Analysis of land development conformity obtained using photogrammetric and remote sensing methods with Geographic Information System ( GIS ) technology

Information regarding spatial management can be obtained using various methods, including satellite and photogrammetric technology. However, different methods give different results in terms of spatial resolution. Nowadays, photogrammetric data used for determining the spatial management usually employs pixels representing 0.25 m; still, the majority of the data are orthophotomaps with pixels representing 0.5 m; whereas the spatial resolution of most commonly used multispectral satellite images ranges from 20 to 30 m (for Landsat and SPOT satellites). This article presents the research results regarding the compatibility of spatial management descriptions based on photogrammetric and satellite sources. A supervised classification was conducted for two units, based on the multispectral satellite images. In this way, raster maps of spatial management were obtained. Also, vector maps of spatial management were made based on the orthophotomaps. Both kinds of description were integrated with each other in the GIS system by placing the data on separate layers. The compatibility analysis was conducted based on the analysis of two main layers representing the data obtained using the two methods. Due to the presence of raster and vector data, the value of each raster pixel was classified, and homogenous sets of data were obtained. The imaging, positioning and degree of generalization of the maps were compared, assuming parameters for the overlaying geometric objects, which was a sort of standardization.


INTRODUCTION
Researches regarding the spatial management of both urban and rural areas have been conducted in Poland for many years.Most of them are propelled by various plans and designs that are meant to organize these areas.As years go by, the methods of data collection change as a result of access to new measurement and computer technologies.Spatial management data are most commonly collected from aerial photographs and their *Corresponding author.E-mail: bartosz.mitka@ur.krakow.pl,Tel: +48 12 662 45 24.Fax: +48 12 662 45 03.Author(s) agree that this article remain permanently open access under the terms of the Creative Commons Attribution License 4.0 International License derivatives (the orthophotomaps) as well as from satellite images.However, different methods give different results in terms of spatial resolution.Photogrammetric products most commonly used for these purposes are orthophotomaps that have the spatial resolution of 0.25 m, while the images delivered by LANDSAT and SPOT satellites have spatial resolution ranging from 20 to 30 m.The use of orthophotomaps for many years has been very successful.It is a perfect solution providing topographical data for many users of various applications.It is current cartometric information about the area for use in managing and planning (Hsieh, 1996;Kerslen and O'Sullivan, 1996).In many countries, it became the basis for the creation of digital cadastral maps (Haumann, 1997;Triglav, 1998), and is now present on web geoportals at least as often as the topographic map; it is also a basic layer of information for Geographic Information System (GIS) (Pyka, 2009).Orthophotomap is also used to create tourist maps (enriched with information important for tourist purposes, such as location of monuments and churches, art galleries and museums, cinemas and theaters and public buildings; there are sketches of the most interesting tourist routes around the city) and the informative-tourist maps (applied to the planning and environmental monitoring, as well as the management of water resources for example, the creation of a water information system for monitoring the distribution of drinking water with integrated GIS using orthophoto, and also for optimization of business, industrial and service companies location) (Kerslen and O'Sullivan, 1996).
The aim of the study was to verify the results of supervised classification for public satellite imaging performed to extract the fundamental ways of land management.It was assumed that as the verification data will serve areas of management determined on the basis of orthophotomap vectorization, the authors of the study accepted the hypothesis, that such verification can be done using tools of the public GIS software packages operating under the Open Source license.
It should be noted that the aim of the elaboration was not to study the possibility of improving the spatial resolution of satellite images by integrating them with air imaging, but only a verification of the results obtained directly from satellite imaging on selected examples.
The following article presents the research on spatial management in two different settlement units-a rural and an urban one.This research allowed the authors to evaluate the compatibility of spatial data collected in different methods, using the GIS analytic tools.

