RO133457A2 - Process for generating green-area and built-up-area dynamics maps based on processing satellite data - Google Patents

Abstract

The invention relates to a process for generating and delivering maps regarding the dynamics of green areas and built-up areas, based on processing of satellite data. The map generating process comprises: optical image acquisition, reading and conversion of optical data, including reading of optical images and of acquisition parameters, applying conversions and saving the results in a suitable file format for the software applications which are used subsequently for image processing, optical data correction by applying several correction types: atmospheric, topographical, radiometric, geometrical corrections, scene connection, i.e. merging adjacent georeferenced images to form an image in which the common covering surfaces are eliminated, creation of a cloud mask which means that, based on the characteristics of clouds, they are confined and eliminated from the image before other processing, computing of normalized difference index as a result of mathematical operations intended to highlight environmental characteristics using two spectral bands in which the tracked element has opposite responses, data classification using predefined classes, in order to highlight activity at the ground surface, carrying out the final map and validation of results. The process for generating the maps is followed by a process of delivering the maps to potential users, by means of a software application.

Description

OHCliX OF THE STATE FOR INVENTIONS Șt ^ 1 ^. Patent application

Nr. .........

Date of filing

Procedure for generating and delivering maps on the dynamics of green areas and built areas based on satellite data processing

The invention relates to a process for generating and delivering maps regarding the dynamics of green areas and built areas based on satellite data processing, respectively based on extracting information from optical images taken from satellite, including for optimizing the method of the "cloud mask" processing chain. .

There are known methods (methods) for generating maps regarding the dynamics of the green areas and the built areas that consist of:

> in-situ determinations (topographic measurements, field observations to characterize the type of coverage: vegetation, constructions)> aerial photograms (photogrammetric cameras embarked on a human pilot flight aircraft, the flight apparatus has powerful internal combustion engines),

The known method for generating maps regarding the dynamics of the green areas and the built areas, respectively the method based on "in-situ determinations" (requires the movement of a human team and of the equipment specialized in the measurement points) has the following disadvantages:

> it is done on small areas,> not all measuring points are easily accessible,> in the area of protected natural areas access can affect biodiversity,> they have a low degree of automation,> their high realization time,> they have a high cost,> their determinations involve consumption of material resources (equipment, fuel for travel, etc.)> the method is not efficient for periodically monitoring the evolution of green areas and built areas,

The known method for generating maps regarding the dynamics of the green areas and the built areas, respectively the method based on "aerial frames" has the following disadvantages:

> noise pollution,> combustion gas emissions,> consumption of fossil fuels (petroleum derivatives),> determinations involve the consumption of material resources (equipment, flight apparatus, fuel for travel, etc.)> shorter duration of comparison with the method based on "in-situ topo-geodesic determinations" and higher compared to the method based on satellite data processing,> lower cost compared to the method based on "in-situ topo-geodesic determinations" and higher compared to the method based on satellite data processing.

There are known methods (methods) for delivering digital maps:

> CD delivery,> electronic delivery (fastupload, wetransfer, ..etc)> websites with specific applications

The known methods for delivering maps have the following disadvantages:

> do not allow the manipulation of cartographic content (for the manipulation of cartographic content it is necessary for the beneficiary to use specialized software packages),> they do not comply with the requirements of the INSPIRE directive (which will allow the interoperability and the exchange of data at European level).

a2017 01125

12/14/2017

The technical problem that the invention solves consists of:

> carrying out a procedure for generating maps on the dynamics of green areas and built areas, based on the processing of optical satellite data, which consists of extracting information from optical images taken from the satellite, including for optimization the method related to the "cloud mask" processing chain, consisting of into the:

S extracting information regarding the classes "vegetation-covered land" (including subclasses: forests, shrubs, grasses, arable land, etc.) and "construction-covered land" (including asphalt, ..etc);

S generating maps regarding the dynamics of the green areas and the built areas,

S map delivery using the PORTAL computer application (the application complies with the INSPIRE directive [code 2004/0175, Directive of the European Parliament and of the Council establishing an infrastructure for spatial information in the European Community, http: //inspire.ec.europa.eul;

> and carrying out a procedure for delivering maps in formats that allow data exchange and interoperability (through electronic means) at European level by integrating a package of tools for manipulating cartographic content and running these tools on the provider's server, the beneficiary receiving by electronic means (on account base and password) cartographic content easy to analyze and interpret.

