KR102530654B1 - Drunk driving vehicle recognition method and intervention system using cctv camera - Google Patents

Abstract

A drunk driving vehicle recognition method and response system using a CCTV camera are disclosed. A drunk driving vehicle recognition method performed by a drunk driving vehicle recognition system includes collecting image data captured using at least one camera; Determining a drunk driving vehicle by analyzing an abnormal driving pattern of the vehicle based on object information detected from the collected image data; and controlling so that warning information is provided to the determined drunk driving vehicle or vehicles around the drunk driving vehicle.

Description

Drunk driving vehicle recognition method and response system using CCTV camera {DRUNK DRIVING VEHICLE RECOGNITION METHOD AND INTERVENTION SYSTEM USING CCTV CAMERA}

The following description relates to a technology for recognizing a drunk driving vehicle.

Among the methods for detecting and responding to drunk driving, existing technologies include methods of obtaining a driver's biometric information with a sensor, measuring the amount of alcohol remaining in the body, determining whether or not drunk driving, and providing a corresponding warning or physically restraining driving.

Conventional technologies are mostly methods of measuring and responding to alcohol concentration through physical contact with a driver in a vehicle. However, if the corresponding sensors are not installed in individual vehicles, or if the driver refuses or avoids the measurement, it is impossible to know whether or not drunk driving and there is no way to stop it.

Reference: KR1017453290000, KR1021446350000

Even if there is no alcohol detection device in each vehicle, it is possible to provide a method and system for recognizing the vehicle based on image data captured by a CCTV camera on the road and determining whether the recognized vehicle is driving under the influence.

It is possible to provide a method and system capable of responding to a drunk driving vehicle by providing warning information on the determined drunk driving vehicle.

A drunk driving vehicle recognition method performed by a drunk driving vehicle recognition system includes: collecting image data captured using at least one camera; Determining a drunk driving vehicle by analyzing an abnormal driving pattern of the vehicle based on object information detected from the collected image data; and controlling so that warning information is provided to the determined drunk driving vehicle or vehicles around the drunk driving vehicle.

The collecting may include detecting object information including a vehicle from the collected image data through a deep learning model for object recognition and tracking, and tracking a change in position of the recognized object information in the detected object information. steps may be included.

The collecting step may include detecting object information including vehicle information, traffic light information, and road information, assigning an ID to vehicle information among the detected object information, and obtaining single object state information and multi-object information from the detected object information. Acquiring object state information including state information, wherein the single object state information includes: vehicle movement speed, vehicle movement direction, vehicle movement speed variation, rapid acceleration or rapid deceleration of the vehicle within unit time frequency, the number of lane changes by a vehicle within a unit time, the number of lane non-compliance by a vehicle within a unit time, the number of lane changes with turn indicators not turned on within a unit time, and signal information of traffic lights, wherein the multi-object state information includes a green signal may include the duration of a stopped state, the speed of a vehicle moving at a red light, and the speed of a moving vehicle when the vehicle changes lanes.

The determining step may include calculating a drunk driving suspicion score for each vehicle through a combination of object state information for each vehicle detected from the collected image data, and calculating a drunk driving suspicion score for each vehicle based on the calculated drunk driving suspicion score. It may include determining whether a set threshold is exceeded.

In the determining step, the image data collected using the first camera is used to calculate the drunk driving similarity of the vehicle through a deep learning model for determining drunk driving, and the second camera is used for vehicles having the drunk driving similarity equal to or greater than a threshold value. and judging by using the vehicle, and adjusting a drunk driving similarity threshold for the determined vehicle.

In the determining step, the collected image data is used to infer traffic law violations and abnormal driving patterns for each vehicle through a deep learning model for determining drunk driving, and traffic laws according to the inferred traffic law violations. Classifying the strength of the violation and classifying the strength of the abnormal driving pattern according to the inferred abnormal driving pattern.

In the determining step, a drunk driving suspicion score is calculated for each vehicle by scoring the intensity of the classified traffic law violation and the intensity of the abnormal driving pattern, and the calculated drunk driving suspicion score is calculated for each vehicle through communication with other edge terminals. Delivering a drunk driving suspicion score for each vehicle, cumulatively calculating the drunk driving suspicion score for each of the delivered vehicles, and determining whether the accumulated drunk driving suspicion score for each vehicle exceeds a preset threshold steps may be included.

