CC BY
13Journal of Siberian Federal University. Engineering & Technologies, 2017, 10(6), 819-827
УДК 004.514, 004.42
Information Support Technique
for Solving Agricultural Land Monitoring Tasks Based on Earth Remote Sensing Data
Ruslan V. Brezhnev and Yuriy A. Maglinets*
Siberian Federal University 79 Svobodny, Krasnoyarsk, 660041, Russia
Received 04.04.2017, received in revised form 12.06.2017, accepted 20.07.2017
The article is dedicated to end users (decision makers) of information support technique in the process of agricultural lands' (AL) remote monitoring with the use of Earth remote sensing (ERS) data. The model of spatial object has been developed, which considers its spatial structure and allows to estimate its state in accordance with preset plan of object development with time.
Keywords: Earth remote sensing data, spatial object model, spatial inhomogeneity, NDVI, agricultural lands, end user, web-GIS, phenophase, agricultural land monitoring, agro-technological measure.
Citation: Brezhnev R.V., Maglinets Yu.A. Information support technique for solving agricultural land monitoring tasks based on earth remote sensing data, J. Sib. Fed. Univ. Eng. technol., 2017, 10(6), 819-827. DOI: 10.17516/1999-494X-2017-10-6-819-827.
© Siberian Federal University. All rights reserved Corresponding author E-mail address: [email protected]
*
Методика информационной поддержки
решения задач агромониторинга
по данным дистанционного зондирования Земли
Р.В. Брежнев, Ю.А. Маглинец
Сибирский федеральный университет Россия, 660041, Красноярск, пр. Свободный, 79
Представлена методика информационной поддержки конечных пользователей (лиц, принимающих решения) в процессе дистанционного мониторинга земель сельскохозяйственного назначения с использованием открытых данных дистанционного зондирования Земли. Разработана модель пространственного объекта, учитывающая его пространственную структуру и позволяющая оценивать его состояние в соответствии с заданным планом его развития во времени.
Ключевые слова: данные дистанционного зондирования Земли, модель пространственного объекта, пространственная неоднородность, NDVI, земли сельскохозяйственного назначения, конечный пользователь, web-ГИС, фенофаза, агромониторинг, агротехнологическое мероприятие.
Introduction
The tasks of space monitoring of spatial objects are solved in a whole range of practical applications: during emergency monitoring, construction, in oil and gas industry, agricultural industry, etc. Image-processing application packages (ENVI, Erdas Imagine, Scanex Imgae Processor, etc.), as well as general purpose geo-information systems (QGis, ArcGis, etc.) and specific purpose ones (e.g., ActiveMap, "Agriculture", Scanex "KosmosAgro") are used as tools for solving such tasks. It is worth mentioning that above-mentioned software tools are designed for specialists in the sphere of geospatial data processing and analysis, while solutions adapted for support of the end user, specialist in a specific subject area, remain underdeveloped. The author considers development of information environment for interaction between end user - agricultural specialist and automated system of agricultural land remote monitoring during setting and solving agricultural land (hereinafter referred to as AL of "field") monitoring tasks.
Problem statement
Study subject is a spatial object - agricultural land. Obvious characteristic of this object is its contour shape. Rectangular shape of the fields with 1:2 aspect ratio is considered to be the best one, or trapeze with min. 60° tapered side angle. If the field has irregular shape, labour productivity deceases, the number of machinery idle runs grows, as well as fuel consumption, all this leads to increase in crop growing cost ratio. In reality, standard shape is not always achieved during field layout. So, agricultural landscapes of Krasnoyarsk Krai are often characterized by small contours, outliers, complex boundaries, small square (less than 100 ha). Next approximation of considered concept if the term "agricultural area" [1-2], i.e., land within agricultural boundaries with homogeneous vegetation.
Apart from shape characteristics, studied spatial object has a vector of iconic features: spectral, textural, metric, topological, etc. During vegetation the field is subject to different influences stipulated
by both natural trends and anthropogenic (technogenic) factor. This results in the change in observed features vector, moreover, some features can be dynamically added to the formed model, by emerging at a given time.
Moreover, during object modelling one more factor shall be considered: due to natural and anthropogenic reasons different components of the studied object evolve in many ways. This explains inhomogeneity of the object structure, and inhomogeneity character is also variable with time.