The process for generating and delivering maps regarding the dynamics of the green areas and the built areas, according to the invention, removes the disadvantages mentioned by the fact that, it consisted in the succession of the following stages:

1. The process for generating maps;

> Acquisition of optical images: consists of satellite programming for the acquisition of new optical images and / or consulting the existing optical image archives - through NASA), ESA, etc; the selection of the acquired data type (spatial resolution, temporal resolution, radiometric resolution) is made according to the field of use.

> Reading and processing of optical data

S includes reading optical images and acquisition parameters,

S applying transformations and saving the results in a file format suitable with the software applications used subsequently for image processing; Among the most common transformations were georeferencing. Through georeferencing, satellite images are brought into a coordinate system. The control points required for this operation are provided by the satellite platform operator or are identified on ready georeferenced maps or images;

S within the transformations of the satellite data also enter those operations that are performed by combining information from several spectral bands of a scene. Thus, mathematical operations (subtraction, addition, multiplication, division) are used to combine and transform the initial bands into new images / products that reflect certain aspects that can only be highlighted in this way. The most known such transformations involve combining the bands, obtaining specific indices (eg:

NDVI, NDWI, EVI etc) or PCA (Principal Component Analysis).

> Optical data correction: depending on the type and resolution of the optical satellite data, it is necessary to apply several types of corrections:

S Atmospheric corrections - is performed using a radiative transfer model that calculates the atmospheric transmittance and radiance for various frequencies. As input parameters, the model uses data related to the atmospheric profile (pressure, temperature, water vapor, ozone), aerosol type, altitude, zenith solar angle and sensor viewing angle.

Ά Topographic corrections - is performed using a numeric altimetric field model (MNAT).

Radi Radiometric corrections - Depending on the sensitivity of the sensor some images have a quality that does not allow a proper analysis (shows distortion) from the radiometric point of view. In order to improve the quality of the images, procedures are used to adapt the dynamics of the radiometry, usually procedures by which the histogram (which in the case of images with low dynamics is concentrated in a limited number of levels) is extended over a sufficiently wide range to ensure the required quality. .

Ά Geometric corrections - it ensures the correction of the geometric deformations due to the way of recording the spectral information.

> Scene connection (optional - is used only if the study area cannot be framed by a single satellite scene). It is the process by which a number of adjacent georeferenced images are glued together to form the same image in which the common cover surfaces are removed. Basic conditions a2017 01125

12/14/2017

to perform this operation are for the images to be georeferenced in the same projection and have the same spatial resolution.

> Creating the cloud mask: the clouds prevent the observation of the Earth's surface with the help of satellites operating in the visible and infrared spectrum. Based on the characteristics of the clouds, they must be delimited and removed from the image before further processing.

> Calculation of normalized indices of differentiation; indices are the result of mathematical operations that aim at highlighting some features of the environment using two spectral bands in which the element pursued has opposite answers. For example, when calculating the differentiated vegetation index, the bands belonging to the near red and infrared spectral area are used, in the first case the light radiation being absorbed by chlorophyll very much, while in the field of infrared chlorophyll it has a high reflectance. For the analysis of the satellite images used (Landsat TM), the open source platform BEAM (http: //www.brockmann-consult.dc/cms/web/beam/) developed by ESA is used.

> Classification of information: supervised classification is used, the method involves assigning pixels from the satellite image of predefined classes to highlight the reality from the ground [forests, uncovered land (eg cleared, arable), grasses, water, built area, .. etc],> Making the final map: The information extracted from the optical satellite images is used in conjunction with other geospatial data (communication networks, localities, hydrography, points of interest, toponymy, etc.) to obtain the final map.

> Validation of results: it is carried out through field measurement campaigns, using appropriate sensors and equipment, or by intercomparison with other higher resolution satellite images or already validated.

