李醒飛1, 2,拓衛(wèi)曉1, 2,邢偉達(dá)1
(1. 天津大學(xué)精密測(cè)試技術(shù)及儀器國(guó)家重點(diǎn)實(shí)驗(yàn)室,天津 300072;2. 深海技術(shù)科學(xué)太湖實(shí)驗(yàn)室,無(wú)錫 214123)
在軌衛(wèi)星平臺(tái)上普遍存在著寬頻帶、低幅值的微角振動(dòng),隨著對(duì)航天相機(jī)探測(cè)能力和遙感圖像空間分辨率要求的不斷提高,微角振動(dòng)對(duì)星載光學(xué)系統(tǒng)成像質(zhì)量的影響更加突出.針對(duì)高頻微角振動(dòng)造成的圖像模糊問(wèn)題,提出使用磁流體動(dòng)力學(xué)(magnetohydrodynamics,MHD)角速度傳感器測(cè)量面陣相機(jī)受到的微角振動(dòng),并根據(jù)面陣相機(jī)角運(yùn)動(dòng)信息構(gòu)建點(diǎn)擴(kuò)散函數(shù),進(jìn)一步采用基于全變差正則化的圖像復(fù)原算法實(shí)現(xiàn)圖像復(fù)原,從而提高圖像質(zhì)量.首先,在分析航天相機(jī)受到微角振動(dòng)干擾產(chǎn)生像移原因的基礎(chǔ)上,建立像移模型,并根據(jù)像點(diǎn)在焦平面上的移動(dòng)軌跡構(gòu)建點(diǎn)擴(kuò)散函數(shù).其次,搭建實(shí)驗(yàn)系統(tǒng)模擬航天相機(jī)在微角振動(dòng)環(huán)境下的成像過(guò)程,在垂直于光軸的方向上施加單頻正弦角振動(dòng)干擾,并使用MHD角速度傳感器進(jìn)行探測(cè).最后,利用基于全變差正則化的圖像復(fù)原優(yōu)化算法對(duì)模糊圖像進(jìn)行復(fù)原,并基于圖像的調(diào)制傳遞函數(shù)對(duì)復(fù)原圖像質(zhì)量進(jìn)行評(píng)價(jià).實(shí)驗(yàn)結(jié)果表明:當(dāng)微角振動(dòng)頻率在20~ 300Hz內(nèi),且產(chǎn)生的像移在13個(gè)像元以下時(shí),復(fù)原圖像的調(diào)制傳遞函數(shù)積分面積可達(dá)到靜態(tài)圖像積分面積的90%以上,圖像細(xì)節(jié)清晰,對(duì)比度得到明顯提升.對(duì)同一頻點(diǎn)進(jìn)行多次重復(fù)實(shí)驗(yàn),結(jié)果表明該方法具有良好的重復(fù)性.相關(guān)實(shí)驗(yàn)結(jié)果表明,基于MHD角速度傳感器的微角振動(dòng)測(cè)量方法和圖像復(fù)原算法相結(jié)合能夠顯著提高航天相機(jī)的成像質(zhì)量.
MHD角速度傳感器;微角振動(dòng);成像質(zhì)量;圖像復(fù)原
衛(wèi)星上動(dòng)量輪等機(jī)械部件的運(yùn)作、姿態(tài)的調(diào)整等都會(huì)產(chǎn)生角振動(dòng).相關(guān)實(shí)驗(yàn)數(shù)據(jù)表明,衛(wèi)星平臺(tái)上普遍存在頻率從0.1~1000Hz、幅值從亞微弧度到幾百微弧度的微角振動(dòng),且幅值會(huì)隨頻率升高逐漸減?小[1-2].隨著遙感技術(shù)的發(fā)展,遙感圖像的空間分辨率不斷提高,如國(guó)外IKONOS、WorldView、GeoEye-2、我國(guó)高分2號(hào)等衛(wèi)星的空間分辨率都達(dá)到了亞米級(jí).對(duì)分辨率要求的提高,使得衛(wèi)星微角振動(dòng)成為影響航天相機(jī)成像質(zhì)量的關(guān)鍵因素[3-4].
為了補(bǔ)償微角振動(dòng)造成的像質(zhì)下降,需要對(duì)圖像運(yùn)動(dòng)模糊機(jī)理和補(bǔ)償方法進(jìn)行深入研究.目前,運(yùn)動(dòng)模糊圖像的補(bǔ)償方式可分為實(shí)時(shí)穩(wěn)像和后期算法補(bǔ)償兩種.實(shí)時(shí)穩(wěn)像系統(tǒng)復(fù)雜度高,對(duì)角振動(dòng)測(cè)量和控制精度的要求都較高,并且會(huì)增加航天相機(jī)的體積和質(zhì)量.而后期算法補(bǔ)償雖然無(wú)法實(shí)現(xiàn)運(yùn)動(dòng)模糊的實(shí)時(shí)補(bǔ)償,但實(shí)現(xiàn)成本低,僅需要航天相機(jī)的姿態(tài)角數(shù)據(jù)便可完成補(bǔ)償,是補(bǔ)償圖像運(yùn)動(dòng)模糊的有效手段.
