*4.1. Field Data Collection*

Figure 6 shows a monitoring vehicle equipped with a camera system. It is a hybrid car equipped with a camera system with control and storage equipment. Control monitoring is performed over defined routes. The number of vehicle plates in residential parking spaces are also recorded, as are traffic signs defining or specifying parking in these parking areas. The result of this recording system is documentation, i.e., photographs in digital form with a mark containing the date and time of the acquisition. Photos are taken five times per second. Stop and start monitoring is automated thanks to the parameters specified for each route.

**Figure 6.** Monitoring car.

The data is generated as the output from the GIS data base of the Brnˇenské komunikace (BKOM) data manager. All control data is managed in the QGIS geographic information system. Spatial static data is stored in the SOBD database; spatial dynamic and attribute data is stored in the MONDATA database to monitor parking rules. A field monitoring unit is a monitoring car equipped with cameras and a software server. Field data and documentation is sent to the server and evaluated. In addition to the monitoring vehicle, data from parking meters are transferred to the database. If a violation of the parking rules is found in the evaluation, the documentation together with the data is sent to the database of potential offenses and the municipal police then deal with the offense. Acquired data are subject to regulation by the EU General Data Protection Regulations (GDPR).

The data server is accessed by data from parking meters (number plates of parking cars and paid parking time) and is used to update the control data of the monitoring vehicle. The SOBD database generates potential offenses data, which is evaluated by the Municipal Police. Police evaluate offenses on the basis of registration plate number (RZ) and other parameters (color, etc.) obtained from image and text recognition and evaluate the quality of the recording material in terms of the legal enforceability of the offense.

An example for displaying the saved track of the monitoring vehicle in the form of TrackPoints in the GIS is shown in Figure 7; an example for displaying TrackPoint attributes from a database is shown in Figure 8.

**Figure 7.** View of saved car positions as TrackPoints—sample from the MONDATA database.


**Figure 8.** TrackPoints attribute listing from MONDATA database via QGIS.

Routes are used to ensure complete coverage of the site. Each route consists of waypoints whose serial numbers provide navigation to the driver as they drive through the streets. The waypoints, as well as triggers, are registered by the system using GNSS positioning. Occasionally, it is necessary to travel several times so that the monitoring vehicle passes through all the streets with parking areas within the paid parking area. This is especially true for one-way streets. The data also includes a street network, which is represented by line elements that contain topographic information. Monitoring triggers and waypoints are designed to lie directly on the line.

TrackPoints are snapped onto lines representing a street network using an algorithm to obtain information from a monitoring trigger or a waypoint. Therefore, if there is a trigger or a waypoint at the origin of the TrackPoint, this information is also written to the TrackPoint. Once identified in the GIS software, you can view the point attributes. It is essential to have an existing unique attribute item within the point (record) identifier.

Data generated during operation of the monitoring vehicle are generated in addition to the camera records and so-called "Tracks" and "TrackPoints", vehicle position records and other data. Individual tracks (rides) contain a set of TrackPoints, i.e., records of the monitoring car's GNSS

position. Furthermore, the track stores information about the driver, the monitoring car, the defined route, the time of the journey, the number of registered registration plates of parked vehicles and more. TrackPoints—that is, individual points—store a specific driving identifier and obtain detailed information that is stored five times per second. This information includes the position, direction of movement, speed, time, and more.

The monitoring vehicle carries out monitoring and documentation based on control data. The data is stored in a database on a server located inside the car. The database can be managed directly on the car's server via remote access, but a database located on the local server of Brno's communication network is used by default and data synchronization of both servers is started after the monitoring car is connected in the garage.

The functionality of the parking system is ensured by checking its compliance with the set rules. These are based on traffic signs and the public decree of the city of Brno. Random checks are carried out by the Municipal Police on the basis of patrols or on the basis of a notice. The systematic technological control is then carried out by the monitoring car, which passes through the streets, falling into the areas of resident parking several times a day.

#### *4.2. Analysis and Methods of Data Evaluation*

The passage of the monitoring vehicle is controlled by an algorithm based on graph theory. It is a modified Chinese postman algorithm (The Chinese Postman Problem) with restrictive conditions resulting from traffic signs, with a required time interval of 30 min to capture repeated data, and normal traffic density parameters.