2. The procedure for delivering maps regarding the dynamics of the green areas and of the areas built to the potential beneficiaries (and for their manipulation) consists in the creation and use of a PORTAL type application (WEB). The PORTAL application allows the loading, handling and delivery of maps in the formats: Shapefile, PostGIS, GeoTIF. The PORTAL computer application was made in compliance with the requirements of the INSPIRE directive [code 2004/0175, Directive of the European Parliament and of the Council establishing an infrastructure for spatial information in the European Community, http: //inspire.ec.europa.eul:

The process according to the invention for generating maps has the following advantages:

> allows investigation on large areas (on the order of hundreds / thousands of kmp),> is completely non-polluting (noise, flue gas) does not affect the biodiversity of the investigated areas,> does not consume fossil fuels,> has a very high degree of automation,> determinations does not involve the consumption of material resources (all determinations are made by satellite),> low duration of implementation,> low cost,> allows to identify the areas for which the information is vitiated by the existence of clouds and to extract these areas from the satellite scene;

The process according to the invention of map delivery (through the PORTAL application) has the following advantages:

> offers a package of facilities that allow the manipulation of the cartographic content in order to facilitate the analysis of the information (does not require the purchase or installation of other scofrware packages),> it exactly respects the requirements of the INSPIRE directive,

Embodiment examples

The following are examples of realizing the invention in relation to 4 maps (fig. 1 fig. 4) regarding the dynamics of the green areas and of the built areas for Bucharest and the surrounding area (40x40 km geographical area) for the period 1990-2012.

FIG. 1 - Map of the extension of the green areas and of the built ones - Bucharest 1990

FIG. 2 - Map of the extension of the green areas and of the built ones - Bucharest 2000

FIG. 3 - Map of the extension of the green areas and of the built ones - Bucharest 2006

FIG. 4 - Map of the extension of the green areas and of the built ones - Bucharest 2012 a2017 01125

12/14/2017 / (,,

FIG. 5 - presents the synthesis information regarding the evolution of the built areas between 1990 and 2012 (the built-up areas occupied an area 78.2 km ² larger than in 1990, the most obvious extensions of the built areas are found in the northern areas and west of Bucharest, less dynamic areas are found in the southern part of Bucharest, as well as some neighboring communes: Berceni, Vidra, Sabareni Dascalu).

The green areas were divided into two categories, forests and urban green areas (parks, public gardens, recreation areas), to highlight more precisely which areas are most affected by urban development.

In the study area surrounding the city of Bucharest, the area covered by forests decreased in the analyzed range by less than 1%, more precisely by 0.98 km ² . Comparing this value (percentage) with the total area that was mapped we can consider that the extension of the built areas did not have a great influence on the areas covered by the forest. The very small percentage of the decrease of these areas can be explained by the not very large extension of the constructions within the area occupied by forests (the extension was realized in the area of viran land and agricultural land).

Another embodiment of the invention is related to data transformation - fig. 6. False-color combination (432) of a Landsat scene (October 4, 2011).

The process for generating and delivering maps regarding the dynamics of the green areas and the built areas, according to the invention, consisted of the succession of the following stages:

1. The process for generating maps consists of:

> Acquisition of optical images: consists of satellite programming for the acquisition of new optical images and / or consultation of existing optical image archives - through NASA, ESA, etc; the selection of the acquired data type (spatial resolution, spectral resolution, temporal resolution, radiometric resolution) is made according to the field of use.

> Reading and processing of optical data

V includes reading of optical images and acquisition parameters,

V applying transformations and saving the results in a file format suitable with the software applications subsequently used for image processing; Among the most common transformations are georeferencing. Through georeferencing, satellite images are brought into a coordinate system. The control points required for this operation are provided by the satellite platform operator or are identified on ready georeferenced maps or images.

> Optical data correction: depending on the type and resolution of the optical satellite data, it is necessary to apply several types of corrections:

V Atmospheric corrections - is performed using a radiative transfer model that calculates the atmospheric transmittance and radiance for various frequencies. As input parameters, the model uses data related to the atmospheric profile (pressure, temperature, water vapor, ozone), aerosol type, altitude, zenith solar angle and sensor viewing angle.

V Topographic corrections - is made using a numeric altimetric field model (MNAT).

V Radiometric corrections - Depending on the sensitivity of the sensor some images have a quality that does not allow a proper analysis (shows distortion) from the radiometric point of view. In order to improve the quality of the images, procedures are used to adapt the dynamics of the radiometry, usually procedures by which the histogram (which in the case of images with low dynamics is concentrated in a limited number of levels) is extended over a sufficiently wide range to ensure the required quality. .

