*3.5. Results: Working out Final Maps*

which was 13.9 m. The value defined the average distance the pedestrian could cover in 10 s, i.e., during the interval between individual images. In our calculations, we assumed that the average walking speed was 5 km/h [28–30]. To avoid the aggregation of polygons located on the opposite **Figure 8.** The map of informal areas of land used by pedestrians: the chorochromatic method (qualitative data). **Figure 8.** The map of informal areas of land used by pedestrians: the chorochromatic method (qualitative data).

values of average intensity only for classified fields, they are considered dasymetric maps. In the research the dot method was used for working out the heat map that presented quantitative data by means of the area method (the dasymetric map). The heat map was created in the software using the *Kernel Density* tool. To work out the *heat map* [39], we used point objects presenting the location of pedestrians. According to the nomenclature of cartographic methodology, to design the heat map (Figure 9), we used a dasymetric mapping method, presenting the average phenomenon intensity only in classified fields, i.e., areas in which the presence of pedestrians was identified. To boost the effectiveness of data presentation, we showed the data in the background of the research area infrastructure, worked out with the use of the chorochromatic method. We calculated pedestrian

density in the research area and provided it in the form of the number of people per 1 m<sup>2</sup>

map of the observed pedestrian movement allowed us to obtain information about the frequency of use of specific research areas by pedestrians (Figure 9). It showed that, despite the existing transport infrastructure, pedestrians use areas that are not meant to be used for walking. Moreover, the map

. The heat

sides of sidewalks, the dividing layer (i.e., the transport infrastructure) was determined. Because of

The presentation of large sets of quantitative data makes it difficult or even impossible to

**Figure 9.** The map of pedestrian density in the research area (continuous scale): the chorochromatic method (qualitative data). **Figure 9.** The map of pedestrian density in the research area (continuous scale): the chorochromatic method (qualitative data).

The use of the heat map compensated for the disadvantage of the dot map, i.e., the lack of differentiation when the dots touch each other or overlap. Therefore, we proposed to present quantitative data using a continuous-scale heat map to obtain information about the intensity of pedestrian traffic in the study area. Unlike the dot map, the heat map presents surfaces with a cumulative number of point features. In the case of this test, it is an area of 1 m<sup>2</sup> . Additionally, we proposed a second version of the heat map (Figure 10), with a more interpretable step scale in four intervals based on Jenks' natural breaks classification method. The ability to specify the accurate location of pedestrians in the images obtained and a large scale of the map resulted in the conclusion that the surface object would make the more accurate representation, allowing us to present the area occupied by an individual person in the way that resembles the actual state. To do that, it was necessary to transform the point representation of pedestrians into the polygon representation. To determine the area occupied by an individual pedestrian, we have specified the value of a buffer by means of which transformation would take place. We assumed that the buffer value of a single point would be 37.5 cm (diameter of 75 cm), which corresponded with the largest length of the pedestrian's footstep (62.5–75 cm) [27]. Then, we marked out the area of the actual land use based on the obtained polygon objects representing pedestrians. The aim of this was to determine the areas informally used by pedestrians. With recording pedestrians at intervals, it is not possible to determine the location of individual pedestrians between the images obtained. In the research, we have assumed that obtaining images at short intervals and the average speed of pedestrians of 5 km/h [28–30] would allow us to determine the aggregate distance between pedestrians captured in individual images. Figure 3 depicts a concept of point-to-polygon transformation of the pedestrian's representation. It shows the different stages of identification and cartographic representation of pedestrians in the research area, as well as the final result presenting the area of the actual land use.

#### *3.2. Obtaining Low-Level Imagery*

In this research, we decided to obtain spatial data by means of the low-level imagery method. The time of recording pedestrians' movement in the research area with the employment of UAVs strictly correlated with the maximum pedestrian movement intensity. A field survey allowed us to determine that pedestrian movement in the selected research area was the most intense between 7:30 and 8:15 on weekdays during the academic year. Thus, we decided to conduct the research with the employment of UAVs specifically during this time. Having considered the size of the research area and the necessity to observe pedestrian movement, we concluded that the multirotor UAV would be the best choice in

terms of obtaining visual data. We used the Tarot X6 platform, equipped with a camera with 16.1 Mpx matrix. Although the platform was equipped with the Global Navigation Satellite Systems (GNSS), we decided that the aerial triangulation process of the imagery obtained would be carried out with the employment of GCPs located in the research area, which allowed us to make the process as accurate as possible [31,32]. The GNSS RTK technique is one of the methods of measuring GCPs [33]. For the measurement, we used the GNSS Trimble R4 Receiver, model 3, with differential corrections provided by the permanent reference station network. To create as accurate photogrammetric analysis of the research area as possible, it was necessary to record it from all sides and to work out the 3D model, using the Stucture-from-Motion (SfM) algorithm [34,35].