進(jìn)行模糊圖像后期算法補(bǔ)償?shù)那疤崾谦@得準(zhǔn)確的航天相機(jī)姿態(tài)角先驗(yàn)信息.目前,航天相機(jī)的微角振動(dòng)測(cè)量方法主要有陀螺儀[5]、線加速度計(jì)[6]和角運(yùn)動(dòng)傳感器等.其中,以動(dòng)力調(diào)諧陀螺儀、液浮陀螺儀為代表的傳統(tǒng)陀螺儀測(cè)量精度較高,但其工作帶寬一般小于20Hz,無(wú)法敏感高頻角擾動(dòng).以激光陀螺儀為代表的光學(xué)陀螺儀精度較高,可達(dá)微弧度量級(jí),帶寬高500Hz,但其質(zhì)量和體積較大[7-8].線加速度計(jì)需要多個(gè)加速度計(jì)進(jìn)行組合測(cè)量,這對(duì)加速度計(jì)的測(cè)量精度和安裝位置精度都提出了較高的要求,并且二次積分時(shí)會(huì)導(dǎo)致高頻誤差的累積.目前,運(yùn)用在衛(wèi)星上的角運(yùn)動(dòng)傳感器主要有基于流體旋轉(zhuǎn)差動(dòng)感應(yīng)式(fluid-rotor differential induction,F(xiàn)DI)的角位置傳感器和基于磁流體動(dòng)力學(xué)(magnetohydrodynamics,MHD)的角運(yùn)動(dòng)傳感器兩種[9].基于FDI原理的角位置傳感器內(nèi)部含有運(yùn)動(dòng)部件,傳遞函數(shù)復(fù)雜,測(cè)量精度依賴于內(nèi)部檢測(cè)元件的性能.相比FDI,基于MHD原理的角運(yùn)動(dòng)傳感器工作原理簡(jiǎn)單,兼具測(cè)量頻帶寬(1~1000Hz)、精度高、低交叉軸靈敏度、內(nèi)部無(wú)機(jī)械運(yùn)動(dòng)部件、可靠性高等優(yōu)勢(shì)[9-10],適用于航天相機(jī)姿態(tài)角的測(cè)量,已在航天器姿態(tài)測(cè)量、視軸穩(wěn)定及圖像的運(yùn)動(dòng)補(bǔ)償?shù)确矫嫒〉脧V泛應(yīng)用[11-12].
MHD角速度傳感器是由美國(guó)Bluehalo公司從20世紀(jì)80年代開(kāi)始研究的,所研發(fā)的型號(hào)中ARS-24精度最高,但體積和質(zhì)量都較大.針對(duì)微角振動(dòng)測(cè)量和抑制系統(tǒng)的需求,先后研發(fā)的型號(hào)有ARS-12、14、15、16等,且已經(jīng)在國(guó)外圖像補(bǔ)償領(lǐng)域取得應(yīng)用.2006年,日本發(fā)射的“先進(jìn)陸地觀測(cè)衛(wèi)星”使用了3個(gè)Bluehalo公司研制的ARS-12G型MHD角速度傳感器正交安裝,并與低頻陀螺儀的輸出數(shù)據(jù)進(jìn)行融合,解決了僅依靠陀螺儀低頻數(shù)據(jù)無(wú)法補(bǔ)償?shù)膱D像條紋噪聲問(wèn)題[13-14].國(guó)內(nèi)自2011年起,先后有蘭州空間技術(shù)物理研究所、天津大學(xué)和上海交通大學(xué)開(kāi)展了MHD角速度傳感器基礎(chǔ)理論以及相關(guān)型號(hào)的研發(fā)[8,15-16].受到國(guó)內(nèi)發(fā)展水平的限制,利用MHD角速度傳感器測(cè)量航天相機(jī)角振動(dòng)并進(jìn)行運(yùn)動(dòng)模糊圖像復(fù)原的研究尚鮮見(jiàn)報(bào)道.
本文首先推導(dǎo)了面陣相機(jī)在微角振動(dòng)環(huán)境下成像時(shí)的像移模型,給出了基于MHD角速度傳感器測(cè)量數(shù)據(jù)的點(diǎn)擴(kuò)散函數(shù)(point spread function,PSF)構(gòu)建方法;然后采用課題組自行研制的MHD角速度傳感器搭建了微角振動(dòng)環(huán)境下的模擬成像系統(tǒng),在垂直于光軸方向加載微角振動(dòng)干擾,并采集模糊圖像.最后采用調(diào)制傳遞函數(shù)(modulation transfer function,MTF)對(duì)基于全變差(total variation,TV)正則化圖像復(fù)原算法的復(fù)原結(jié)果進(jìn)行評(píng)價(jià).