V Geometric corrections - ensures the correction of the geometric deformations due to the way of recording the spectral information.

> Scene connection (optional - is used only if the study area cannot be framed by a single satellite scene). It is the process by which a number of adjacent georeferenced images are glued together to form the same image in which the common cover surfaces are removed. The basic conditions for performing this operation are for the images to be georeferenced in the same projection and have the same spatial resolution.

> Creating the cloud mask; clouds prevent the observation of the earth's surface by the satellites operating in the visible and infrared spectrum. Based on the characteristics of the clouds, they must be delimited and eliminated from the image before further processing (in Annex I we presented the processing chain "creation of the cloud mask") of 2017 01125

12/14/2017> Calculation of normalized indices of differentiation: indices are the result of mathematical operations that aim at highlighting some characteristics of the environment using two spectral bands in which the element pursued has opposite answers. For example, when calculating the differentiated vegetation index, the bands belonging to the near red and infrared spectral area are used, in the first case the light radiation being absorbed by chlorophyll very much, while in the field of infrared chlorophyll it has a high reflectance. For the analysis of the satellite images used (Landsat TM), the open source platform BEAM (http://www.brockmann-consult.de/cms/web/bearn/ ') developed by ESA is used.

> Classification of information: supervised classification is used, the method involves assigning pixels from the satellite image of predefined classes to highlight the reality from the ground [forests, uncovered land (eg cleared, arable), grasses, water, built area, .. etc],> Making the final map: The information extracted from the optical satellite images is used in conjunction with other geospatial data (communication networks, localities, hydrography, points of interest, toponymy, etc.) to obtain the final map.

> Validation of results: it is carried out through field measurement campaigns, using appropriate sensors and equipment, or by intercomparison with other higher resolution satellite images or already validated.

2. The procedure for delivering maps regarding the dynamics of the green areas and of the areas built to the potential beneficiaries (and for their manipulation) consists in creating and using a PORTAL type application (WEB) in compliance with the INSPIRE directive [code 2004/0175, Directive of the European Parliament and of the Council establishing an infrastructure for spatial information in the European Community, http: //inspire.ec.europa.eu1:

The PORTAL application allows the loading, handling and delivery of maps in the formats: Shapefile, PostGIS, GeoTIF.

The PORTAL computer application was made in compliance with the following requirements (technical design specifications):

1. General requirements (GEN)

No. Description 0 1 JAN-O 10 The SISat geoportal must include geospatial information relevant to the study area and the research objectives. GEN-020 The SISat geoportal must be accessible on the Internet. GEN-030 The SISat geoportal must be compatible with the specifications and principles of the European Directive INSPIRE. GEN-040 The SISat geoportal must be compatible with the specifications and principles of the European Earth Monitoring Program (Copemicus, former GMES) GEN-050 The products of the SISAT service must be accessible online through a friendly graphical interface.

2. Data Requirements (DAT)

No. Description 0 1 DATE 10 The SISat geoportal must be able to assimilate data in raster and vector format. DAT-020 Geospatial data should be published in standard formats (eg GeoTiff, HDF, CSV, GML, KML). DAT-030 Geospatial data should be published in coordinate systems used in Romania (eg WGS84 - EPSG: 4326; Stereo70 - EPSG: 31700). DAT-040 Attributes associated with vector data must use unitary conventions. DAT-050 The quantitative attributes associated with the vector data should be expressed in the same units of measurement.

