Computer vision based ITS solutions

Publications

Computer vision based ITS solutions

ITS FUNCTIONALITY IMPLEMENTED TODAY

Undoubtedly, the understanding of what ITS is will constantly change, refine, expand, supplement. ITS must correspond to the level of development of infrastructure technologies and related areas in order to be implemented in modern realities, while quickly absorbing the fruits of technological progress.

The degree of controllability of the system is determined by 3 levels. It is impossible to move to the next level without mastering the previous one:

  1. Collection of information. We can manage the system only if we have complete, current and reliable data about it.
  2. Adequate response to a changing situation. In this case, receiving information that the system has left its optimal state, we return it to this state with control actions.
  3. Forecasting the situation. This is the next level of management. When, on the basis of the information received and the accumulated experience, we can anticipate the development of events and take measures to prevent it in advance, even before a critical situation arises.

If we translate all this into the ITS plane, then, taking into account the technologies available today, we will receive the following requirements for the functionality of modern intelligent transport systems:

1. Мониторинг дорожного движения, выполняющий следующие функции в автоматическом режиме 24/7 с точностью не менее 95%:

1. Сбор и определение всех параметров дорожного движения, необходимых для определения уровня обслуживания (приказы Минтранс РФ 114,479)
2. Сбор данных о событиях

1. Детекция и классификация инцидентов

1. Посторонние предметы на проезжей части
2. Остановка автомобиля на магистрали/ мосту и пр.
3. Пешеходы, животные на автомагистрали/ мосту и пр.
4. ДТП

2. Контроль соблюдения ПДД

2. На основе полученных данных формируется управленческие воздействия для оптимизации дорожного движения

1. Адаптивное регулирование светофорных объектов
2. Управление транспортными потоками на основе информации об инцидентах
3. Рекомендации водителям об оптимальном скоростном режиме

3. Прогнозирование дорожной ситуации

1. Определение параметров дорожного движения на ближний/средний/дальний горизонт на основе текущей ситуации и вероятности возникновения событий, резко влияющих на дорожную обстановку

1. ДТП
2. Мероприятия
3. Дорожные работы
4. И пр.

2. Выработка возможных сценариев компенсации дестабилизирующих факторов

The developments of the TrafficDate company are aimed at implementing precisely these ITS functions. This article discusses the implementation of traffic monitoring in the TrafficData software.

TRAFFIC MONITORING

The Federal Law “On Traffic Management in the Russian Federation and on Amendments to Certain Legislative Acts of the Russian Federation” dated December 29, 2017 No. 443-FZ prescribes regular monitoring of road traffic. To implement these measures on a national scale, it is necessary to create an Automated Traffic Monitoring System. Below are the principles of building such a system based on TrafficData.

Today, the network of outdoor CCTV cameras is becoming ever denser. With the proliferation of UAVs, it became possible to film interchanges and large intersections entirely, which makes it possible to track the movement of vehicles in all directions at once in all directions by one video. The development of computer vision and machine learning technologies makes it possible to turn all these cameras into high-tech, trainable multi-sensors and to abandon expensive specialized equipment for collecting traffic data. The traffic monitoring system can now be implemented on the basis of conventional video surveillance cameras.

Figure: 1. Analysis of traffic flows using UAVs.
Figure: 1. Analysis of traffic flows using UAVs.

Flow rate. Traffic intensity is determined by direction. Directions are set by the alignments indicated by the user in the video. There are 3 types of doors available: entrance, exit and through.

Stream composition. The analysis of the composition of the traffic flow, supported in TrafficData, takes into account the requirements of both the order of the Ministry of Transport of the Russian Federation No. 479 dated December 26, 2018, and the SP 34.13330.2012 standards for country roads, and SP 396.1325800.2018 for urban roads. To provide all this diversity, TrafficData recognizes 23 types of vehicles and pedestrians.

Travel speed. The video image determines:

  • Instantaneous speed at each point of the trajectory;
  • Average speed in the section between sections;
  • Speed with 85%, 95% security.
Figure: 2. Visualization of speed data using an interactive heat map in the TrafficData Air interface.
Figure: 2. Visualization of speed data using an interactive heat map in the TrafficData Air interface.

Density of movement. The traffic density for a road section is determined by the formula:

cars reduced to passenger cars, per kilometer,
where:
— the average time interval between the given cars; — the average speed of movement of vehicles on the road section.

Queue length and traffic delays. TrafficData allows you to determine the length of the queue and the time spent by the car in the queue in the considered direction, as well as the parking time using the video image.

A delay in traffic at an intersection is the time that a participant loses when crossing the intersection. For this, zones are set in TrafficData Land, when cars or pedestrians hit which it is considered that a traffic participant has arrived at the intersection. As soon as the car has entered this zone, the time counter turns on and turns off when the exit (red) line is crossed. Thus, the time spent on crossing the intersection in each direction is recorded (Fig. 3).

Figure: 3. Determination of the queue length and delay time in TrafficData Land by specifying polygons before the stop line, where the delay time is determined.
Figure: 3. Determination of the queue length and delay time in TrafficData Land by specifying polygons before the stop line, where the delay time is determined.

 

When detecting a delay in difficult traffic, for example, on a stretch or roundabout, the delay counter turns on when the car enters the queue and turns off when leaving the queue or crossing the red line. The criterion for getting a car into the queue is the distance between the cars in the light no more than the length of the car (Fig. 4).

Figure: 4. Determination of the queue length and delay time in TrafficData Air (cars in the queue are united by yellow lines, the labels change color from black to red, the time spent in the queue is displayed).
Figure: 4. Determination of the queue length and delay time in TrafficData Air (cars in the queue are united by yellow lines, the labels change color from black to red, the time spent in the queue is displayed).

Calculated motion parameters. Further, on the basis of the collected data for the traffic sections, the calculated traffic parameters are determined:

  • Service level;
  • Congestion indicator;
  • Time index;
  • Buffer index.

Data bank formation. Traffic data can be determined from downloadable video fragments or from a video stream in real time. In the second case, the connection to the IP surveillance cameras is carried out using the RTSP link (which is implemented in the TrafficData Live software). Based on the collected data, a database for the traffic section is formed. For ease of use, it is possible to display data on traffic parameters in Excel format in the reporting documents according to the forms regulated by the order of the Ministry of Transport of the Russian Federation No. 114 dated April 18, 2019. The Automated Traffic Monitoring System presented in this article is already functioning at all regulated intersections in the city of Tobolsk.

DEVELOPMENT OF ITS FUNCTIONS

In addition to the social functions discussed in this article (the so-called Intelligent Transport Systems in the book by Roman Dushkin. – Moscow: DMK Press, 2020. – 282 p.), Modern ITS has a number of others, implemented using computer vision technology. These include supporting functions. For example, on the basis of regular data on the intensity of traffic flows and a mathematical model of the dependence of the resource of the structure on the traffic intensity, it is possible to estimate the resource of artificial structures, as well as to determine the turnaround time of the asphalt concrete pavement. But this is already a topic for the next article …