Characteristics of research objects
One of the researched objects was Kasinka Mala, in the community of Mszana Dolna (Figures 1 and 2), located in a valley, in the mountains of Beskid Wyspowy.Its area covers both the bottom of the valley, as well as its slopes.The diversity of terrain is the main factor contributing to its scenic qualities.
The next research object was Miechow, an urban unit (Figure 3).Miechow is a city located in southern Poland, in Malopolskie Voivodeship.It is the head of a rural and urban community in the Miechow District.Miechow has congested, in most cases low housing with dominating single family houses, while blocks of flats are in the minority.In the centre of town there is a medieval market square surrounded by historic buildings.Major parts of the community are arable lands.

Vectorization of utilized areas basing on an orthophotomap
In order to determine the utilized areas, an on-screen vectorization of the orthophotomaps was performed using Bentley Microstation V8i software, for three selected sheets of the orthophotomaps.Two of those sheets depicted the area of the Kasinka village.One of them showed the agricultural and wooded areas and the other covered the center of Kasinka, with typical rural housing.The third sheet covered the suburbs and a part of the center of Miechow.The sheets selected in this fashion represent various forms of spatial management in rural areas and small towns.All of the sheets have the resolution of 0.25 m per pixel, same DTM error used for creation of the orthophotomap equal to ±0.6 m and similar geometric precision of the orthophotomap of about 0.3 m.Thus, the data can be assumed to be homogeneous in terms of spatial precision and geometric resolution.However, the orthophotomaps of Kasinka involve colors, while for Miechow they are monochromatic.This results in possible differences in visual interpretation of green and agricultural areas.However, this fact is not relevant for the results, as for Miechow we are dealing mostly with urban and suburban areas, which involve an insignificant amount of agricultural lands.The characteristics of the source data taken from the Documentation Centre of Geodesy and Cartography (CODGIK) is shown below.In order to generate the orthophotomaps, color aerial photos were employed.These were taken using the UltraCam Xp camera with the pixel size of 0.2 m.Also, the data obtained from aerotriangulation was used, which had been collected as a part of "Modernization and actualization of Land Parcel Identification System (LPIS) databases" project, for OB9.

Orthophotomap
The precision of the digital terrain model: ±0.6 m (for flat terrain with the average inclination <2°); ±0.9 m (for undulating and mountainous terrain).

Technology of orthophotomap creation:
The orthophotomap was created using external orientation elements of the images, defined during aerotriangulation for a block, whose range corresponds to  The map was generated using 1:13 000 images taken in March and April, 2003.The images were scanned with the pixel size of 14 µm.The map is also based on aerotriangulation data and DTM performed as a part of 49/DEG/2004/2613 project.The Digital Terrain Model precision is: ±0.6 m.

Technology of orthophotomap creation:
The orthophotomap was created using external orientation elements of the images, defined during aerotriangulation for a block, whose range corresponds to a 5 LPIS object.The DTM in TTN structure covering the whole block was employed.The orthophotomap was created as a part of description covering the map sheets mentioned above.The orthorectification was performed using bilinear method in OrthoPro software.The tonal alignment was performed using PhotoShop, I/RAS C. Mosaicing and division into modules was performed in OrthoPro software.

Verification of geometric precision of the orthophotomap:
Number of used points: 5 Average error: 0.26 m

Determination of spatial management based on the orthophotomap
The following kinds of land use were defined: i) Asphalt and dirt roads; ii) Built-up areas; iii) Waters; iv) Forests, wooded areas v) Areas covered with shrubs; vi) Agricultural lands; vii) Meadows, grasslands, green areas; viii) Other areas Vectorization was performed with the precision of ±1 pixel, which is equivalent to 30 cm in the researched area.Having in mind the terrain pixel of the satellite images, the areas determined using the orthophotomap can be used as a reference of precision of satellite image based classification.However, problems may occur during the photointerpretation of elements shown on the orthophotomap.Below, the most significant ones are listed.i) Some of the borders between areas are not clear; ii) High intensity of trees and shrubs obstructing the identification of the boundaries of areas under the crowns of trees and shrubs; iii) Difficulty in distinguishing agricultural lands and grasslands because of the time in which the photos were taken.iv) The radial movement of crowns of trees and roofs of buildings makes it difficult to identify the boundaries of areas -in particular, on the borders of dense forests; v) The definition of a built-up area.