to 2017 01125

12/14/2017

3. Functionality Requirements (FUN)

No. Description 0 1 FUN-010 The SISat geoportal must allow downloading restrictions of the data sets. FUN-020 The SISat geoportal must allow the selection and display of information on temporal criteria. FUN-030 The SISat geoportal must allow the use of EPSG definitions / codes for coordinate systems. FUN-040 The SISat geoportal must allow the addition of new data sources. FUN-050 The SISat geoportal must allow the traceability of the information. FUN-060 The SISat geoportal must include a CSW service. FUN-070 The SISat geoportal must include a WMS service. FUN-080 The SISat geoportal must include a WCS service. FUN-090 The SISat geoportal must include a KML service. FUN-100 The SISat geoportal must allow users to access geospatial information using a WMS interface. FUN-110 The SISat geoportal must allow users to access geospatial information using a WCS-like interface. FUN-120 The SISat geoportal must allow users to access geospatial information using a KML interface. FUN-130 The default coordinate system will be Stereographic 1970 (Stereo70 / EPSG: 31700). FUN-140 SISAT standard geospatial services (WMS, WCS, WFS) must support at least four coordinate systems: WGS84, UTM34 / 35, Stereo 70, Spherical Mercator coordinate system used by international mapping service providers such as Google Maps, Bing Maps , OpenStreetMap, etc.). FUN-150 The SISat geoportal must allow the definition of user groups. FUN-160 The SISat geoportal must include a catalog for metadata. FUN-170 The "publisher" users will have the right to publish metadata about the data sets in the SISat geoportal. FUN-180 The users of "administrator" type will have administrative rights on the SISAT catalog. FUN-190 Regular users will have the right to view metadata. FUN-200 The metadata catalog will allow you to search by name of the datasets. FUN-210 The metadata catalog will allow you to search by name of the institution that created a particular data set. FUN-220 The metadata catalog will allow you to search by keywords. FUN-230 The metadata catalog will allow you to search using the geographic extension of the data sets. FUN-240 The metadata catalog will allow you to search by temporal criteria. FUN-250 The metadata catalog will allow you to preview the data sets identified by the user with the help of search functions. FUN-260 The SISat geoportal must include a web mapping service. FUN-270 The web mapping service will allow access of geospatial information from standard SISAT data services (WMS, WFS, WCS) FUN-280 The graphical interface of the web mapping service will include components for accessing predefined WMS services. FUN-290 The graphical interface of the web mapping service will allow interactive selection of the layers to be displayed from the predefined WMS services. FUN-300 The web mapping service will allow integration with popular mapping services (eg Google Maps, Bing Maps, OpenStreetMap) FUN-310 The SISat geoportal must include a WPS invocation service. FUN-320 The graphical interface of the application will include functionality for calling the WPS service and displaying the results generated by it. FUN-330 The WPS service must include time series processing capabilities.

to 2017 01125

12/14/2017 γ o

4. Security Requirements (SEC)

No. Description 0 1 SEC-O 10 The SISat geoportal will be secured through a "firewall". SEC-020 The SISat Geoportal will be configured in such a way as to allow access only to authorized users in certain sections of the web portal. SEC-030 The transfer of user login credentials will be done through a secure connection. SEC-040 The SISat geoportal will allow secure access to confidential geospatial information layers.

5. Hardware Requirements (HAR)

No. Description 0 1 HAR-O 10 The SISat geoportal will be installed on a single server with x86 or x86-64 architecture. HAR-020 SISat service performance testing will be performed on a standard server. HAR-030 The SISat Geoportal will be accessible via the Internet.

6. Software Requirements (SOF)

No. Description 0 1 SOF-OIO The SISat geoportal will be implemented on a server running a GNU / Linux operating system. SOF-020 The SISat geoportal will be implemented exclusively with open source free software (FOSS) SOF-030 The SISat geoportal will include server-side programming languages (eg Python, PHP). SOF-040 The SISat Geoportal will allow you to use the GCC compiler. SOF-050 The SISat Geoportal will allow you to use the GDAL / OGR library. SOF-060 The SISat Geoportal will allow you to use the ImageMagick library SOF-070 The SISat Geoportal will allow you to use the GRASS GIS software suite SOF-080 The SISat Geoportal will allow you to use the OTB (Orfeo Toolbox) suite of programs SOF-090 The SISat Geoportal will allow you to use the SAGA GIS software suite SOF-100 The SISat Geoportal will allow you to use the SEXTANTE software suite SOF-110 The SISat Geoportal will allow you to use the GEOS library

To load the data in the PORTAL application, three sets of working instructions have been defined, one set of instructions for each Shapefile, PostGIS or GeoTIF format.