本文所用MHD角速度傳感器實(shí)物照片如圖1(b)所示,其測(cè)量帶寬為2~1000Hz,質(zhì)量小于200g,頻率響應(yīng)曲線如圖2所示.
圖1?MHD角速度傳感器工作原理和實(shí)物照片
將面陣相機(jī)的成像過(guò)程看作小孔成像,以透鏡組的光心為坐標(biāo)原點(diǎn),建立相機(jī)坐標(biāo)系,如圖3所示,其中軸與相機(jī)光軸重合.相機(jī)繞軸轉(zhuǎn)動(dòng)時(shí),像點(diǎn)在焦平面上做,相同角振幅下,對(duì)像質(zhì)的影響遠(yuǎn)遠(yuǎn)小于繞另外兩個(gè)軸旋轉(zhuǎn)時(shí)產(chǎn)生的影響. 繞軸的角振動(dòng)會(huì)產(chǎn)生水平方向上的像移,繞軸的角振動(dòng)則會(huì)導(dǎo)致豎直方向上的像移.考慮到兩者的分析方法相同,本文僅對(duì)相機(jī)繞軸轉(zhuǎn)動(dòng)時(shí)引起的像移進(jìn)行分析和補(bǔ)償.
圖2?MHD角速度傳感器頻率響應(yīng)曲線
圖3?以光心為原點(diǎn)建立的坐標(biāo)系
根據(jù)三角函數(shù)公式可得
由式(2)可以看出,角振動(dòng)引起的像移不僅與轉(zhuǎn)動(dòng)角度和焦距有關(guān),還與物點(diǎn)和光軸夾角有關(guān),圖像邊緣受到的影響比中心更大.
圖5?通過(guò)統(tǒng)計(jì)采樣間隔構(gòu)建PSF
本節(jié)通過(guò)設(shè)計(jì)實(shí)驗(yàn),模擬面陣相機(jī)受到微角振動(dòng)干擾時(shí)的成像過(guò)程,并對(duì)拍攝到的模糊圖像進(jìn)行復(fù)原.實(shí)驗(yàn)流程如圖6所示,角振動(dòng)源受到正弦激勵(lì)后,會(huì)產(chǎn)生一個(gè)正弦角振動(dòng),成像系統(tǒng)受到角振動(dòng)的影響,成像模糊.使用MHD角速度傳感器測(cè)量成像系統(tǒng)所受角振動(dòng),積分得到角位移,并進(jìn)一步計(jì)算像移軌跡,構(gòu)建PSF.最后使用圖像復(fù)原算法進(jìn)行模糊圖像復(fù)原,并對(duì)復(fù)原圖像和模糊圖像進(jìn)行MTF分析,評(píng)價(jià)補(bǔ)償效果.
圖6?實(shí)驗(yàn)方案流程
實(shí)驗(yàn)系統(tǒng)的模型如圖7(a)所示,實(shí)物圖如圖7(b)所示.該實(shí)驗(yàn)系統(tǒng)中,成像設(shè)備為面陣工業(yè)相機(jī),相關(guān)參數(shù)列于表1.
圖7?實(shí)驗(yàn)系統(tǒng)
表1?面陣相機(jī)相關(guān)參數(shù)
Tab.1?Parameters of the area scan camera
實(shí)驗(yàn)中選用的角振動(dòng)激勵(lì)源是105-AVT單軸高頻角振動(dòng)系統(tǒng)(簡(jiǎn)稱角振動(dòng)臺(tái)),無(wú)負(fù)載時(shí),該角振動(dòng)臺(tái)可以提供5~2500Hz、轉(zhuǎn)角范圍±5°的角振動(dòng).面陣相機(jī)和MHD角速度傳感器剛性固連在角振動(dòng)臺(tái)上,當(dāng)上位機(jī)給角振動(dòng)臺(tái)一個(gè)單頻正弦激勵(lì)信號(hào)時(shí),角振動(dòng)臺(tái)發(fā)生轉(zhuǎn)動(dòng),同時(shí)MHD角速度傳感器持續(xù)輸出相機(jī)角速度,并由16位高性能NI數(shù)據(jù)采集卡進(jìn)行采集后,傳送至上位機(jī).系統(tǒng)穩(wěn)定運(yùn)行后,相機(jī)開(kāi)始曝光,對(duì)靶標(biāo)成像,得到模糊靶標(biāo)圖像,并傳輸至上位機(jī),實(shí)驗(yàn)靶標(biāo)采用ISO 12233標(biāo)準(zhǔn)分辨率測(cè)試卡.