Detection of local object inhomogeneities using satellite images allows to estimate uniformity of crop emergence, its degradation, as well as the level of fulfillment of the planned agro-technological measures. However, carried out literature review [3-5] showed that the issues of local spatial inhomogeneities analysis during organization of spatial object monitoring systems haven't been sufficiently addressed so far.
Conceptual model of spatial object "AL"
The object is characterized by the following groups of characteristics significant for organization of monitoring system.
1) Spatial localization (precise geographical coordinates of the object, with account of topological peculiarities).
2) Name and variety of the crop of the current field season.
3) Seasonal vegetation process schedule as evidenced by the change of crop phenological phases.
4) Possible abnormalities (in connection with phenological phases), such as non-emergence, soil overwetting, drought, infestation, etc.
5) Technogenic burden, represented by planned and implemented agro-technological measures (ATM).
6) Crop characteristics (such as integral state, biomass, leaf area, etc.).
Principal goals of agricultural land monitoring are observation of seasonal vegetation process, localization of abnormal manifestations, tracing of ATM and their results.
Evety individual crop 00 correlates with the model of its development in time under the influence af natuoal and technovenin fattori in the form oS stetn-transitlon diagram G, (see Fig. 1):
Gi=<F„R>, (1)
where.0 = SfJ- asft of field states,=f-a set ofstate transitions;
R=fi1UR2UR3, (2)
R1 - set of operators, which describe time sequencing (natural phenophase change), R2 - set of operators, which describe abnormal phenomena in vegetation community development, R3 - set of operators, iwhich simulate ATM implementation. The state can be simple (plant phenophase) and composite (significant changes within one phenophase). Every state is characterized by a set of features described above.
From the point of aerospace monitoring organization, it is important to single out a set of features, detected on the object digital images, and find connection between the values of observed features on the image, on the one part, and F set elements, on the other part.
stm States (f) of field under the influence of factors (r) 1
Fig. 1. Amodel ofmonitoringobject changein timeunderthe influenceofnaturalandtechnogenicfactors
The article considers the case when an observed object is fixed in space (spatial boundaries of agricultural contourareprecisely known).Such assumption isexplainedby the fact that field boundaries have low oariabttitycn ttmo, asoheaarw lfmoiedXy no^turalXxi^i^^^os ^o^a^i^^o^Xii; ^squaMer, soade, l^cibulnXo^o e^^^Oinhe, watero, tfiMfr, tree^andsbo^iii, ^^e.l,o\i^c^]ia^eitablished history of use and cadastral division of territories.
After fixing field boundaries during monitoring task setting, it is necessary to determine spectral ahdtadtacal imxgefeafuoetmsMo theeontour esaentMfor maoutoringoiiJaounation Red,mr, green, biur eonmnr^sonit their rombinahions (¡¡w, NDVI, PVI, EVI, bAVI inbicer,a^(ifarousebui fOT^^n^i'rrt oU vogotatiimyommuyttifs cXasoeterisdics. Tpxturad iea-mea nun Ico additianutlyintenrated an toll detail images.
In general, image area s, which is interpreted as a field in the context of above-stated limitations, is charaoCerizeXhy veatorofiuamcfyaturei X u (ah- Xhe aidueof' ^i^iii-vMo^al foaoueein an average numerie v^loe wilh rese>ock to cH pixe(nalnos reiatine ou oeM. ReeebreXer c^rrie^^i^t in pnXert [6, 0lshownigncxrrelationbetween wearurements on the images and vegetarian cover noeperttee. This allows to define g reference function synthesis problem, which compares sj image location in multidimensional feature space and observed AL state f.
Aid aX ove1srab)PreObrf loimage areas with homogeneous structure. However, studies of vegetation dynamics en -Pi fields of Sunpobuztmonn DistrknoУKralnpe arskKrai (Xbato lSeX oolds in ahe period from 2013 to 2016) showed that only 30% of agriculture contours exhibit homogeneity property throughout the entire vegetation season.
For e,-, bharacterinod by inters inkamo^neity before caotyi nn ond1nteopoeiericn of-tearta of the field corresponding to the images, it is necessary to perform s i segmentation in wj area, which meets homogeneity criteria. In contrast to sj, which space coordinates are fixed, wj position in space requires localization. Moreover, feature vector X, made for wj, apart from iconic features, it is necessary to include geometrical characteristics, which physical interpretation shows position and shape of inhomogeneities within the field.