The providing of the warning information may include generating warning information to be delivered to a driver or pedestrian of a vehicle through at least one warning method selected from among a physical warning method, a visual warning method, and an audible warning method.

The providing of the warning information may include: 1) activating a negative or embossed rumble strip through a controller at a location on the road where the determined drunk driving vehicle passes to generate warning vibrations for the determined drunk driving vehicle; 2) a method of transmitting a warning message using a laser projection device in front of the road where the determined drunk driving vehicle passes, 3) displaying a drunk driving warning message on a traffic signboard on the road where the determined drunk driving vehicle passes A warning method of any one of a display method, 4) a method of turning on a warning light on the road where the determined drunk driving vehicle passes, and 5) a method of generating a warning message through a sound device on the road where the determined drunk driving vehicle passes. It may include the step of selecting.

The drunk driving vehicle recognition system includes: an image collection unit that collects image data captured using at least one camera; a drunk driving determination unit for determining a drunk driving vehicle through analysis of an abnormal driving pattern of the vehicle based on object information detected from the collected image data; and a warning control unit that controls warning information to be provided to the determined drunk driving vehicle or vehicles around the drunk driving vehicle.

Even without using a physical device, it is possible to automatically identify a drunk driving vehicle through analysis of the vehicle's abnormal driving pattern.

By providing a physical, visual, and auditory warning to a vehicle (driver) determined to be drunk driving, it alerts nearby drivers and pedestrians to prevent traffic accidents.

1 is a diagram for explaining the configuration of a drunk driving vehicle recognition system according to an embodiment.
2 is a block diagram illustrating a configuration of an edge terminal according to an exemplary embodiment.
3 is a flowchart illustrating a method of recognizing a drunk driving vehicle in an edge terminal according to an exemplary embodiment.
4 is a diagram for explaining an operation between edge terminals according to an embodiment.
5 and 6 are examples for explaining that warning information is provided to a drunk driving vehicle according to one embodiment.

Hereinafter, an embodiment will be described in detail with reference to the accompanying drawings.

1 is a diagram for explaining the configuration of a drunk driving vehicle recognition system according to an embodiment.

The drunk driving vehicle recognition system may include an edge terminal 100 , a server 110 and a camera unit 120 . The edge terminal 100, the server 110, and the camera unit 120 may each transmit and receive data through communication. In this case, the drunk driving vehicle recognition system may or may not include the server 110 .

The edge terminal 100 may collect image data captured by the camera unit 120 . The edge terminal 100 may detect object information from collected image data, and may determine a drunk driving vehicle by analyzing an abnormal driving pattern of the vehicle based on the detected object information. The edge terminal 100 may control warning information to be provided to the determined drunk driving vehicle or vehicles surrounding the drunk driving vehicle. For example, the edge terminal 100 may set the strength of the warning information according to the determined drinking level of the drunk driving vehicle.

The server 110 may exist on-site or at a place other than the site, receive data transmitted from the edge terminal 100, and provide warning information to the vehicle or surrounding vehicles. For example, the server 110 may control an external device (eg, a traffic light) related to a road based on a control command transmitted from the edge terminal 100 . Accordingly, a physical, visual, and audible warning is issued to a vehicle (driver) determined to be drunk driving so as to alert drivers and pedestrians around them and to prevent accidents. In addition, the server 110 may store data transmitted from the edge terminal 100, and may bring external data and deliver it to the edge terminal 100.

The camera unit 120 may include one or more cameras. The camera unit 120 is installed at a specific location to take pictures of a road including an intersection and a pedestrian path including the road. The camera unit 120 may correspond to at least one camera for taking pictures of the road, CCTV including a camera function, and the like. For example, at least one or more cameras may take images in different directions, but partially overlap each other. Alternatively, it may be photographed using a camera capable of rotating 360 degrees.

2 is a block diagram illustrating a configuration of an edge terminal according to an exemplary embodiment, and FIG. 3 is a flowchart illustrating a method of recognizing a drunk driving vehicle in an edge terminal according to an exemplary embodiment.

The processor of the edge terminal 100 may include an image collection unit 210, a drunk driving determination unit 220, and a warning control unit 230. Components of the processor may be representations of different functions performed by the processor according to control commands provided by program codes stored in the edge terminal. The processor and components of the processor may control the edge terminal to perform steps 310 to 330 included in the method for recognizing a drunk driving vehicle of FIG. 3 . At this time, the processor and components of the processor may be implemented to execute instructions according to the code of the operating system and at least one program code included in the memory.