為了能準(zhǔn)確得到曝光的起始時(shí)刻,相機(jī)采用外觸發(fā)采集模式,觸發(fā)信號(hào)由上位機(jī)給出.為了使靶標(biāo)在不同曝光時(shí)長(zhǎng)內(nèi)都能成明亮的像,在靶標(biāo)后面增加一個(gè)亮度可調(diào)的面陣LED光源進(jìn)行照明.同時(shí)為了減少外界振動(dòng)對(duì)實(shí)驗(yàn)的干擾,以上設(shè)備全部安裝在氣浮隔振平臺(tái)上.
2.2.1?TV正則化算法
圖像復(fù)原問(wèn)題是數(shù)字圖像處理領(lǐng)域的重要研究課題.1992年,Rudin等[17]將Tikhonov正則化中的圖像2范數(shù)約束更改為圖像梯度和約束,提出了著名的ROF模型用于圖像去噪.該算法能夠在去除噪聲的同時(shí)保護(hù)圖像邊緣細(xì)節(jié)[18],后被大量用于模糊圖像的復(fù)原中.
TV正則化復(fù)原優(yōu)化算法將圖像復(fù)原問(wèn)題轉(zhuǎn)換為求解無(wú)約束最優(yōu)化問(wèn)題,即
為了快速求解該最優(yōu)化問(wèn)題,已有諸多學(xué)者提出了多種算法,如分裂Bregman算法[19]、快速TV(fast total variation deconvolution,F(xiàn)TVd)算法[20]等,這些算法都致力于減少計(jì)算復(fù)雜度,提高運(yùn)行速度.其中分裂Bregman算法[19]通過(guò)拆分算子將復(fù)雜的優(yōu)化問(wèn)題拆分為多個(gè)子問(wèn)題,然后應(yīng)用Bregman迭代法求解多個(gè)最優(yōu)化子問(wèn)題.
式(9)可分解為對(duì)以下2個(gè)子問(wèn)題進(jìn)行求解.
根據(jù)二維收縮定理可得出式(11)的閉合解為
分裂Bergman算法通過(guò)以下步驟迭代求解最優(yōu)化問(wèn)題.
2.2.2?圖像質(zhì)量評(píng)價(jià)方法
圖像的MTF可以用來(lái)評(píng)價(jià)光學(xué)系統(tǒng)的成像質(zhì)量,它能反映出成像系統(tǒng)對(duì)不同空間頻率目標(biāo)的解像力[21-23].傳統(tǒng)的MTF測(cè)量方法需要分別測(cè)量成像系統(tǒng)對(duì)不同空間頻率黑白線條的解像力,過(guò)程復(fù)雜,花費(fèi)時(shí)間長(zhǎng).而傾斜刃邊(slanted edge,SE)法只需要選取一張圖像上的一個(gè)傾斜邊緣,通過(guò)過(guò)采樣細(xì)化刃邊的灰度值變化,得到邊緣擴(kuò)散函數(shù)(edge spread function,ESF),將ESF微分可得到線擴(kuò)散函數(shù)(line spread function,LSF),LSF的傅里葉變換即為完整的MTF曲線,過(guò)程如圖8所示.ISO 12233標(biāo)準(zhǔn)分辨率測(cè)試卡上提供了一個(gè)小角度的傾斜邊緣用于MTF?檢測(cè).
圖8?傾斜刃邊法獲取MTF過(guò)程
為了驗(yàn)證TV正則化圖像復(fù)原算法對(duì)微角振動(dòng)導(dǎo)致的模糊圖像的復(fù)原效果,選擇維納濾波算法和RL(Richardson-Lucy)算法與TV正則化算法進(jìn)行比較,這兩種經(jīng)典的非盲復(fù)原算法分別在頻域和空域取得了理想的圖像復(fù)原結(jié)果.給角振動(dòng)臺(tái)施加一個(gè)32Hz的正弦激勵(lì),產(chǎn)生10個(gè)像元的像移,采集模糊圖像.使用3種方法對(duì)同一模糊圖像進(jìn)行復(fù)原,噪聲方差取0.001,RL算法循環(huán)20次,復(fù)原結(jié)果如圖9所示.從圖9中可以看出:維納濾波復(fù)原圖像可分辨出文字和數(shù)字輪廓,但噪點(diǎn)被放大,圖像顆粒感嚴(yán)重;而RL算法出現(xiàn)明顯振鈴現(xiàn)象;TV正則化算法能夠抑制噪聲,平坦區(qū)域未出現(xiàn)明顯振鈴,數(shù)字輪廓和文字恢復(fù)較好.綜上,基于TV正則化的圖像復(fù)原優(yōu)化算法在本實(shí)驗(yàn)中的表現(xiàn)優(yōu)于其他兩種算法,故實(shí)驗(yàn)中選擇TV正則化算法作為模糊圖像復(fù)原的方法.