It was important to provide security during UAV raids. The observation of pedestrian movement was conducted in accordance with the existing provisions of law. In the study, the priority was to ensure safety for the operator, the UAV platform, and most of all for people in the vicinity during the flight mission using UAV technology. Bearing in mind safety and legal regulations, it became impossible to fly directly over pedestrians. This was mainly due to the inability to obtain images using the classic photogrammetric flight path based on photogrammetric series. In order to achieve the aim of the study, we have determined that it is necessary to register all pedestrians moving in the study area at the same time. We also assumed that the registration of pedestrian traffic would be carried out from one observation post. Such an approach in the conducted study forced the necessity to obtain oblique photos. We used an unmanned aerial vehicle (UAV)—a multirotor platform—which ensured stable registration of the research area from one observation position. The photogrammetric images obtained were processed in the Agisoft Metashape Professional software. The GCP-based aerial triangulation process resulted in obtaining the average RMSE value for all GCPs at the level of 0.27 m. Orthophotomaps were exported from Agisoft Metashape Professional for all the images obtained in the GeoTIFF format to identify pedestrians recorded in the research area. During the recording of pedestrian movement in the research area, we obtained 275 images in total.

#### *3.3. Identification and Vectorization of Pedestrians*

The orthophotomaps, created for all images, were then imported to the GIS software. In the study, we will use the ESRI ArcMap software to process spatial data and develop thematic maps. In this software, we vectorized pedestrians recorded in individual images to determine their exact situational coordinates (X; Y). It was difficult to determine the accurate vectorization place of each pedestrian, as pedestrians were constantly moving. We specified that the place of projecting the center of gravity of each pedestrian in the research area plane would constitute the vectorization point (Figure 4); this was conducted for the 275 images obtained at 10-s intervals.

#### *3.4. Cartographic Visualization of Pedestrians' Location*

In our attempt to meet research objectives on designing maps that use adequate mapping techniques for presenting the real land use by pedestrians, we adopted the following order of creating intermediary maps:


Mapping techniques were listed in terms of data type (geometry of objects).

Vectorization conducted in the research allowed one to obtain quantitative data presenting the location (X; Y) of pedestrians in the research area. The most natural way to present such data is to use dots [36]. For presenting quantitative data we used the dot method. Points as 0-dimensional objects are represented with a pair of coordinates [24], as presenting a point object on the map through direct physical representation of point geometry would make such object invisible. To read its location, it was necessary to use the additional attribute, the dot ratio. When designing our dot map, we assumed the dot width (diameter) of 0.5 mm, because it is the smallest symbol size recognizable by the map user [37]. The dot created is a cartographic representation of a point feature, and the geometric center of the dot corresponds to the coordinates of the point it represents. On the previously prepared map (Figure 5) pedestrians were represented as objects in the shape of a small circle, so that it is visible to map users. This is only the cartographic representation of a point object on the map. Quantitative data were presented in the background of the map showing the infrastructure of the research area (Figure 2). Such combination of methods of cartographic presentation made it possible to present both qualitative and quantitative data. As a result, researchers obtained a map for the creation of which two methods, the dot method and the chorochromatic map were used. In Figure 5 we demonstrated the spatial layout of all the recorded pedestrians (quantitative data) in the background of the infrastructure of the research area (qualitative data).

In the research. we decided to mark out the areas incorrectly used by pedestrians, i.e., informal land use. Pedestrians, who were moving outside the specified transport infrastructure, were considered informal land users. To single such people out of all recorded pedestrians, we carried out subtraction on the quantitative data set. Such activity was possible thanks to the use of the *Erase* tool, which allowed to localize only those pedestrians outside the designated transport infrastructure. In Map 6 we marked pedestrians using land in the informal way. As in Figure 5, for presenting quantitative (pedestrians) and qualitative (infrastructure) data, we used two mapping methods, i.e., the dot method and chorochromatic mapping.

A small research area and the UAV technology employed allowed us to obtain accurate location of individual pedestrians. We concluded that the representation of pedestrians by means of the dot mapping method was not adequate. It was related to the opportunity to work out large-scale cartographic visualizations that can faithfully reflect pedestrians in the research area. The intermediary objective of the research was to determine the area of the land that pedestrians used incorrectly, i.e., informal areas. We specified that it was necessary to carry out the point-to-polygon transformation of pedestrians' representation.