The processor may load a program code stored in a program file for a method for recognizing a drunk driving vehicle into a memory. For example, when a program is executed in the edge terminal, the processor may control the edge terminal to load a program code from a program file into a memory under control of an operating system. At this time, each of the image collection unit 210, the drunk driving determination unit 220, and the warning control unit 230 executes a command of a corresponding part of the program code loaded into the memory to execute subsequent steps 310 to 330. There may be different functional representations of the processor for

In step 310, the image collection unit 210 may collect image data captured using at least one camera. For example, the image collecting unit 210 may simultaneously or non-simultaneously collect image data photographed using a plurality of cameras. The image collection unit 210 may input each image data to the deep learning model, or synchronize a plurality of image data to generate one image data and input the generated image data to the deep learning model. can The image collecting unit 210 may detect object information including a vehicle from the collected image data through a deep learning model for object recognition and tracking, and track a change in position of the recognized object information in the detected object information. . The image collection unit 210 detects object information including vehicle information, traffic light information, and road information, assigns an ID to vehicle information among the detected object information, and uses single object state information and multi-object state information from the detected object information. Object state information including information may be obtained.

In step 320, the drunk driving determination unit 220 may determine a drunk driving vehicle by analyzing an abnormal driving pattern of the vehicle based on object information detected from the collected image data. The drunk driving determination unit 220 calculates a drunk driving suspicion score for each vehicle through a combination of object state information for each vehicle detected from the collected image data, and calculates a drunk driving suspicion score for each vehicle based on the calculated drunk driving suspicion score. It may be determined whether or not a set threshold is exceeded. The drunk driving determination unit 220 calculates the drunk driving similarity of the vehicle through a deep learning model for drunk driving determination using the image data collected using the first camera, and the second camera for the vehicle with the drunk driving similarity equal to or greater than a threshold value. It is possible to discriminate a vehicle by using and adjust a drunk driving similarity threshold for the discriminated vehicle. The drunk driving determination unit 220 infers traffic law violations and abnormal driving patterns for each vehicle through a deep learning model for drunk driving discrimination based on the collected image data, and infers traffic laws according to the inferred traffic law violations. The intensity of the violation may be classified, and the intensity of the abnormal driving pattern may be classified according to the inferred abnormal driving pattern. The drunk driving determination unit 220 calculates a drunk driving suspicion score for each vehicle by scoring the intensity of the classified traffic law violation and the intensity of the abnormal driving pattern, and determines each vehicle calculated through communication with other edge terminals. The drunk driving suspicion score for each vehicle is transferred, the drunk driving suspicion score for each vehicle is accumulated, and it is determined whether the accumulated drunk driving suspicion score for each vehicle exceeds a preset threshold. .

In step 330, the warning control unit 230 may control warning information to be provided to the determined drunk driving vehicle or vehicles around the drunk driving vehicle. The warning control unit 230 may generate warning information to be delivered to a driver or pedestrian of a vehicle through at least one warning method selected from among a physical warning method, a visual warning method, and an audible warning method. The warning control unit 230 is configured to: 1) generate a warning vibration for the determined drunk driving vehicle by activating a negative or embossed rumble strip through a controller at the location of the road where the determined drunk driving vehicle passes; 2 ) A method of delivering a warning message using a laser projection device in front of the road where the determined drunk driving vehicle passes, 3) A method of displaying a drunk driving warning message on a traffic signboard on the road where the determined drunk driving vehicle passes, 4) Discrimination One warning method can be selected from among a method of turning on a warning light on the road where the drunk driving vehicle passes, and 5) a method of generating a warning message through a sound device on the road where the determined drunk driving vehicle passes.

4 is a diagram for explaining an operation between edge terminals according to an embodiment.

Image data may be obtained as the road is photographed by CCTV. Road image data captured by CCTV in real time may be collected, and the captured road image data may be accumulated for a predetermined period. In this case, the CCTV may include at least one camera, and may additionally include sensor devices such as laser, lidar, and radar in addition to the camera. Alternatively, the camera and sensor device may be separately installed outside the CCTV and transmit/receive data through wireless communication.