限于角振動(dòng)臺(tái)的帶負(fù)載能力,僅對(duì)20~300Hz的角振動(dòng)進(jìn)行模擬.給高頻角振動(dòng)臺(tái)施加20~300Hz的單頻正弦激勵(lì),采集不同頻率下的模糊圖像,并使用TV正則化優(yōu)化算法進(jìn)行圖像復(fù)原.限于篇幅,文中僅給出角振動(dòng)頻率為40Hz、110Hz、233Hz以及267Hz的圖像復(fù)原結(jié)果,實(shí)驗(yàn)參數(shù)列于表2,模糊圖像和復(fù)原圖像如圖10~圖13所示.
圖9?32Hz正弦激勵(lì)下的不同算法圖像復(fù)原結(jié)果
表2?40 Hz、110 Hz、233 Hz、267 Hz角振動(dòng)實(shí)驗(yàn)條件
Tab.2 Experimental conditions at 40,110,233,and 267Hz angular vibrations
從視覺(jué)上判斷,40Hz、110Hz、233Hz和267Hz 4種頻率的正弦角振動(dòng)都使模糊圖像出現(xiàn)了“重影”現(xiàn)象,像質(zhì)退化嚴(yán)重.利用TV復(fù)原算法復(fù)原后,重影消失,邊緣輪廓清晰分明,圖像的對(duì)比度和銳度都有了明顯的提升,圖像整體質(zhì)量得到明顯改善.
為了更進(jìn)一步評(píng)價(jià)圖像的復(fù)原質(zhì)量,使用傾斜刃邊法獲得40Hz、110Hz、233Hz和267Hz頻率下靜態(tài)圖像、模糊圖像和復(fù)原圖像的MTF曲線,如圖14所示.從圖14中可以看出,復(fù)原圖像的MTF曲線整體高于模糊圖像的MTF曲線,且與靜態(tài)圖像的MTF曲線接近,特別在低頻部分,復(fù)原圖像的 MTF 曲線甚至高于靜態(tài)圖像,圖像輪廓得到了更好的復(fù)原,而高頻部分下降較快,截止頻率低于靜態(tài)圖像,但仍在可接受范圍內(nèi).
圖10?40Hz正弦激勵(lì)下的圖像復(fù)原結(jié)果(a=1.00)
圖11?110Hz正弦激勵(lì)下的圖像復(fù)原結(jié)果(a=1.38)
圖12?233Hz正弦激勵(lì)下的圖像復(fù)原結(jié)果(a=1.15)
圖13?267Hz正弦激勵(lì)下的圖像復(fù)原結(jié)果(a=1.38)
在MTF分析中,圖像MTF曲線與坐標(biāo)軸圍成的面積稱為調(diào)制傳遞函數(shù)面積(MTFA),是一個(gè)用來(lái)描述光學(xué)系統(tǒng)整體分辨能力的物理量.為了進(jìn)一步量化評(píng)價(jià)圖像質(zhì)量的改善程度,將上述MTF曲線進(jìn)行積分,積分范圍分別為0~150cycle/mm(奈奎斯特頻率),所得結(jié)果列于表3.
從表3中可以看出,40Hz、110Hz、233Hz、267Hz的角振動(dòng)產(chǎn)生的像移使成像系統(tǒng)的整體圖像質(zhì)量下降嚴(yán)重,圖像復(fù)原后,復(fù)原圖像整體質(zhì)量與靜態(tài)圖像的整體質(zhì)量相近,且高于模糊圖像的整體質(zhì)量2倍以上,復(fù)原效果良好,與直接觀察得出的結(jié)論一致.
圖14 40Hz、110Hz、233Hz和267Hz靜態(tài)圖像、模糊圖像和復(fù)原圖像的MTF曲線比較
表3 40Hz、110Hz、233Hz、267Hz角振動(dòng)下的MTFA
Tab.3 MTFA at 40,110,233,and 267Hz angular vibrations
表4?不同角振動(dòng)頻率下的實(shí)驗(yàn)結(jié)果
Tab.4?Experimental results at different angular vibration frequencies
對(duì)多個(gè)角振動(dòng)頻率下模糊圖像的復(fù)原結(jié)果進(jìn)行比較,可以看出,對(duì)20~300Hz角振動(dòng)導(dǎo)致的圖像模糊,圖像復(fù)原算法能進(jìn)行較好的復(fù)原,復(fù)原后的圖像整體質(zhì)量都有了一定的提升.