To summarize, during AL description as a spatial object, homogeneous and inhomogeneous objects shall be distinguished. The former are characterized by spatial and time coordinates, vector of informative iconic features X and reference function g, which allows to interpret AL state based on GPS survey data. For the second group of objects it is necessary to additionally find segmentation into homogeneous areas, each of them characterized by the same features as the first group objects and, in addition, a set of geometrical characteristics.
Type representation has been developed on the basis of conceptual model:
M0=<K,ID,A,N,T >, (3)
where K is object contour coordinates, ID is object identifier or field number, A is a vector of values of geometrical features, N is a vector of values of iconic features, T is a measurement time. This representation isthe basisforobject-relationalmodelofALobjectin the system.
Organization of adialoguefor remote monitoringtask setting for end user
Remote monitoring task setting is considered as a task of generation and resolution of information inquiries on the current (actual) state of the studied objects observed on satellite image. Due to the focuson end user(EU),technique shall ensure thefollowingpossibilities:
• Means of organization of a dialogue for information inquiry generation, which must provide for EU minimum number of steps in the process of object initial state building up, specifying only that information in which EU is competent.
• Information inquiry resolution scheme, which, in general, involves search for current satellite data,itsprocessing andanalysis.
• Means of organization of a dialogue for information inquiry results assessment, which must visualize to the user current state of the object (Sf) with account of its structure.
The dialogue for information inquiry generation starts with EU authorization process, followed by automatic identification of agricultural producer by the system and display of the boundaries of the fields used by him for EU. Moreover, the map of agricultural contours of the enterprise is displayed.
EU has a possibility of specifying monitoring area by localizing the object or a group of objects in space by means of multiple selection of their contours (Fig. 2a)). In other words, EU generates information inquiry in the system, assigning initial set of features, which allow the system to localize precisely monitoring area -{kj, t , id,}, kt e K, tt e T, idt e ID.
In response to the inquiry, the system determines the necessary elements of the vectors A, N of the model (3), which shall be calculated and updated; the relevant calculations are performed.
Then the system offers to carry out time localization of the object. EU can set a precise point of time localization for update of the object state, or initiate scheduled update in automatic mode. The schedule is based on the model (1). The system considers timeframe of planned events and activates generation of information inquiry. At the same time EU has a possibility of changing localization of data update point, and, accordingly, influence the process of object model update.
End user information inquiry resolution algorithm
The algorithm solves the problem of EU information inquiry resolution on the current state of spatial objects. Algorithm preconditions are completion of initial feature space generation in above-described dialogue between EU and automatic system. Let's consider main steps of the algorithm.
Obtaining ERS data from open sources [8, 9]. United States Ground Survey - USGS resource, where Landsat, Aqua, Terra, Sentinel, etc. programs' data are published, is used as ERS data source.
Algorithm description is displayed in the example of Landsat-8 data search, mastery and processing on a localized territory in automatic mode.
ERS data processing and interpretation. This process includes the following sequence of steps:
1. Unpack downloaded archive and extract Red (BandRed) and NIR (BandNIR) channels for NDVI calculation.
2. Combine BandRed and BandNIR and save as a single image in GeoTIFF format.
3. Open qualitychannelCSawii^andrepresentit asatwo-dimensional array.
4. Dclecminenon-infonmative imagepixels ana1 represent tliam asa two-dimensional array.
5. Deleteeoniinfarmative pixele of tlin channeOs BaadRed e BandMR. BandRcD[mw, col] = 0, fiacidbrow, noii = t). Dxlntira occursby dedncRion of a set of pixels tf f^i^annn^g,^ chante^represented aa a twoodiRonaional array from a sei of pixs;ls of BandRed andatawinnffiChanneSSqaisouapresented asawoxCimenséonal aeeays: ARidtASQA\ Anir\Abqa. This allows to improve the process of object pixel structure analysis.
6. Cadp ihe .maga bya nttveotor maak of ihe abjects. Qopprngahowsio receive image segment which ceores^ods te predefiuieU objort. Tliis atep ea chiaes volume unp nimo oHmage processing by preeeseinh oul\a segmenS oHnteeest.