In FIG. 4 , an operation of transmitting/receiving data through wireless communication between each edge terminal when determining a drunk driving vehicle in an edge terminal will be described. For example, a CCTV, an edge terminal, and additional devices (eg, a controller, a laser projector, etc.) may exist on a road where a vehicle passes. At this time, each of the edge terminal 1 (101) and the edge terminal 2 (102) is installed in a specific location, and analyzes data generated in a specific location (specific area) to determine a drunk driving vehicle.

Since edge device 1 101 and edge device 2 102 perform the same operation, edge device 1 101 will be described as an example. Edge terminal 1 (101) may collect video data (input video) captured through CCTV (410). At this time, a vehicle driving on the road may be included in the image data (input image).

Edge terminal 1 (101) may recognize object information from the collected image data (402). Edge terminal 1 (101) can detect and track object information including vehicle information (vehicle number), traffic light information (vehicle traffic light, pedestrian traffic light, etc.), road information (road, lane, center line, crosswalk, etc.) there is. At this time, the edge terminal 1 (101) may separately recognize static object information including lanes, roads, traffic lights, etc., and dynamic object information including vehicles, pedestrians, and the like.

In detail, the edge terminal 1 101 may detect object information including a vehicle through a deep learning model for object recognition and tracking from collected image data. Edge terminal 1 (101) may track a location change (location tracking) of detected object information. Edge terminal 1 101 may obtain object state information including single object state information and multi-object state information from the sensed object information. For example, the edge terminal 1 (101) is the vehicle's moving speed, the vehicle's moving direction (zigzag, reverse driving, etc.), the amount of change in the vehicle's moving speed, the frequency of rapid acceleration or deceleration of the vehicle within a unit time, and the vehicle within a unit time. object state information including the number of lane changes, the number of lane non-compliance of vehicles within unit time, the number of lane changes without direction indicators turned on within unit time, and signal information of traffic lights (red, yellow, green, blinking, etc.) there is.

Edge terminal 1 101 may determine a drunk driving vehicle by analyzing an abnormal driving pattern of the vehicle based on object information detected from the collected image data. Edge terminal 1 (101) can determine a drunk driving vehicle based on a rule, and can determine a drunk driving vehicle based on deep learning.

As an example, a rule-based drunk driving determination operation will be described. Edge terminal 1 (101) calculates a drunk driving suspicion score for each vehicle through a combination of object state information for each vehicle detected from the collected image data, and the calculated drunk driving suspicion score for each vehicle is preset. It can be determined whether or not the threshold is exceeded. For example, edge terminal 1 101 may calculate single object information and multi-object state information. At this time, the single object state information includes the vehicle's moving speed, the vehicle's moving direction (zigzag, reverse driving, etc.), the amount of change in the vehicle's moving speed, the frequency of sudden acceleration or deceleration of the vehicle within a unit time, and the number of lane changes of the vehicle within a unit time. , duration of vehicle and lane crossing within unit time (time of driving while stepping on the lane), number of lane changes without direction indicators turned on within unit time, signal information of traffic lights (red, yellow, green, blinking, etc.) included can do. The multi-object state information may include the duration of a green signal but not moving and a stopped state, a red signal but speed information of a vehicle moving while violating a signal, and the speed of a moving vehicle when changing lanes.

In other words, edge terminal 1 101 may calculate a drunk driving suspicion score for a vehicle through a combination of state information of a corresponding object from image data. Edge terminal 1 101 may normalize object state information. For example, edge terminal 1 101 may normalize object state information to a value between 0 and 1. At this time, the edge terminal 1 101 may assign a weight to more dangerous object state information according to specific conditions such as road conditions or time information/weather information. Edge terminal 1 101 may calculate a total score of object state information weighted per unit time. Edge terminal 1 101 may determine the vehicle as a drunk driving vehicle when the result of calculating the state information of the object exceeds a preset threshold. Alternatively, a score may be designated for each attribute information for each single object state information or multi-object state information. For example, in the edge terminal 1 ( 101 ), the number of lane changes without direction indicators turned on within a unit time may be designated as 3 points per time. Edge terminal 1 101 may determine the vehicle as a drunk driving vehicle when the result of calculating the state information of the object exceeds a preset threshold.