為驗(yàn)證提出的方法在各頻點(diǎn)的重復(fù)性,任意選取20~300Hz內(nèi)的10個(gè)頻點(diǎn),保持某一頻點(diǎn)的激勵(lì)幅值不變,進(jìn)行模糊成像實(shí)驗(yàn),重復(fù)采集該頻點(diǎn)的10張模糊圖像.分別對(duì)每一個(gè)頻點(diǎn)的10張模糊圖片進(jìn)行圖像復(fù)原,并對(duì)復(fù)原圖像、模糊圖像和靜態(tài)圖像進(jìn)行MTF分析,計(jì)算MTFA值.
限于篇幅,僅給出175Hz角振動(dòng)激勵(lì)條件下的重復(fù)實(shí)驗(yàn)結(jié)果.在175Hz頻率下,像點(diǎn)產(chǎn)生7個(gè)像元的像移,靜態(tài)圖像MTFA為117.63,其10次模糊圖像結(jié)果MTFA和復(fù)原圖像MTFA列于表5.實(shí)驗(yàn)過(guò)程中,曝光時(shí)長(zhǎng)確定,但曝光開(kāi)始時(shí)刻的角振動(dòng)相位無(wú)法控制,使得曝光時(shí)間內(nèi)測(cè)得的角振動(dòng)在非整周期部分存在隨機(jī)性,導(dǎo)致點(diǎn)擴(kuò)散函數(shù)稍有差別,但像移的最大幅值未發(fā)生變化.
從表5的實(shí)驗(yàn)結(jié)果可以看出,在同一角振幅下,盡管曝光開(kāi)始時(shí)刻角振動(dòng)相位的隨機(jī)性使模糊圖像的MTFA出現(xiàn)了波動(dòng),但復(fù)原圖像的MTFA達(dá)到了靜態(tài)圖像MTFA的90%以上,可以證明本文提出的方法在各頻點(diǎn)處重復(fù)實(shí)驗(yàn)時(shí),都能取得較好的效果.
表5?175 Hz角振動(dòng)重復(fù)性實(shí)驗(yàn)結(jié)果
隨著遙感技術(shù)的發(fā)展,對(duì)航天相機(jī)的空間分辨率要求越來(lái)越高,微角振動(dòng)引起的像移對(duì)像質(zhì)的影響不可忽視.本文提出使用MHD角速度傳感器測(cè)量航天相機(jī)的微角振動(dòng),并通過(guò)基于TV正則化的圖像復(fù)原算法完成模糊圖像復(fù)原,提高圖像質(zhì)量.通過(guò)搭建模擬成像系統(tǒng),模擬了面陣相機(jī)受到單頻正弦微角振動(dòng)時(shí)的成像過(guò)程,并使用MHD角速度傳感器對(duì)成像系統(tǒng)受到的微角振動(dòng)進(jìn)行探測(cè).圖像復(fù)原結(jié)果表明:20~300Hz實(shí)驗(yàn)頻率下,復(fù)原后圖像的MTFA都提高到了靜態(tài)圖像的90%以上,圖像質(zhì)量得到了明顯提高.基于MHD角速度傳感器探測(cè)相機(jī)角振動(dòng)的圖像復(fù)原方法能夠有效地補(bǔ)償微角振動(dòng)造成的圖像運(yùn)動(dòng)模糊.
限于實(shí)驗(yàn)條件,本文僅對(duì)繞相機(jī)軸的角振動(dòng)進(jìn)行了模擬,但實(shí)際成像時(shí),同時(shí)存在繞其他兩軸的角振動(dòng),同樣需要進(jìn)行測(cè)量與補(bǔ)償.為了更好地復(fù)原圖像,需要進(jìn)一步使用多個(gè)MHD角速度傳感器對(duì)繞相機(jī)3個(gè)軸的角振動(dòng)進(jìn)行測(cè)量,并構(gòu)建更準(zhǔn)確的PSF.目前使用的TV正則化算法雖然方法成熟,噪聲抑制能力強(qiáng),但運(yùn)算量大、運(yùn)行速度過(guò)慢,仍需尋找更快速的復(fù)原算法,實(shí)現(xiàn)實(shí)時(shí)圖像補(bǔ)償.并且在算法運(yùn)行過(guò)程中,尚有需要人工干預(yù)部分,應(yīng)進(jìn)一步完善文中提出的方法,構(gòu)建能自動(dòng)尋找最優(yōu)參數(shù)的自適應(yīng)復(fù)原算法.
[1] 孫?陽(yáng),薛棟林,鄭立功,等. 航天器平臺(tái)在軌振動(dòng)對(duì)遙感成像質(zhì)量的影響[J]. 光學(xué)技術(shù),2013,39(6):553-558.
Sun Yang,Xue Donglin,Zheng Ligong,et al. Satellite platform on-orbit vibration’s influence to image quality of remote sensing[J]. Optical Technique,2013,39(6):553-558(in Chinese).
