Analysis of the Potential Use of Unmanned Aerial Vehicles and Image Processing Methods to Support Road and Parking Space Management in Urban Transport

Abstract

Progressive urban density affects city centers especially and results in growing congestion, lack of parking spaces, and increasing environmental costs of transportation, causing increased air pollutant emissions and noise. These phenomena reduce the attractiveness of the city and result in a degradation of the quality of life for its residents. In light of these phenomena, there is a clear need for intelligent management of urban space using new technologies that would be complementary to existing intelligent transportation systems. Expanding information resources obtained from mobile cameras will have a positive impact on increasing the efficiency of transportation management and use of limited space in city centers. It will also have an impact on reducing external transport costs and increasing the quality of logistics services provided in the city. The main aim of the paper is to develop a concept of a transport management system in cities using mobile vision systems mounted on unmanned aerial vehicles. The model will concern the cases of lane occupation by freight vehicles and the analysis of parking spaces in the city in order to improve their management. The results of the developed model will contribute to the automation of the parking space management process and increase the efficiency of the use of city parking space resources.

1. Introduction

The city is a system of social, legal, functional and physiognomic structures that together form its spatial structure. All functions of the city—residential, commercial or recreational—are carried out in space. It is easy to notice that the processes of implementing the city’s functions are associated with the constant movement of people and material goods in the urban space, necessary for the course of these processes. Thus, they generate streams of flow of both people, information and products. The physical layout of the functional structures forces their users to make efforts in terms of relocation and its associated costs. The efficiency of the city’s functioning in the real sphere is related to the degree of route complexity, travel methods and the efficiency of traveling through space and is one of the dimensions of the residents’ quality of life [1]. Proper functioning and development of cities is influenced by the condition of the transport system and the level of amenities for residents and visitors. Keeping the balance between an efficient transport system and the comfort of living in the city is becoming a growing challenge.

Emerging technologies have a major social impact, resulting in changes in many aspects of everyday life. A growing worldwide trend is the use of innovative solutions using Internet of Things (IoT) technology for public services. An extension of the IoT concept, proposed and developed since the early 2000s, is applying this technology to means of transport. We are then referring to the so-called Internet of Vehicles (IoV) concept, which is a development of the earlier Vehicular Ad-hoc Networks (VANETs) system. IoV technology significantly extends the VANET concept by solving some of the main problems found in traditional networks. These improvements include improved coordination between different vehicles that move at a distance from each other, scalability, availability of information, etc. In the IoV concept, each unit in the network can connect wirelessly with the global Internet as well as roadside devices, other vehicles, drivers, passengers and even pedestrians. In addition to information exchange, Internet connectivity provides flexibility in extending the scale of the IoV network [2]. Considering how many everyday objects and devices have access to the global Internet, we can now talk about a new concept that treats this issue more broadly and includes all these devices and technologies under one name: the Internet of Everything (IoE). IoE combines existing concepts related to smart devices connected to the global internet. The increasing number of devices connected to each other via the Internet or short-range wireless networks produces large amounts of data. This data contains information from which, with the help of processing algorithms, it is possible to obtain knowledge about the surrounding environment. This set of information and its acquisition, storage and processing is known as Big Data. These tools are increasingly being used in smart cities and this raises a further challenge for city authorities and residents as to awareness of what purposes this data can be used for. Emerging IoT tools, usually used independently, dedicated to specific applications should be integrated with each other as much as possible in the future. The exchange of data between them can add significant value for the management of smart cities [3,4].

Effective transport management is essential to ensuring high-quality and reliable transport services, which are an important component of well-functioning smart cities. Traffic congestion is a serious problem in the transport systems of growing cities, with negative consequences and a negative impact on the comfort of urban life. Ensuring continuous and collision-free traffic flow of vehicles requires appropriate management of both the control infrastructure and the parking space, which is most often located in the immediate vicinity of roads. These problems have long been known and discussed within the literature [5,6] and inadequate management of parking space causes a number of varied problems, the main of which is lane occupancy, which increases congestion and increases the risk of collisions with other vehicles trying to avoid a parked delivery truck.

In this paper the concept of extending the existing methods of transport management included in Intelligent Transport Systems is proposed, along with the analysis of the road situation of parked and in-motion vehicles, using unmanned aerial vehicles, as well as image processing and analysis methods.