As another example, a drunk driving determination operation based on deep learning will be described. Edge terminal 1 101 may generate a unique vehicle ID for vehicle information among detected object information, and may assign the generated unique vehicle ID to the vehicle information. Edge terminal 1 (10) may discriminate the vehicle information to which the unique vehicle ID is assigned (430). For example, the edge terminal 1 101 may calculate the drunk driving similarity of the vehicle using the image data collected using the first camera through a deep learning model for determining drunk driving. Edge terminal 1 (101) may use the second camera to discriminate against a vehicle whose similarity to drunk driving is greater than or equal to a threshold value. Edge terminal 1 (101) may adjust a drunk driving similarity threshold with respect to the vehicle judged by the second camera. For example, edge terminal 1 101 may adjust the drunk driving similarity threshold equal to or lower than a preset threshold. Edge terminal 1 101 may determine a drunk driving vehicle when the drunk driving similarity exceeds a threshold value.

More specifically, edge terminal 1 (101) can infer whether traffic laws have been violated and abnormal driving patterns for each vehicle through a deep learning model for determining drunk driving using the collected image data (440, 450). . For example, a data set of situation information (normal situation, violation situation, imitation situation, etc.) for determining drunk driving may be trained in a deep learning model in advance. Collected image data may be input to such a deep learning model. Through the deep learning model, it is possible to determine whether the video data has violated or not violated traffic laws, and whether it is an abnormal driving pattern or not. In addition, the type of traffic law violated can also be determined.

At this time, using each deep learning model, traffic law violations may be inferred, and abnormal driving patterns may be inferred. Alternatively, traffic law violations and abnormal driving patterns may be inferred using one deep learning model. In addition, as whether or not traffic law violation is inferred, the intensity of the traffic law violation may be classified using a deep learning model (441). As the abnormal driving pattern is inferred, the intensity of the abnormal driving pattern may be classified using a deep learning model (442). The severity classification of traffic law violations refers to situations in which greater penalty points are awarded in violation of traffic laws (eg, speed violation exceeding 20 km/h vs. exceeding 60 km/h). Abnormal driving pattern inference does not violate traffic laws, but means anomaly detection, a pattern that is not commonly seen. For example, most vehicles drive only in a straight line, but in the case of a very small number of vehicles, reverse driving, zigzag driving, stopping in the middle of the road, and waiting without starting at a green light may fall under abnormal driving pattern inference.

The edge terminal 1 (101) may calculate the drunk driving suspicion score for each vehicle by scoring the intensity of the classified traffic law violation and the intensity of the abnormal driving pattern (460). In this case, scores may be set according to the strength of the inferred traffic law violation and the strength of the abnormal driving pattern. Edge terminal 1 (101) may transmit the calculated drunk driving suspicion scores for each vehicle to edge terminal 2 (102) through communication with other edge terminals (480). At this time, the drunk driving suspicion scores for each vehicle calculated through the D2D communication unit 480 may be shared. Edge terminal 1 (101) may cumulatively calculate drunk driving suspicion scores for each vehicle (460). In other words, the edge terminal 1 (101) may accumulate drunk driving suspicion scores of each vehicle calculated within the area of the edge terminal 1 (101). In addition, edge terminal 1 (101) may accumulate drunk driving suspicion scores of each vehicle received from other edge terminals. Edge terminal 1 (101) may determine whether the accumulated drunk driving suspicion scores for each vehicle exceed a preset threshold (470). If the accumulated drunk driving suspicion score for each vehicle exceeds a preset threshold, the edge terminal 1 101 may determine a vehicle exceeding the threshold as a drunk driving vehicle.

5 and 6 are examples for explaining that warning information is provided to a drunk driving vehicle according to one embodiment.

Warning information may be provided to a driver or pedestrian of a vehicle through one or more warning methods selected from among a physical warning method, a visual warning method, and an audible warning method. Accordingly, a route is given to a driver determined to be drunk driving through a physical warning method, a visual warning method, or an audible warning method to alert drivers and pedestrians around them and to prevent accidents.

The physical warning method is a method of forming a vibration in the vehicle through the operation of the negative/positive conversion device on the road. Referring to FIG. 5 , a concave or embossed rumble strip may be activated through a controller at a location on a road where a drunk driving vehicle passes, and a warning vibration may be generated for the drunk driving vehicle.

