WO2019103197A1 - System for predicting traffic accident on basis of artificial intelligence and method therefor - Google Patents

Abstract

The present invention relates to: a system for predicting a traffic accident on the basis of artificial intelligence, the system using image recognition and deep learning techniques so as to predict a vehicle, a person or a road that are at high risk for an accident, and transmitting a predicted result to a control center so as to quickly take follow-up actions, thereby enabling an accident to be prevented in advance; and a method therefor. The system for predicting a traffic accident on the basis of artificial intelligence is implemented by comprising: a vehicle system, which recognizes an object from an image captured through a camera, analyzes the recognized object with an artificial intelligence-based accident probability algorithm so as to predict an accident probability, and transmits accident prediction data to a control center when the occurrence of an accident is predicted as a result of the prediction; and the control center, which analyzes the accident prediction data transmitted from the vehicle system so as to transmit action information to authorities according to the degree of emergency, learns the accident probability by using the accident prediction data, and transmits the learning result to the vehicle system so as to update the artificial intelligence-based accident probability algorithm.

Description

Artificial intelligence based traffic accident prediction system and method

The present invention relates to an artificial intelligence-based traffic accident prediction system and method thereof, and more particularly, to a traffic accident prediction system and method using artificial intelligence based on image recognition and deep learning, The present invention relates to an artificial intelligence-based traffic accident prediction system and a method thereof, which can prevent an accident in advance by promptly taking a follow-up action.

The rapid development of automotive technology has led to an increase in the number of vehicles worldwide, and on the other hand, the vehicle has become a convenient civilization, but on the other hand it has become a tool that can cause car accidents and harm lives and property.

As the number of vehicles increases, the importance and necessity of traffic information has been continuously increased. Various systems have been proposed to collect and provide traffic information according to the necessity. In order to detect traffic accidents and take quick follow-up measures, An accident detection system has also been proposed.

[0006] The techniques proposed in the past for a traffic accident detection system are disclosed in Patent Documents 1 to 3 below.

The prior art disclosed in Patent Document 1 is composed of an optical cable fixed to a guardrail installed on a road, a controller capable of transmitting and receiving a frequency provided at a start point and an end point of a guardrail provided with an optical cable, and a control head connected to the controller through wired / wireless communication In case of a traffic accident, the control headquarters can confirm the occurrence of a traffic accident in real time, and can precisely check the location of an accident when a traffic accident occurs, so that it can respond quickly to the structure of life and the accident recovery.

In addition, the prior art disclosed in Patent Document 2 includes a traffic accident-causing element sensing unit provided inside a vehicle to detect a traffic accident-causing element and then transmit a sensed signal to the outside, A signal transmission unit for transmitting position information to the sensing signal transmitted from the traffic accident-causing element sensing unit, and a sensing signal generated from the insured vehicle subscribed to the insurance product of the insurer, It implements the detection signal providing service system of the traffic accident causing element including the insurance company server. With this configuration, the signaling unit that receives the signal of detecting the traffic accident-causing element in the vehicle during operation provides the insurer with insurance premium from the insurance company, and the insured who insured the insurance provides the traffic accident inducing element So that the occurrence of traffic accidents can be prevented.

Further, in the conventional technique disclosed in Patent Document 3, when a vehicle is detected that continues beyond a predetermined speed toward an intersection after the vehicle signal lamp of the first diagram of the intersection is changed to a red signal, the other second or second And the vehicle signal lamp of the third road is maintained as a red signal for a predetermined time and then the vehicle of the first roadway is converted into a straight or left turn signal after passing through the intersection so as to violate the signal at the intersection and run toward the intersection By detecting an overspeed vehicle and actively delaying or changing the vehicle traffic light at an intersection, it is possible to prevent traffic accidents due to speeding and signal violation at an intersection.

[Prior Art Literature]

(Patent Document 1) Korean Patent Laid-Open No. 10-2010-0072931 (2010.07.01) (Traffic Accident Detection System and Detection Method)

(Patent Document 2) Korean Patent Laid-open Publication No. 10-2013-0026538 (published on Mar. 13, 2013) (System and Method for Providing Detection Signal of Traffic Accident Indicating Element)

(Patent Document 3) Korean Patent Laid-open No. 10-2016-0092959 (published on 2016.08.05) (Method and system for preventing traffic accident at intersection due to speeding and signal violation)

However, the general traffic accident detection system and the conventional technology described above detect the occurrence of a vehicle accident, and it is impossible to predict a traffic accident before the occurrence of a traffic accident and to take a follow-up action.