The structure of the article is as follows: in the second chapter, selected aspects of urban space management are presented with an emphasis on vehicle parking spaces and the hazards that result from inadequate management of parking spaces and how this affects traffic flows. The third chapter presents the concept and the methods of its implementation with the use of various information and communication technology (ICT) tools and unmanned aerial vehicles. The fourth chapter contains the results of the experiments carried out, and the fifth chapter provides conclusions from the current research and planned further projects in this direction.

2. Selected Aspects of Urban Space Management

Contemporary metropolises are characterized by different degrees of complexity of infrastructure, which should be effectively developed, modernized and adapted to the needs of the citizens. Many modern cities use the concept of smart cities for their management, which is based on reorganising all spheres of city life in such a way as to manage the city as efficiently and sustainably as possible through the creation and implementation of modern information and telecommunications technologies. Along with the development of these technologies, the number of devices that can supply the city management system with extended data is growing. This data can be used to collect, process, analyse and share information for residents. The multitude of wireless technologies in the urban space, such as GNSS (Global Navigation Satellite System), Wi-Fi, Bluetooth, NFC (Near-field communication) and RFID (Radio-frequency identification) causes problems in the management of this heterogeneous data, so the smart city concept must involve the creation of efficient information management mechanisms based on intelligent ICT systems with the ability to process large data sets reliably and securely. As statistics show, the number of people living in cities is increasing. This is associated with additional challenges of effective management, taking into account sustainable development. This is possible thanks to the implementation of emerging technologies and intelligent systems. Information and communication technologies should be used effectively to improve the quality, productivity and interactivity of city services. In order to reduce the consumption of resources and the cost of living in cities, it is necessary to create reliable social and communication links between citizens and government structures [7].

In the literature on the subject, there is no one consistent definition of a smart city; nevertheless, among several concepts, common features can be found. First, the implementation of modern communication technologies is the basis of information systems oriented at providing services to residents. An example is intelligent transport systems, the aim of which is to “maximize the use of the existing road network through more effective traffic control and management and its optimization in relation to the strategic goals in the area of the broadly understood transport system.” [8]. Secondly is stakeholder participation in creating the city through active participation in socio-economic initiatives. An example may be the Urban Lab—“it is an instrument (...) of cooperation between municipal authorities and residents (...), enterprises (...) and scientific entities (...), aimed at improving the quality of life of residents through innovative solutions to identified problems (...) and generating additional value using municipal resources.” [9]. Third is sustainable development manifested in low-carbon technologies and rational use of resources. Particularly noteworthy is the aspect of the high quality of life of the inhabitants and the harmony between the natural and anthropogenic environment. People living in the city expect their existence to provide them with satisfaction and thus a better quality of life than outside the city.

The Smart City Model was developed by a team led by Rudolf Giffinger from the Vienna University of Technology, presented in the 2007 Smart Cities report. The Ranking of European medium-sized cities was performed [10]. This indicator is based on 31 factors grouped into six characteristics: smart economy; smart people; smart governance; smart mobility; smart environment and smart living. Each of the characteristics covers a specific aspect of the city functioning as the living environment of its dwellers. The aspect of urban transport is perceived in this model via three factors. The first is local, regional and international accessibility, expressed in terms of ease of travel and ability to reach particular destination. The second factor is a transport system that should be safe, sustainable and innovative, not only for active travellers but also for residents, so that it does not cause excessive external costs. The third factor is the availability of ICT infrastructure, allowing inhabitants usage of modern communication methods and decision making [11,12,13,14]. This factor is crucial for the information society, in which the quality of life is closely related to access to information.

Becoming a smart city can take many forms depending on the type of metropolis. For example, smart concepts will be implemented in modern and young cities (e.g., cities of the United Arab Emirates), but a completely different approach is needed in European countries, where cities were established in ancient times. One of the concepts of building a city development strategy is the use of a model to analyse the sentiment of residents and the implementation of urban space management related to the research results. Local authorities are constantly innovating in cities, encouraging multinational companies and individuals to pursue the concept of sustainable development related to renewable energy and provide products and services for municipalities according to sustainable policy This applies not only to industry, but also to the broadly understood awareness of individual users, i.e., energy consumers. Local authorities of many cities encourage the use of urban transport, in parallel with the introduction onto city streets of fully electric vehicles (EV), or at least hybrid ones, in order to reduce emissions of harmful substances into the atmosphere (pollution and noise). Another challenge in this regard is for the authorities of modern cities to ensure preferential conditions to encourage companies from the energy sector to invest in cities. Public managers should take into account the multi-criteria decision-making problems faced by energy companies when deciding on a location in the city. [15,16].

