PROPERTY AND REMOVAL OF JITTER IN BEIJING-1 SMALL SATELLITE PANCHROMATIC IMAGES

Transcription

1 PROPERTY AND REMOVAL OF JITTER IN BEIJING-1 SMALL SATELLITE PANCHROMATIC IMAGES Qiong Ran a,b,, Yaobin Chi a,b, Zhiyong Wang b a IRSA Institute of Remote Sensing Applications, Chinese Acaemy of Science,100101,Beiing,China b BLMIT Beiing Lanview Mapping Information Technology Co. Lt.,100096,Beiing,China Corresponing kateeme@163.com Commission I, WG-I/6 KEY WORDS: Beiing-1 Small Satellite, Panchromatic Image, Jitter Removal, Jitter Property ABSTRACT: Attitue of the Small satellite is prone to be affecte by external isturbing factors ue to its small inertia. As for Beiing-1 small satellite, itter is inuce an exhibite in the images taken while the outsie isturbance is intense. The itter is manifeste as agge lines an eges in the images, leaing to istortion of geometric property of the obects. In this paper, properties of the Beiing-1 small satellite itter are explore base on analyses of affections of attitue itter with iverse frequency an irections to the imaging process. Accoringly an image base itter removal metho is brought up, the propose metho removes itter effectively an can be introuce into practical prouction process. 1. INTRODUCTION Small satellite has gaine universal attention ue to remarkable avantages incluing flexibility an low cost, an has become a tren for eveloping of future satellites. However, owning to massive application of lightweight materials, inertia of the small satellite is greatly reuce, leaing to low attitue stability an frequent imaging errors. Beiing-1 small satellite was launche into the space as a member of the Disaster Monitoring Constellation in October The panchromatic sensor onboar has 4-meter resolution an 24-kilometer swath. Till now, images acquire from the satellite have been wiely use in agriculture, hyrology, city planning, relic protection, 2008 Olympic Game relate observation an many others. However, at times when the attitue is quite unstable, the satellite trembles an causes itter of the onboar camera, thus leaing to abnormality of the acquire images. Analyses have been one on exploring isciplinarian of the itter an its affections on the imaging process [1-5], however, few researches has been one on manifestation an removal of the ittere images. In this paper, manifestation an regulation of the itter in images is explore an research on removal of the itter is conucte. 2.1 Affections of Jitter with Different Frequency Jitter of any kin can be ecompose into a series of sine an cosine waves. Therefore the analyses below are base on consieration of simple sine waves. Quality reuction status of ittere image varies with the ratio of integration time t an T, where T represents perio of the itter. If t is larger than T, the itter is calle high frequency itter. One or more itter cycles are present uring the integration time, the high frequency itter expans iameter of the blur circle an causes blurring of the image (Figure 1(a)). If t is smaller than T, the itter is calle low frequency itter. In this case, the itter can be taken as linear movement ae to the imaging process. Motion of the blur circle causes egraation of the image. Displacement of every blur circle is ecie by the place of pixel integration time on the itter wave. The value fluctuates in certain ranges. If swing of the itter is rather large, so is the isplacement value, the pixel is likely to be eviate from the original place (Figure 1). Displacement value T 2. JITTER AFFECTIONS ON IMAGING PROCESS During the on-orbit operation of the satellite, any motivation of the components can cause itter response of the camera. When attitue of the satellite is isturbe, the attitue compensation moule is motivate. In this course, mechanical movement of the apparatus an activation of electronical system may lea to itter, even resonance, of the camera. In operation, the itter is stochastic, it can be of various frequency an irections an have iverse affections on the imaging process. Time (a) t 929