Following is the processing chain creating the "cloud mask" (optimization by the method related to the processing chain "cloud mask"), for generating and delivering maps on the dynamics of green areas and built areas based on satellite data processing, in relation to Figures 7 ... 13 which represent:

- Figure 7 - Scene I. Grayscale image obtained by using all channels, for extracting clouds. Date: July 10, 2010

- Figure 8 - Scene I. Overlay of the two preliminary masks, PCM and PSM

- Figure 9 - The final mask for Scene I. Travel parameters are lenm - 10 pixels, len_M = 30 pixels, ^ = -550, ^ft - ^M = -35o

- Figure 10 - Scene I. The contours of clouds (black lines) and shadows (white lines) over a representation in gray shades of the original scene

- Figure 11 - Scene II. Date July 20, 2010. Parameters: $ = 223, = 0.9857, 11, = 1.1, len m = 30 pixels, len_M = 65 pixels, ^a - ^m = -60o, ^a - ^M = -30o a2017 01125

12/14/2017

- Figure 12 - Scene III. Date: July 29, 2010. Parameters: = 232, = 0.9901, ^Y = 10, ^μ = 1.1, len_m = 10 pixels, len_M = 25 pixels, ^a - ^m = -55o, ^a - ^M = -35o

- Figure 13 - Scene IV. Date: August 21, 2010. Parameters: = 235, ^ = 0.9915, 10, ^μ = 1.1, lenm = 10 pixels, len_M = 30 pixels, ^a - ^m = -55o, ^a - ^M = -35o

- Figure 14 - Scene V. Date: September 15, 2010. Parameters: ^ = 245, ^ = 0.9256, ^ = 15, 1.1, len_m = pixels, len_M = 100 pixels, ^a - ^m = -70o, ^a - ^M = -45o

The clouds prevent the observation of the Earth's surface by the satellites operating in the visible and infrared spectrum. Based on the characteristics of the clouds, they must be delimited and removed from the image before further processing.

Clouds are the main disadvantage of using optical images, especially in areas with high nebula throughout the year. However, images containing cloud formations should not be neglected in the analyzes, they can be of real use if the clouds, together with the corresponding shadows, can be identified and subsequently their effects can be compensated, either by elimination or by order corrections. spectral and radiometric.

There are numerous algorithms developed in this regard, with more or less notable achievements. The methods are varied and most often involve detailed knowledge of data auxiliary to satellite images.

Perhaps the most well-known algorithm of this kind is the one implemented by NASA to assign Landsat images to scores related to the percentage of cloud coverage of a scene - Automatic Cloud Cover Assesment (ACCA). It is mainly based on the application of certain thresholds, calculation of relationships between bands and specific indices. The bands used are 2, 3,4, 5 and 6. The simplicity of the algorithm is probably what determined the use of this algorithm despite some weaknesses, such as:

> detection problems in cold areas, where the landscape generally has a high reflectance;

> imperfect discrimination between clouds and snow;

> insensitive to hot clouds;

> the performance has to suffer in case of small angles of sun elevation.

The algorithm for identifying clouds and their shadows consists of the following steps:

1. Extraction of a Preliminary Cloud Mask (MNP) containing all objects that can be passed as candidates for the cloud posture;

2. Extraction of a Preliminary Shadow Mask (MUP) containing all the dark areas and which are candidates for shadow quality;

3. Finally, only those pairs of cloud-shadows that are coherent together are added, and some areas with spectral characteristics very close to what a cloud might represent;

the whole process can be automated, but a semi-automatic method is preferred so that the operator can have greater control over the whole process.

It is considered a satellite image with several spectral bands.

The following notation is used:

f-kxCxAT where:

R 'is the set of discrete values of the spectral intensity in each band,

L is the number of rows,

C is the number of columns N is the number of bands (channels).

ΡΓΪη ⁵ (· ''), also written as Sn, is understood as n spectral channel and by marking all the spectral values in position (i, j).

a2017 01125

12/14/2017

In order to obtain derived images in different grayscale tones, the linear operator described below is used:

Is "-p (n)

Gray (S, p) = You--.

n = l where:

p is a real number,

N-dimensional parameter.

/ „(*) =

Preliminary Cloud Mask (PCM)

The spectral signature of the clouds is approximately the same in all bands (high reflectance), with the exception of thermal infrared (TIR) bands, where clouds appear in dark tones (low reflectance).

To force a unit display, a reverse color operator can be introduced:

R- S _n if the band n is TIR,

Otherwise.

you

A simple method of obtaining a cloud mask is the classification of pixels using a lower segmentation threshold:

PCM (l (S), p, 0) = Gray (l (S), p)> 0.