[2] 霍紅慶,馬勉軍,李云鵬,等. 衛(wèi)星微角顫振高精度測(cè)量技術(shù)[J]. 傳感器與微系統(tǒng),2011,30(3):4-6,9.
Huo Hongqing,Ma Mianjun,Li Yunpeng,et al. High precision measurement technology of statellite’s angle micro vibration[J]. Transducer and Microsystem Technologies,2011,30(3):4-6,9(in Chinese).
[3] Lin Li,Li Wang,Li Yuan,et al. Micro-vibration suppression methods and key technologies for high-precision space optical instruments[J]. Acta Astronautica,2021,180:417-428.
[4] Roques S,Jahan L,Rougé B,et al. Satellite attitude instability effects on stereo images[C]//2004 IEEE International Conference on Acoustics,Speech,and Signal Processing. Piscataway,USA,2004:477-480.
[5] Fan Mo,Xie Junfeng,Liu Yuxuan. Vibration model of Ziyuan3 satellites considering frequency changing[J]. Journal of Vibration and Control,2020,26(17/18):1484-1502.
[6] Zhang Guo,Guan Zhichao. High-frequency attitude jitter correction for the Gaofen-9 satellite[J]. The Photogrammetric Record,2018,33(162):264-282.
[7] 邊志強(qiáng),張?健,牛升達(dá). 慣性技術(shù)在航天器中的應(yīng)用及展望[C]// 前沿技術(shù)與慣性技術(shù)的融合與應(yīng)用研討會(huì)文集. 大連,中國(guó),2021:428-438.
Bian Zhiqiang,Zhang Jian,Niu Shengda. The application and prospect of inertial technology in spacecraft[C]// Proceedings of the Symposium on Integration and Application of Frontier Technology and Inertial Technology. Dalian,China,2021:428-438(in Chinese).
[8] 吳文韜,李醒飛,劉?帆,等. 一種衛(wèi)星微角振動(dòng)高精度測(cè)量方法[J]. 航天器工程,2020,29(1):16-22.
Wu Wentao,Li Xingfei,Liu Fan,et al. A high-precision measurement method for satellite micro-angular vibration[J]. Spacecraft Engineering,2020,29(1):16-22(in Chinese).
[9] 霍紅慶,馬勉軍,李云鵬,等. MHD 角速度傳感器在航天任務(wù)中的應(yīng)用[J]. 真空與低溫,2011(2):114-120.
Huo Hongqing,Ma Mianjun,Li Yunpeng,et al. The application of MHD angular rate sensor in aerospace[J]. Vacuum and Cryogenics,2011(2):114-120(in Chinese).
[10] Laughlin D R. A magnetohydrodynamic angular motion sensor for anthropomorphic test device instrumentation[J]. SAE Transactions,1989:1648-1682.
[11] Laughlin D R,Smith D. Development and performance of an angular vibration sensor with 1—1000 Hz bandwidth and nanoradian level noise[C]// Free-Space Laser Communication and Laser Imaging. San Francisco,USA,2002:208-214.
[12] Lee D V,Velinsky S A. Analysis and experimental verification of a three-dimensional noncontacting angular motion sensor[J]. IEEE/ASME Transactions on Mechatronics,2007,12(6):612-622.
[13] Iwata T. Precision geolocation determination and pointing management for the advanced land observing satellite(ALOS)[C]// 2003 IEEE International Geoscience and Remote Sensing Symposium. Piscataway,USA,2003:1845-1848.
[14] Iwata T,Kawahara T,Muranaka N,et al. High-bandwidth attitude determination using jitter measurements and optimal filtering[C]// AIAA Guidance,Navigation,and Control Conference. Chicago,USA,2009:6311.
[15] 吳建銘,李建勛,張明舉. 角隨機(jī)顫振傳感器導(dǎo)電流體環(huán)數(shù)學(xué)模型推導(dǎo)[J]. 測(cè)控技術(shù),2018,37(7):69-73.
Wu Jianming,Li Jianxun,Zhang Mingju. Mathematical model of conducting fluid ring for freedom angular flutter speed sensor[J]. Measurement & Control Technology,2018,37(7):69-73(in Chinese).
[16] 周海佳,李得天. MHD傳感器敏感結(jié)構(gòu)地面溫度特性數(shù)值分析[J]. 西北工業(yè)大學(xué)學(xué)報(bào),2018,36(增):110-116.
Zhou Haijia,Li Detian. Numerical analysis of temperature sensitivity for sensitive mechanism in MHD sensor on ground[J]. Journal of Northwestern Polytechnical University,2018,36(Suppl):110-116(in Chinese).
[17] Rudin L I,Osher S,F(xiàn)atemi E. Nonlinear total variation based noise removal algorithms[J]. Physica D:Nonlinear Phenomena,1992,60(1/2/3/4):259-268.