2 The International Archives of the Photogrammetry, Remote Sensing an Spatial Information Sciences. Vol. XXXVII. Part B1. Beiing 2008 Displacement value 3. PROPERTIES OF BEIJING-1 SMALL SATELLITE JITTER To facilitate exploration of itter characteristics, ata employe in our research is the Level 1 prouct with raiometric correction one. t Figure 1. Affection of ifferent frequency itter on imaging process, where (a) represents high frequency itter with t > T, represents low frequency itter when t < T 2.2 Affections of Jitter with Different Directions Satellite itter can be separate into three irections, along optic axis, along track an across track. Here image quality affections of itter in ifferent irections are iscusse. When the itter is along optical axis, change of the blur circle iameter can be expresse as Df = H f a Where f = focal length H = height of the orbit a = swing of the itter wave D = aperture of the camera T For itter along track, the equation is as follows. Time (1) In Beiing-1 small satellite images, itter performs as agge eges, perio of the ag is about 8 to 9 lines, the frequency approximates 200 HZ. Perio of the itter is not only larger than pixel integration time, but also larger than line integration time. It is low frequency itter. In visual observation, the agge eges mainly express the evastating across track itter, the itter can be assume as similar linear movement of pixels on the whole line, which causes isplacements of pixels to their original positions an lea to agge eges an linear obects shown in the image. As perio of itter is far larger than pixel integration time, isplacements of pixels on the same line can be taken as one value. To prove applicability of the above assumptions in itter removal process, the moel is testifie with Beiing-1 panchromatic image containing linear obects to see if itter performance in the image oes conform to the hypothesis. As itter shows clearly on linear obects an eges, the runway from an airrome is chosen to evaluate property of the itter. Properties obtaine from an airome are taken as reference for correction of the whole line. The process is as follows. First, the ittere runway is segmente using proper threshol, then DN centroi of the segmentation result is calculate as central position of the runway, after that, linear regression is one with x y value of the central position an resiual value is calculate. The runway can be restore to be linear by taking the resiual as isplacement value in horizontal shifting of relevant lines. With applying the isplacement value to shifting an sampling of whole lines of the image, itter that affiliates to all linear obects an eges are remove (Figure 2). Therefore it can be confirme that the assumption of the itter in Beiing-1 panchromatic image as uniform isplacement of pixels on lines is vali. f a = c+ vt = c+. /4. t (2) H T Where v = motion spee of the image obects. Jitter in along track irection mainly changes v value an isturbs the imaging process. (a) For the fact that in satellite imaging process, H is far larger than f, the change of is inapparent, along optical axis an along track itter has little impact on quality of the acquire images. Jitter across track causes iameter expansion of the blur circle, if swing of the itter is large to some extent, pointing obect of the CCD sensor will change an cause geometric errors in imaging process. Affection of across track itter can be evastating an is the most remarkable. Figure 2. Jitter evaluation with image of an airrome, where (a) represents the ittere image, represents the image after shifting an sampling of whole lines 930

3 The International Archives of the Photogrammetry, Remote Sensing an Spatial Information Sciences. Vol. XXXVII. Part B1. Beiing 2008 The achieve isplacement from the runway is shown in Figure 3, the isplacement value ranges from -1 to 1, fluctuates with the isciplinarian similar to that of a sinusoi, except that the frequency an swing is stochastic. Similar conclusion can also be rawn from experiments with other linear obects. Displacement value(pixel) Line (a) Figure 3. Displacement value of each line Uniform are isplacements of pixels on the same line, manifestation of itter affiliate to linear obects in ifferent irections varies. Hence line itter performance with image containing multi-irection linear obects is simulate as valiation (Figure 4). Consulting the figure above, perio of the itter is set to be 8.3 lines an swing is assume to be one pixel. Displacement value of line i can be calculate with equation 3. 2π i li ( ) 1.0*sin( ) 8.3 = (3) The simulation result explains the variational performance of itter on lines with iversifie irection. Jitter shows clearly in vertical an near-vertical lines. However, as irection of the line approaches horizontal, isplacement of pixels on the line is somewhat compensate by their tenency. Jitter graually faes away as lines fall horizontal, in horizontal lines itter can not even be seen visually. Performance of itter in the simulate image is the same as that in Beiing-1 panchromatic images. Figure 4. Jitter simulation with image containing multiirection linear obects, where (a) is the original image, is the image acquire with line itter ae the first grounp the secon group the thir group It can be conclue from the above analyses that itter in the Beiing-1 panchromatic image can be moelle as varie horizontal isplacements of the lines. Thus by proper horizontal shifting of the lines, itter in the image can surely be remove. The key factor lies in proper assessment of the line isplacement values. Correlation Line isplacement 4. JITTER REMOVAL The aacent lines in 4-meter resolution image can be consiere as highly correlate ue to the high sampling rate. If the lines are shifte, their correlation value egraes. This theory fits the correlation of lines in unittere Beiing-1 panchromatic image. Figure 5 shows variation of the line correlation values against isplacement changes between aacent unittere image lines. In the figure, three groups of lines an their correlation variation curves are rawn, showing egrae correlation with increasing isplacement while the absolute isplacement value is less than 40. It implies that the isplacement value can be evaluate by egraation conition of the correlation. Figure 5. Correlation between unittere aacent lines The itter in Beiing-1 panchromatic images range from -1 to 1. In this range, the correlation istribution is a one-peak curve. Figure 6 shows correlation variations of nine groups of ittere image lines, the peaks swing from -1 to 1, the isplacement values with correlation on the peak can be taken as line isplacement values that can be use to rectify the itter. Therefore, least square fitting of the curves can be use to get the isplacement values with the highest correlation. 931

