3.1. Locality
When the locality of interest is about to be chosen, a landscape with wide ranging features would be an appropriate choice [
19]. A landscape part spectrum (an open country, forest areas, low built-up areas, habitation) is an ideal choice for the orthophoto-images post processing method for the working environment testing. Accordingly, the wider Solinky surroundings, which form the southern part of Žilina, were chosen. The measured area boundaries were defined as the main settlement catchment area. We may understand the settlement itself and the chosen non-urban part that has been traditionally connected with the settlement, under this term. The settlement extent and dimensions are adequate for the research purposes for the modern data acquisition and postprocessing testing. In the
Figure 1, the basic examined landscape features are depicted based on their use/purpose(s). The examined area includes several landscape features or types, starting with the high built-up areas, segment A (determined by boundaries of the housing estate) represents the settlement itself. The less built-up areas represented by the suburbs, segment D (determined by boundaries of cadastral area), respectively by the gardening areas. Segment E (defined by boundaries of area used for gardening in cadaster of real estates) and various natural feature types used for recreational and agricultural purposes, segments B, C and F (differenced visually for segments B and C and by the boundaries of agricultural usage for segment F). The single segments purpose and percentage are stated in the
Table 7 a graphically depicted in the
Figure 2.
The examined locality overall area was 244.2 ha. For the following processing purposes, the correct measuring way and Unmanned Aerial Vehicle (UAV) method is necessary. The measuring method and measured data processing is listed in the following chapter section.
3.2. Aerial Photogrammetry by UAV
In the process of creating a detailed orthographic photograph of the earth’s surface were used only aircrafts. With the current trend of using Unmanned Aerial Vehicles (UAVs) in various applications, unmanned aerial vehicles have found application in this area as well [
20]. For the purpose of research and experiment was used UAV DJi Mavic Pro. The experiment consisted of the following phases:
The flights were flown in the controlled area of Žilina Airport (LZZI) at a time when the area is in the category ATZ (Aerodrome Traffic Zone) and outside the descent planes GNSS [
21], and ILS approach and within the applicable rules of flying and safety procedures [
22] with UAV according to Decision no. 2/2019 of 14 November 2019. The front image overlaps of 75% and side overlaps of 65% were used in the imaging, the imaging speed was set to 8 m·s
−1. The total flight altitude was 120 m AGL (Above Ground Level) with direct visual contact of the operator. The total flight time was 27 min, during which 2 batteries were used, 487 photos were taken and an area of 115 acres (0.46 square kilometers) was covered.
For images collection was used default camera of DJI Mavic pro drone with following parameters:
Sensor: 1/2.3” (CMOS), Effective pixels: 12.35 M (Total pixels:12.71 M),
Lens: FOV 78.8° 26 mm (35 mm format equivalent) f/2.2, distortion < 1.5% Focus from 0.5 m to ∞,
ISO Range: video: 100–3200, photo: 100–1600,
Image Size: 4000 × 3000.
The total resolution while maintaining the current flight parameters was 3.3 cm of real distance in the image per pixel. In terms of the selection of the season and time during the day for the implementation of the flight plan [
23], the period was selected according to the following parameters. The first most important parameter was imaging with the elimination of shadows [
24] from buildings and vegetation. This fact could be avoided by two factors. By choosing a period with the sun as high as possible above the horizon and choosing light meteorological conditions. Aerial photography close to the date of the longest day provided optimal conditions. In terms of meteorological influence, the day was chosen when there were high cirrus cloud conditions in the sky. The cloud type provided lighting conditions and light diffusion that produced softer shadows than on a clear day. In addition to meteorological conditions and forecast [
25] for chosen landscape area, the level of developed vegetation in the period without flowers and with leaves was also taken into account. The last factor was the period between precipitation and severe drought. It was the extremes in the supply of water to the ecosystem that caused problems in the subsequent evaluation of the area. One extreme was the formation of soaked soil [
26] and the other was the sore dry vegetation. Both cases belong to an environment with vegetation, but by their nature during the evaluation they would be included in an area without vegetation.
To complete the whole area of research, aerial work has been divided into two polygons. The first polygon was described in
Figure 3 and its parameters were defined above. Due to the area of the second polygon (
Figure 4), its relief and character, the flight and imaging conditions were adapted to the total flight time. While in the polygon covering the built-up area, the flight altitude was set to 120 m (AGL); when shooting the current polygon, it was 150 m (AGL). Again, all safety conditions were observed during the shooting in accordance with Regulation 2/2019 of the Transport Office of the Slovak Republic in the current conditions of ATZ (Aerodrome Traffic Zone) Žilina. The total aerial photography time of the second polygon was 58 min and a total of 4 batteries were used. The current area was recorded on 1159 images. The resolution of the polygon with the given parameters and a side overlap of 65% and a front overlap of 75% was 2.5 cm per pixel.
The final clipped area used in the next sections is displayed on the
Figure 5.