The visual warning method is a method of delivering a message using a laser projecting device in front of the driver or turning on/off a traffic signboard or a warning light (next to a camera device) to notify the driver. In the case of a laser projection device, a laser image may be projected in real time in front of the vehicle by tracking the location of the vehicle. Referring to FIG. 6 , a warning message may be delivered using a laser projecting device to a road ahead of a drunk driving vehicle. For example, a warning message may be delivered to a drunk driving vehicle using a laser projecting device including a 3-axis rotary joint. The laser projection device may project a warning message onto the road within a preset range based on the location where the laser projection device is installed. When the laser projecting device approaches a range capable of projecting a drunk driving vehicle, a warning message may be delivered in the form of a projected image. In this case, a warning message may be projected and displayed on the road based on the speed information of the drunk driving vehicle. As another example, a drunk driving warning message may be displayed on a traffic electronic board on a road where a drunk driving vehicle passes. As another example, a warning light may be turned on on a road where a drunk driving vehicle passes. For example, the warning light may be turned on in various ways, such as a blinking form or a form of lighting different colors.

The audible warning method is a method of generating a warning sound using a sound device near the driver (eg, a speaker for the visually impaired or a dedicated alarm device next to a camera). A warning message may be generated through a sound device on a road where a drunk driving vehicle passes. For example, the size of the warning message may be generated differently depending on the drinking level of the drunk driving vehicle.

The devices described above may be implemented as hardware components, software components, and/or a combination of hardware components and software components. For example, devices and components described in the embodiments may include, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA) , a programmable logic unit (PLU), microprocessor, or any other device capable of executing and responding to instructions. A processing device may run an operating system (OS) and one or more software applications running on the operating system. A processing device may also access, store, manipulate, process, and generate data in response to execution of software. For convenience of understanding, there are cases in which one processing device is used, but those skilled in the art will understand that the processing device includes a plurality of processing elements and/or a plurality of types of processing elements. It can be seen that it can include. For example, a processing device may include a plurality of processors or a processor and a controller. Other processing configurations are also possible, such as parallel processors.

Software may include a computer program, code, instructions, or a combination of one or more of the foregoing, which configures a processing device to operate as desired or processes independently or collectively. The device can be commanded. Software and/or data may be any tangible machine, component, physical device, virtual equipment, computer storage medium or device, intended to be interpreted by or provide instructions or data to a processing device. can be embodied in Software may be distributed on networked computer systems and stored or executed in a distributed manner. Software and data may be stored on one or more computer readable media.

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. Program commands recorded on the medium may be specially designed and configured for the embodiment or may be known and usable to those skilled in computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floptical disks. - includes hardware devices specially configured to store and execute program instructions, such as magneto-optical media, and ROM, RAM, flash memory, and the like. Examples of program instructions include high-level language codes that can be executed by a computer using an interpreter, as well as machine language codes such as those produced by a compiler.

As described above, although the embodiments have been described with limited examples and drawings, those skilled in the art can make various modifications and variations from the above description. For example, the described techniques may be performed in an order different from the method described, and/or components of the described system, structure, device, circuit, etc. may be combined or combined in a different form than the method described, or other components may be used. Or even if it is replaced or substituted by equivalents, appropriate results can be achieved.

Therefore, other implementations, other embodiments, and equivalents of the claims are within the scope of the following claims.

Claims (12)

