DEVELOPMENT OF A MODEL OF CONTROL AND ANALYSIS OF ROAD TRAFFIC USING THE ALGORITHM OF NEURAL NETWORKS
Accidents on road transport cause huge material and moral damage to society as a whole and to individual citizens. From year to year in the Russian Federation as a result of road accidents more than 270 thousand people are killed and injured. Road safety (road safety) is an integral part of the national objectives of ensuring personal safety, solving demographic, social and economic problems, as well as improving the quality of life and promoting the development of regions.
One of the most important tasks of the center of traffic management (gku TSODD) – prevention of offenses in the field of traffic, as well as reducing the number of road accidents associated with traffic violations. Successfully solve this problem can be installed on the roads of the country automatic complexes photo-video fixation.
The relevance of the chosen topic and the importance of this research is confirmed by the content of these regulations. In particular, among the tasks approved in the passport of the National project “Safe and high-quality roads” contains the following items: “Development of methods of redistribution of locations of cameras photo-video recording of traffic violations” with a period of implementation to 01.10.19, as well as a consistent annual (until 2024) increase in the number of fixed cameras photo-video recording on roads of various values up to 211% of the base number.
The purpose of this research is to assess the impact of different features of the road on the number of violations and search for criteria for determining the places with the largest number of violations and the allocation of these criteria, taking into account the coefficients of influence on the number of violations.
Scientific novelty. As part of this work, a study was conducted on the impact of various factors on the number of administrative offenses in the field of traffic. It was analyzed the statistics of offenses and road accidents in the Russian Federation and the Penza region on the basis of  and . In accordance with this and the diagram from  the most significant factors are identified:
visibility of the road (slopes, turns);
the condition of the road surface;
the presence of markup;
width and number of bands;
additional speed limits in this area;
time of day and seasonality;
distance from the settlement;
purpose of the road (Federal/local)
the presence of the dividing strip.
For a visual representation of the phenomenon under study, a model was developed (figure 1), which is a directed graph G = (V, E), where V is a finite set of vertices: N, F1,F2,F3,F4,F5,F6,F7,F8,F9,F10,F11, F12 where the vertex N is the number of violations in the area over a period of time, and the vertices F1-F12 correspond to the pre-selected factors.
Figure 1. Graph model of the studied dependence
The arcs on the graph represent the relationships of the influence of F1-F8 factors on the number of violations.
Dashed rectangles on the model highlight groups of factors similar in their “nature”. The factors of the first group can be described as the geometric parameters of the road, set at the design stage of the road section or in the process of its subsequent repair or expansion. The second group of factors can be conventionally combined with the concept of “road surface characteristics”, which can change over time due to external influences. The group of factors of the third group will be called “weather-time”. These are factors of random external influence, which do not depend on a person.
Dotted arcs indicate the possible influence of factors on each other, both on individual parameters and on the selected groups as a whole.
We will consider the problem of choosing the final input parameters of the model from the point of view of solving the problem of determining the potentially effective location of the complex of automatic photo-video recording of violations. Then, at this stage, we can say about the inexpediency of taking into account the factors of the weather-time group. This is due to the fact that the fixation complexes work constantly, and do not turn off depending on the time of day or weather conditions, and the weather phenomena themselves are more probabilistic in nature and can not be set as an input parameter for forecasting. Also, within the framework of this work, the factors of “Distance from the settlement” and “Additional speed limits” were also decided to be excluded from the final set and left for consideration in the framework of further additional research and improvements to expand and improve the model of this subject area.
Then, more than 2 million data from a number of complexes of photo-video fixing on the Penza region were analyzed and samples on these data were prepared.
As components of the final model, in the framework of the current work, the factors from the model in figure 1 were identified, except for the above. Their empirical estimates were matched and numerical coefficients assigned.