KR102327478B1 - Risk assessment inference method and system based on geographical danger point for safe route to school - Google Patents

Abstract

The present invention relates to a geographical danger point-based risk assessment inference method and system for a safe route to a school. The geographical danger point-based risk assessment inference system for a safe route to a school in accordance with the present invention comprises: a geohash conversion unit that converts geographic data of input data composed of geographic data and feature data into geohash data; a data fusion unit that fuses the geohash data and the feature data; a one-hot encoder that pre-processes the fused data; a risk index model unit that calculates a risk index using the geographic data; and an RNN model unit that is trained using the pre-processed data and the risk index and calculates a predicted risk index when a route feature is input.

Description

RISK ASSESSMENT INFERENCE METHOD AND SYSTEM BASED ON GEOGRAPHICAL DANGER POINT FOR SAFE ROUTE TO SCHOOL

The present invention relates to a geographical risk point-based risk assessment reasoning method and system that provides a Safe Routes To School (SRTS) through a risk reasoning approach to predict a route risk index.

Road traffic accidents are one of the most widespread and most important problems worldwide.

According to World Health Organization statistics, road traffic accidents are the second leading cause of death among young people between the ages of 5 and 29 worldwide, with over 1.2 million people killed and approximately 50 million injured in road traffic accidents each year, according to World Health Organization statistics. are said to be suffering

A road accident or hazard is usually due to five factors, including people (drivers or pedestrians), environment (weather conditions), infrastructure (road design), traffic conditions (traffic congestion), and vehicle-related factors (vehicle age or size). known to occur.

Various methods have been studied to reduce road hazards and accident rates over the past few decades, and many researchers are working to improve vehicle design and control performance to lower risk levels.

In addition, various road evaluation programs are being used in the redesign of roads to minimize the risk of the road, and they are trying to propose an intelligent solution that can reduce the risk of risk based on the past or current sensing data of a specific road area.

Accordingly, in recent years, several methods have been developed to reduce traffic problems, and Geographic Information System (GIS) and machine learning algorithms are being used to monitor and control risks on the road.

Geographic Information System (GIS) has the advantage of more effectively visualizing large-scale data during decision-making and analysis processes, and GIS-based maps are helping to find collision points, danger points, and hazardous areas of highways and highways.

In addition, recently, an approach based on HMM (Hidden Markov Model) for destination and route prediction has been developed, and an algorithm for predicting driver routes and destinations is provided. The HMM-based prediction algorithm uses the last visited road link as an input parameter, and uses GPS to provide real-time visual prediction risk through the client application.

In the road traffic risk assessment, the steering angle and speed of the vehicle may be analyzed to predict the vehicle's movement for a certain period of time, and the predicted path of adjacent vehicles may be calculated. At this time, the HMM algorithm calculates each vehicle's steering angle and road safety level to provide continuous monitoring of road traffic risk.

However, these conventional techniques merely predict the danger on the road, and there is a problem in that specific and accurate information on a route that can be used safely and conveniently to the user cannot be provided.

The present invention has been devised to solve the above problem, and in the present invention, a risk inference approach method for predicting a route risk index using geographic information including a route and a danger point We aim to provide Safe Routes To School (SRTS) through

A geographic risk point-based risk assessment inference system for a safe isometric road according to an embodiment of the present invention for solving the above-mentioned problem, among the input data consisting of geographic data and feature data, Geohash conversion unit that converts the city (geohash) data; a data fusion unit that fuses the geohash data and the feature data; a one-hot encoder for pre-processing the fused data; a risk index model unit for calculating a risk index using the geographic data; and an RNN model unit that is trained using the pre-processed data and the risk index, and calculates a predicted risk index when a path feature is input.

According to another embodiment of the present invention, the input data may be composed of geographic danger point coordinates and risk level of a danger point of a school route, which are collected through a survey.

According to another embodiment of the present invention, the data fusion unit may integrate the geohash data and the feature data including gender, transportation means or risk type into one data set.

According to another embodiment of the present invention, the risk index model unit may calculate the risk index (Risk Index) by calculating the distance between the dangerous point (Danger Point) and the route point (Route Point) of the geographic data.

According to another embodiment of the present invention, the RNN model unit may be configured based on Long-Short Term Memory (LSTM) to calculate the predicted risk index and the risk level of the path.