Color composite images
Supervised classification uses so-called "training fields" -selected parts of the image (their boundaries are defined in the form of a special vector file), which can be considered as representative for each category of land use, forming patterns of classes.This allows the interpreter "supervise" the course of the automatic classification process, based on all available information, acquired not only from the contents of the image, but also other sources (cartographic materials, aerial photos, text descriptions, field observations, etc).During classification for selected training fields and for each spectral ranges there are determined the main statistical parameters such, as: average, minimum and maximum values and additionally: standard deviation, variance and covariance.Then the image is analyzed, so that his every pixel is classified to the appropriate class.
The assignment of a pixel or group of pixels to a class (e.g. the deciduous forest) on the basis of the registration of only one channel (e.g.channel 1) is difficult or impossible.However, information on the second channel makes it easier to distinguish.Information from other channels increases the chance of correctly, automatically classifying the image contents.Therefore, supervised classification is the method that can be used to automatically carried out generalizations of the image according to the pattern adopted by the user.
There are three basic methods of classification: a) The shortest distance method, b) The maximum probability method, c) The parallelograms method.
During the classification using the 'shortest distance method,' the pixel is assigned to the category for which the distance to the calculated mean value for the class is the smallest.However, the variability of pixel values for each class may vary significantly (for example the deciduous forest is characterized by high volatility and the desert areas by the low one), which means that there might arise errors in classification.To avoid this factor, so called standardized distances (normalized) are introduced, which are calculated by dividing distances from the mean by the standard deviation for a particular class.
During classification using the 'maximum probability method' for each pixel, the probability of belonging to each class is calculated and the pixel is assigned to the category for which the probability is the biggest.For each class, the average value and variancecovariance matrix are calculated which allows the assignment of the pixel to the class.
The fastest but also the least reliable method is the one that is virtually used only in didactics, the 'method of parallelograms'.It consists in defining for each distinguished category and spectral rangethe limits of the range of pixel values.For the analyzed two spectral channels, the pixels classified to the class must be found within the rectangle.When analyzing the three channels it will be a prism.
One of the most basic ways of getting information out of the measurement images is their direct interpretation.However, the image analysis and identification of the objects may be very time consuming.In order to make the process easier, and to allow getting more information out of the image, the image is saved in different spectral channels.Putting together three such images, saved in proper ranges of electromagnetic radiation, allows creating a color composite satellite image.The color composite images used in this study were created from three spectral extracts taken by Landsat 7 satellite using the multispectral scanner ETM+ in the years 2004 and 2009.
TM multispectral scanner records in seven spectral bands: four of these include the visible and near-infrared radiation, two -medium infrared range, one records the thermal radiation, thereby enabling the collection of information even in the dark.Imaging made with the scanner in the visible and near-infrared radiation have the resolving terrain power 30 × 30 m and allow the mapping of phenomena even in scale 1:25000, and terrain resolving power of infrared thermal radiation is 120 × 120 m (Mularz et al., 2000).
Color composite images are often used as the main source of information about an area.Despite their lower resolution when compared to panchromatic images, they are much more useful.The color composer images make the object recognition easier, due to their having colors and tones instead of shades of grey.Moreover, depending on the spectral ranges the composite image employs, other environmental objects of the area are exposed.
The composition of colour formed of channels 4, 5, 3 is primarily used to assess the vegetation areas coverage; it further emphasizes borders between regions used in different ways.A clear difference between the urban area and undeveloped ones can be observed.Moreover, we can clearly see the road network and development directions of urban space, differences in the layout and surface of arable land and permanent grasslands.
Channel 3 -detailed differentiation of vegetation Channel 4 -distinguishing between surface waters Channel 5 -estimating the moisture content in plants and soils.
Figures 4 and 5 present color composite images created using channels 4, 5 and 3.They are used mostly to distinguish between types of terrain coverage.