Traffic congestion in urban areas often occurs in areas of intense economic development. Transport system planners and managers face the problems of effectively implementing rules of conduct depending on the traffic situation, and for this it is necessary to properly identify the problems. Availability of information for transport planning is the main condition for finding any solution. The last few decades have seen the use of technical solutions to acquire the data needed for transport planning using a number of independent devices. Many traffic parameters such as traffic flow, vehicle speed, capacity, density, safety and infrastructure availability have to be taken into account to collect information about the current traffic situation. All of these factors are required in effective transport planning to ensure efficient urban space management [17].

In the context of reducing congestion, attention should be paid to the characteristics of Smart Mobility, especially factors such as local transport accessibility and sustainable, innovative and safe transport systems. These issues are closely related to the area of city logistics operations, thus mainly with the problem of supplying retail and service points and last kilometre deliveries. Last mile delivery aspects are considered in the context of Sustainable Urban Logistics. This is a concept belonging to Smart Logistics, defined as a combination of modern technologies in the service of administration and human activities. It creates the ability to anticipate problems and minimize their negative impact on a given area of the city. The main advantages of this concept are the effective coordination of resources and their supplies to achieve the assumed goals by eliminating communication barriers between all points of the supply chain [18]. The large building density of the city centres, and thus the accumulation of points where mainly commercial and recreational functions are carried out, causes the concentration of the largest traffic flows in this area. Paradoxically, allocating a certain part of the road lane along the entire length of the street for parking spaces leads to the intensification of the phenomenon of congestion. The mechanism of this phenomenon is as follows—the lack of a parking space for loading/unloading activities results in the fact that the drivers performing these activities occupy the right lane, drastically reducing the road capacity and increasing the risk of an accident. According to Iwan et al. [19], unloading bays should be considered as an environmentally friendly, efficient measure which supports urban delivery systems. The advantage of this measure is reduction of congestion in city centres. Freight vehicles parked directly on streets during the unloading activity significantly increase congestion. Unloading bays are located to support the logistics of goods in cities in this way so as not to interfere with road traffic. There is another advantage, in that delivery vehicles do not cause excessive pollution of the environment by driving in search of a place to unload, resulting in additional fuel or power consumption in ICEVs and EVs, or additional costs and downtime related to traffic jams.

“Based on the cellular automata simulation, the analysis showed that application of unloading bays in the studied sections in the city of Szczecin, Poland increased the traffic fluidity by 8% on average. This allows reduction of pollutant emissions on average by:

• 4% in the case of carbon monoxide (CO);
• 5% in the case of hydrocarbons (HC);
• 4% in the case of nitrogen oxides (NO_X).” [19]

Therefore, in almost every medium-sized and large city, there is a shortage of space, prompting municipal authorities to use tools to limit the intensity of traffic flows. Such methods are:

• restriction of entry to the city centre—manifested in activities such as the introduction of tolls or temporary limitation of accessibility (e.g., setting specific delivery times) or limiting the type of vehicles allowed for traffic (e.g., clean transport zones);
• limiting the time spent in the centre (e.g., paid parking zones).

The latter solution is commonly used to force vehicle rotation in the limited number of surface car park lots, due to insufficient space. The city authorities are trying to adjust the rate for using the parking space so that drivers leave their vehicles for as short a time as possible. According to the authors, the key is to determine the parameters of road infrastructure use, such as:

• degree of rotation;
• degree of use of the car park lot (capacity);
• frequency of road lane occupancy.

The determination of these parameters will allow, first of all, better management of this space, and secondly, it will enable the dissemination of this data among interested parties, according to the smart city concept. This does not mean that data will be provided directly to users (data on the number of free parking spaces). The degree of turnover or occupancy should be seen as an indicator to be used in strategic decisions regarding the designation of parking zones (differentiated in terms of cost and parking availability), the degree of progression of parking tariffs (how much more expensive the subsequent hours of parking are), or traffic organisation, as well as the future planning of new and the reorganisation of existing parking spaces [20,21].

