KR102540794B1 - Accident detection system - Google Patents

Abstract

This embodiment relates to a road accident detection system. A road accident detection system according to an aspect includes a collection unit including a sound collection unit collecting sound information on a road and a traffic information collection unit collecting traffic information on the road; an accident analysis unit that determines whether or not there is an accident on the road through the information collected by the collection unit; And an accident processing unit that bubbles up in the information detected by the accident analysis unit and provides a treatment plan, wherein the accident analysis unit analyzes the acoustic information collected by the sound collection unit and detects an accident through a primary algorithm. It includes a deep learning model, the deep learning model is made through a CNN (Convolution Neural Network), and when an accident is detected through the first algorithm, a second algorithm is made based on the information collected by the traffic information collection unit. .

Description

Road accident detection system {Accident detection system}

This embodiment relates to a road accident detection system.

When many cars drive on the road, traffic accidents occur due to various causes, resulting in human life and property damage. In particular, in areas where vehicles are traveling at high speed, such as highways, there is a very high risk of secondary accidents after the first accident, so a system capable of notifying subsequent vehicles of the occurrence of an accident is required.

Recently, an accident detection system through video has been widely used, but cases in which accidents are detected incorrectly due to special situations such as weather conditions and fire smoke are increasing. The accident detection system through video itself can deliver immediate road information, but due to the above problems, an accident detection system that compensates for the disadvantages of the video accident detection system is required.

The present invention has been proposed to improve the above problems, and provides a road accident detection system capable of detecting whether an accident has occurred by utilizing various information collected from the road and quickly securing a countermeasure for the accident. is in providing

The road accident detection system according to the present embodiment includes a collection unit including a sound collection unit that collects sound information on the road and a traffic information collection unit that collects traffic information on the road; an accident analysis unit that determines whether or not there is an accident on the road through the information collected by the collection unit; And an accident processing unit that bubbles up in the information detected by the accident analysis unit and provides a treatment plan, wherein the accident analysis unit analyzes the acoustic information collected by the sound collection unit and detects an accident through a primary algorithm. It includes a deep learning model, the deep learning model is made through a CNN (Convolution Neural Network), and when an accident is detected through the first algorithm, a second algorithm is made based on the information collected by the traffic information collection unit. .

According to this embodiment, in addition to sound information, various information such as road traffic information may be utilized as information for accident detection, so that accidents on the road may be identified and processed more quickly and accurately.

1 is a block diagram of a road accident detection system according to an embodiment of the present invention;
2 is a diagram showing the data flow of a road accident detection system according to an embodiment of the present invention.
3 is a conceptual diagram for explaining the installation and operation of a traffic information collection unit on a road according to an embodiment of the present invention;
4 is an exemplary diagram for explaining classification of sound information of a sound collecting unit according to an embodiment of the present invention;
5 is a diagram for explaining a process of converting a sound source into two-dimensional information through a fast Fourier according to an embodiment of the present invention.
6 is a diagram for explaining a process of converting a sound source into two-dimensional information through a logmel spectogram according to an embodiment of the present invention;
7 is a diagram for explaining an EAD process according to an embodiment of the present invention;
8 is a diagram for explaining an accident detection algorithm through CNN according to an embodiment of the present invention.
9 is a graph for explaining accident detection through traffic flow speed information according to an embodiment of the present invention.
10 is a graph for explaining accident detection through information on the number of vehicles in a tunnel according to an embodiment of the present invention.
11 is a table for explaining a process of determining an accident through traffic flow speed information and traffic volume information on a road according to an embodiment of the present invention.
12 is a table showing test results of accident detection through a road accident detection system according to an embodiment of the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

However, the technical idea of the present invention is not limited to some of the described embodiments, but may be implemented in a variety of different forms, and if it is within the scope of the technical idea of the present invention, one or more of the components among the embodiments can be selectively implemented. can be used by combining and substituting.

In addition, terms (including technical and scientific terms) used in the embodiments of the present invention, unless explicitly specifically defined and described, can be generally understood by those of ordinary skill in the art to which the present invention belongs. It can be interpreted as meaning, and commonly used terms, such as terms defined in a dictionary, can be interpreted in consideration of contextual meanings of related technologies.

Also, terms used in the embodiments of the present invention are for describing the embodiments and are not intended to limit the present invention. In this specification, the singular form may also include the plural form unless otherwise specified in the phrase, and when described as “at least one (or more than one) of A and (and) B and C”, a combination of A, B, and C Can include one or more of all possible combinations.

Also, terms such as first, second, A, B, (a), and (b) may be used to describe components of an embodiment of the present invention.