Classification of satellite image contents
The method of supervised classification of elements presented on the image allows a much greater interference of the user in the process of class creation.They are created by the user depending on the required precision, direction, and the goal of classification.This results from the fact, that in this method the pixels are assigned to the classes, and not the other way round.The principle  of this method is using the color composite image created using the least correlated channels to create the so called training fields.These are the areas that were chosen and assigned to particular classes.Thus, the supervised classification uses the terrain data in the process of class creation and assigning image elements to them.
The precision of classification depends mostly on the operator.The training fields must represent areas that are characteristic for particular kind of usage and should not contain elements assigned to another class.Thus, while selecting training fields, an area connected with a particular class should be selected with precision.
After indicating the fields, statistical parameters were calculated and were later used in the process of classification.Every training field has an average value of brightness, standard deviation and also minimal and maximal value.
In order to obtain the most representative classification results, the number of classes was reduced to five.For Kasinka Mala area, the following areas were defined (Figure 6): -Forests and wooded areasgreen -Green areasbrown -Agricultural landsyellow  -Watersblue -Built-up areas -red Table 1 shows the list of numbers of pixels used in the definition of the training fields for supervised classification of the Kasinka Mala area.
In case of Miechow the following classes were defined (Figure 7): -Forests and wooded areasgreen -Green areasbrown -Agricultural landsyellow -Built-up areasred -Roadsblack There are four basic methods of conducting supervised classification: the parallelepiped method, the minimum distance method, the Mahalanobis distance method and maximum probability method (Eastman, 1992).
In the study area, several attempts of supervised classification were carried out using: the shortest distance, maximum probability and parallelograms methods.Satellite images from different time periods were used.For the selected areas (Kasinka Mala, Miechow), the shortest distance method gave the most promising results.
The minimum distance method was applied in the study.The spatial distribution of the pixels forms clusters that can be enclosed within an ellipse.Classes created in this way have different density and are created from the pixels of training fields described by the statistical parameters.The closer it is to the center of the class (ellipse), the higher the density.Also, the pixel brightness tends to the average value when going towards the center.During classification the pixels are analyzed in terms of Euclidean (geometric) distance from the centers of defined classes.The inspected element will be assigned to the nearest group.The distance is calculated from the formula (Adamczyk and Będkowski, 2005): (1) where: Eigeometric distance of inspected pixel from the i th class Nnumber of classes jspectrum channel id ginspected pixel location μi th class center location Supervised classification can also be the source of data regarding the spatial management of the Earth's surface.The freedom in class creation and the possibility of taking into account many terrain factors during classification is the main advantage of this method.
Because of the simplicity of the process and relatively high precision of the results it is a good alternative for traditional 2 1 () methods of satellite image interpretation.Supervised classification can be used in various stages of spatial planning, landscape architecture, investigating rate of change of urban areas, agricultural lands, forests, and many others.