[18] 張?彬,于欣妍,朱永貴,等. 圖像復(fù)原優(yōu)化算法[M]. 北京:國(guó)防工業(yè)出版社,2019.
Zhang Bin,Yu Xinyan,Zhu Yonggui,et al. Optimal Algorithm for Image Restoration[M]. Beijing:National Defense Industry Press,2019(in Chinese).
[19] Osher S,Burger M,Goldfarb D,et al. An iterative regularization method for total variation-based image restoration[J]. Multiscale Modeling & Simulation,2005,4(2):460-489.
[20] Wang Yilun,Yang Junfeng,Yin Wotao,et al. A new alternating minimization algorithm for total variation image reconstruction[J]. SIAM Journal on Imaging Sciences,2008,1(3):248-272.
[21] 孫崇尚,王?琦,丁亞林,等. 基于傾斜刃邊法航空相機(jī)像移的調(diào)制傳遞函數(shù)測(cè)量[J]. 光學(xué)學(xué)報(bào),2014,34(12):108-113.
Sun Chongshang,Wang Qi,Ding Yalin,et al. Modulation transfer function measurement of image motion of aerial camera based on slanted-edge method[J]. Acta Optical Sinica,2014,34(12):108-113(in Chinese).
[22] Estribeau M,Magnan P. Fast MTF measurement of CMOS imagers using ISO 12233 slanted-edge methodology[C]// Optical Systems Design. St Etienne,F(xiàn)rance,2003:243-252.
[23] Wong S,Jassemi-Zargani R. Predicting Image Quality of Surveillance Sensors[M]. Ottawa:Defence Research and Development Canada,2014.
Blurred Image Restoration of Area Scan Camera Based on Magnetohydrodynamics Angular Rate Sensor
Zhao Ling1, 2,F(xiàn)ei Rong1,Li Xingfei1, 2,Tuo Weixiao1, 2,Xing Weida1
(1. State Key Laboratory of Precision Measuring Technology and Instruments,Tianjin University,Tianjin 300072,China;2. Taihu Laboratory of Deep Sea Technology and Science,Wuxi 214123,China)
Wide-band and low-amplitude micro-angular vibrations are common in orbiting satellite platforms. With the continuously improving requirements for the detection capability of aerospace cameras and the spatial resolution of remote sensing images,micro-angular vibrations can have a serious impact on the imaging quality of a satellite optical system. To solve the problem of image blur caused by high-frequency micro-angular vibrations,a magnetohydrodynamics(MHD)angular rate sensor was proposed to measure the micro-angular vibrations of an area scan camera by constructing a point spread function based on the angular motion information of the area scan camera and using the image restoration algorithm based on total variation regularization for image restoration,thereby improving the image quality. First,the reasons for the image movement caused by the micro-angular vibration interference of the aerospace camera were analyzed,the image movement model was developed,and the point spread function was constructed based on the movement trajectory of the image point on the focal plane. Second,an experimental system was designed to simulate the imaging process of the aerospace camera in a micro-angular vibration environment,and the single-frequency sinusoidal micro-angular vibration disturbances were applied perpendicular to the optical axis and detected using the MHD angular rate sensor. Finally,the blurred image was restored using the total variation regularization image restoration algorithm,and the restoration image quality was evaluated based on the modulation transfer function. The experimental results demonstrated that when the micro-angular vibration frequency lay between 20 and 300Hz and the resulting image movement was less than 13 pixels,the modulation transfer function area of the restored image could be increased to more than 90% of the static image with clear image details and a significantly improved contrast. After repeating the experiments on the same micro-angular vibration frequency several times,the image restoration results confirmed a good repeatability of the proposed method.The relevant experimental results showed that the combination of the micro-angular vibration measurement method based on the MHD angular rate sensor and the image restoration algorithm can significantly improve the imaging quality of the aerospace camera.
magnetohydrodynamics(MHD) angular rate sensor;micro-angular vibration;image quality;image restoration
10.11784/tdxbz202205051
TP751.1
A
0493-2137(2023)08-0796-11
2022-05-27;
2022-10-02.
趙?苓(1984—??),女,博士,副教授,lingzhao84@126.com.Email:m_bigm@tju.edu.cn
李醒飛,lixf_mhd@163.com.
國(guó)家自然科學(xué)基金資助項(xiàng)目(61733012,62203322);深海技術(shù)科學(xué)太湖實(shí)驗(yàn)室“揭榜掛帥”項(xiàng)目(2022JBGS03001);中國(guó)博士后科學(xué)基金資助項(xiàng)目(2022M712372).
the National Natural Science Foundation of China(No.61733012,No.62203322),the Enlisting and Leading Program of the Taihu Laboratory of Deep Sea Technology and Science(No.2022JBGS03001),China Postdoctoral Science Foundation (No.2022M712372).
(責(zé)任編輯:孫立華)