A geohazard point-based risk assessment inference method for a safe equivalence road according to an embodiment of the present invention includes a geohash transforming unit, among the input data consisting of geographic data and feature data, the geohash ( geohash) a first step of converting data; A second step in which a data fusion unit fuses the geohash data and the feature data: a third step in which a One-Hot encoder pre-processes the fused data; The fourth step of calculating the risk index (Risk Index) by the risk index model unit using the geographic data, and the RNN model unit is trained using the preprocessed data and the risk index, and when a path feature is input, the predicted risk index a fifth step of calculating;

According to another embodiment of the present invention, the first step may further include; receiving the input data composed of the geographical danger point coordinates and the risk level of the danger point of the school route;

According to another embodiment of the present invention, in the second step, the data fusion unit may integrate the geohash data and the feature data including gender, transportation means or risk type into one data set.

According to another embodiment of the present invention, in the fourth step, the risk index model unit calculates the distance between the dangerous point (Danger Point) and the route point (Route Point) of the geographic data to calculate the risk index (Risk Index) can be calculated.

According to another embodiment of the present invention, the fifth step comprises: configuring the RNN model unit based on Long-Short Term Memory (LSTM); The RNN model unit is trained using the preprocessed data and the risk index; and calculating a predicted risk index and a risk level of a path when a path feature is input to the trained RNN model unit.

In the present invention, a safe route to school (SRTS) is established through a risk inference approach that predicts a route risk index using geographic information including routes and danger points. has the effect of providing.

1 is a diagram for explaining a geographical risk point-based risk assessment inference system using student questionnaire data for safe commuting according to an embodiment of the present invention.
2 is a diagram for explaining a risk index model unit according to an embodiment of the present invention.
3 is a diagram for explaining a risk assessment inference method based on a geographical risk point for a safe isometric road according to an embodiment of the present invention.
4 is a diagram for explaining an RNN learning method according to an embodiment of the present invention.

Hereinafter, an exemplary embodiment of the present invention will be described in detail with reference to the accompanying drawings. However, in describing the embodiment, if it is determined that a detailed description of a related known function or configuration may unnecessarily obscure the gist of the present invention, a detailed description thereof will be omitted. In addition, the size of each component in the drawings may be exaggerated for explanation, and does not mean the size actually applied.

Safe Routes To School (SRTS) can improve children's health and well-being, reduce traffic congestion near schools, and improve the quality of life in communities by providing safe walking and biking to and from school.

Most studies have found that road infrastructure has a major impact on the risk of traveling by vehicle and on foot. In particular, there are black spots and commonly known intersections on roads.

A black spot is a specific location where the risk of accidents and potential collisions for all road users is higher than on other roads. The risk and number of hazards on a route are usually determined by traffic conditions such as traffic composition, vehicle flow or volume, and during peak hours the level of risk and the number of accidents increase, affecting not only vehicle drivers, but also pedestrians.

Based on this data, a Geographic Information System (GIS) provides functional capture, storage, manipulation, analysis and management capabilities. Geospatial information can track not only things, activities, and events, but also where these things, activities, and events exist or occur. Therefore, a safer route can be found by eliminating risks based on predictable parameters with sufficient data.

Recently, machine learning algorithms are being used to determine future occurrences or to derive optimal results in various fields. Long-Short Term Memory (LSTM)-based Recurrent Neural Networks (RNNs) are data has been successfully applied.

In the present invention, a safe route to school (SRTS) is established through a risk inference approach that predicts a route risk index using geographic information including routes and danger points. would like to provide

1 is a diagram for explaining a geographical risk point-based risk assessment inference system using student questionnaire data for safe commuting according to an embodiment of the present invention, and FIG. 2 is a risk index model according to an embodiment of the present invention. It is a figure for demonstrating a part.

Hereinafter, a geographical hazard point-based risk assessment inference system using student questionnaire data for safe commuting according to an embodiment of the present invention will be described with reference to FIGS. 1 and 2 .

As shown in FIG. 1, the geographic risk point-based risk assessment inference system using student questionnaire data for safe equivalence according to an embodiment of the present invention includes a geohash transformation unit 110, a data fusion unit 120, It is configured to include a One-Hot encoder 130 , an RNN model unit 140 , and a risk index model unit 150 .