In addition, although the prior art disclosed in Patent Document 2 can detect a traffic accident element in advance to determine a traffic accident, this Patent Literature 2 also predicts an accident probability based on artificial intelligence, It is impossible to predict the traffic accident in advance and take quick action in cooperation with the control center.

Accordingly, the present invention has been proposed in order to solve all the problems occurring in the related art as described above, and it is an object of the present invention to predict a vehicle, a person and a road with high accident risk by using image recognition and deep learning The present invention relates to an artificial intelligence-based traffic accident prediction system and a method thereof, which can prevent an accident in advance by promptly transferring a traffic accident to a control center.

In order to achieve the above object, the artificial intelligence-based traffic accident prediction system according to the present invention recognizes an object from an image photographed by a camera and analyzes the recognized object with an artificial intelligence-based accident probability algorithm A vehicle system for predicting the probability of an accident and transmitting accident prediction data to the control center when the accident is predicted as a result; Analyzing the accident prediction data transmitted from the vehicle system, transmitting the action information to the related organization according to the urgency, learning about the accident probability using the accident prediction data, transmitting the learning result to the vehicle system, And a control center for updating the intelligence-based accident probability algorithm.

In the above, the vehicle system includes an image recognition unit for photographing through a camera to acquire an image; An image data storage unit for storing image data recognized by the image recognition unit; An object recognition unit for recognizing an object from the image data stored in the image data storage unit, analyzing the recognized object information using an artificial intelligence-based accident probability algorithm to calculate an accident probability, predicting an accident based on the calculated accident probability value, And an accident predicting unit for transmitting the accident prediction data to the control center when it is predicted that an accident has occurred.

The vehicle system may further include a wireless transmission unit for converting the accident prediction data output from the accident prediction unit into wireless data and transmitting the wireless data to a remote control center.

The accident predicting unit may include an artificial intelligence-based accident probability processor for recognizing object information from the image data and analyzing the recognized object information using an artificial intelligence-based accident probability algorithm to calculate an accident probability; A comparison and analysis unit for comparing the accident probability value calculated by the artificial intelligence based fault probability processing unit with a reference value and outputting the comparison result; And a normal presence / absence determination unit for performing an accident prediction based on the comparison result output from the comparison / analysis unit and transmitting the accident prediction data to the control center if the prediction result indicates that an accident has occurred.

In the above, the artificial intelligence-based incident probability processing unit extracts objects in the image data using a CNN (Convolution Neutral Networks) algorithm, and analyzes the extracted objects to calculate an incident probability.

Wherein the control center comprises a control center radio receiving unit for receiving the accident prediction data transmitted from the vehicle system; A control center data storage for storing the accident prediction data received by the control center radio receiver; An accident probability determination unit for analyzing the accident prediction data stored in the control center data storage unit to calculate an accident probability and determining an urgency based on the extracted accident probability.

In addition, the control center may further include an emergency request transmitting unit for selecting an affiliated institution according to the degree of urgency determined by the accident probability determining unit, and then transmitting urgent request information for an action to the selected related institution.

The control center may further include an accident probability learning unit for learning through deep learning using the received accident prediction data and updating an artificial intelligence based accident probability algorithm in the vehicle system based on the learning result, do.

According to an aspect of the present invention, there is provided an artificial intelligence-based traffic accident prediction method, comprising: (a) recognizing an object from an image captured through a camera in a vehicle system; Estimating an accident probability by analyzing with an accident probability algorithm, and transmitting accident prediction data to a control center when an accident occurrence is predicted as a result; (b) analyzing the accident prediction data transmitted from the vehicle system at the control center and transmitting the action information to the related organization according to the degree of urgency, learning about the accident probability using the accident prediction data, And updating the artificial intelligence-based accident probability algorithm by transmitting to the vehicle system.

The step (a) includes: (a1) capturing an image through a camera to acquire and store an image, and recognizing an object from the stored image data; (a2) analyzing the recognized object information with an artificial intelligence-based accident probability algorithm to calculate an accident probability; (a3) predicting an accident on the basis of the calculated accident probability value, and transmitting the accident prediction data to the control center when it is predicted that the accident occurred; (a4) receiving the accident probability algorithm update data from the control center, updating the artificial intelligence-based accident probability algorithm.

