UA92541C2 - Method to increase spatial fragmentation of hyperspectral aerospace imageries on the base of subpixel redistribution of spectral component particles - Google Patents

Abstract

The invention concerns processing of the hyperspectral imagery from onboard hyperspectral systems for Earth remote sensing.The method performs the spectral endmembers unmixing, reallocation the endmembers subpixel fractions using their spatial neighborhood and mixing the said reallocated endmembers fractions. As result the enhanced spatial resolution hyperspectral imagery is composed.The method provides more detailed terrain rendering on hyperspectral images and therefore refines the interpretation possibilities.

Description

A well-known method of image processing for their conver- The theoretical basis of this possibility is a simulation (analogue - Opiiea Ziaiiez Raiepi Mo mapping by a model of spatial relations of opti- 7,352,910) |8I), according to which the physical characteristics of signals in the pixels of the aerospace image are determined on the image objects of the scene, such as smearing of the real distribution of signal-forming materials, texture, etc., vector materials are drawn over them in the same pixels, which explains the increase with interpolation, further according to the spatial scheme of fig. 1. As a rule, the natural and artificially increased distribution of the physical characteristics of the objects of aerospace imaging subordinates and restores the enlarged image. based on the principle of compactness, i.e. subpixel

The disadvantage of this method is the ambiguity of parts of the same material in adjacent pixels. Determination of the physical characteristics of objects is most likely spatially located scenes, in particular, their material, according to images, also in adjacent subpixels. This principle may require the involvement of additional data, for example, to be described by a certain vector field, which includes the spectral characteristics of materials. broods over the pixel portion of the spectral

Multi- and hyperspectral images of the component by its connections with pixel particles make it possible to more confidently determine the materials of the same spectral component in other objects of the scene by their spectral characteristics - image pixels. kami A known method of increasing the spatial disparity of the multispectral aerospace field is determined by the specific applied images based on the classification of the spectral model - Gibbs spatially symmetric, signatures of objects (analogue - a patent of Ukraine for a wine- as in the prototype, window with a given kernel transition Me84877) I9Y), which performs resampling of the vortex, inverse squares, exponential (i.e., proportional increase with oversampling or other. If the external vector field over the position) of the multispectral aerospace imaging pixel of the image is determined (fig. . 2), the addition of low resolution to the spatial distance makes it possible to redistribute the sub-pixel frequency differences of the multispectral aerospace tki ai, I-1...n, in the middle of the pixel in such a way that the high-resolution image by placing the "center of gravity" of all n of sub-pixel particles shifted sification of spectral signatures of elements in of the direction of the external field vector by the amount of discrimination based on a set of spectral signatures of its intensity. At the same time, the objects of the scene must be executed. limitation on storage of the full spectrum share-

The disadvantage of this method is the need for a small component of the image in the entire pixel: a hatospectral aerospace image with a high spatial resolution. i-th n)

The large dimensions of the ground projection of the pixels of aerospace images lead to the possibility After the redistribution of the subpeak has been carried out, several different materials, soil particles, for all 5 spectral components, causing significant identification errors (10). For Eke-(Ek), )-1...t, present in the current pixel, where it is possible to separate hyperspectral images - the number of spectral channels in hyperspectralizing different materials within one pixel image, the full optical signal E; in i- by separating mixtures of spectral compo- subpixels can be found as a spectral nent (11-13). Many methods have been built on this, a mixture with weighting coefficients of the corresponding sub-detection and identification of objects on hyperspectral images, for example (14, 151. PdTeek

The closest to the proposed method (2) is the method of spectral mixing by convention. the spatial smoothing method (prototype - Opiea Mixing (d) is successively performed for all Viaiese Raiepi Mo 7,200,243) 161, which obtains the spectral channels of all n subpixels and creates a hyperspectral image, on it a hyperspectral aerospace separation of spectral mixtures is effectively obtained component image of subpixel spatial resolution, the spectral characteristics of which - to, are stored in the spectral library, further pro redistribution of particles of spectral components, with a spatially symmetric model, which ultimately provides a more detailed reproduction of objects. Enhanced hyperspectral imaging is required. the sequence of operations of the method is shown in fig. 3.

The disadvantage of this method is that it does not provide an increase in the spatial resolution of the resulting image - block 1, in each pixel of this existing hyperspectral image. brewing is carried out in one of the possible ways

But the principle of redistribution of spectral particles - the separation of mixtures of spectral components - the component that is the basis of the prototype, block C, the spectral characteristics of which are preserved - can also be used and increased in the spectral library - block 2, and in this way the spatial resolution of hyperspectral ae is obtained fractions of spectral components - space images. block 4 in hyperspectral image pixels,

further, according to the selected model of spatial relations 7. Gose V.R., Vgapsiyupe M., the pixel fractions of the spectral components in each pixel are calculated - blocks 5.5... 5" and therefore Ziagez Raiepi Mo 6,332,044. - Osetreg 18, 2001. - they themselves set the vector field of influence for 11 pixel parts of all available spectral components - blocks 6, 6". .. 6", according to this vector Itade rgosezzipd teo, itade rgosevzvipd fields redistribute the particles of spectral arragace, apa itade epiagdipd teShtoi / Opiea component in the subpixels of each hyperspec- Ziagez Raiepi Mo 7,352,910. - Argii 1, 2008. - 2or. images - blocks 7, 9. Popov M.O., Stankevich S.A. The method of increasing 7007", and thus obtaining subpixel divisions of the spatial resolution of the multispectra of all spectral components - block 8, on final aerospace images based on classics they perform mixing of redistributed fications of spectral signatures of objects /Patent of particles of spectral components in subpixels - Ukraine for invention Me84877. - 10.12.2008. - 146. block 9, as a result of which hyperspectra are obtained - 10. Popov M.O., Stankevich S.A. The study of aerospace imaging of increased space-equivalent spatial resolution of multi-toroidal resolution - block 10. of spectral aerospace images for sum-

The application of the proposed method can be useful in the interpretation of hyperspectral works of the National Scientific and Research Aerospace Imagery of Natural and Artificial Center of Defense Technologies and Military Security of DZ33 objects. of Ukraine, 2006. - Issue 4(33). - P.109-117.