Transformation was understood as a change of the mapping technique, i.e., the transformation of the geometry of the element used in cartography, which would allow us to determine the area that pedestrians occupy outside the transport infrastructure specified. The average length of a human footstep is between 62.5 to 75 cm, depending on sex and type of walk [27]. We assumed that the area occupied by an individual moving pedestrian would be equal to a circle that is 75 cm (the maximum length of a human footstep) in diameter. That meant that each point object representing an individual pedestrian was been surrounded by a buffer of 37.5 cm, counting from the values of coordinates describing the location of the point. After transformation, one person was represented by means of a polygon object with the geometry of a circle occupying the area of 0.44 m<sup>2</sup> (Figure 7). The ArcMap *Bu*ff*er* tool was used to transform point features into area features. Transformation of the point representation of a pedestrian into the polygon representation resulted in the change in data type (from quantitative to qualitative data). The map presenting the area occupied by pedestrians outside the transport infrastructure (Figure 7) was worked out with the use of the chorochromatic mapping method. The buffer value adopted in the research constitutes just an example and can be modified, depending on needs and the assumptions of the research.

#### *3.5. Results: Working out Final Maps*

Working out intermediary maps (Figures 5–7) made it possible to create additional cartographic analyses that effectively demonstrate the actual land use by pedestrians. Since we distinguished two mapping techniques, the following map types were worked out:


The transformation of point objects into polygon objects (the representation of pedestrians on the map) allowed us to determine the area occupied by individual pedestrians in the research area. The area in question refers only to pedestrians recorded in individual images and does not determine the land use area. When obtaining images at specific intervals, it is impossible to obtain data on the location of individual pedestrians in the research area between the states recorded. However, the large number of recorded pedestrians in individual images of the same area and visibly repetitive pedestrian routes in the research area make it possible to obtain data on the location of pedestrians at the time between individual images. In the research, individual images recording the research area were obtained at 10-s intervals, which was related to the observation of the dynamic phenomenon. To determine the actual size of informal land use areas in the analyzed research area, we decided to aggregate (generalize) the polygons created. For this purpose, the *Aggregate Polygons* cartographic generalization tool available in the ArcMap software was used. During the process of defining aggregate variables we determined the aggregation distance representing individual pedestrians, which was 13.9 m. The value defined the average distance the pedestrian could cover in 10 s, i.e., during the interval between individual images. In our calculations, we assumed that the average walking speed was 5 km/h [28–30]. To avoid the aggregation of polygons located on the opposite sides of sidewalks, the dividing layer (i.e., the transport infrastructure) was determined. Because of this aggregation, we obtained the area of the informal land use by pedestrians in the research area (Figure 8). In addition, the repeatability of pedestrian traffic along the transport infrastructure, recorded on interval images from Low-Level Aerial Imagery and carrying out cartographic generalization consisting in the aggregation of surface objects presenting individual pedestrians on images made at different times, allowed for the estimation of their paths, and obtaining information about the location of pedestrians in the time between single images. For working out the map, we used the chorochromatic mapping method that presents qualitative data on land use. Additionally, aggregation allowed us to generalize areal data presenting the area occupied by individual pedestrians and to obtain the total area of informal land use (Figure 8).

The presentation of large sets of quantitative data makes it difficult or even impossible to interpret the represented phenomenon correctly. To effectively present the results of the quantitative data set study, we used the *heat map* [23]. In traditional cartographic methodology maps presenting quantitative data are considered cartograms if they cover the entire area [38]. However, if they show values of average intensity only for classified fields, they are considered dasymetric maps. In the research the dot method was used for working out the heat map that presented quantitative data by means of the area method (the dasymetric map). The heat map was created in the software using the *Kernel Density* tool. To work out the *heat map* [39], we used point objects presenting the location of pedestrians. According to the nomenclature of cartographic methodology, to design the heat map (Figure 9), we used a dasymetric mapping method, presenting the average phenomenon intensity only in classified fields, i.e., areas in which the presence of pedestrians was identified. To boost the effectiveness of data presentation, we showed the data in the background of the research area infrastructure, worked out with the use of the chorochromatic method. We calculated pedestrian density in the research area and provided it in the form of the number of people per 1 m<sup>2</sup> . The heat map of the observed pedestrian movement allowed us to obtain information about the frequency of use of specific research areas by pedestrians (Figure 9). It showed that, despite the existing transport infrastructure, pedestrians use areas that are not meant to be used for walking. Moreover, the map makes it possible to draw a conclusion that such phenomenon occurs very frequently in some parts of the research area.