These terms are only used to distinguish the component from other components, and the term is not limited to the nature, order, or order of the corresponding component.

In addition, when a component is described as being 'connected', 'coupled' or 'connected' to another component, the component is not only directly connected to, coupled to, or connected to the other component, but also the component It may also include cases of being 'connected', 'combined', or 'connected' due to another component between the and other components.

In addition, when it is described as being formed or disposed on the "top (top) or bottom (bottom)" of each component, the top (top) or bottom (bottom) is not only when two components are in direct contact with each other, but also It also includes cases where one or more other components are formed or disposed between two components. In addition, when expressed as “up (up) or down (down)”, it may include the meaning of not only the upward direction but also the downward direction based on one component.

1 is a block diagram of a road accident detection system according to an embodiment of the present invention, FIG. 2 is a diagram showing the data flow of the road accident detection system according to an embodiment of the present invention, and FIG. 3 is an embodiment of the present invention It is a conceptual diagram for explaining the installation and operation of a traffic information collection unit on a road according to , FIG. 4 is an exemplary diagram for explaining sound information classification of a sound collection unit according to an embodiment of the present invention, and FIG. 5 is an embodiment of the present invention. 6 is a diagram for explaining a process of converting a sound source into 2D information through a fast Fourier, and FIG. 6 explains a process of converting a sound source into 2D information through a logmel spectogram according to an embodiment of the present invention. FIG. 7 is a diagram for explaining an EAD process according to an embodiment of the present invention, FIG. 8 is a diagram for explaining an accident detection algorithm through CNN according to an embodiment of the present invention, and FIG. 10 is a graph for explaining accident detection through information on the number of vehicles in a tunnel according to an embodiment of the present invention, and FIG. 11 is a graph for explaining accident detection through traffic speed information according to an embodiment of the present invention. A table for explaining a process of determining an accident through traffic flow speed information and traffic volume information on a road according to an embodiment of the present invention, and FIG. 12 is a test result of accident detection through a road accident detection system according to an embodiment of the present invention. is a table showing

1 to 12, a road accident detection system 10 according to an embodiment of the present invention includes a collection unit 100 that collects various information on the road and determines whether there is an accident on the road through the collected information. It may include an accident analysis unit 200 to do, and an accident processing unit 300 to provide and perform a handling plan when an accident occurs.

The collection unit 100 may include a sound collection unit 110 that collects sound information on the road and a traffic information collection unit 120 that collects traffic information on the road.

The sound collection unit 110 may collect sound information on the road. The sound collecting unit 110 may include a microphone, an audio codec, a microprocessor, and Ethernet.

The microphone may convert sound heard from the road into an electrical signal. The microphone may be referred to as a sound collector 111. The sound collector 111 may be installed on a road.

The audio codec transmits a signal to the accident analysis unit 200, and through this, the analog signal collected by the microphone can be converted into digital information so that the analysis of the signal transmitted from the accident analysis unit 200 is easy. there is.

The microprocessor controls peripheral chips, collects data, performs calculations, and receives and stores sound digital data received from the audio codec in a memory. In addition, the microprocessor may perform a CPU role of the sound collection unit 110 to control Ethernet.

The Ethernet enables communication between the sound collection unit 110 and the accident analysis unit 200 and between the sound collection unit 110 and an external server through TCP/IP.

The traffic information collecting unit 120 may collect traffic information on the road. The traffic information on the road may include speed information of vehicles traveling on the road and traffic volume information on the road. Here, the traffic volume information on the road may be the number of vehicles running within a specific section.

The traffic information collection unit 120 may include a speed sensor 121 . As shown in FIG. 3 , the speed sensors 121 may be provided in plural and spaced apart from each other within a specific section. Separation distances between the plurality of speed sensors 121 may be different from each other. For example, the plurality of speed sensors 121 include a first speed sensor 121-1, a second speed sensor 121-2, a third speed sensor 121-3, and a fourth speed sensor 121-1. 4), when the fifth speed sensor 121-5 is included, the separation distance between the first speed sensor 121-1 and the second speed sensor 121-2, the second speed sensor 121 -2) the separation distance between the third speed sensor 121-3, the separation distance between the third speed sensor 121-3 and the fourth speed sensor 121-4, the fourth speed sensor 121 -4) and the fifth speed sensor 121-5 may be different from each other.

Vehicle speed information for each section collected by the plurality of speed sensors 121 may be collected.

The speed sensor 121 may detect the speed of a vehicle traveling on the road using microwave, infrared, laser, or image methods. Among them, microwave, infrared, and laser-type speed sensors transmit radio waves to approaching vehicles and calculate the speed of the moving vehicle by calculating the return time after being reflected by the vehicle. The image speed sensor compares images (image frames) before and after a certain period of time to measure the moving distance of the vehicle, and then divides the distance by time to calculate the speed.