The output of data evaluation is the identification of an offense against parking rules and generation of the necessary documentation on the offense for administrative proceedings via the municipal police. A general mathematical model for evaluating the parking rule is applied for evaluating the data and identifying the offense. The last two items are the subject of this article. The model concept is based on a set of rules on parking that have been approved by the competent administrative authority. Mathematically, we can express the model by function P. It is a Boolean function that depends on the specific area (grounds) A and looks like this:

$$P\_A(r, l, \mathbf{t}) = R \ge (L \cap T) \tag{1}$$

where

*r* ∈ *R* is the vehicle registration plate number,

*l* ∈ *L* is the parking area (parking place),

*t* ∈ *T* is the date and time of parking.

Index *A* of function *P* in Equation (1) means that there are different parking rules in each area. If the *PA* function is True, then the vehicle with the registration plate r is parked at l at a given time t in accordance with the rules, If the *PA* is False, then a rule violation is suspected. The aim is to address the control of the established parking rules in the given areas *A* and their localities *L*. The checks will be in two ways:


The monitoring is then carried out on the basis of this data. They determine at which point which camera starts recording and navigate the driver's route. The monitoring results (tracking and TrackPoint records) are transferred to the MONDATA database after connecting the control car in the garage to the computer network. At the same time, the process of automated evaluation of scanned photographic documentation is performed by means of scripts, which ensure the comparison between the individual parking places on the traversed route and attributes of polygons from the SOBD database at the time of scanning. During this process, the photographic documentation showing parking spaces that were not regulated by the resident parking system at the time the monitoring car passed by, i.e., outside the time stamp on the traffic sign indicating the residential parking area, will be deleted. These are mainly parking places where the reserved parking regime is valid for a part of the day, and parking places for disabled people, places of entry, sections of block cleaning, etc. The evaluation uses the position and time of the control vehicle at the time of recording. This information is part of the scanned data (TrackPoints).

#### **5. Results**

After completing this data evaluation process, the next phase of comparison is followed, where the remaining photo documentation is subjected to the automated license plate recognition of the parked vehicles, and then these license plates are compared with the SOBD database, which includes information on vehicle registrations parked in regulated resident parking lots. The insertion of their car's license plate into the system is done by drivers when parking in a parking meter or mobile application.

The documentation (in which the registration number of the parked car is not stored) in the SOBD database, together with the identification data of the record and photographic documentation of the traffic sign to which the potential offense was related, is submitted to the Municipal Police for review. If it is found to be a misdemeanor, it is sent to the Department of Transport Administration Activities of the City of Brno (MMB). The time between two consecutive passes of the monitoring vehicle is a minimum of 30 min. If the same parked vehicle is identified on both records, it is subject to analysis on the legitimacy of its parking.

Technical support of the system (service of monitoring vehicles and data server administration) is provided by Eltodo, Praha, CR, the stock company.

In Figure 9a–e is an example of the output of the web documentation of suspected offenses against residential parking generated from the SOBD system for the municipal police. Some data are anonymized for privacy reasons (GDPR). The key part is the photo documentation (Figure 9d) and text with image analysis of the identified license plate (Figure 9b) of the vehicle during the first and second passage. The Municipal Police can choose from more photographs than the specified ones to be included in the offense report. It has a similar structure to this document, but is supplemented by vehicle data obtained from the vehicle register and its owner and specification of the legal provision of the offense. This protocol shall be sent to the owner of the vehicle through the competent authority.

For undeniable image documentation, it is important that in both series of photographs taken after a minimum of 30 min of each pass, a traffic sign indicating the parking area (Figure 9c) is identified based on the satellite location of the route with the marker kept in that place in the SOBD database. In terms of GNSS localization requiring a free horizon for satellite signal reception, it has often been the case in dense urban areas that localization accuracy has been reduced or even GNSS signal reception has been interrupted, causing problems with mobile camera recording. The incompleteness of the image data requires a repeated pass of the same street. In Figure 9d is an example of an incomplete image sequence at the first pass (lack of a traffic sign image) that has necessitated a subsequent second pass, and a subsequent third pass after an offense was identified 30 min later.