7. CalculaCy ^j\^otr^iei^at^rt;q - aagetation éndex NDVi baret on ehtoblained image:
^7/ = (4)
Nie+eed
where NIR 3s a nqar Mrared band, aiisl Red mnanx rod crsCoui". ("ea a xesull ^etiiiiiye channet wiïh the values NDVI (Nmacfwt is yroated (Fig. 2-»..
8. Analyye object in^nene^^. Sxcmentxtren of ihe objec, tex)ural areaslnsxecuteg by BandNDVI basnd an aree marting using threshold meateS) ln thie cnntrxt aaightnese tuneticytraxtformation operator is preset:
TH:/(x,y) s(x,y),
p! where ^ < /(x,;y) < T+1, (5)
h(r, y) = j Ào where/(x,y) < T'a, (. Àfc_i where /(Sa^ > Tfc_i
where s(x, y) is a se^ented image, K esa neimbor of segmentation areas, h(QT X, ..^¿e are segmented ateas) marks, T0,Tb. ..,7^ ace thresholnvalues, arr^nge^scottiae; T[<Shl< Tkd. Tha valursof segmented areas' points correspond to threshold values: T = 0.05, 0.1, 0.15, 0.2, ..., 0.7. As a result an image with averaged valuesiscreated, let's denoteit as BandNDVI.
+: Filterthe imape Band^-, IF12 2a). FihiaCionic me.u fxa pdae .om.mg on the image which appears, id any case, at a resuil oR seumantcCian. This ellaws toexaludsinsigmPuantiuhomogeneous area) an aiie impge wiih + si:ze <t:l teas tCan 2 pieiek 2or Lsndsatr8 aets. Mechad filter need for digital imageprocessingisapplied forfiltration:
x* = meeOi^, ...y«). (6)
Asaresult BandFiUer image iscreated.
10. Calculatemetriccharacteristics -forevery spatialinhomogeneityon BandFHter image.
S quare:
5= llYi=1(Xi+Xi+1)(Yi-Yi+1)l, (7)
where {(Xii)}, i = 12, ...,n is a sequence of coordinates of adjacent area peaks,
Perimeter:
P = I?=1li, (8)
waeae / is "t^a^ae^g^tli ja^^^sfd^, i = 1,2,... ,n isa number of area sides.
11. Makevectormapofspatialinhomogeneitiesoftheobject, Fig. 2d.
Conclusion
The article considers problem setting technique for remote monitoring of spatial object based on ERS data, developed by the authors, which generates object feature vector, precisely localizing searcharea, which enables end user to setinformationinquirygoalwith a minimum number of dialogue steps. Technique is based on a conceptual model of spatial object "AL" with account of its? spatial structure. This allows to estimate object state change with time in automatic mode bared on feature values' tracing with account of local inhomogeneties. Model implementation on the basis of object-relational approach has been created. End user information inquiry resolution algorithm on remote monitoring of spatial obj ect based on ERS data has been developed as part oO the technique.
c)
d)
i*
ST
Vll
c) d)
Fig.2. ttesuttsof algoriOhmatepcexecotion:a)preset object"AL"No.l03, b) calculated NDVI index, c) atoraged and eetesed oygmenl; yy NDVI velues, d) dighliyhted inliosreogenytite foe prerea obieot
Specified results are used in Web-GIS components of agricultural land monitoring system [10-12]. The system was practically tested during research of AL growth dynamics in Sukhobuzimsky District of Krasnoyarsk Krai.
References
[1] Распоряжение Правительства РФ от 30 июля 2010 г. № 1292-р «О Концепции развития государственного мониторинга земель сельскохозяйственного назначения и земель, используемых или предоставленных для ведения сельского хозяйства в составе земель иных категорий, и формирования государственных информационных ресурсов об этих землях на период до 2020 года» [RF Government Order of July 30, 2010 № 1292-r "On the Concept of development of public monitoring of agricultural land and land used or provided to agriculture as a part of other categories of land, and the formation of state information resources on these lands until 2020 of the year" (in Russian)]
[2] Дибихин К.Ю., Коротков М.А. Распознавание элементов подстилающей поверхности сельскохозяйственного полигона, Многопрофильный университет как региональный центр образования и науки, Оренбург, Оренбургский государственный университет, 2009, 587-591 [Dibihin K.Yu., Korotkov M.A. Recognition of the elements of the underlying surface of agricultural landfill, Multidisciplinary University as a regional center of education and science, Orenburg, Orenburg State University, 2009, p.p. 587-591 (in Russian)]
[3] Nikos Pelekis, Babis Theodoulidis, Ioannis Kopanakis and Yannis Theodoridis Literature review of spatio-temporal database models, Cambridge University Press New York, NY, USA, Journal The Knowledge Engineering Review. Volume 19, Issue 3 September 2004, 235-274.