The layout of parking spaces within the parking lot is equally important. Parking lots consist of a manoeuvring area—for entry, exit and siting the vehicles—and parking space designated for stoppage or charging electric vehicles. It is recommended to place the parking of trucks parallel to the axis of movement, while parking of passenger cars mainly should be arranged according to an oblique angle of 45–60° [22]. In city centres, usually with congeste buildings, this type of parking space is applied, mostly due to the fact that parking lots are located along streets between the traffic lane and the pavement.

3. Materials and Methods

Despite the variety of different types of sensors and devices included in intelligent transportation systems, there is no such system for evaluating the parking situation of vehicles in different parts of the city to continuously analyse this situation and lane occupancy of delivery trucks. This paper proposes the concept of an automatic analysis system using unmanned aerial vehicles, for the purpose of evaluating the rotation of cars in the parking zone in the study area and the time of lane occupancy during the unloading and loading of goods, as part of urban logistics. An additional advantage of such a system is that it can be flexibly adapted to current needs without the need to interfere with existing infrastructure.

Figure 1 shows a schematic of the proposed system. The system integrates several information and communication technologies for video data transmission and storage, information processing and automatic situation assessment using image processing algorithms. This concept focuses on using existing image processing methods widely reported in the literature, both static and moving camera methods mounted on UAVs [23,24,25,26,27,28,29]. It should be noticed that most existing detectors based on image processing are fixed or mounted on vehicles.

Two main scenarios were conducted to study the traffic situation and its impact on current vehicle traffic flows. The first concerned lane occupancy by delivery trucks and how the average speed of vehicles on a road section changed before and during lane occupancy, and the second scenario concerned parking space occupancy in selected parts of the city and the rotation of cars at different times of the day. Additionally, each of the mentioned scenarios contained several subcategories. This was intended to show the possibility of using UAVs flexibly for current needs and involved only the application of appropriate image processing algorithms.

The concept is to be able to automatically launch, inspect after predefined points and land the UAV. For the purpose of continuous monitoring, it is necessary to have a docking station for inductive charging of the drone and a minimum of two drones that will perform alternate inspections. It should be noted that there are restrictions on the performance of flights, defined by the aviation law specific to particular countries. In the case of Poland, drone operator licenses and notification of the operation to PANSA (Polish Air Navigation Services Agency) are required to perform flights. Conducting the analysis was preceded by preliminary research on how to fly over urban roads. The influence of UAV flight altitude on the correctness of image recognition was determined. It was shown that correct results are obtained for heights from 40 to 60 m above the ground. This gives the possibility to analyse up to 10 lanes simultaneously [30].

Scenario 1. Lane occupancy detection and vehicle speed estimation.

It is a common practice to occupy a lane for loading/unloading of goods. Often this happens when there is space on the roadside. The reason for this is that there is a tacit acceptance of such practices and the operators, knowing that they will not face any consequences, repeatedly practice such actions. This is highly dangerous to the on-going traffic, where at the moment of loading/unloading operations, two lanes suddenly become one, without any prior warning to the drivers. An additional hazard is lane blocking in close proximity to intersections and pedestrian crossings.

During the study, lane occupancy was observed for several to over ten minutes, up to 5 times per hour at different locations on the same study street.

The algorithms used to detect lane occupancy and study vehicle behaviour were implemented in the Visual Studio environment using the Open Computer Vision library—OpenCV [31]. The types of transformations used are illustrated in the diagram shown in Figure 2.

Figure 3 shows the video processing steps from the original image through binarization, background subtraction, morphological filters, blob finding and the result of vehicle motion analysis. This method provides the ability to detect objects in motion in a specified area and to detect objects that are treated as obstacles in the path of vehicle motion. It is possible to estimate the speed of vehicles and the number of passing cars. The discussion of the test outcomes is presented in Results.

Scenario 2. Parking space occupancy analysis.

Possibilities of using automatic UAV flights for inspection of parking spaces in the centre of urban agglomeration were investigated (Figure 4). Flights were realized in automatic mode based on predefined points—located in the middle of intersections, the start and the end of the street along which the parking lot is located. The aim was to determine the level of utilization of the parking lot (number of actually parked vehicles in relation to the designed capacity) and the level of rotation (exchange of parked vehicles). This type of data can be helpful to identify places with particularly intensive use, which can be used to properly calculate parking fees and to detect irregularities.