(**d**)

**Figure 9.** *Cont.*

**Figure 9.** (**a**) Documentation of suspected offense (header). (**b**) Documentation of suspected offense—identification of vehicle registration number (anonymized). (**c**) Documentation of suspected offense—identification of the parking area traffic sign from the traffic sign catalog in SOBD. (**d**) Documentation of suspected violation—video sequences of transit records. (**e**) Documentation of suspected violation—localization of parking area (purple), parking area traffic signs (black dot) and vehicles suspected of offense (short black line) from QGIS.

At the same time, the parking area is identified, and in the map section (Figure 9e), the specific location of the standing vehicle suspected of a parking violation (short black line) and the corresponding traffic sign indicating the parking area (black dot) is marked.

In the initial phase of trial operation, the system showed up to 60%, in terms of legal enforceability, unrecognizable offenses. This high percentage was subjected to reverse analysis and was found to be due to several factors:


The need for system calibration was triggered by the uncertainty of the starting trigger location resulting from the uncertainty of vehicle positioning by GNSS due to the considerable obscurity of the horizon in dense buildings. This resulted in a shift in the image sequence in which not all the necessary objects were recorded for correct image evaluation (e.g., missing an appropriate vertical traffic sign or inappropriate camera angle). The problem was solved by changing the position of the trigger points in the control data by adjusting the orientation of the cameras on the vehicle, and refining the position was solved by using an Inertial Measurement Unit (IMU).

In Table 1, the statistics document the benefits of system calibration by comparing the percentage of qualitative data improvement by the offense demonstrability parameter. After the calibration of the system, the quality of the monitored data was improved by 25% in terms of demonstrable offenses.


**Table 1.** Statistics of evaluation of qualitative improvement in the provability of the offense.

Table 2 shows in detail the structure of the percentage of the individual parameters in the proportion of data excluded due to the non-verifiability of the offense, which are still subject to improvement.


**Table 2.** Structure of data discarded by the system due to the non-verifiability of the offense.

The seemingly high percentage of inadequate documentation for offense proceedings is due to a large number of limiting factors regarding the legal unquestionability of the offense. Failure to comply with any of them makes the offense ineligible.

#### **6. Discussion**

The main objectives of this project were to prevent long-term (in many cases unauthorized) parking of vehicles in the center of Brno city, thus blocking the parking capacity, and ensuring a sufficient number of parking spaces for Brno residents and visitors by increasing the turnover of parking, which was accomplished by the introduction of the parking system.

The whole technology is fully automated; the human factor enters the process in the initial phase of setting and modifying the parameters and then at the end in evaluating offenses in administrative proceedings. The benefit of the technology lies in the automated processing of a large number of records that could not be evaluated manually. Another benefit is to free the city center from long-term parked vehicles and increase parking availability.

Our solution utilizes the integration of image data referencing position and time by GNSS satellite technology, while utilizing database resources and Open Source GIS. The methods of graph theory and image recognition were used in algorithmization. In related works (Chapter 2), the methods of the Internet of Things (IoT) [5,22], were used in particular, while Machine Learning (ML) [9,12], and image processing were used [11,12], in other contexts than in our solution. The use of a modified Chinese postman algorithm (The Chinese Postman Problem) has proved to work best for our project.

From the perspective of those interested in parking, information is available on the website or in the mobile application about the availability of parking spaces at the place of need. The availability of parking information is now a frequently used IoT-based application.

The system of static cameras is used for the evaluation of vehicle entrances to the center through a mechanical barrier. For monitoring barrier-free parking spaces in urban conditions, the method of mobile monitoring has clearly proved its worth. The use of static cameras requires multi-angle shots, generating increased demands on the number and installation of cameras, but the success rate of image recognition is high and is up to 89% [33–35].

The illegibility of the license plate is due to multiple factors, e.g., a dirty or missing license plate or one scanned at an unfavorable angle. The technology has only been operational for one year and is constantly being evaluated and improved. The main aim of the system is to force drivers to respect the parking rules.