[4] Introduction to Spatio-Temporal Databases [Электронный ресурс]. Режим доступа: http:// www-users.cs.umn.edu/~lmani/spatial/HW5-B4.pdf#page=4&zoom=auto,-100,695.
[5] Nandal R. Spatio-Temporal Database and Its Models: A Review. R. Nandal. IOSR Journal of Computer Engineering (IOSR-JCE). 2013. Vol. 11, Issue 2. P. 91-100.
[6] Сапрыкин Е.И., Сладких Л.А., Кулик Е.Н. Оценка состояния посевовзерновых культур по данным дистанционного зондирования Земли, Журнал Интерэкспо Гео-Сибирь, Выпуск № 1, Том 4, 2015, 40-46 [Saprykin E.I., Sladkih L.A., Kulik E.N. Assessment of grain crops based on remote sensing data, Journal Interekspo Geo-Siberia, Issue 1, Vol. 4, 2015, 40-46 (in Russian)]
[7] Шукилович А.Ю., Федотова Е.В., Маглинец Ю.А. Применение сенсора modis для оперативного мониторинга земель сельскохозяйственного назначения, Журнал Сибирского федерального университета, Серия Техника и технологии, 2016, 9(7), 1035-1044 [Shukilovich A.Yu., Fedotova E.V. and Maglinets Yu.A., Using Modis Data for Agricultural Areas Monitoring, Journal of Siberian Federal University. Engineering & Technologies, 2016, 9(7), 10351044 (in Russian)]
[8] Маглинец Ю. А. и Брежнев Р. В., Развитие средств автоматизации приема и обработки спутниковой информации региональной системы ДЗЗ СФУ, Современные проблемы дистанционного зондирования Земли из космоса, 2011, Том 8, № 3, 120-128 [Maglinets Yu.A. and Brezhnev R.V., Automation development of satellite information reception and processing from regional Earth remote sensing system of Siberian Federal University, Modern problems of Earth remote sensing from the Space, 2011, Vol. 8, No. 3, 120-128 (in Russian)]
[9] Мальцев Е., Маглинец Ю., Брежнев Р., Соснин А., Курносов В. Технология динамического расширения набора атрибутов для поиска пространственных изображений в базе данных, Распознавание образов и анализ изображений 2011, Том 21, № 1, 35-40 [Maltsev E., Maglinets Y., Brejnev R., Sosnin A., Kurnosov V. The technology of dynamic extension of attribute set for the search of satellite images in a data base, Pattern Recognition and Image Analysis, 2011, Vol. 21, No. 1, 35-40 (in Russian)]
[10] Брежнев Р.В., Маглинец Ю.А., Мальцев Е.А., Перфильев С.Е., Сидоров А.Ю., Цибульский Г.М., Шокол А.С. Прогаммно-технологическая инфраструктура информационной поддержки решения задач территориального управления, Журнал Сибирского федерального университета 5 (3), 340-352 (2012) [Brezhnev R.V., Maglinets Yu.A., Mal'tsev E.A., Perfil'ev S.E., Sidorov A.Yu., Tsibul'skii G.M. and Shokol A.S. Software technological infrastructure for information support of territorial control problems, Journal of Siberian Federal University, Engineering & Technologies, 5 (3), 340-352 (2012) (in Russian)]
[11] Маглинец Ю. А., Мальцев Е. А. И Цибульский Г. М. Многоцелевая геоинформационная система управления территориями Енисейского меридиана, Распознавание образов и анализ изображений , 2012, Том 22, № 2, с. 318-322 [Maglinets Y. A., Mal'tsev E. A., and Tsybul'skii G. M. Multipurpose geoinformation management system of Yenisei meridian territories, Pattern Recognition and Image Analysis, 2012, Vol. 22, No. 2, pp. 318-322 (in Russian)]
[12] Brezhnev R.V., Mal'tsev E.A. An Ontological Spatial Monitoring System for Agricultural Land Monitoring, Pattern Recognition and Image Analysis, 2015, Vol. 25, No. 2, pp. 201-208