(B) comprises: (b1) receiving and storing accident prediction data transmitted from the vehicle system; (b2) analyzing the stored accident prediction data to calculate an accident probability, and determining an urgency based on the extracted accident probability; (b3) selecting an affiliated institution according to the determined degree of urgency, and then transmitting urgent request information for an action to the selected affiliated institution; (b4) learning through deep learning using the stored accident prediction data, and updating an artificial intelligence-based accident probability algorithm in the vehicle system based on the learning result.

According to the present invention, by using the image recognition and deep learning technique, it is possible to predict a vehicle, a person and a road with a high risk of an accident, transmit the prediction result to the control center, There is an advantage that it can be prevented.

1 is a block diagram of an artificial intelligence-based traffic accident prediction system according to the present invention;

FIG. 2 and FIG. 3 are flowcharts illustrating a method for predicting a traffic accident based on artificial intelligence according to the present invention.

Hereinafter, an artificial intelligence-based traffic accident prediction system and method according to a preferred embodiment of the present invention will be described in detail with reference to the accompanying drawings.

FIG. 1 is a block diagram of an artificial intelligence-based traffic accident prediction system according to a preferred embodiment of the present invention. The object is recognized from an image captured through a camera, and an artificial intelligence-based accident probability algorithm is applied to the recognized object (100) for predicting the probability of an accident and transmitting the predicted accident prediction data to the control center when the occurrence of the accident is predicted as a result of the prediction, and an accident prediction data transmitted from the vehicle system (100) And a control center 200 for transmitting action information, learning about an accident probability using the accident prediction data, and transmitting the learning result to the vehicle system 100 to update an artificial intelligence-based accident probability algorithm .

The vehicle system 100 includes an image recognition unit 10 that captures an image through a camera and acquires an image, an image data storage unit 20 that stores image data recognized by the image recognition unit 10, And estimates the accident on the basis of the calculated accident probability value. If the predicted result is predicted as an accident occurrence, the accident prediction data is obtained A wireless transmission unit 40 for converting the accident prediction data output from the accident prediction unit 30 into wireless data and transmitting the wireless data to a remote control center 200, .

The accident predicting unit 30 includes an artificial intelligence-based accident probability processing unit 31 for recognizing object information from the image data and analyzing the recognized object information with an artificial intelligence-based accident probability algorithm to calculate an accident probability, A comparison / analysis unit 32 for comparing the accident probability value calculated by the artificial intelligence-based incident probability processing unit 31 with a reference value and outputting the comparison result; And a normal presence / absence determination unit (33) for transmitting the accident prediction data to the control center (200) if it is predicted that an accident has occurred.

The control center 200 includes a control center radio receiving unit 50 for receiving the accident prediction data transmitted from the vehicle system 100, a control unit 50 for storing the accident prediction data received by the control center radio receiving unit 50, An accident probability determination unit 70 for calculating an accident probability by analyzing the accident prediction data stored in the center data storage unit 60 and determining the urgency based on the extracted accident probability, , And an emergency request transmission unit (80) for selecting an affiliated institution according to the degree of urgency determined by the accident probability determination unit (70), and then transmitting urgent request information for an action to the selected related institution.

The control center 200 preferably includes an accident probability learning unit 90 that learns through deep learning using the received accident prediction data and updates an artificial intelligence based accident probability algorithm in the vehicle system based on the learning result, .

The operation of the artificial intelligence-based traffic accident prediction system according to the preferred embodiment of the present invention will be described in detail as follows.

First, when the vehicle starts to be driven, the vehicle system 100 acquires an image by photographing it through a camera in the image recognition unit 10. [ Here, it is assumed that the camera is installed at various positions of the vehicle to acquire an image around the vehicle, but in the present invention, it is assumed that the camera is installed in front of the vehicle to photograph the vehicle traveling direction to acquire images.

The image data recognized by the image recognition unit 10 is stored through the image data storage unit 20.

Next, the accident predicting unit 30 recognizes an object from the stored image data, calculates an accident probability by analyzing the recognized object information by an artificial intelligence-based accident probability algorithm, and calculates an accident probability based on the calculated accident probability value And transmits the accident prediction data to the control center 200 when it is predicted that the accident occurred. Here, the prediction of the accident by analyzing the image data can be performed in real time or at regular intervals.