Through the calibration process of the mobile parking monitoring system, the quality of data was improved from 28% to 35% in terms of provable offenses. We note that this percentage is made up of excellent data in terms of meeting all the necessary conditions for successful enforcement of offenses. In [11], a similar improved success rate from 24% to 43% is reported. The error rate is very individual and depends on local conditions.

Brnˇenské Komunikace company, Brno, CR is the operator of the resident parking system, but the evaluation is automated and camera records are only available to the municipal police. This complicates the correct setting of control data, as GIS operators cannot access the monitoring car documentation for privacy reasons (according to GDPR), and the municipal police initiative is the only response for correcting settings.

The electronic availability of parking rule information via smart communication devices, which most people use, contributes to the acceptance of the whole system and its increasing popularity. A similar system of regulated parking also operates in the capital of the Czech Republic, Prague.

#### **7. Conclusions**

The article introduced a system for the intensive use of parking spaces. It is an automated system based on monitoring the situation in parking places using a mobile camera set equipped with data storage and means for data transfer to the central database. The data is evaluated in a suitable GIS-based software according to parking rules. It is this evaluation that is a critical part of the whole system. In practice, there have been cases where a complaint was sent to municipal police on parking rules violation, which was not fully justified. Therefore, the further development of the described system will be aimed at improving the evaluation process in order to minimize unauthorized cases. Detailed information about the parking system is very well presented on the web or in the form of a smartphone application [36], which lists all the possible parking situations in Brno.

Prior to the introduction of the system, the parking yield was around 30% compared to the current situation after the introduction of the parking system.

In the city center, where the entrances are regulated by a bolt system, the use of static cameras is effective, while for the control of parking in the street network it is cheaper and more flexible in terms of video recording to use a mobile camera system on a vehicle.

The main benefit of the technical solution was the integration of image, location and time data acquired by a moving vehicle equipped with a camera system in combination with GNSS navigation technology.

After the introduction of a monitored parking system, the overall rate of violations of parking rules assessed from the records of offenses in the monitored areas decreased. The statistics of the municipal police show that, in 2019, parking was unjustified, on average, 10% of the time, with a monthly variability of 17% to 7%. Mobile data collection has a number of disturbances affecting the quality of data, and for the correct evaluation of offenses it is necessary to have flawless data. All formulated research goals were fulfilled and the GIS environment proved to be usable both for data integration and for their subsequent analysis and efficient distribution of purpose outputs.

The continuation of the research will consist of improving the data quality in statistically identified problematic parameters, especially license plate recognition, the problem of changes resulting from temporary traffic signs, and traffic in reduced visibility.

The introduction of a regulated parking system in the center of Brno and in the surrounding areas contributed significantly to improving the availability of parking for residents, subscribers and visitors. The system of parking houses and car parks will be gradually extended in the city of Brno, along with the modernization of the transport network. The new parking system offers many advantages. The goal is not to collect money purposefully, but to achieve effective renewal of parking, thus allowing more people to park and deal with the necessary issues. The limitation makes sense where residents have a problem with parking or where there is a day-to-day overpressure. In most of Brno, where a new system was introduced, regulation is only introduced at night. In the most critical places (the first two zones), there is regulation even during the day.

The City of Brno is working to make the most of the use of information technology in conjunction with smart communication devices and smart applications to improve city life.

With the introduction of the automated parking system in the city of Brno as the last part of the SOBD system, one of the important databases for the future possibility of intelligent communication between vehicles, solved in the project C-ROADS Czech Republic, was created.

**Author Contributions:** P.K. provided support materials and is the administrator of the BKOM company parking system, D.B. elaborated the literature review and system model, J.B. wrote the introductory section, a chapter on experimental results, discussion and conclusion, O.Š. conducted the overall editing of the article the a professional translation. The roles of D.B. and J.B. are in the professional consultancy of GIS technology in support of BKOM. All authors have read and agreed to the published version of the manuscript.

**Funding:** This paper was elaborated with the support of European project C-ROADS and with the theoretical cooperation of Brno University of Technology, Specific Research Project FAST-S-18-5324 and FAST-J-20-6374.

**Acknowledgments:** Thanks to all reviewers for their suggestive comments.

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

#### **References**


© 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/).