APPLICATION OF THE SPATIAL DATA ANALYSIS METHOD
The aim of this stage was comparison of two layers obtained from the satellite and photogrammetric methods using geoprocessing tools from the GIS software.To perform the job, the whole software OpenSource was used: operating system OpenSUSE 12.1 and Desktop GIS including: Quantum GIS 1.7.4 and GRASS 6.4.
Comparison of the images originating from two sources could be performed under the condition of application of the same projection for different materials.In order to do this each photo was given an attribute EPSG of the projection and each photo was converted to the "2000" coordinates system (EPSG 2178) which is used in Poland for geodetic works (Figure 8).
Coordinate system "2000" for the whole geographic length is divided into four strips at intervals of 3° of length, which gives accuracy <7 cm on the edge of strips.After conversion, the images overlapped making the mosaic of lines, most often not overacting areas of similar use.It was necessary to convert to vector model, because in this model it is possible to link many descriptive attributes with each polygon representing the kind of land use (Coen, 2002;Hughes and Forsyth, 2006).For the orthophotomap, the vectorization was performed manually, while for color satellite image, values of pixels for all color channels were transformed to common feature value using the formula: (sat_02@1*7+sat_02@2*17+sat_02@3*37+sat_02@4*5 3)/(sat_02@1 + sat_02@2 + sat_02@3 + sat_02@4) (2) The formula had to join channels and preserve unique values of new pixels, therefore prime numbers were used as multiplier (Stones and Matthew, 2002), (Salata, 2008).Supervised classification of satellite image was performed in RGB colors, which caused pixels of particular color channels to have extreme values 0 or 255, so in numerator the value of any channel was multiplied by the prime numberthe others were equal zero.Denominator was divided by total value of pixels in all channels of the raster.Thus, the image presented in Figure 9, in which values of pixels created groups corresponding with particular land use was obtained.The image was vectorized automatically in the next stage.
The next stage was automatic vectorization using functions enclosed in GRASS software, namely function "r.to.vect":
-b --Do not build vector topology (use with care for massive point export) -q --Quiet -Do not show progress --overwrite --Allow output files to overwrite existing files --verbose --Verbose module output --quiet --Quiet module output After obtaining vector form with linked attributes table, which is a matrix of pixels values for areas of particular use, each area was assigned descriptive value: agricultural areas, waters, built-up, and so on.The talk is about areas, because adjacent pixels of the same value created one area, without borders inside (Figure 10).At this stage it became possible to join thematic levels, their mutual analysis, multiplying objects situated on two layers.It became possible to separate areas which are unique in a scene of sum of sets of attributes for orthophotomap and satellite layers (Stones and Matthew, 2002).Each of separated areas possesses unique set of attributes, describing the land use way certified on the basis of orthophotomap and satellite image.Attributes were named "*_ortho" and "*_sat", and they are products of consistent and inconsistent terrain functions.Presently, state area was calculated for each area and set together using application in Quantum GIS named GroupStat.Areas, where land uses are consistent have similar values of attributes, for example."water_ortho"; "water_sat"; "buildings_ortho" and "buildings_sat".Inconsistent configurations mean inaccuracy in interpretation of land use of areas by analyzed methods (Salata, 2012).