It should be pointed out that the algorithm was not used to detect individual characteristics of particular vehicles, so the method is not designed to control parking fees. Control of parking fee payment is possible based on data from terminals located by the parking lot or payment systems in applications. This method, however, allows an assessment only in quantitative terms, while the drone image allows for quantitative (comparison of the planned number of vehicles to the number of parking fee payments) and qualitative analysis, for example, considering the location of vehicles and the spaces between or next to them, which is particularly important when parking parallel to the direction of traffic. In addition, it allows assessment of rotation at the parking lot, i.e., the frequency of replacing vehicles. In other words, UAV images allows spatial analyses of parking lots.

Only the occupancy and rotation of vehicles parking in an area were detected. Two inspection methods were investigated. The first assumed continuous monitoring based on a moving video image, while the second method consisted of taking pictures periodically and comparing them with each other, to detect the turnover in a given space and the percentage occupancy of the entire parking area. The moving image method failed with background subtraction algorithms and even with optical flow and panoramic background subtraction. One more drawback of this solution is the problem that roads and parking lots are not in perpendicular alignment to the camera view. Therefore, it was decided to use a sequence of images, which were then aggregated and plotted on a real map of the area. This gave the possibility of measuring the total area of the parking lots and, with the help of object feature detection algorithms, investigation of the empty areas. Figure 4 shows a sample of the image sequence and a measurement of the parking area on a selected portion of the curved road using object feature finding algorithms associated with the parking lot surface.

4. Results

To simulate the concept of using UAVs in both scenarios, the following simplifications were used. Due to the lack of access to a docking station to charge the UAVs, continuous drone operation was simulated by alternating two flying units. A total of over 4 h of video footage was collected. It was also decided to use the least computationally complex image processing and analysis algorithms, so that inspection would be possible without involving high-performance graphical workstations. Therefore, the use of artificial intelligence and deep learning methods was not considered.

In scenario 1, vehicle parking situations directly affecting traffic flows were investigated on a selected road section.

Table 1. Sample measurements in scenario one, divided into with and without lane occupancy.

Lane Occupancy						Without Lane Occupancy
No.	type	Fv	t [s]	s (m)	v (km/h)	type	Fv	t (s)	s (m)	v (km/h)
1	car	210	7.0000	80	41.1429	car	150	5.0000	80	57.6000
2	freight	280	9.3333	80	30.8571	car	220	7.3333	80	39.2727
3	car	262	8.7333	80	32.9771	car	178	5.9333	80	48.5393
4	car	261	8.7000	80	33.1034	freight	240	8.0000	80	36.0000
5	car	218	7.2667	80	39.6330	car	250	8.3333	80	34.5600
6	freight	215	7.1667	80	40.1860	car	205	6.8333	80	42.1463
7	freight	262	8.7333	80	32.9771	car	150	5.0000	80	57.6000
8	car	265	8.8333	80	32.6038	freight	200	6.6667	80	43.2000
9	car	233	7.7667	80	37.0815	car	190	6.3333	80	45.4737
10	car	237	7.9000	80	36.4557	car	200	6.6667	80	43.2000
11	car	222	7.4000	80	38.9189	car	180	6.0000	80	48.0000
12	car	240	8.0000	80	36.0000	car	200	6.6667	80	43.2000
13	freight	390	13.0000	80	22.1538	car	175	5.8333	80	49.3714
14	freight	460	15.3333	80	18.7826	car	180	6.0000	80	48.0000
15	car	410	13.6667	80	21.0732	freight	270	9.0000	80	32.0000
16	car	435	14.5000	80	19.8621	car	180	6.0000	80	48.0000
17	car	290	9.6667	80	29.7931	car	220	7.3333	80	39.2727
18	car	270	9.0000	80	32.0000	car	245	8.1667	80	35.2653
19	car	260	8.6667	80	33.2308	freight	220	7.3333	80	39.2727
20	car	400	13.3333	80	21.6000	car	205	6.8333	80	42.1463
				MS	31.5216				MS	43.6060

Where: Fv—Number of frames vehicle visibility; t—time of visibility; s—distance; v—sectional speed measurement; MS—Mean Speed of vehicles.

shows the results of automatic sectional vehicle speed measurements in the cases of with lane occupation and without lane occupation. The studied road section was 80 m long and had two lanes in one direction.

For the purpose of determining vehicle movement parameters, technical image parameters such as the number of displayed frames per second (Fv) and time of vehicle presence in the image were used. On this basis, an estimated segmental speed measurement (v) was automatically obtained.