3.3. Postprocessing
The post processing contains various steps. The sample of postprocessing is presented in
Figure 6 and it includes steps such as definition of location, which is measured (in our case using the UAV method), the final data must be edited and corrected, for next usage it has to be georeferenced [
27,
28,
29,
30]. After the preprocessing, data are able to be used for the aims, that are described below.
After the flight over the area, images processing and orthophoto-images creation, the orthophoto-images were used as the base for the next calculations. Individual orthophoto-images were reclassified in the program ArcMap environment using the toolbar Image Classification, specifically ISO Cluster Unsupervised Classification, as it meets our requirements for the locality of interest orthophoto-images division in the best way. For correct elimination of graphic errors such as shadows, before Cluster Unsupervised Classification, Principal Component was used.
The
Principal Components tool is used to transform the data in the input bands from the input multivariate attribute space to a new multivariate attribute space whose axes are rotated with respect to the original space. The axes (attributed) in the new space are uncorrelated. The main reason to transform the data in a principal component analysis is to compress data by eliminating redundancy [
16,
17,
18]. The result of using the tool is a multiband raster with the same number of bands as the specified number of components (one band per axis or component in the new multivariate space). The first principal component will have the greatest variance, the second will show the second most variance not described by the first, and so forth.
Principal Components requires the input bands to be identified, the number of principal components into which to transform the data, the name of the statistics output file, and the name of the output raster [
19]. The output raster will contain the same number of bands as the specified number of components. The number of components is the same number mentioned earlier. Each band will depict a component [
31].
The result of the
Principal Components reclassification (post-processing) is displayed in
Figure 7. Without any other necessary editing it there is possible to see the differences between basic landscape structures such as built-up areas (mostly in yellow), permanent grasslands (mostly in blue), forest areas (mostly in green) and arable areas (mostly in purple).
ISO Cluster Unsupervised Classification performs the reclassification of the input orthophoto-image using the
Maximum Likelihood Classification tools. This algorithm is based on two principles: The cells in each class sample in the multidimensional space being normally distributed Bayes’ theorem of decision making [
20]. To assign each cell to one of the classes the tool takes into consideration both the variances and covariances of the class signatures. If the distribution of a class sample is normal, a class can be characterized by the mean vector and the covariance matrix. These conditions enable determination of the cells membership to the class.
Within the
ISO Cluster Unsupervised Classification application in the ArcMap environment [
31] after choosing the georeferenced orthophoto-image [
32,
33], it is necessary to define the number of classes into which the orthophoto-image is about to be classified. The
Minimum class size which defines the minimum number of cells in a valid class must be predefined too, together with the
Sample interval. This is an interval used for sampling. The number of the intervals within the postprocessing of every single orthophoto-image had to be adopted according to the orthophoto-image content [
12]. During the postprocessing the orthophoto-image was divided into the irregular shapes which delimit the landscape structure type. For explanation, in the section formed by the housing development with various roof types, respectively with the various facade color solutions, the misinterpretation may have occurred according to the wrong class determination as it is automatized. If low numbers of classes were used, some valuable structures could fade away in the settlement territory. In the
Figure 8, the reclassified classes of the segment F are possible to be seen. After many experiments the most appropriate number of reclassified classes for built-up areas is 50 and for mostly monocultural areas is 25. The orthopohoto reclassification into four basic landscape structures is displayed in the
Figure 9.
According to the reclassification into four basic landscape structures, the area of each land structure was calculated and graphical displayed in the
Figure 10. The last step of the post image processing is the calculation of index
CES, not only for whole examined area, but also in the separate segments The pixel classification was used due to working with UAV data presented by the orthophoto-image. For the first
CES formula, there are no values for the categories, they are only divided into two groups, one is in final calculation the numerator, the other is the denominator, the final
CES index presents the ratio of focused area. The index may reach values from 0 up to infinity. For the second
CES formula, indexes for single landscape type structures were given as 0.20 for arable areas, 0.50 for temporary grasslands, 0.80 for built-up areas and 1.50 for forest areas. In this calculation, the areas percentage values are necessary. The
CES2 values may reach up to infinity as well. Within the
CES3 calculation, the index for temporary grasslands had to be classified differently for segments A, D, E, where its value was 0.50 which presents garden areas and in segments B, F the index value was 0.62 which presents meadows. The overall
CES3 value of examined locality was calculated as weighted arithmetic mean presented by value 0.59 for temporary grasslands. The index value for built-up areas was given as 0.00. for arable areas 0.14. In the
CES3 calculation, the index value for forest lands was given as 0.63 according to the tree types predominance in the examined areas. Within the
CES4 calculation, the built-up areas are defined as areas with no importance with index 0.00. The index value for arable areas was set as 1.50 that represents the compromise between the areas with low and very low importance. There is not a clearly defined area which can be called a small or large arable land. The index value for meadows is set as 3.00 and for forest areas 4.00 whereas the trees creating forests cannot be defined as natural with index 5.00.
According to the
CES calculations using various formulas, single segments reached different
CES values, which are object of interpretation. The final
CES values are shown in the
Table 8 below.