For example, the artificial intelligence-based accident probability processing unit 31 of the accident predicting unit 30 recognizes object information from the stored image data by using an artificial intelligence-based accident probability algorithm, and then analyzes the recognized object information, . Here, it is desirable to use the artificial intelligence-based accident probability algorithm using the CNN (Convolution Neural Networks) algorithm. The CNN algorithm consists of a convolution layer, a pooling layer, and a feed forward layer. The convolution layer is a layer for extracting meaningful features in layers when implementing a filter to extract convolution features. The pooling layer is a layer that performs sub-sampling to reduce the feature because there are many pixels due to the nature of the image. The feed forward layer is a layer that classifies using features extracted from the Convolution layer and the Pooling layer. That is, if the window or mask of 3 by 3 or more is repeatedly performed on the entire image, an appropriate result can be obtained according to the coefficient values of the mask. In this way, objects are extracted from the image. Then, the extracted object is analyzed to calculate the accident probability. To obtain feature extraction and topology invariance, we perform filter and sub-sampling and repeat this process to obtain global from the local feature. Simply select the largest stimulus in each window. Multiple passes through this process (convolution + sub-sampling) will provide a global feature that can represent the entire image. By learning these features, you will have a strong cognitive ability in topology change, and use this to calculate the probability of an accident.

Next, the comparative analysis unit 32 compares the accident probability value calculated by the artificial intelligence-based accident probability processing unit 31 with a predetermined reference value, and outputs an error value, which is the difference, as a comparison result.

Then, the normal presence / absence determiner 33 compares the error value, which is a comparison result output from the comparison / analysis unit 32, with a setting value for determining whether the error is normal or not. If the error value is less than the set value, And deletes the acquired image data. Since the deletion of unnecessary image data does not require the use of a large-capacity memory, the system implementation cost can be reduced. On the other hand, if the error value is larger than the set value, it is predicted that an accident occurs. If it is predicted that an accident has occurred, the predicted data is generated and transmitted to the wireless transmitting unit 40.

The wireless transmitting unit 40 converts the accident prediction data output from the accident predicting unit 30 into wireless data and transmits the wireless data to the remote control center 200.

The control center 200 receives the accident prediction data transmitted from the vehicle system 100 through the control center wireless receiving unit 50 and the control center data storage unit 60 receives the accident prediction data from the control center wireless receiving unit 50 And stores the received accident prediction data.

Next, the accident probability determiner 70 calculates the accident probability by analyzing the accident prediction data stored in the control center data storage 60, and determines the urgency based on the extracted accident probability. For example, the accident probability determiner 70 calculates an accident probability by adding the presently stored accident probability error value and existing data, and if the calculated accident probability value is equal to or greater than a predetermined reference value, do. Here, the emergency number is the magnitude of the accident probability value predicted. The larger the value, the more urgent the situation is, the smaller the smaller, the less urgent the situation becomes.

According to this urgency, the emergency request transmission unit 80 selects an appropriate agency. For example, if the current vehicle has a high probability of accident due to drowsiness, drunken driving or other abrupt driving, the police can be selected as the relevant authority to take immediate action when the emergency degree becomes high, and the relevant image data is sent to the police station for quick action .

In addition, if the current road condition has a high probability of accident, send the image data to the relevant agencies concerned with the road (for example, Ministry of Land, Korea Highway Corporation) and take appropriate measures.

In another aspect of the present invention, the control center 200 learns through deep learning using the accident prediction data received through the accident probability learning unit 90, To the vehicle system 100, update information for updating the AI-based accident probability algorithm. When the update information is received, the vehicle system 100 updates the artificial intelligence-based accident probability algorithm using the updated information, thereby improving the accuracy of the accident probability prediction.

FIG. 2 and FIG. 3 are flowcharts illustrating an artificial intelligence-based traffic accident prediction method according to a preferred embodiment of the present invention, wherein (a) (S101 to S108) (S101 to S108) of analyzing the object with an artificial intelligence-based accident probability algorithm to predict an accident probability, and when the accident occurrence is predicted, the accident prediction data is transmitted to the control center (200) (200) analyzes the accident prediction data transmitted from the vehicle system (100) and transmits the action information to the related organizations according to the degree of urgency, learns about the accident probability using the accident prediction data, (S201 to S207) to the vehicle system 100 and updating the AI-based accident probability algorithm.

The method includes the steps of: (a1) capturing an image through a camera to acquire and store an image, and recognizing an object from the stored image data (S101); (a2) (A3) estimating an accident on the basis of the calculated accident probability value, transmitting the accident prediction data to the control center 200 when it is predicted that the accident occurred, (S103 to S106); (a4) receiving the accident probability algorithm update data from the control center 200, updating the artificial intelligence-based accident probability algorithm (S107 To S108).