The traffic information collection unit 120 measures the speed of each vehicle traveling on the road through the speed sensor 121 to determine the traffic flow and traffic volume on the road.

The traffic flow on the road through the traffic information collection unit 120 may be regarded as an average speed of vehicles passing through a specific section for a certain period of time (a minute). For example, if the current traffic flow speed is the average speed of vehicles passing in a minute, the average speed of the next period is considered to be the average speed of (1/n minutes, a+1/n minutes). Here, the values of a and n may be set in various ways according to the road environment and time zone.

The traffic information collecting unit 120 may also include a microprocessor and Ethernet. Therefore, the degree of vehicle speed detected by the speed sensor 121 is transmitted to the microprocessor, and the microprocessor calculates the average speed for the time defined above (a minute) and calculates the average speed for a certain period of time (1/n minute). The average speed for the time (a minute) defined for each may be transmitted to the accident analysis unit 200.

The traffic information collection unit 120 may include a traffic volume detection sensor 125 . The traffic detection sensor 125 may detect the number of vehicles present in a specific section. For example, when the traffic information collection unit 120 is disposed in a tunnel, the traffic volume detection sensor 125 is installed at the top of the tunnel entrance, and measures the vehicle entering the tunnel entrance, such as whether or not the vehicle has passed through the tunnel entrance for each lane. Can detect logarithms. The traffic detection sensor 125 may include a camera.

On the other hand, using the average speed for each section collected by the speed sensor 121 of the traffic information collection unit 120, the expected time required for the vehicle passing through the current entrance to reach the exit can be measured.

For example, as shown in FIG. 3, a specific section on the road is divided into a total of n sections, the average speed of the 11th section having a section distance m ₁ is v ₁ , and similarly, the nth section having a section distance m _n Assuming that the average speed of is v _n , the expected time for the vehicle that has passed through the entrance to arrive at the exit can be estimated as m ₁ /v ₁ +m ₂ /v ₂ ++m _n /v _n . Therefore, the time required for the vehicle to pass through the specific section can be measured through the expected time and the traffic volume detection sensor 125 disposed at the entrance side in the specific section.

Therefore, the traffic information collection unit 120 determines the number of vehicles existing in a specific section on the road and the vehicles passing through the exit of the specific section at the corresponding time through the speed sensor 121 and the traffic detection sensor 125. can detect the number of

Similarly, the traffic volume detection sensor 125 may be installed at the upper end of an exit within a specific section to detect the number of vehicles actually exiting the exit.

On the other hand, the number of vehicles actually passing through a specific section through the traffic detection sensor 125 disposed on the exit side and the number of speed sensors 121 and the traffic detection sensor 125 disposed on the entrance side are estimated. When the difference in the number of vehicles exiting a specific section exceeds a certain range, it may be determined that the vehicles on the road are congested due to an unexpected situation.

Accordingly, the accident analysis unit 200 may analyze whether an accident has occurred by analyzing sound information on the road.

Meanwhile, the traffic information collection unit 120 may include an individual vehicle recognition unit that recognizes each individual vehicle, and may detect where each vehicle is located on the road through the individual vehicle recognition unit. In this case, a tag to be detected by the individual vehicle identification unit may be attached to each vehicle.

The accident analysis unit 200 analyzes accidents through processes of collection, processing, analysis, and external link interface.

In the collecting process, the acoustic data received from the collecting unit 100 may be transmitted to the collecting server 210 . Due to this, the collection server 210 can deliver data to each process required for data processing.

Each of the above processes may include DBSAVE, StreamGW, and ADSS (Automated Decision Support System).

In the DBSAVE, sound and speed data information is stored in the database 220 every minute.

The StreamGW, as its own server, can deliver collected real-time sound and speed information to the client server 260.

The ADSS analyzes the received sound and speed data to detect an accident, transmits the detected data to GWIF, and stores the detected data in the database 220.

The GWIF is in charge of control information input from an external server and all interfaces of transactions, and can perform controller setting and data storage functions.

In summary, when the data received from the collecting unit 100 is transmitted to each process, each process can perform functions of storage, real-time streaming, and accident detection through the received acoustic data. In addition, the accident-detected data is stored in the database 220 through an external connection interface, and the management server extracts the stored data from the database 220 and delivers it to the administrator.

Meanwhile, the road accident detection system 10 may include an integrated interface 250 for integrating detected information and accident information.

The algorithm through the deep learning model 240 of the accident analysis unit 200 will be described.