Preliminary cloud-shadow mask (PSM)

In order to obtain an initial shade mask, only shades of gray constructed using a special parameter are used:

if band n is TIR, then R is otherwise.

In this way, only the spectral channels where shadows appear. Moreover, in the other bands, the shadows appear as black dots (areas with low reflectance). Based on this reasoning, a new classification for the creation of a preliminary shading mask is applied:

PSM (s, p ', y) = Gray (s, p') <y, where is the maximum value of the segmentation threshold.

segmentation

A mask of clouds and one of shadows were obtained but no individual clouds and shadows were identified yet.

In order to make a correlation between the elements within the scene, it is necessary to calculate a segmentation of the preliminary masks.

The bole masks, such as PCM and PSM, are easy to segment by simply identifying and connecting areas with values of one and zero. To extract them, a mechanism based on the principle of cross-expansion is used (a pixel is added to a certain collection of pixels if it is contiguous vertically or horizontally with one of the respective collections).

After segmentation, the objects (clouds and potential shadows) from the mask collections are located in a similar way to the detection made by the human eye [5].

a2017 01125

12/14/2017

A segmentation algorithm based on two preliminary cloud masks has been described, one including the other: PCM (S, p, _S ) _< PCA / (5, p, A-5) _with Ae (0, l) _{factor extension} .

The sign <denotes that the first mask is included in the second. Specifically, each cloud in the first mask is contained by a cloud in the second mask. Each cloud is obtained by accretion processes by adding new adjacent pixels. This method guarantees that only a smaller part of the scene is classified as cloud only if it has very light spots within it and that an authentic cloud will be extracted to less dense areas on the edges.

Similarly, for the shadow segmentation, pairs of masks were built,

PSM {s, p, /) _<PSAf (s, p> A '/), _with mj f _ac tor the extension of> ^1.

Cloud mask

It was concluded that there is no perfect correlation between clouds and their shadows in a satellite image, for the following reasons:

the shadows can be partially or completely covered by the cloud that generates them or by another cloud; topography (especially in mountain areas) can change the shape of the shadow;

clouds within the same scene can be found at different heights above the ground and, by projection, the distance between them and their shadows may vary;

the density of the clouds determines the level of opacity and implicitly and the distribution and clarity of the shade;

The regular characteristics in the case of cloud-shadow relationship are brief:

Any shadow resembles a form with a cloud or part of a cloud;

The shadows are smaller in size than the clouds that generate them;

All the shadows are placed in the same direction as the corresponding clouds, and the distance between the clouds and the shadows does not vary significantly.

These principles and limitations have contributed to the definition of the algorithm.

Controlled variations for projections are allowed by using five parameters: lenrn - the minimum cloud-shadow distance (pixels), len_M - the maximum cloud-shadow distance (pixels), ^a - ^m - the minimum angle between the horizontal and the cloud-shadow direction (o) , a_M _ _a gjjj _u j _max i _m between the horizontal and the direction of the shadow (o), ovl - the minimum percentage of overlap (%).

For each potential shadow a generator cloud is searched in proximity, in polar coordinates, defined as radial distance in the interval [len m, len_M] and the angle in the interval [ ^a - ^m , ^a - ^].

The search is discrete, by moving the shadow with a round number of pixels on the horizontal and vertical axes. If more than one egg (%) of the pixels under the shifted shadow belong to a certain cloud, then both the shadow and the cloud are confirmed and marked as valid.

If the candidate cloud object is larger than a given number of pixels, P, then it is automatically considered as a cloud, without any confirmation from a corresponding shadow.

This action may have poor results in winter scenes, as snow-covered areas may be mistaken for clouds. In this case the ambiguity can be eliminated by reducing the critical number of pixels P, a2017 01125

12/14/2017 U, ΐή

RESULTS

The following are some examples of results obtained by implementing the algorithm in MatLab for five Landsat scenes within a time series (Scenes I-V), with the spatial resolution of 30 m.

The target area is the Great Island of Braila, on the Danube, Romania. Image size parameters are L = 1128, C = 1093, N = 7 for all scenes, with the sixth band being in the thermal infrared domain. The spectral resolution is R = 255.