The geohash conversion unit 110 converts the geographic data 101 into geohash data among the input data 105 including the geographic data 101 and the feature data 102 .

At this time, the collected input data 105 is composed of spatial data expressed in two dimensions, that is, longitude and latitude, and is composed of geographic hazard point coordinates and hazard level of the danger point of the school path collected through a survey. can be

Such two-dimensional data is adjusted to a shorter single string value through the geohash transform unit 110 . The geohash algorithm used here provides a hierarchical grid-based model of the earth whose locations are represented as Base32 strings.

In order to use more accurate and consistent data in data processing, transformed geohash data and feature data are integrated through data fusion of the data fusion unit 120 .

Data fusion is used to increase accuracy and streamline overall operational processes, helping to combine data from multiple sources or related databases.

The data fusion unit 120 may integrate the geohash data and the feature data including gender, transportation means or risk type into one data set.

The fused data in this way is input to the One-Hot encoder 130 .

One-Hot encoder 130 provides a data preprocessing process that allows categorical variables to be converted into binary parameters to provide better performance to machine learning algorithms.

As such, the data encoded by the one-hot encoder 130 is input to the RNN model unit 140 to train a Long-Short Term Memory (LSTM).

The RNN model unit 140 is trained using the preprocessed data and the risk index.

In addition, the RNN model unit 140 trained in this way can calculate a predicted risk index when a path feature is input, and more specifically, the RNN model unit 140 uses a Long-Short Term Memory (LSTM). Based on it, it is possible to calculate the predicted risk index and the risk level of the route.

On the other hand, the risk index model unit 150 calculates a risk index (Risk Index) using the geographic data, at this time the risk index model unit 150 through the route point (Route Point) and the risk point (Danger Point) exponents can be calculated.

Referring to FIG. 2 , the risk index model unit 150 uses two parameters, a risk point and a path point.

At this time, the parameters of latitude and longitude of each danger point and route point are expressed using (Dlat(j), Dlng(j)) and (Rlng(i), Rlng(i)), respectively.

Latitude and longitude play an essential role in many fields such as aerospace engineering, logistics, and transportation, but one of the most useful uses is to measure the distance between two locations. In this way, the distance calculation for latitude and longitude provides the shortest, fastest and most optimal route between two locations.

According to the present invention, both the distance between the danger point (Danger Point) and the route point (Route Point) are calculated.

The risk index model unit 150 may calculate a route risk index using Equation 1 below.

[Equation 1]

Equation 1 is an equation based on the actual distance between the risk point and the path point, and may calculate a route risk index.

3 is a diagram for explaining a risk assessment inference method based on a geographical risk point for a safe isometric road according to an embodiment of the present invention, and FIG. 4 is a diagram for explaining an RNN learning method according to an embodiment of the present invention. .

Hereinafter, with reference to FIGS. 3 and 4 , a method for inferring risk assessment based on a geographical risk point for a safe equivalence road according to an embodiment of the present invention will be described.

Correct selection of input and output data is critical to performing path hazard index analysis.

The input data is based on geospatial risk point coordinates of school route risk points and risk levels collected from student surveys. Data set coordinates like this are organized in a two-dimensional format, including latitude and longitude locations.

In addition, the output data is the risk index and the risk level of the pathway.

The geohash conversion unit 110 converts the geographic data into geohash data among input data including geographic data and feature data. Geohash data is efficient at encoding latitude and longitude-based geographic locations to short digits and characters.

The geohash conversion unit 110 provides a path geohash sequence and danger point data in a geohash data format.

The transformed geohash data and feature data including hazard type, traffic and student gender are integrated into one data set by the data fusion unit 120 . The consolidated data set includes categorical variables such as gender, transport name, hazard type, and geohash-based hazard location locations.

These different ways of variable data confuse machine learning models, and to avoid this, data must be encoded. Accordingly, the one-hot encoder 130 pre-processes the fused data, and the pre-processed one-hot encoded data uses the binary format of Long-Short Term Memory (LSTM)-based Recurrent Neural Networks (RNNs). It is structured in a format that can be easily understood by a computer.