References 11. Stankevich SA, Shklyar SV. Classification 1. Kononov V.Y., Stankevich S.A. Comparator of covered landscapes on hyperspectral informational assessment of digital astro-astrocosmic images based on high- and low-resolution space images of mixtures of spectral components of Tavricheskoye Study Notes //Problem! Management and Informatics, 2006. - National University named after YOU. Vernadsko- Mob. - P.106-115. th, 2004. - Vol. 17(56). - Mo 2. - P.88-95. 12. Z,ipepipe .M., TotrkKkipv 5., MeMagop-Vhom/p 2. Eremeev V.A., Mordvintsev I.N, Platonov K.5. Meipoa g zeiIesiipd hergezepiayme epatetreg

N.G. Modern hyperspectral sensor and sotropepiv goth gresiga! daia /Opiei biaiez Raiiepi methods of processing hyperspectral data Mo 7,221,798.- Mau 22, 2007.-11 r.

Researches of the Earth from space, 2003. - Mo 6. - 13. Zipepipe ).M., Totrkipv 5., MeMagop-Vhom/p

P.80-90. K.5. Meitpoay g zeiIesiipud hergezepiayme epatetbreg

Z. Ztsyp Kh. daga /Opinei 5iayez Raiepi zraya! gezoishiop epnpapsetepi /Opigei 5iagez Raiepi Mo 7,321,691. - December 22, 2008. - 11

Mo 7,019,777. - May 28, 2006. - 19 years old. 14. PP S.-R. Meinoy g Nuregzresiga! itadegu 4. 7papd U. buvieet apa teinoa og itade gTiviop ehrioyayop apa rihe! zresiga! iptihipd /Opinea /Opnei 5iasez Raiepi Mo 7,340,099. - Magsn 4, 2008. Ziasez Raiepi Mo 6,665,438. - Oesetreg 16, 2003. - - gr. 23 years old 5.5ZsYshSheg 9.M., Zsprpeg O.A., Nomaga ..a. 15. PP S.-B. Meinoy yug iagdei aeiesiop apa

АІдогіптіс (есппідие юг іпсгеазіпуд Пе зрайаї! асийу ідепійісайоп Бу изіпуд ргохітйу ріхе! іпіоптайоп ог а юсаї ріапе агтау еїІесіго-оріїс ітадіпд 5узіет /Опней 5іагез Раїєпі Мо 6,940,999. - Зеріетре" 6, /Опінеа 5іагез Раїепі Мо 7,248,751. - ушу 24, 2007 - 2005. - 15 years 1 year 16. Keipap OM., NKapa 8.5. Zresiga! tikhishige 6. Satid 9.3. Vezoishiop epnapsetepi yug itadez rgosev5 sopayopea ru 5rayaPu-vtooiy ragpyopipod zYugei ip a dayaraze / Opiei iafez Raiepiga Mo / Raigepiei epe Mo 7,200,243. - Ari! 3, 2007. - 7,123,780. - Osyubeg 17, 2006. - 12. 22.

E I Z i N : ; ' і і і N nin p p VA A

N N I g ::

E and 3 th : th k H 1 : th th ! i Z Z N : i : i N Ya dents ptktiya uk three deti j I N ; N

N N ' Tolya VT s OO a and sov CHI ! і шыни нна я в ИНН в повно Д В у у ов кы ШИ и PO A ВН: пы кю и AK и и ку и Н и Б я и выны! by him and by him shi I tk : a aa N : : we : : : Y ! N : : OO Z IN : : I : : it i KK Zh nn I p I i ' N ' : OZya dya N | : | : ! i I i i | : ! ! I and and: I

I: N | I And and | N | i N kn a a a a a a a p PI A A V y 1 y Y i N i i N I N !

Fig. 1

N ; : Her name is N N : Z i ! and and! : and : and

Fi. Ge « « : N

She : ; : : Z : : 3 : ! : schu ' ' : H 3 x H po my p ni Vin Ne p n i DES : 15 : : 4 th N

E : Z H and : | and ! ! and and : and and we and I and

N shchi Z Zah i i i | zu i i : Z

N. Y. N

: and we n VOK 1 irig. 2

Kn Shi Sho nn denta rin - steette z z sya z. : Go t

Si visa y 5/6 B 7 8 91 kny NA "6 g Ya u sho, o Yad | Ivan Zla boyah sen y ! z ii same i po t 2. Yaki t : Y ts EY

I Kon Konya Y styu yutyuyutyutyuttey : mokko kva N

Get et - from still Czech s

DK m for zhlllyanya re p I "fig. WITH

Computer typesetting by A. Ryabko Signature Circulation 26 approx.

Ministry of Education and Science of Ukraine

State Department of Intellectual Property, str. Urytskogo, 45, Kyiv, MSP, 03680, Ukraine

SE "Ukrainian Institute of Industrial Property", str. Glazunova, 1, Kyiv - 42, 01601