PCM is obtained after a type transformation:

(S „S ₂ , S ₃ , S ₄ , S ₅ , S ₆ , S ₇ ,) -> (S ,, S ₂ , S ₃ , S ₄ , S ₅ , / i -S ₆ , S ₇ ,)

For segmentation of clouds in Scene I a vector p = (1, 1, 1, 1, 1, 3, 1), ^ = 235 and = 0.9 was used.

In Figure 7, the grayscale image Ρ, Α δ) _{a T} j T j _nuta using these parameters.

The shadows were obtained with a vector p * = (1, 1, 1, 1, 1, 0, 1), 15 and = 1.1.

An image in shades of flu> A7) _{es TRENDS} generated in Figure 7.

In Figure 7 the two preliminary masks (cloud and shadow mask) are superimposed (white represents clouds, gray shadows, and the rest is black). Many of the preliminary shadows are in fact simple objects with low reflectance, such as water surfaces.

Several tests were performed on different scenes and the percentages of error were evaluated. Both types of errors (omission and exaggeration) remained for 95% for cloud and over 90% for shadows.

Claims (3)

CLAIM Procedure for generating and delivering maps on the dynamics of green areas and built areas based on satellite data processing, based on extracting information from optical images taken from satellite, including for optimization the method related to the "cloud mask" processing chain characterized by that, it consists of in the sequence of the following stages: 1. The process for generating maps; > Optical image acquisition: depending on the area of use, spatial resolution, temporal resolution, radiometric resolution will be selected. > Reading and transforming optical data • S includes reading optical images and acquisition parameters, ν 'applying transformations and saving the results in a file format suitable with the software applications subsequently used for image processing; > Optical data correction: depending on the type and resolution of the optical satellite data, it is necessary to apply several types of corrections: S Atmospheric corrections - is performed using a radiative transfer model that calculates the atmospheric transmittance and radiance for various frequencies. As input parameters, the model uses data related to the atmospheric profile (pressure, temperature, water vapor, ozone), aerosol type, altitude, zenith solar angle and sensor viewing angle. S Topographic corrections - is made using a numeric altimetric field model (MNAT). S Radiometric corrections - procedures for adapting the dynamics of radiometry are used to improve image quality. S Geometric corrections - ensures the correction of the geometric deformations due to the way of recording the spectral information. > Scene connection (optional - is used only if the study area cannot be framed by a single satellite scene). It is the process by which a number of adjacent georeferenced images feel glued together to form a single image in which the common cover surfaces are removed. > Creating the cloud mask: the clouds prevent the observation of the Earth's surface with the help of satellites operating in the visible and infrared spectrum. Based on the characteristics of the clouds, they must be delimited and removed from the image before further processing. > Calculation of normalized indices of differentiation; indices are the result of mathematical operations that aim at highlighting some features of the environment using two spectral bands in which the element pursued has opposite answers. For example, when calculating the differentiated vegetation index, the bands belonging to the near red and near infrared spectral area are used, in the first case the light radiation being absorbed by chlorophyll very much, while in the field of infrared chlorophyll it has a high reflectance. > Classification of information; It involves assigning the pixels in the satellite image to predefined classes to highlight the reality from the ground [forests, uncovered land (eg cleared, arable), the built area, ..etc],> Making the final map; The information extracted from the optical satellite images is used in conjunction with other geospatial data (communication networks, localities, hydrography, toponymy, etc.) to obtain the final map. > Validation of results: it is carried out through field measurement campaigns, using appropriate sensors and equipment, or by intercomparison with other higher resolution satellite images or already validated. 2. The procedure for delivering the maps to the potential beneficiaries (and for manipulating them) consists in the creation and use of a PORTAL (WEB) computer application; the application allows the loading, handling and delivery of the maps in the formats: Shapefile, PostGIS, GeoTIF and was made in compliance with the requirements of the INSPIRE directive [code 2004/0175, Directive of the European Parliament and of the Council establishing an infrastructure for spatial information in the European Community, http: //inspire.ec.europa.eu1; 3. The method related to the chain processing "cloud mask", respectively algorithms for identifying clouds and their shadows: > preliminary cloud mask removal (PCM),> preliminary cloud-shadow mask (PSM) removal,> segmentation,> final cloud mask removal