The lane occupation in this case occurred at a short distance from the intersection and pedestrian crossings. The average speed of vehicles forced to evade the delivery truck was about 31.5 km/h, while the speed was about 43.6 km/h when vehicles were moving smoothly. The number of vehicles recorded in these two cases differed by about 23% in favour of the lane-block free situation. In addition to the obvious inconveniences associated with the occupation of the lane, it should be mentioned that the risk of collision of vehicles trying to bypass the van standing on the road during unloading increases at this point, which also poses a threat to it and especially to the person handling the loading/unloading processes.

The distribution of traffic flows when a lane is occupied shows a large accumulation of cars within the obstruction and a significant reduction in vehicle speed. Figure 5 illustrates vehicle flow situations during lane occupation and free flowing vehicles. The points were read based on the pixel data of the centres of the quadrilaterals, determined by the moving object recognition algorithm. The measurement points were collected at equal time intervals, hence a smaller distance between the points means a reduction in the speed of vehicle movement. A reduction in speed in the immediate vicinity of an obstacle, i.e., a delivery truck during loading and unloading, can be clearly observed.

In scenario 2, parking space occupancy measurements were implemented using object feature detection algorithms and a geographic information system georeferencing algorithm (Figure 6). Initially, areas of interest need to be defined, so that during image processing there is no situation where an unwanted area is analysed that may distort the final results. Samples of the parking lot pavement are necessary to start the analysis. The more diverse the samples, the more accurate the test results. Drone flights were conducted over a selected 2-week period from Monday to Sunday. Sampling (video data collection) took place from 6 a.m. to 9 am, from 2 pm to 4 pm with an interval of 30 min, and from 5 p.m. to 10 p.m. with an interval of one hour. Each drone flight lasted about 1.5 min and collected about 20 high-resolution images each time, which were then processed and analysed.

Similar to scenario one, two drones were used to simulate continuous operation. It is possible to inspect the terrain of parking spaces assuming that the surveyed area is well imaged beforehand for subsequent automatic analysis. For a selected section of the city, the occupancy rate of parking spaces was investigated, which averaged 78% occupancy during the weekday between 6 a.m. and 4 p.m. and had high vehicle turnover. In contrast, between 5 p.m. and 10 p.m., the average occupancy rate reached 93% with very low vehicle turnover. It was observed that inspecting the area after dark with standard video cameras is practically impossible. Figure 6 illustrates an example of the performance of the parking lot surface feature detection algorithm during a continuous flight, which was used to calculate the free space between vehicles.

The situation on public holidays showed a low turnover of vehicles and a very high occupancy percentage of approximately 94% in the measured time. Figure 7 summarises the parking occupancy percentages for the study period by hours between 6 a.m. and 10 p.m. The inspection carried out during the selected hours provides an opportunity to designate hours during the week with lower occupancy in order to potentially designate time-segregated loading and unloading zones for freight vehicles.

The site selected for the survey in no case showed occupancy below 55% of the available space regardless of the hour of measurement on working days and 75% on non-working days.

The data obtained should not be considered as a source of dynamic information directed to drivers, e.g., in the form of a presentation of the number of available or occupied parking spaces. The reason for this is the aforementioned limitation of UAVs such as flight time or weather conditions. The possibility of using drones as tools to collect kerbside parking statistics at selected time intervals to build a parking space management strategy should be considered. Instantaneous, dynamic information on the number of vacant spaces requires continuous observation, so this study may allow the selection of sites for future ITS devices and fixed cameras for continuous monitoring.

The experimentation with the use of unmanned aerial vehicles and image analysis algorithms has proven that it is possible to apply them in a changing and dynamic environment and, in particular, in places not equipped with other devices and methods from the ITS field. The use of mobile cameras and the use of fit-for-purpose image processing and analysis methods can assist in parking space management and, in the long term, urban space management, in particular during land-use decision-making processes and decisions on future investments.

5. Discussion

The proposed concept of using unmanned aerial vehicles to analyse selected urban traffic situations showed that it is possible to use drones as mobile video camera transporters in places that are inaccessible to other Intelligent Transport Systems devices. It has also been shown that, using appropriate image processing and analysis methods, it is possible to automatically analyse the congestion status of roads and parking spaces and their impact on other traffic participants.