ANALYSIS OF COMPATIBILITY OF SATELLITE IMAGES WITH THE ORTHOPHOTOMAP BASED SPATIAL MANAGEMENT CLASSIFICATION
Because the geometric resolution of the orthophotomap is two orders of magnitude higher than the resolution of satellite images (0.30 m per pixel for the orthophotomap, 30 m per pixel for the satellite images), the spatial management obtained from the orthophotomap vectorization was taken as a reference for evaluation of the satellite image based classification.
For Kasinka area (rural area in mountainous terrain) the following results were obtained (Tables 2 and 3).Tabulation presented below presents inconsistencies in mutual interpretation of orthophotomaps and satellite imagery for each class obtained from the orthophotomaps.The decisive factor in the direction of the analysis is primarily much higher resolution of the air  model, and also the possibility to determine a larger number of classes of land use.Presentation of Table 1 area approach should be more specific using the presentation of the percentages of selected noncompliances.Built-up areas in dispersed settlements do not reflect well on satellite imaging and are spread over many classes marked with the suffix "_sat": buildings_sat (19.8%), forests_sat (0.9%), green_sat (32.0%), rural_sat (21.0%) and water_sat (26.4).Presentation of relative discrepancies in Figure 11, shows the areas where the interpretation should be as high as possible: "forests_ortho"  "forests_sat" -compliance level 61.5%,"green_ortho"  "green_sat"low compliance level 33.3%, "rural_ortho"  "rural_sat" -compliance level only 6.0%, "water_ortho"  "water_sat"compliance level 70.6%.On the examined object, some classes are also interpreted differently than one would assume.These are: agricultural areas interpreted in 80.8% as green areas, and green areas ("green_ortho") interpreted as: "water_sat" 29.0%, "rural_sat" -14.4%, "green_sat" -33.3%, "forests_sat" -2.9% and "buildings_sat" -20.3%.Linear features (roads) have been identified as scattered randomly and evenly over all satellite classes.Table 2 presents spatial data in terms of the absolute value of the surface in hectares.The table shows that areas with small-sized area, close to the limit  of satellite imaging resolution do not image there properly or do not at all, e.g."roads_ortho" or "bushes_ortho" -the correctness of the interpretation of a few percent level.The increase in the surface of the compact area significantly improves efficiency of the imaging -the correctness of the interpretation at the level of 61.5% ("rural_ortho" "rural_sat").In Table 3, the color blue indicates classes of objects that should show the highest values of the coefficients of conformity in relative terms between them, but only forests proved to be clearly interpreted classes on two types of images: aerial and satellite.Red highlights definitely misclassified areas, and brown -classes ambiguous even in aerial photographs.These precise classes could take the high values, but remain at a level of 11.50 to 23.31% of satellite imaging incorrect classification.Analysis of the data above shows, that the compatibility between the orthophotomap and the satellite image is satisfactory only for forests (Figure 12).Built-up areas were mainly in three kinds of land areas: housing, agricultural lands, and wastelands (in total 80.6% of the area classified as built-up).This is a result of scattered nature of housing in the area as well as the presence of fields and other non-built-up areas in very close neighborhood (Figure 13).The areas classified on the satellite images as green areas are majorly agricultural lands and wastelands (together 86.0% of classified area).It happens, because the green areas are not much distinguishable from agricultural lands during vegetation period (Figure 14).
What is interesting for the area is the result of agricultural lands.Because of the mountainous terrain and the inclination of the slopes, large forests were classified as agricultural lands.This is an example of possible errors of the supervised classification algorithm, and a proof that the results need supervision (Figure 15).The result of water classification is also interesting (Figure 16).In the researched area, the border between the areas without trees and forests was classified as water.Also, the areas of low vegetation along roads and streams were assigned to this class.For the area of Miechow (urban area located on flat terrain) the following results were obtained (Tables 4 and 5).In Table 4 the list of overlapping areas with various methods of use is presented.To correctly interpret the list, analysis should       resolution of 30 m, so in this case unilateral dependence of misinterpretation on the side of this material can be concluded.
The table presents the interpenetration of classes of built-up land uses determined in the table as "buildings_ortho" into three categories of satellite imaging, with the fact that the largest percentage (82.9%)accounted for the category labeled "buildings_sat" and areas "forests_sat" and "rural_sat" respectively 8.2 and 8.8% of the area.Areas covered with bushes are visible on the orthophotomap in more than half of the surface and distributed on satellite imaging to built-up areas (61.2%), forests (24.9%) and agricultural areas (13.9%).Forest areas in Miechow occupy a small area (1.89 ha),while almost all (93.1%) were classified as built-up areas on satellite imaging.The distribution of green areas on satellite imaging is as follows: 56.1% as "building_sat", 9.0% a "forests_sat" and 34.9% as "rural_sat".A similar situation exists for roads, wasteland and water -objects with a linear charactermost often are interpreted as built-up areas "buildings_sat" respectively: 82.0, 88.2 and 72.7%.Agricultural areas on orthophotomap andsatellite imaging show interpretation compliance at the level of 67.3% and water areas in a high proportion (27.3%) are interpreted as "forests_sat".
Table 5 shows the matrix of relationships between the objects acquired from two sources: orthophotomap and satellite imaging blue color indicated theoretically the same nature of areas, with high relative compliance.For forest areas it takes a very low rate (8.8%), while the correlation is higher for built-up areas (52.91%) and poor for agricultural areas (22.18%).The color red indicates undesirable variation in the interpretation of test methods, namely: 47.57% of built-up areas (buildings_ortho) has been interpreted as forests (forests_sat), 18.86% as rural areas (rural_sat).Class "bushes_ortho", "roads_ortho" and "rural_ortho" are not directly visible in mapping satellite; therefore they were not marked in the table and are differently allocated to the classes of satellite imagery.
Analysis of the data above shows, that the results were not satisfactory for any of the classes.This conclusion has been confirmed visually (Figure 17).Additionally, for the analyzed area there was no way of defining on the satellite images the remaining development classes marked on the orthophotomap.