4 The International Archives of the Photogrammetry, Remote Sensing an Spatial Information Sciences. Vol. XXXVII. Part B1. Beiing 2008 Correlation the first group the secon group the thir group the forth group the fifth group the sixth group the seventh group the eighth group the ninth group Line isplacement Figure 6. Relation of correlation to line isplacement = Jittere image ( x, < li (, ), li ( + x, ) > ) > 1 1 = x Least square curve fitting max f ( x) 1 For the nee to conuct small error curve fitting aroun the peak, the samples use shoul be in the peak vicinity. So, only five groups of correlation of line an -1 with isplacement ranging from -2 to 2 are use in the curve fitting. The groups are fitte to a quaratic curve. The fitte equation an the esire isplacement value are as follows. Horizontal sampling Deittere image ++ 2 f ( x) = a* x + b* x + c = x = b /2* a max f( x) To avoi curve fitting errors, the absolute value of acquire is constraine to be smaller than 1, conitione correction is performe uring estimation, the is substitute by isplacement of the preceing line -1 if the value is greater than 1. As the itter is moelle to be horizontal shifting cause by eviate observation, the image can be restore by position correction of each pixel, thus horizontal shifting is performe accoring to, the iea is shown in Figure 7, for simplification, the linear pixel unmixing is aopte, each pixel is consiere as linear combination of the original two aacent pixels. (4) Figure 8. Process of itter removal After processing, itter in the images is remove, shape of the eges are restore. The processe image regains geometric information of the obects an achieves better visual effect. The images before an after processing are shown in Figure 9. l L Figure 7. Horizontal sampling the sampling equation is as follows (a) Li (, ) = * li (, ) + (1 )* li ( + 1, ) (5) The whole process of itter removal is shown in Figure

5 The International Archives of the Photogrammetry, Remote Sensing an Spatial Information Sciences. Vol. XXXVII. Part B1. Beiing 2008 For few researches has been one on satellite itter an its manifestation in images, property of the itter in Beiing-1 small satellite has to be achieve via assumptions an experiments, thus leaing to the solution. The itter removal metho brought up here has been teste with various ata sets an is prove to be effective, besies, its aaptiveness in large image treatment an batch processing facilitates the application in inustrialize prouction process. REFERENCES [1]Algrain, Marcelo C., Determination of attitue itter in small satellites. Center for AeroSpace Information (CASI) NASA, USA Figure 9. Processe results, where (a) is the ittere image, an is the image after eitter processing [2]John Molnar, Mike Garnek,1993. UARS In-Flight Jitter Stuy for EOS. NASA Contractor Report [3]Chao fan,yincai Li, etc,2007.influence analysis of buffeting onimage quality of TDICCD Camera. ACTA PHOTONICA SINICA,36(9),pp CONCLUSIONS [4]Pengbo Wang, Yinqing Zhou,etc,2006.Effect of satellite attitue itter on spotlight SAR imaging quality.raar science an technology, vol4(5),pp High resolution imaging requires attitue of the satellite to be extraorinarily stable. Small satellite, owning to the small inertia, is har to keep still uner outsie isturbance. Usually, extra umper is fixe onboar to counteract the affection. Even so, itter that can not be eliminate by the attitue compensation evices still exists. Groun image processing methos have to be resorte to remove the itter. [5]Chengwei Yu, Derong Che, etc,2004.influence of satelliteattitue itter on LASIS imaging quality. Optoelectronic engineering, 31(5),pp

6 The International Archives of the Photogrammetry, Remote Sensing an Spatial Information Sciences. Vol. XXXVII. Part B1. Beiing