The use of the heat map compensated for the disadvantage of the dot map, i.e., the lack of differentiation when the dots touch each other or overlap. Therefore, we proposed to present quantitative data using a continuous-scale heat map to obtain information about the intensity of pedestrian traffic in the study area. Unlike the dot map, the heat map presents surfaces with a cumulative number of point features. In the case of this test, it is an area of 1 m<sup>2</sup> . Additionally, we proposed a second version of the heat map (Figure 10), with a more interpretable step scale in four intervals based on Jenks' natural breaks classification method. *ISPRS Int. J. Geo-Inf.* **2020**, *9*, x FOR PEER REVIEW 13 of 17

**Figure 10.** The map of pedestrian density in the research area (graduated scale): the chorochromatic **Figure 10.** The map of pedestrian density in the research area (graduated scale): the chorochromatic method (qualitative data).

#### method (qualitative data). **4. Conclusions and Discussion**

**4. Conclusions and Discussion** In this article, we demonstrated a method of researching and visualizing actual land use. The research was conducted based on low-level aerial imagery obtained from UAVs. The UAV technology allowed us to observe and record pedestrian movement in the research area under analysis. As pedestrians were recorded, it was challenging to design a raid with the UAV platform to ensure maximum security for them. It was then necessary to meticulously observe the analyzed urbanized area in search of possible location of photogrammetric raid stations. Determining them with precision helped us accurately capture the entire research area. It mattered also in terms of opportunities to carry out the process of aerial triangulation of the imagery obtained and to achieve possibly accurate value of pedestrians' location in the research area. Modelling static objects by means of the UAV technology, as commonly described in the literature, allowed us to extend the methodology of observing pedestrian movement as a dynamic phenomenon. The methodology of designing maps presenting the actual land use and considering the areas used by pedestrians based on UAV images can be used in other areas while meeting two conditions. The first condition is that there should be relatively high pedestrian traffic in the transport infrastructure and in nonpedestrian In this article, we demonstrated a method of researching and visualizing actual land use. The research was conducted based on low-level aerial imagery obtained from UAVs. The UAV technology allowed us to observe and record pedestrian movement in the research area under analysis. As pedestrians were recorded, it was challenging to design a raid with the UAV platform to ensure maximum security for them. It was then necessary to meticulously observe the analyzed urbanized area in search of possible location of photogrammetric raid stations. Determining them with precision helped us accurately capture the entire research area. It mattered also in terms of opportunities to carry out the process of aerial triangulation of the imagery obtained and to achieve possibly accurate value of pedestrians' location in the research area. Modelling static objects by means of the UAV technology, as commonly described in the literature, allowed us to extend the methodology of observing pedestrian movement as a dynamic phenomenon. The methodology of designing maps presenting the actual land use and considering the areas used by pedestrians based on UAV images can be used in other areas while meeting two conditions. The first condition is that there should be relatively high pedestrian traffic in the transport infrastructure and in nonpedestrian zones. The second condition is the ability to use the UAV platform with safety for people and appropriate technical conditions.

zones. The second condition is the ability to use the UAV platform with safety for people and appropriate technical conditions. Data obtained in this way made it possible to create thematic maps to present the actual use of Data obtained in this way made it possible to create thematic maps to present the actual use of a small, urbanized area, including areas where pedestrians changed the form of use.

a small, urbanized area, including areas where pedestrians changed the form of use.

raid and constantly observing the entire research area. A selection of time intervals between images

To be able to obtain data on land use and the location of pedestrians through the observation of

To be able to obtain data on land use and the location of pedestrians through the observation of their movement, it was necessary to select the appropriate UAV platform. A study of the subject made it possible to choose a multirotor UAV as a highly effective tool, maintaining one position during a raid and constantly observing the entire research area. A selection of time intervals between images was also a highly significant factor. Taking into consideration the average speed of pedestrians during walking, we established that 10-s intervals are optimal for recording pedestrian movement.

It was possible to determine the area occupied by individual pedestrians in the research area thanks to the change of method, as the dot method (Figure 5) turned into the area method through the transformation of point objects into polygon objects based on the adopted basis of the buffer value (Figure 7). The value of the buffer diameter is a proposal based on the human stride length while walking [27]. It can be modified depending on the observed dynamic phenomenon; for example the visualization of runners. The cartographic generalization, consisting in the aggregation of polygon objects presenting individual pedestrians, allowed us to group individual polygon objects in informal land use areas (Figure 8). We presented a concept of the representation of the point-to-polygon transformation of pedestrians in Figure 3.

We worked out mapping methods adjusted to vectorial point-to-polygon transformation of pedestrians' representation and indicated validity of using qualitative methods (the range method) and quantitative methods (the dot method, the dasymetric version of the heat map, Figure 9) other than just the chorochromatic method.