Conclusions
Performing comparison by GIS methods, with application of geoprocessing and precise tools classifying areas on basis of the attribute tables enables quick and half automatic obtainment of quantitative data related to forms of land use.Based on the geometric models resulting from intersection of borders of similar classification, but using different methods, it became possible to gain analytic material, providing that comparison of effects of these methods is relevant.
Analytic accuracy is very high.Factors disturbing obtained accuracy are connected with mutual location of source data: orthophotomap and multispectral composition.The second aspect of performed analysis revealed the necessity of looking for adequate combination of channels and adjusting training fields for performing supervised classification.It turns out that for some areas (for instance forests) choice of channels set brings expected results, while for correct designation other forms of land use (covered with bush or wastelands) some other sets are needed.
Quantitative analysis shows, that bigger areas have a chance to have good reflection on satellite images, with pixel of dimension 30×30 m; then on the edges of those areas appear differences related to the location of the border and it depends on the size of pixel.On areas, where forms of land use are smaller than pixel dimension of satellite image, it is necessary to adjust set of channels from the composition after deep observation of land use based on orthophotomap.From the results, it can be concluded that high compatibility of satellite image based classification with the areas determined based on orthophotomap can be achieved only for forests.In other cases, the results diverged, which means the results supervised classification should be interpreted with caution, as they may be unreliable.

Figure 4 .
Figure 4. Color composite image created using the channels 4, 5 and 3, for the area of Kasinka Mala.Source: own research.

Figure 5 .
Figure 5. Color composite image created using the channels 4, 5 and 3, for the area of Miechow.Source: own research.

Figure 8 .
Figure 8. Choice and assignment of projection for layers in Quantum GIS program.Source: own research.

Figure 9 .
Figure 9.One channel image as an effect of joining color channels.Source: own research.

Figure 10 .
Figure 10.Methodology of spatial analysis performing: (a) Layer of vectorized orthophotomap, (b) Layer of vectorized areas classification on the basis of satellite image, (c) Overlapping layers from Figures a and b, and (d) Product of compared layers together with merging their attributes.Source: own research.

Figure 11 .
Figure 11.Relative compliance interpretation of classes use areas.Source: own research.

Figure 12 .
Figure 12.Compatibility of forest classification.Source: own research.

Figure 13 .
Figure 13.Compatibility of built-up area classification.Source: own research.

Figure 14 .
Figure 14.Compatibility of classification of green areas.Source: own research.

Figure 15 .
Figure 15.Compatibility of classification of agricultural lands.Misclassification of a forest as an agricultural land.Source: own research.

Figure 16 .
Figure 16.Compatibility of water classification.Source: own research.

Figure 17 .
Figure 17.Results of supervised classification compared to the orthophotomap.Source: own research.

Table 1 .
The list of numbers of pixels used in the study in Kasinka Mala.
Source: own research.
Source: Own research.

Table 3 .
Compatibility percentage for satellite classes.
Source: Own research.

Table 5 .
Compatibility percentage for satellite classes.
Source: own research.