On the other hand, the risk index model unit 150 calculates a risk index (Risk Index) using the geographic data, at this time the risk index model unit 150 through the route point (Route Point) and the risk point (Danger Point) exponents can be calculated.

Thereafter, the RNN model unit 140 is trained using the preprocessed data and the risk index, and the trained RNN model unit 140 may calculate a predicted risk index when a path feature is input.

Referring to FIG. 4 , in the RNN model unit 140, input and output parameters are x and y, respectively.

Recurrent Neural Networks (RNNs) provide one of the most powerful algorithms for sequential data processing, and Long-Short Term Memory (LSTM) is defined as a computational unit (memory cell) that can replace conventional artificial neurons. Networks can be effectively coupled with these memory cells to provide high predictive performance.

As described above, geohash data and feature data are converted into a One-Hot encoded data format by the One-Hot encoder 130 .

At this time, the input layer connected to all RNN hidden neurons is composed of 38, and the output is connected to the layer fully connected to the path hazard point and path hazard index.

According to the present invention through such configurations, a safe school commuting route (SRTS: Safe Routes to School).

In the detailed description of the present invention as described above, specific embodiments have been described. However, various modifications are possible without departing from the scope of the present invention. The technical spirit of the present invention should not be limited to the above-described embodiments of the present invention, and should be defined by the claims as well as the claims and equivalents.

101: geographic data
102: feature data
105: input data
110: geohash conversion unit
120: data fusion unit
130: One-Hot encoder
140: RNN model unit
150: risk index model part

Claims (10)

a geohash conversion unit that converts the geographic data into geohash data among input data consisting of geographic data and feature data;
a data fusion unit that fuses the geohash data and the feature data;
a one-hot encoder for pre-processing the fused data;
a risk index model unit for calculating a risk index using the geographic data; and
an RNN model unit that is trained using the preprocessed data and the risk index, and calculates a predicted risk index when a path feature is input;
Geographic hazard point-based risk assessment inference system for safe commuting, including
The method according to claim 1,
The input data is
A geographic hazard point-based risk assessment inference system for safe escalation, which consists of the geographic hazard point coordinates and hazard levels of the hazard points of the school route, collected by a survey.
The method according to claim 1,
The data fusion unit,
A geographic hazard point-based risk assessment inference system for safe equivalence that integrates the geohash data and the feature data including gender, mode of transport or risk type into one data set.
The method according to claim 1,
The risk index model unit,
A geographic risk point-based risk assessment inference system for safe equivalence that calculates the distance between the Danger Point and the Route Point of the geographic data to calculate the Risk Index.
The method according to claim 1,
The RNN model unit,
A geographic hazard point-based risk assessment inference system for a safe ascent that is configured based on LSTM (Long-Short Term Memory) and calculates the predicted risk index and risk level of the route.
A first step of converting the geographic data into geohash data, among the input data consisting of geographic data and feature data, by a geohash conversion unit;
A second step in which the data fusion unit fuses the geohash data and the feature data:
a third step of pre-processing the fused data by a one-hot encoder;
A fourth step of calculating a risk index (Risk Index) by the risk index model unit using the geographic data, and
a fifth step of calculating a predicted risk index when the RNN model unit is trained using the pre-processed data and the risk index, and a path feature is input;
Geographic hazard point-based risk assessment inference method for safe escalation, including
7. The method of claim 6,
The first step is
receiving the input data consisting of geographic danger point coordinates and risk level of a danger point of a school route;
A geographical hazard point-based risk assessment inference method for safe escaping further comprising a.
7. The method of claim 6,
The second step is
A geographical hazard point-based risk assessment inference method for safe equivalence in which the data fusion unit integrates the geohash data and the feature data including gender, means of transport or risk type into one data set.
7. The method of claim 6,
The fourth step is
The risk index model unit calculates the distance between the danger point (Danger Point) and the route point (Route Point) of the geographic data to calculate the risk index (Risk Index) based on a geographical risk point-based risk assessment inference method for safe equivalence.
7. The method of claim 6,
The fifth step is
The RNN model unit is configured based on LSTM (Long-Short Term Memory);
The RNN model unit is trained using the preprocessed data and the risk index; and
calculating a predicted risk index and a risk level of a route when a route feature is input to the trained RNN model unit;
Geographic hazard point-based risk assessment inference method for safe escalation, including

Images