In the drunk driving vehicle recognition method performed by the edge terminal,
Collecting image data captured using at least one camera;
Determining a drunk driving vehicle based on object information about the vehicle detected from the collected image data and object state information about the vehicle; and
Controlling so that warning information is provided to the determined drunk driving vehicle or vehicles around the drunk driving vehicle
including,
The step of determining the drunk driving vehicle,
calculating a drunk driving suspicion score for the vehicle by combining object information about the vehicle and object state information about the vehicle;
Transmitting the calculated drunk driving suspicion score to another edge terminal communicating with the edge terminal;
Receiving a drunk driving suspicion score for the vehicle calculated by the other edge terminal from the other edge terminal;
accumulating the drunk driving suspicion score calculated by the edge terminal and the drunk driving suspicion score received from the other edge terminal; and
Determining whether the vehicle is a drunk driving vehicle based on the accumulated drunk driving suspicion score
Including, drunk driving vehicle recognition method. delete delete According to claim 1,
The collecting step is
Detecting object information including a vehicle from the collected image data through a deep learning model for object recognition and tracking, and tracking a positional change of the recognized object information in the detected object information.
A drunk driving vehicle recognition method comprising a. According to claim 4,
The collecting step is
An object that detects object information including vehicle information, traffic light information, and road information, assigns an ID to vehicle information among the detected object information, and includes single object state information and multi-object state information from the sensed object information Steps to acquire status information
including,
The single object state information includes the vehicle's moving speed, the vehicle's moving direction, the amount of change in the vehicle's moving speed, the frequency of sudden acceleration or deceleration of the vehicle within a unit time, the number of lane changes of the vehicle within a unit time, and the speed of the vehicle within a unit time. It includes the number of lane non-compliance, the number of lane changes with turn indicators not turned on within a unit time, and signal information of traffic lights,
The multi-object state information includes the duration of a state in which the vehicle is stopped at a green light, the speed of a vehicle moving at a red light, and the speed of a vehicle moving when the vehicle changes lanes.
A drunk driving vehicle recognition method, characterized in that. According to claim 1,
Determining whether the vehicle is a drunk driving vehicle,
Determining whether the accumulated drunk driving suspicion score exceeds a preset threshold
A drunk driving vehicle recognition method comprising a. According to claim 1,
The step of determining the drunk driving vehicle,
The image data collected using the first camera is used to calculate the drunk driving similarity of the vehicle through a deep learning model for determining drunk driving, and the second camera is used to judge the vehicle for which the drunk driving similarity is higher than a threshold value. and adjusting a drunk driving similarity threshold for the judged vehicle.
A drunk driving vehicle recognition method comprising a. According to claim 1,
The step of determining the drunk driving vehicle,
The collected image data is used to infer traffic law violations and abnormal driving patterns for the vehicle through a deep learning model for determining drunk driving, classify the intensity of traffic law violations according to the inferred traffic law violations, Classifying abnormal driving pattern intensity according to the inferred abnormal driving pattern
A drunk driving vehicle recognition method comprising a. According to claim 8,
The step of calculating the drunk driving suspicion score,
Calculating a drunk driving suspicion score for the vehicle by further using the classified traffic law violation intensity and abnormal driving pattern intensity
A drunk driving vehicle recognition method comprising a. According to claim 1,
Providing the warning information,
Generating warning information to be delivered to a driver or pedestrian of a vehicle through at least one warning method selected from among a physical warning method, a visual warning method, and an audible warning method.
A drunk driving vehicle recognition method comprising a. According to claim 10,
Providing the warning information,
1) A method of generating a warning vibration for the determined drunk driving vehicle by activating an intaglio or embossed rumble strip through a controller at a road location where the determined drunk driving vehicle passes, 2) a method of generating a warning vibration for the determined drunk driving vehicle; A method of delivering a warning message using a laser projection device in front of the road where the drunk driving vehicle passes, 3) A method of displaying a drunk driving warning message on a traffic electronic board on the road where the determined drunk driving vehicle passes, 4) A method of displaying a drunk driving warning message on the road where the determined drunk driving vehicle passes, 4) The determined drunk driving warning message Selecting one of the warning methods among a method of turning on a warning light on the road where the drunk driving vehicle passes, and 5) a method of generating a warning message through a sound device on the road where the determined drunk driving vehicle passes.
A drunk driving vehicle recognition method comprising a. In the edge terminal,
An image collection unit for collecting image data captured using at least one camera;
a drunk driving determination unit for determining a drunk driving vehicle through analysis of an abnormal driving pattern of the vehicle based on object information detected from the collected image data; and
Warning control unit for controlling the provision of warning information to the determined drunk driving vehicle or vehicles around the drunk driving vehicle
including,
The determination unit,
Combining object information about the vehicle and object state information about the vehicle to calculate a drunk driving suspicion score for the vehicle,
Transmitting the calculated drunk driving suspicion score to another edge terminal communicating with the edge terminal;
Receiving a drunk driving suspicion score for the vehicle calculated by the other edge terminal from the other edge terminal,
Accumulating the drunk driving suspicion score calculated by the edge terminal and the drunk driving suspicion score received from the other edge terminal,
Determining whether the vehicle is a drunk driving vehicle based on the accumulated drunk driving suspicion score
Edge terminal characterized by.

Images