The results of the two independent research scenarios carried out provide the opportunity to draw out a number of interesting insights and provoke thought about the utilitarian use of combined unmanned aerial vehicles technologies, image processing algorithms and traffic flow investigation methods. The results of scenario one concern the analysis of lane occupation by other vehicles (mainly trucks during the process of unloading goods) and the impact on traffic flows of other road users. A reduction in the speed of vehicles avoiding congestion on the road is shown, as well as an increase in vehicle congestion where the vehicle traffic stream is narrowed to one lane. Numerical information obtained from the image data of mobile cameras allows automatic analysis of traffic situations.

Measurements of parking space occupancy and car circulation require continuous monitoring and the results can be used for enhanced parking space management. One scenario for the use of UAV measurements is to assess the need to designate unloading bays available at certain hours of the working day (e.g., between 7 a.m. and 4 p.m.), and at other times of the day and on non-working days these bays would become regular parking spaces available to all residents. It has been observed that there is a need for flexibility in the treatment of urban spaces according to needs that change dynamically during the day and the week. The methods used to measure parking space occupancy can serve as a complement to existing methods included in Intelligent Transport Systems or can be used in locations without accessible ICT infrastructure. The advantage of the presented concept is that there is no need to interfere with existing road infrastructure and there is a high degree of mobility and flexibility in the selection of the analysis location.

6. Conclusions

The study was conducted to analyse the possibility of using unmanned aerial vehicles as a concept for a system for automatic analysis of traffic situations, taking into account bad parking practices that affect the flow of vehicle traffic, as well as the safety of all road users. During the research, a number of conclusions and recommendations were made that would need to be implemented in future work to develop the proposed concept. First, the concept of automatic analysis of vehicle parking situations is possible to implement assuming the following limitations resulting from the technologies used. The operating time of UAVs on a single battery, which is about 25 min, forces the use of platforms that allow charging the aircraft batteries between inspections or replacing them. Continuous human supervision to operate such a drone is also necessary, for reasons of flight safety and the safety of other airspace users. It is necessary to adapt the survey scenarios to the applicable aviation laws of the location and, if necessary, to obtain the appropriate authorization to control the UAVs. Secondly, the optimal flight ceiling (40–60 m) is above the tree crowns, which makes the use of this method impossible in locations with a high density of trees. Similarly, in case of bad weather conditions—rain or strong wind—the use of a drone is impossible

In model terms, the proposed concept extends the possibilities of studying traffic flows in a flexible way without interfering with the existing infrastructure. The way parking spaces are used strongly influences the actual traffic situation in cities. Unloading bays, mentioned earlier, are an effective solution to the problem of unloading in the right lane, provided that the rules of use are respected by users. Aerial inspection may contribute to better control of compliance in this area.

An important element of the presented research is image processing and analysis algorithms. Since the early 1980s, numerical methods have been developed in the field of computer graphics and computer vision, so nowadays there are technologies available on the market that provide all possibilities for analysing vision images in real time. The plethora of available techniques can be a problem to be aware of when deciding which methods to use. The presented concept and previous studies show that the use of simple methods can allow the analysis of traffic parameters such as lane occupancy, vehicle speed, number of vehicles, area of the study region, etc. For these purposes, simple algorithms of background subtraction and finding characteristic features of objects are sufficient, although it is known that, in order to increase the reliability of recognizing objects in moving images, it would be necessary to use available databases of various types of vehicles and to apply at least Haar classifiers for reliable analysis. For years, the problem in the field of image analysis associated with working under changing lighting conditions has been known. Sudden changes in lighting or shading of objects by other objects introduce measurement inaccuracies that must be taken into account.

In the case of transport systems, forecasting is most often concerned with estimating the volume of future traffic, passenger or freight traffic in an existing or planned transport networks. There is a need to take into account the problems of parking space occupancy and related impediments when planning transport networks and deciding on changes in traffic organisation, as well as planning temporary traffic organisation due to current difficulties. The above-mentioned problems may be solved with the help of innovative solutions proposed in this article, i.e., the application of a research approach consisting of the use of ICT tools, which enable flexible conducting of analyses with a different level of detail in relation to different transport infrastructure objects. Existing services that are part of the broadly defined Intelligent Transport Systems, using devices such as traffic detectors, scanners, radars and inductive loops, enable the provision of extensive data sets that can serve as input to the development of models for planning and managing transport networks, which are an integral part of well-functioning smart cities. The proposed concept extends these capabilities with additional data collected in real-time using mobile devices with mounted cameras. The information obtained through image processing can complement data from stationary sources, especially in dynamic situations such as changes in traffic organisation, street reconstruction and others.