The step (b) includes the steps of (b1) receiving and storing the accident prediction data transmitted from the vehicle system 100 (S201), (b2) calculating the accident probability by analyzing the stored accident prediction data, (S203 to S203); (b3) selecting the relevant agency according to the determined degree of urgency, and then transmitting urgent request information for the action to the selected related authority (S204); (b4) learning through deep learning using the stored accident prediction data, and updating the artificial intelligence-based accident probability algorithm in the vehicle system based on the learning result (S205 to S207).

The artificial intelligence based traffic accident prediction method according to the preferred embodiment of the present invention will be described in detail as follows.

First, when the vehicle starts to run in step S101, the vehicle system 100 captures an image through the camera in the image recognition unit 10 and acquires the image. It is assumed that the camera is installed at the front of the vehicle to acquire images by photographing the traveling direction of the vehicle.

The image data recognized by the image recognition unit 10 is stored through the image data storage unit 20.

Next, in step S102, the accident predicting unit 30 recognizes the object from the stored image data using the artificial intelligence-based accident probability algorithm, and analyzes the recognized object information to calculate the accident probability. Here, it is desirable to use the artificial intelligence-based accident probability algorithm using the CNN (Convolution Neural Networks) algorithm. The CNN algorithm consists of a convolution layer, a pooling layer, and a feed forward layer. The method of calculating the probability of an accident using the CNN algorithm is described in detail in the system of FIG. 1, and will not be described here.

Next, the comparison and analysis unit 32 compares the accident probability value calculated by the artificial intelligence-based accident probability processing unit 31 with a predetermined reference value in step S103, and outputs the error value as the comparison result.

Next, the normal presence / absence determining unit 33 compares the error value, which is a comparison result outputted from the comparison / analysis unit 32, with a setting value for determining whether or not the normal state is present in step S104. If the error value is equal to or smaller than the set value It is judged as normal and the process goes to step S106 to delete the acquired image data. Since the deletion of unnecessary image data does not require the use of a large-capacity memory, the system implementation cost can be reduced.

On the other hand, if the error value is larger than the set value, it is predicted that an accident occurs. If it is predicted that an accident has occurred, the process proceeds to step S105 to generate accident prediction data and transmits the data to the remote control center 200. [

In step S107, when the update information is received, the vehicle system 100 proceeds to step S108 and updates the AI-based accident probability algorithm based on the update information to improve the accuracy of the accident probability prediction.

Meanwhile, the control center 200 receives the accident prediction data transmitted from the vehicle system 100 through the control center wireless receiving unit 50 in step S201, and the control center data storage unit 60 stores the accident prediction data transmitted from the control center wireless And stores the accident prediction data received by the receiving unit 50.

Then, the accident probability determiner 70 calculates the accident probability by analyzing the accident prediction data stored in the control center data storage 60 in step S202, and determines the urgency based on the extracted accident probability. For example, if the calculated accident probability value is equal to or greater than a preset reference value, the occurrence probability of an accident is predicted and the urgency is confirmed. Here, the emergency number is the magnitude of the accident probability value predicted. The larger the value, the more urgent the situation is, the smaller the smaller, the less urgent the situation becomes.

According to this urgency, the emergency request transmission unit 80 selects an appropriate agency. For example, if the current vehicle has a high probability of accident due to drowsiness, drunken driving or other abrupt driving, the police may be selected as the relevant authority to immediately take measures to deal with the accident, and the corresponding image data is transmitted to the police station So that they can take action.

If the current state of the road is high, the process goes to step S205 to transmit the image data to the relevant authority (for example, Ministry of Land, Korea Highway Corporation) related to the road, and take appropriate measures.

Meanwhile, as another feature of the present invention, the control center 200 performs learning through deep learning using the accident prediction data received through the accident probability learning unit 90 in step S206. In step S207, Based on the intelligence-based incident-probability algorithm in the vehicle system to the vehicle system 100 to update the artificial intelligence-based incident-probability algorithm to improve the accuracy of the accident-probability prediction.

As described above, the present invention analyzes vehicles and areas with high possibility of accidents through cameras of vehicles and roads, finds vehicles or people with abnormal behaviors, unusually violates roads, and transmits them to related agencies to take measures Thereby preventing accidents in advance.

Meanwhile, research on the present invention has been conducted as part of the technology development project of the ICT convergence industry of the Ministry of Science, Technology, Information and Communication Technology and Information and Communication Technology Promotion Center (R7118-16-1002, Organized by Korea Electronics and Telecommunications Research Institute, Research period: 2016.04.01 ~ 2019.12.31.)

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments, but, on the contrary, It is obvious to those who have.