In the accident analysis unit 200, the sound source collected in real time can be divided into sound source data of a certain time unit (n seconds). Accordingly, as shown in FIG. 4 , specific sound source data is generated, and the next sound source may be generated again after a specific time period (m seconds) smaller than the frame length (n seconds). For example, if the first sound source includes sound information from 0 to n seconds, the next sound source may include sound information from m seconds to n+m seconds.

The sound sources classified as above may be converted into two-dimensional information. Conversion into 2D information may be performed by a fast Fourier transform as shown in FIG. 5 or a log-mel spectrogram as shown in FIG. 6 . According to FIGS. 5 and 6 , sound information may be converted into a two-dimensional image having time as an X-axis and a frequency band as a Y-axis.

An event activity detection (EAD) process may be added to filter whether or not a specific event is included before inputting the converted result into an algorithm. As shown in FIG. 7, the EAD processor may compare the information value of the matrix generated by the transformation with the previous frame and the current frame, and determine whether or not the information value of the matrix is increased by a specific multiple to be input into the model. Here, the specific multiple means a value of about 1.05 to 1.15 and can be changed according to the tunnel environment or the type of event to be detected.

When the EAD process is applied, the load on the server can be reduced by about 50% compared to when all existing frames are included in the model.

As shown in FIG. 8 , the accident analysis unit 200 may analyze accidents through a Convolution Neural Network (CNN), which is a convolutional neural network. A frame imaged through conversion may be input to the CNN algorithm.

The deep learning model 240 of the accident analysis unit 200 may include a primary algorithm. First, the output values obtained through the primary algorithm may include the type of detected event, the intensity of collected acoustic information, and the degree of similarity of the detected event.

The structure and parameters of the primary algorithm, events to be classified, and the like can be changed according to the environment and structure of the road. In general, crash sounds, sudden braking sounds, environmental sounds, and the like are events to be classified, and horn sounds, sirens, and the like may be additionally included.

When an event is detected by the first algorithm, the detected information can be transmitted to the detection server 230, and the detection server 230 can perform a second algorithm to verify the detection information according to the first algorithm. .

In the second algorithm, the type of detected event, the intensity of the collected sound information, the similarity (probability) of the detected event, which are the output values obtained through the first algorithm, and the traffic flow speed information and traffic volume provided by the traffic information collection unit 120 The information can be used to verify the results of the first-order algorithm.

If an accident is detected in the first algorithm, the second algorithm can be made through the traffic flow speed information. The secondary algorithm through the traffic flow speed information can be performed by a process of determining whether the traffic flow speed of the section of the time zone where the accident occurred and the section of the previous time zone significantly decreased. This is in consideration of the fact that, in the event of an accident, the traffic flow of following vehicles would be obstructed, thereby reducing the traffic flow speed.

Here, as shown in FIG. 9, a significant decrease in traffic flow speed means a speed slower than the 95% confidence interval of the average speed for one hour before the accident.

The secondary algorithm may be performed through traffic information. When the number of vehicles currently in the tunnel is significantly greater than the expected number of vehicles in the tunnel using the average speed through the traffic volume sensing device, it can be determined that the vehicle cannot pass through the tunnel due to an accident.

As shown in FIG. 10 , a significant increase in traffic volume means a number of vehicles greater than the 95% confidence interval of the expected number of vehicles in a specific section of the current road.

Therefore, through the secondary algorithm in the detection server 230, the results of the primary algorithm of the deep learning model 240 determine whether or not the traffic flow speed significantly decreased, the extent of the decrease rate, and the significant increase and amount of increase in the traffic volume in the tunnel. Additionally, it is possible to definitively detect whether an accident has occurred.

As a specific example of the second algorithm, as shown in FIG. 11, when an accident is detected by the first algorithm, a significant decrease in traffic flow speed and an increase in traffic volume on the road are all detected. An “accident confirmation” message is displayed. And, if only one of the two items is detected, “Accident warning” is detected.

In summary, the road accident detection system 10 according to an embodiment of the present invention primarily detects whether an accident has occurred with a primary algorithm using the acoustic information collected by the acoustic collecting unit 110 through CNN, and thereby Accordingly, when a situation is detected, it is possible to definitively detect whether an accident has occurred by a secondary algorithm through the traffic information collected by the traffic information collection unit 120.

When the accident analysis unit 200 determines that an accident has occurred, detection information is collected through the integrated interface 250 .

The manager server 260 receives the event information detected from the corresponding interface, gives the road operator sound data detected by the primary algorithm, and detects accidents using traffic flow and changes in traffic volume detected by the secondary algorithm as well as information. can be passed later.