The suggested method of new visualizations consisted in the point-to-polygon transformation of the representation of pedestrians [18]. We believe it is the way to enrich the process of designing land use maps that so far have focused on conveying spatial data in the traditional way in the form of chorochromatic maps. We are not trying to suggest the best method, as each of these mapping techniques depicts different features of the phenomenon [40]. Hence, the effectiveness of these maps depends on the aspect of the spatial phenomenon they analyze. The employment of the principles of static map design helps one to create a clear and transparent cartographic image [22].

Quantitative data, such as the location of individual pedestrians on the map, can be depicted by means of the dot method. However, if multiple point objects occur next to one another, the effectiveness of the map diminishes and it becomes less clear [41]. To boost the effectiveness of the map, such quantitative data can be presented by means of the heat map (specific variant of the choropleth map), which makes it possible to present the number of pedestrians per given area.

The point-to-polygon transformation of pedestrian representation constitutes a shift from quantitative to qualitative data. Thanks to the generalization of qualitative data in the area occupied by individual pedestrians, it is possible to depict the actual land use. However, such data presentation fails to provide information on magnitude of the occurring phenomenon, as opposed to heat maps. In our opinion, the actual use of land, considering the places for which pedestrians changed the form of use, is a good way to combine several mapping techniques to present the quantitative and qualitative aspects of this spatial phenomenon. We decided to add the second version of the heat map (Figure 10) with a more interpretable step scale [17,22,38,42].

The research conducted and the cartographic analysis designed constitute a suggestion how pedestrians recorded in the research area by UAVs can be represented. So far, researchers have attempted to develop a visualization of participants in a mass event based on low-level aerial imagery [18]. In their research, they proposed a set of animations that presented the distribution of participants in a mass event (dynamic objects) in the research area at specific times of the event. Additionally, the animations show the main points of interest of the participants of the mass event (static objects). Dynamic and static objects were presented using an orthophotomap, a dot map and a map of buffers then assessed by experts in terms of the effectiveness. Most of the pedestrians in the study cited sat, but were not in motion. Methods of presentation can be developed and modified, and then tested on users [18]. Sets of data resulting from such studies can be used for creating visualizations in 3D and 4D, which were not the subject matter of the research but have a great potential for perspective and

temporal geovisualizations [7]. Broad sets of quantitative data, obtained by means of UAVs, can be also used for designing animated cartographic visualizations [43].

Currently, new methods of cartographic presentation can be analyzed more accurately in terms of effectively conveying data through eye-tracking studies carried out on users [44,45]. Recognition and tracking of human trajectories is a valuable issue due to many aspects of everyday life. A method commonly used for this purpose was satellite receivers [46]. In the past, attempts were made to determine human movement due to the epidemiological risk and the possibility of transmitting serious diseases through movement and contact with other people [47]. Data obtained with the use of basic GPS receivers, which were equipped with students moving around the campus. However, the basic GPS receiver does not allow to obtain accurate data on the location of individual people, unlike the UAV platform and the obtained images fitted into the coordinate system based on the GCPs. In addition, the proposed method of tracking pedestrians with the use of a GPS receiver requires that each tracked person be equipped with it, which may cause unnatural behavior. The method we propose using the UAV platform allows for the observation of people moving without affecting their behavior. As such, it was possible to determine the actual land use by pedestrians and to design a large-scale thematic map to present the phenomenon.

**Author Contributions:** Conceptualization, Maciej Smaczy´nski, Beata Medy´nska-Gulij; Methodology, Maciej Smaczy ´nski, Beata Medy ´nska-Gulij; Software, Maciej Smaczy ´nski; Validation, Maciej Smaczy ´nski and Łukasz Halik; Formal Analysis, Beata Medy´nska-Gulij; Investigation, Maciej Smaczy´nski, Beata Medy´nska-Gulij and Łukasz Halik; Resources, Maciej Smaczy´nski; Data Curation, Maciej Smaczy´nski; Writing-Original Draft Preparation, Maciej Smaczy ´nski, Beata Medy ´nska-Gulij; Visualization, Maciej Smaczy ´nski, Beata Medy ´nska-Gulij; Supervision, Beata Medy´nska-Gulij; Project Administration, Maciej Smaczy ´nski; Funding Acquisition. All authors have read and agreed to the published version of the manuscript.

**Funding:** This paper is the result of research on visualization methods carried out within statutory research in the Department of Cartography and Geomatics, Faculty of Geographical and Geological Sciences, Adam Mickiewicz University in Pozna ´n, in Poland.

**Conflicts of Interest:** The authors declare no conflict of interest.

## **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