The present invention is applied to a technique for improving the running safety of an autonomous vehicle by allowing an accident probability to be predicted using the image recognition and artificial intelligence accident probability algorithm and taking measures by using it.

Claims (11)

As a system for predicting traffic accidents based on artificial intelligence, A vehicle system for recognizing an object from a camera image, analyzing the recognized object with an artificial intelligence-based accident probability algorithm to predict an accident probability, and transmitting accident prediction data to a control center when an accident occurrence is predicted ; And Analyzing the accident prediction data transmitted from the vehicle system, transmitting the action information to the related organization according to the urgency, learning about the accident probability using the accident prediction data, transmitting the learning result to the vehicle system, And a control center for updating the intelligence-based accident probability algorithm. The vehicle system according to claim 1, wherein the vehicle system comprises: an image recognizing unit for acquiring an image by photographing through a camera; An image data storage unit for storing image data recognized by the image recognition unit; An object recognition unit for recognizing an object from the image data stored in the image data storage unit, analyzing the recognized object information using an artificial intelligence-based accident probability algorithm to calculate an accident probability, predicting an accident based on the calculated accident probability value, And an accident predicting unit for transmitting the accident prediction data to the control center when it is predicted that an accident has occurred. The system of claim 2, wherein the vehicle system further includes a wireless transmission unit for converting the accident prediction data output from the accident prediction unit into wireless data and transmitting the wireless data to a remote control center. The accident predicting unit may include an artificial intelligence-based accident probability processor for recognizing object information from the image data and analyzing the recognized object information using an artificial intelligence-based accident probability algorithm to calculate an accident probability; A comparison and analysis unit for comparing the accident probability value calculated by the artificial intelligence based fault probability processing unit with a reference value and outputting the comparison result; And a normal presence / absence determination unit for performing an accident prediction based on the comparison result output from the comparison / analysis unit and transmitting the accident prediction data to the control center if the prediction result is that an accident occurred, system. The artificial intelligence-based accident probability processing unit may extract an object in the image data using a CNN (Convolution Neural Networks) algorithm, and analyze the extracted object to calculate an accident probability. Prediction system. The control center according to claim 1, wherein the control center comprises: a control center radio receiving unit for receiving the accident prediction data transmitted from the vehicle system; A control center data storage for storing the accident prediction data received by the control center radio receiver; An accident probability determination unit for calculating an accident probability by analyzing the accident prediction data stored in the control center data storage unit and determining an urgency based on the extracted accident probability, . [Claim 6] The system of claim 6, wherein the control center further comprises an emergency request transmission unit for selecting the relevant organization according to the degree of urgency determined by the accident probability determination unit, and then transmitting the emergency request information for the action to the selected related organization Intelligent Traffic Accident Prediction System. The control center may further include an accident probability learning unit for learning through deep learning using the received accident prediction data and updating an artificial intelligence based accident probability algorithm in the vehicle system based on the learning result A traffic accident prediction system based on artificial intelligence. As a method for predicting traffic accidents based on artificial intelligence, (a) An object is recognized from an image taken through a camera in a vehicle system, and an artificial intelligence-based accident probability algorithm is applied to the recognized object to predict an accident probability. To a center; And (b) analyzing the accident prediction data transmitted from the vehicle system at the control center and transmitting the action information to the related organization according to the degree of urgency, learning about the accident probability using the accident prediction data, And updating the artificial intelligence-based accident probability algorithm by transmitting to the vehicle system the artificial intelligence-based traffic accident prediction method. [12] The method of claim 9, wherein the step (a) comprises: (a1) capturing an image through a camera to acquire and store an image, and recognizing an object from the stored image data; (a2) analyzing the recognized object information with an artificial intelligence-based accident probability algorithm to calculate an accident probability; (a3) predicting an accident on the basis of the calculated accident probability value, and transmitting the accident prediction data to the control center when it is predicted that the accident occurred; (a4) receiving the accident probability algorithm update data from the control center, updating the artificial intelligence-based accident probability algorithm. The method of claim 9, wherein step (b) comprises: (b1) receiving and storing accident prediction data transmitted from the vehicle system; (b2) analyzing the stored accident prediction data to calculate an accident probability, and determining an urgency based on the extracted accident probability; (b3) selecting an affiliated institution according to the determined degree of urgency, and then transmitting urgent request information for an action to the selected affiliated institution; (b4) learning through deep learning using the stored accident prediction data, and updating the artificial intelligence-based accident probability algorithm in the vehicle system based on the learning result. Traffic accident prediction method.

Images