Accident display lights capable of displaying two-color (red, orange, or yellow) lights are installed at regular intervals on the inner wall of a road such as a tunnel, and the road operator can operate them for each section.

Here, as the section, a distance between sound collection periods may be set as a section, a distance between traffic flow velocity collection periods may be set as a section, or an arbitrary section may be set.

If the secondary algorithm results in an accident confirmation, the accident display light in the section to which the sound collector that detected the accident belongs can be turned on in red, and the previous section can be turned on in orange.

If the secondary algorithm shows an accident warning as a result, the accident display light in the section to which the sound collector that detected the accident belongs can be turned on in orange.

In the above method, it is possible to inform the cars entering the accident zone of the attention request message through a change in the color of the lighting, which can have a higher identification effect than a message delivered through a message such as VMS.

Additional processing measures include adjusting the facilities 310 (eg, VMS (Variable Message Sign), LCS (Line Control System), emergency broadcasting, etc.) associated with the road, and using the facilities to event In the event of an accident, secondary accidents that may occur subsequently can be prevented.

In addition, it is possible to rescue people due to emergency rescue activities by disseminating the contents of the event as soon as an event occurs to the related police station, fire station, road management team, etc.

In the above, even though all the components constituting the embodiment of the present invention have been described as being combined or operated as one, the present invention is not necessarily limited to these embodiments. That is, within the scope of the object of the present invention, all of the components may be selectively combined with one or more to operate. In addition, terms such as 'include', 'comprise' or 'having' described above mean that the corresponding component may be present unless otherwise stated, and thus exclude other components. It should be construed as being able to further include other components. All terms, including technical or scientific terms, have the same meaning as commonly understood by a person of ordinary skill in the art to which the present invention belongs, unless defined otherwise. Commonly used terms, such as terms defined in a dictionary, should be interpreted as being consistent with the contextual meaning of the related art, and unless explicitly defined in the present invention, they are not interpreted in an ideal or excessively formal meaning.

The above description is merely an example of the technical idea of the present invention, and various modifications and variations can be made to those skilled in the art without departing from the essential characteristics of the present invention. Therefore, the embodiments disclosed in the present invention are not intended to limit the technical idea of the present invention, but to explain, and the scope of the technical idea of the present invention is not limited by these embodiments. The protection scope of the present invention should be construed according to the claims below, and all technical ideas within the equivalent range should be construed as being included in the scope of the present invention.

Claims (5)

a collecting unit including a sound collecting unit that collects sound information on the road and a traffic information collecting unit that collects traffic information on the road;
an accident analysis unit that determines whether or not there is an accident on the road through the information collected by the collection unit; and
Based on the information detected by the accident analysis unit, including an accident processing unit that provides a handling plan,
The accident analysis unit includes a deep learning model for detecting an accident through a primary algorithm by analyzing the acoustic information collected by the acoustic collection unit,
The deep learning model is made through a Convolution Neural Network (CNN),
When an accident is detected through the first algorithm, a second algorithm is performed based on the information collected by the traffic information collection unit,
The traffic information collection unit includes a speed sensor for detecting the speed of a vehicle running on the road, an entrance traffic information collection unit disposed at an entrance side of a specific section, and an exit traffic information collection unit disposed at an exit side of a specific section. include,
The specific section is divided into a plurality of sections,
The speed sensor is provided in plurality and disposed in each of the plurality of sections,
The expected passage time of the specific section is detected through the average speed detected by the speed sensor of each of the plurality of sections and the section distance of each of the plurality of sections,
The difference between the number of vehicles exiting the specific section through the exit traffic information collection unit and the number of vehicles exiting the specific section estimated through the entrance traffic information collection unit and the expected passage time of the specific section A road accident detection system in which an unexpected situation within the specific section is detected when it is determined that a certain range is exceeded.
According to claim 1,
The road accident detection system of each of the plurality of sections is different from each other.
According to claim 1,
After detecting an accident through the primary algorithm, when both a decrease in traffic flow speed in a specific section on the road through the traffic information collection unit and an increase in traffic volume in a specific section in the road are both detected, the road accident detection system detects the accident confirmation.
According to claim 1,
After detecting an accident through the primary algorithm, an accident is detected when either a decrease in the traffic flow speed of a specific section of the road or an increase in the traffic volume of the specific section of the road is detected through the traffic information collection unit Road accident detection system.
According to claim 4,
It is installed on the road and includes an accident display light that displays lights of different colors,
When an accident is confirmed in a specific section of the road, red light is turned on in the accident display light in the specific section of the road,
When an accident is detected in a specific section of the road, a road accident detection system in which a color different from the red is turned on in an accident display light in the specific section of the road.

Images